1
0
mirror of https://github.com/golang/go synced 2024-10-07 01:21:21 -06:00
go/src/pkg/reflect/value.go

2524 lines
71 KiB
Go
Raw Normal View History

// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package reflect
import (
"math"
"runtime"
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
"strconv"
"unsafe"
)
const bigEndian = false // can be smarter if we find a big-endian machine
const ptrSize = unsafe.Sizeof((*byte)(nil))
const cannotSet = "cannot set value obtained from unexported struct field"
// TODO: This will have to go away when
// the new gc goes in.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
func memmove(adst, asrc unsafe.Pointer, n uintptr) {
dst := uintptr(adst)
src := uintptr(asrc)
switch {
case src < dst && src+n > dst:
// byte copy backward
// careful: i is unsigned
for i := n; i > 0; {
i--
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
*(*byte)(unsafe.Pointer(dst + i)) = *(*byte)(unsafe.Pointer(src + i))
}
case (n|src|dst)&(ptrSize-1) != 0:
// byte copy forward
for i := uintptr(0); i < n; i++ {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
*(*byte)(unsafe.Pointer(dst + i)) = *(*byte)(unsafe.Pointer(src + i))
}
default:
// word copy forward
for i := uintptr(0); i < n; i += ptrSize {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
*(*uintptr)(unsafe.Pointer(dst + i)) = *(*uintptr)(unsafe.Pointer(src + i))
}
}
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// Value is the reflection interface to a Go value.
//
// Not all methods apply to all kinds of values. Restrictions,
// if any, are noted in the documentation for each method.
// Use the Kind method to find out the kind of value before
// calling kind-specific methods. Calling a method
// inappropriate to the kind of type causes a run time panic.
//
// The zero Value represents no value.
// Its IsValid method returns false, its Kind method returns Invalid,
// its String method returns "<invalid Value>", and all other methods panic.
// Most functions and methods never return an invalid value.
// If one does, its documentation states the conditions explicitly.
//
// A Value can be used concurrently by multiple goroutines provided that
// the underlying Go value can be used concurrently for the equivalent
// direct operations.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
type Value struct {
// typ holds the type of the value represented by a Value.
typ *rtype
// val holds the 1-word representation of the value.
// If flag's flagIndir bit is set, then val is a pointer to the data.
// Otherwise val is a word holding the actual data.
// When the data is smaller than a word, it begins at
// the first byte (in the memory address sense) of val.
// We use unsafe.Pointer so that the garbage collector
// knows that val could be a pointer.
val unsafe.Pointer
// flag holds metadata about the value.
// The lowest bits are flag bits:
// - flagRO: obtained via unexported field, so read-only
// - flagIndir: val holds a pointer to the data
// - flagAddr: v.CanAddr is true (implies flagIndir)
// - flagMethod: v is a method value.
// The next five bits give the Kind of the value.
// This repeats typ.Kind() except for method values.
// The remaining 23+ bits give a method number for method values.
// If flag.kind() != Func, code can assume that flagMethod is unset.
// If typ.size > ptrSize, code can assume that flagIndir is set.
flag
// A method value represents a curried method invocation
// like r.Read for some receiver r. The typ+val+flag bits describe
// the receiver r, but the flag's Kind bits say Func (methods are
// functions), and the top bits of the flag give the method number
// in r's type's method table.
}
type flag uintptr
const (
flagRO flag = 1 << iota
flagIndir
flagAddr
flagMethod
flagKindShift = iota
flagKindWidth = 5 // there are 27 kinds
flagKindMask flag = 1<<flagKindWidth - 1
flagMethodShift = flagKindShift + flagKindWidth
)
func (f flag) kind() Kind {
return Kind((f >> flagKindShift) & flagKindMask)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// A ValueError occurs when a Value method is invoked on
// a Value that does not support it. Such cases are documented
// in the description of each method.
type ValueError struct {
Method string
Kind Kind
}
func (e *ValueError) Error() string {
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
if e.Kind == 0 {
return "reflect: call of " + e.Method + " on zero Value"
}
return "reflect: call of " + e.Method + " on " + e.Kind.String() + " Value"
}
// methodName returns the name of the calling method,
// assumed to be two stack frames above.
func methodName() string {
pc, _, _, _ := runtime.Caller(2)
f := runtime.FuncForPC(pc)
if f == nil {
return "unknown method"
}
return f.Name()
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// An iword is the word that would be stored in an
// interface to represent a given value v. Specifically, if v is
// bigger than a pointer, its word is a pointer to v's data.
// Otherwise, its word holds the data stored
// in its leading bytes (so is not a pointer).
// Because the value sometimes holds a pointer, we use
// unsafe.Pointer to represent it, so that if iword appears
// in a struct, the garbage collector knows that might be
// a pointer.
type iword unsafe.Pointer
func (v Value) iword() iword {
if v.flag&flagIndir != 0 && v.typ.size <= ptrSize {
// Have indirect but want direct word.
return loadIword(v.val, v.typ.size)
}
return iword(v.val)
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// loadIword loads n bytes at p from memory into an iword.
func loadIword(p unsafe.Pointer, n uintptr) iword {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Run the copy ourselves instead of calling memmove
// to avoid moving w to the heap.
var w iword
switch n {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
default:
panic("reflect: internal error: loadIword of " + strconv.Itoa(int(n)) + "-byte value")
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case 0:
case 1:
*(*uint8)(unsafe.Pointer(&w)) = *(*uint8)(p)
case 2:
*(*uint16)(unsafe.Pointer(&w)) = *(*uint16)(p)
case 3:
*(*[3]byte)(unsafe.Pointer(&w)) = *(*[3]byte)(p)
case 4:
*(*uint32)(unsafe.Pointer(&w)) = *(*uint32)(p)
case 5:
*(*[5]byte)(unsafe.Pointer(&w)) = *(*[5]byte)(p)
case 6:
*(*[6]byte)(unsafe.Pointer(&w)) = *(*[6]byte)(p)
case 7:
*(*[7]byte)(unsafe.Pointer(&w)) = *(*[7]byte)(p)
case 8:
*(*uint64)(unsafe.Pointer(&w)) = *(*uint64)(p)
}
return w
}
// storeIword stores n bytes from w into p.
func storeIword(p unsafe.Pointer, w iword, n uintptr) {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Run the copy ourselves instead of calling memmove
// to avoid moving w to the heap.
switch n {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
default:
panic("reflect: internal error: storeIword of " + strconv.Itoa(int(n)) + "-byte value")
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case 0:
case 1:
*(*uint8)(p) = *(*uint8)(unsafe.Pointer(&w))
case 2:
*(*uint16)(p) = *(*uint16)(unsafe.Pointer(&w))
case 3:
*(*[3]byte)(p) = *(*[3]byte)(unsafe.Pointer(&w))
case 4:
*(*uint32)(p) = *(*uint32)(unsafe.Pointer(&w))
case 5:
*(*[5]byte)(p) = *(*[5]byte)(unsafe.Pointer(&w))
case 6:
*(*[6]byte)(p) = *(*[6]byte)(unsafe.Pointer(&w))
case 7:
*(*[7]byte)(p) = *(*[7]byte)(unsafe.Pointer(&w))
case 8:
*(*uint64)(p) = *(*uint64)(unsafe.Pointer(&w))
}
}
// emptyInterface is the header for an interface{} value.
type emptyInterface struct {
typ *rtype
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
word iword
}
// nonEmptyInterface is the header for a interface value with methods.
type nonEmptyInterface struct {
// see ../runtime/iface.c:/Itab
itab *struct {
ityp *rtype // static interface type
typ *rtype // dynamic concrete type
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
link unsafe.Pointer
bad int32
unused int32
fun [100000]unsafe.Pointer // method table
}
word iword
}
// mustBe panics if f's kind is not expected.
// Making this a method on flag instead of on Value
// (and embedding flag in Value) means that we can write
// the very clear v.mustBe(Bool) and have it compile into
// v.flag.mustBe(Bool), which will only bother to copy the
// single important word for the receiver.
func (f flag) mustBe(expected Kind) {
k := f.kind()
if k != expected {
panic(&ValueError{methodName(), k})
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
}
// mustBeExported panics if f records that the value was obtained using
// an unexported field.
func (f flag) mustBeExported() {
if f == 0 {
panic(&ValueError{methodName(), 0})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
if f&flagRO != 0 {
panic("reflect: " + methodName() + " using value obtained using unexported field")
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
}
// mustBeAssignable panics if f records that the value is not assignable,
// which is to say that either it was obtained using an unexported field
// or it is not addressable.
func (f flag) mustBeAssignable() {
if f == 0 {
panic(&ValueError{methodName(), Invalid})
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Assignable if addressable and not read-only.
if f&flagRO != 0 {
panic("reflect: " + methodName() + " using value obtained using unexported field")
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
if f&flagAddr == 0 {
panic("reflect: " + methodName() + " using unaddressable value")
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
}
// Addr returns a pointer value representing the address of v.
// It panics if CanAddr() returns false.
// Addr is typically used to obtain a pointer to a struct field
// or slice element in order to call a method that requires a
// pointer receiver.
func (v Value) Addr() Value {
if v.flag&flagAddr == 0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect.Value.Addr of unaddressable value")
}
return Value{v.typ.ptrTo(), v.val, (v.flag & flagRO) | flag(Ptr)<<flagKindShift}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Bool returns v's underlying value.
// It panics if v's kind is not Bool.
func (v Value) Bool() bool {
v.mustBe(Bool)
if v.flag&flagIndir != 0 {
return *(*bool)(v.val)
}
return *(*bool)(unsafe.Pointer(&v.val))
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Bytes returns v's underlying value.
// It panics if v's underlying value is not a slice of bytes.
func (v Value) Bytes() []byte {
v.mustBe(Slice)
if v.typ.Elem().Kind() != Uint8 {
panic("reflect.Value.Bytes of non-byte slice")
}
// Slice is always bigger than a word; assume flagIndir.
return *(*[]byte)(v.val)
}
// runes returns v's underlying value.
// It panics if v's underlying value is not a slice of runes (int32s).
func (v Value) runes() []rune {
v.mustBe(Slice)
if v.typ.Elem().Kind() != Int32 {
panic("reflect.Value.Bytes of non-rune slice")
}
// Slice is always bigger than a word; assume flagIndir.
return *(*[]rune)(v.val)
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// CanAddr returns true if the value's address can be obtained with Addr.
// Such values are called addressable. A value is addressable if it is
// an element of a slice, an element of an addressable array,
// a field of an addressable struct, or the result of dereferencing a pointer.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// If CanAddr returns false, calling Addr will panic.
func (v Value) CanAddr() bool {
return v.flag&flagAddr != 0
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// CanSet returns true if the value of v can be changed.
// A Value can be changed only if it is addressable and was not
// obtained by the use of unexported struct fields.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// If CanSet returns false, calling Set or any type-specific
// setter (e.g., SetBool, SetInt64) will panic.
func (v Value) CanSet() bool {
return v.flag&(flagAddr|flagRO) == flagAddr
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Call calls the function v with the input arguments in.
// For example, if len(in) == 3, v.Call(in) represents the Go call v(in[0], in[1], in[2]).
// Call panics if v's Kind is not Func.
// It returns the output results as Values.
// As in Go, each input argument must be assignable to the
// type of the function's corresponding input parameter.
// If v is a variadic function, Call creates the variadic slice parameter
// itself, copying in the corresponding values.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Call(in []Value) []Value {
v.mustBe(Func)
v.mustBeExported()
return v.call("Call", in)
}
// CallSlice calls the variadic function v with the input arguments in,
// assigning the slice in[len(in)-1] to v's final variadic argument.
// For example, if len(in) == 3, v.Call(in) represents the Go call v(in[0], in[1], in[2]...).
// Call panics if v's Kind is not Func or if v is not variadic.
// It returns the output results as Values.
// As in Go, each input argument must be assignable to the
// type of the function's corresponding input parameter.
func (v Value) CallSlice(in []Value) []Value {
v.mustBe(Func)
v.mustBeExported()
return v.call("CallSlice", in)
}
func (v Value) call(op string, in []Value) []Value {
// Get function pointer, type.
t := v.typ
var (
fn unsafe.Pointer
rcvr iword
)
if v.flag&flagMethod != 0 {
t, fn, rcvr = methodReceiver(op, v, int(v.flag)>>flagMethodShift)
} else if v.flag&flagIndir != 0 {
fn = *(*unsafe.Pointer)(v.val)
} else {
fn = v.val
}
if fn == nil {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect.Value.Call: call of nil function")
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
isSlice := op == "CallSlice"
n := t.NumIn()
if isSlice {
if !t.IsVariadic() {
panic("reflect: CallSlice of non-variadic function")
}
if len(in) < n {
panic("reflect: CallSlice with too few input arguments")
}
if len(in) > n {
panic("reflect: CallSlice with too many input arguments")
}
} else {
if t.IsVariadic() {
n--
}
if len(in) < n {
panic("reflect: Call with too few input arguments")
}
if !t.IsVariadic() && len(in) > n {
panic("reflect: Call with too many input arguments")
}
}
for _, x := range in {
if x.Kind() == Invalid {
panic("reflect: " + op + " using zero Value argument")
}
}
for i := 0; i < n; i++ {
if xt, targ := in[i].Type(), t.In(i); !xt.AssignableTo(targ) {
panic("reflect: " + op + " using " + xt.String() + " as type " + targ.String())
}
}
if !isSlice && t.IsVariadic() {
// prepare slice for remaining values
m := len(in) - n
slice := MakeSlice(t.In(n), m, m)
elem := t.In(n).Elem()
for i := 0; i < m; i++ {
x := in[n+i]
if xt := x.Type(); !xt.AssignableTo(elem) {
panic("reflect: cannot use " + xt.String() + " as type " + elem.String() + " in " + op)
}
slice.Index(i).Set(x)
}
origIn := in
in = make([]Value, n+1)
copy(in[:n], origIn)
in[n] = slice
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
nin := len(in)
if nin != t.NumIn() {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect.Value.Call: wrong argument count")
}
nout := t.NumOut()
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// Compute arg size & allocate.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// This computation is 5g/6g/8g-dependent
// and probably wrong for gccgo, but so
// is most of this function.
size, _, _, _ := frameSize(t, v.flag&flagMethod != 0)
// Copy into args.
//
// TODO(rsc): This will need to be updated for any new garbage collector.
// For now make everything look like a pointer by allocating
// a []unsafe.Pointer.
args := make([]unsafe.Pointer, size/ptrSize)
ptr := uintptr(unsafe.Pointer(&args[0]))
off := uintptr(0)
if v.flag&flagMethod != 0 {
// Hard-wired first argument.
*(*iword)(unsafe.Pointer(ptr)) = rcvr
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
off = ptrSize
}
for i, v := range in {
v.mustBeExported()
targ := t.In(i).(*rtype)
a := uintptr(targ.align)
off = (off + a - 1) &^ (a - 1)
n := targ.size
addr := unsafe.Pointer(ptr + off)
v = v.assignTo("reflect.Value.Call", targ, (*interface{})(addr))
if v.flag&flagIndir == 0 {
storeIword(addr, iword(v.val), n)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
} else {
memmove(addr, v.val, n)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
off += n
}
off = (off + ptrSize - 1) &^ (ptrSize - 1)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Call.
call(fn, unsafe.Pointer(ptr), uint32(size))
// Copy return values out of args.
//
// TODO(rsc): revisit like above.
ret := make([]Value, nout)
for i := 0; i < nout; i++ {
tv := t.Out(i)
a := uintptr(tv.Align())
off = (off + a - 1) &^ (a - 1)
fl := flagIndir | flag(tv.Kind())<<flagKindShift
ret[i] = Value{tv.common(), unsafe.Pointer(ptr + off), fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
off += tv.Size()
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
return ret
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// callReflect is the call implementation used by a function
// returned by MakeFunc. In many ways it is the opposite of the
// method Value.call above. The method above converts a call using Values
// into a call of a function with a concrete argument frame, while
// callReflect converts a call of a function with a concrete argument
// frame into a call using Values.
// It is in this file so that it can be next to the call method above.
// The remainder of the MakeFunc implementation is in makefunc.go.
func callReflect(ctxt *makeFuncImpl, frame unsafe.Pointer) {
ftyp := ctxt.typ
f := ctxt.fn
// Copy argument frame into Values.
ptr := frame
off := uintptr(0)
in := make([]Value, 0, len(ftyp.in))
for _, arg := range ftyp.in {
typ := arg
off += -off & uintptr(typ.align-1)
v := Value{typ, nil, flag(typ.Kind()) << flagKindShift}
if typ.size <= ptrSize {
// value fits in word.
v.val = unsafe.Pointer(loadIword(unsafe.Pointer(uintptr(ptr)+off), typ.size))
} else {
// value does not fit in word.
// Must make a copy, because f might keep a reference to it,
// and we cannot let f keep a reference to the stack frame
// after this function returns, not even a read-only reference.
v.val = unsafe_New(typ)
memmove(v.val, unsafe.Pointer(uintptr(ptr)+off), typ.size)
v.flag |= flagIndir
}
in = append(in, v)
off += typ.size
}
// Call underlying function.
out := f(in)
if len(out) != len(ftyp.out) {
panic("reflect: wrong return count from function created by MakeFunc")
}
// Copy results back into argument frame.
if len(ftyp.out) > 0 {
off += -off & (ptrSize - 1)
for i, arg := range ftyp.out {
typ := arg
v := out[i]
if v.typ != typ {
panic("reflect: function created by MakeFunc using " + funcName(f) +
" returned wrong type: have " +
out[i].typ.String() + " for " + typ.String())
}
if v.flag&flagRO != 0 {
panic("reflect: function created by MakeFunc using " + funcName(f) +
" returned value obtained from unexported field")
}
off += -off & uintptr(typ.align-1)
addr := unsafe.Pointer(uintptr(ptr) + off)
if v.flag&flagIndir == 0 {
storeIword(addr, iword(v.val), typ.size)
} else {
memmove(addr, v.val, typ.size)
}
off += typ.size
}
}
}
// methodReceiver returns information about the receiver
// described by v. The Value v may or may not have the
// flagMethod bit set, so the kind cached in v.flag should
// not be used.
func methodReceiver(op string, v Value, methodIndex int) (t *rtype, fn unsafe.Pointer, rcvr iword) {
i := methodIndex
if v.typ.Kind() == Interface {
tt := (*interfaceType)(unsafe.Pointer(v.typ))
if i < 0 || i >= len(tt.methods) {
panic("reflect: internal error: invalid method index")
}
m := &tt.methods[i]
if m.pkgPath != nil {
panic("reflect: " + op + " of unexported method")
}
t = m.typ
iface := (*nonEmptyInterface)(v.val)
if iface.itab == nil {
panic("reflect: " + op + " of method on nil interface value")
}
fn = unsafe.Pointer(&iface.itab.fun[i])
rcvr = iface.word
} else {
ut := v.typ.uncommon()
if ut == nil || i < 0 || i >= len(ut.methods) {
panic("reflect: internal error: invalid method index")
}
m := &ut.methods[i]
if m.pkgPath != nil {
panic("reflect: " + op + " of unexported method")
}
fn = unsafe.Pointer(&m.ifn)
t = m.mtyp
rcvr = v.iword()
}
return
}
// align returns the result of rounding x up to a multiple of n.
// n must be a power of two.
func align(x, n uintptr) uintptr {
return (x + n - 1) &^ (n - 1)
}
// frameSize returns the sizes of the argument and result frame
// for a function of the given type. The rcvr bool specifies whether
// a one-word receiver should be included in the total.
func frameSize(t *rtype, rcvr bool) (total, in, outOffset, out uintptr) {
if rcvr {
// extra word for receiver interface word
total += ptrSize
}
nin := t.NumIn()
in = -total
for i := 0; i < nin; i++ {
tv := t.In(i)
total = align(total, uintptr(tv.Align()))
total += tv.Size()
}
in += total
total = align(total, ptrSize)
nout := t.NumOut()
outOffset = total
out = -total
for i := 0; i < nout; i++ {
tv := t.Out(i)
total = align(total, uintptr(tv.Align()))
total += tv.Size()
}
out += total
// total must be > 0 in order for &args[0] to be valid.
// the argument copying is going to round it up to
// a multiple of ptrSize anyway, so make it ptrSize to begin with.
if total < ptrSize {
total = ptrSize
}
// round to pointer
total = align(total, ptrSize)
return
}
// callMethod is the call implementation used by a function returned
// by makeMethodValue (used by v.Method(i).Interface()).
// It is a streamlined version of the usual reflect call: the caller has
// already laid out the argument frame for us, so we don't have
// to deal with individual Values for each argument.
// It is in this file so that it can be next to the two similar functions above.
// The remainder of the makeMethodValue implementation is in makefunc.go.
func callMethod(ctxt *methodValue, frame unsafe.Pointer) {
t, fn, rcvr := methodReceiver("call", ctxt.rcvr, ctxt.method)
total, in, outOffset, out := frameSize(t, true)
// Copy into args.
//
// TODO(rsc): This will need to be updated for any new garbage collector.
// For now make everything look like a pointer by allocating
// a []unsafe.Pointer.
args := make([]unsafe.Pointer, total/ptrSize)
args[0] = unsafe.Pointer(rcvr)
base := unsafe.Pointer(&args[0])
memmove(unsafe.Pointer(uintptr(base)+ptrSize), frame, in)
// Call.
call(fn, unsafe.Pointer(&args[0]), uint32(total))
// Copy return values.
memmove(unsafe.Pointer(uintptr(frame)+outOffset-ptrSize), unsafe.Pointer(uintptr(base)+outOffset), out)
}
// funcName returns the name of f, for use in error messages.
func funcName(f func([]Value) []Value) string {
pc := *(*uintptr)(unsafe.Pointer(&f))
rf := runtime.FuncForPC(pc)
if rf != nil {
return rf.Name()
}
return "closure"
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// Cap returns v's capacity.
// It panics if v's Kind is not Array, Chan, or Slice.
func (v Value) Cap() int {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Array:
return v.typ.Len()
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Chan:
return int(chancap(v.iword()))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Slice:
// Slice is always bigger than a word; assume flagIndir.
return (*SliceHeader)(v.val).Cap
}
panic(&ValueError{"reflect.Value.Cap", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Close closes the channel v.
// It panics if v's Kind is not Chan.
func (v Value) Close() {
v.mustBe(Chan)
v.mustBeExported()
chanclose(v.iword())
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Complex returns v's underlying value, as a complex128.
// It panics if v's Kind is not Complex64 or Complex128
func (v Value) Complex() complex128 {
k := v.kind()
switch k {
case Complex64:
if v.flag&flagIndir != 0 {
return complex128(*(*complex64)(v.val))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return complex128(*(*complex64)(unsafe.Pointer(&v.val)))
case Complex128:
// complex128 is always bigger than a word; assume flagIndir.
return *(*complex128)(v.val)
}
panic(&ValueError{"reflect.Value.Complex", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Elem returns the value that the interface v contains
// or that the pointer v points to.
// It panics if v's Kind is not Interface or Ptr.
// It returns the zero Value if v is nil.
func (v Value) Elem() Value {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Interface:
var (
typ *rtype
val unsafe.Pointer
)
if v.typ.NumMethod() == 0 {
eface := (*emptyInterface)(v.val)
if eface.typ == nil {
// nil interface value
return Value{}
}
typ = eface.typ
val = unsafe.Pointer(eface.word)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
} else {
iface := (*nonEmptyInterface)(v.val)
if iface.itab == nil {
// nil interface value
return Value{}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
typ = iface.itab.typ
val = unsafe.Pointer(iface.word)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
fl := v.flag & flagRO
fl |= flag(typ.Kind()) << flagKindShift
if typ.size > ptrSize {
fl |= flagIndir
}
return Value{typ, val, fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Ptr:
val := v.val
if v.flag&flagIndir != 0 {
val = *(*unsafe.Pointer)(val)
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// The returned value's address is v's value.
if val == nil {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
return Value{}
}
tt := (*ptrType)(unsafe.Pointer(v.typ))
typ := tt.elem
fl := v.flag&flagRO | flagIndir | flagAddr
fl |= flag(typ.Kind() << flagKindShift)
return Value{typ, val, fl}
}
panic(&ValueError{"reflect.Value.Elem", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Field returns the i'th field of the struct v.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Struct or i is out of range.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Field(i int) Value {
v.mustBe(Struct)
tt := (*structType)(unsafe.Pointer(v.typ))
if i < 0 || i >= len(tt.fields) {
panic("reflect: Field index out of range")
}
field := &tt.fields[i]
typ := field.typ
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Inherit permission bits from v.
fl := v.flag & (flagRO | flagIndir | flagAddr)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Using an unexported field forces flagRO.
if field.pkgPath != nil {
fl |= flagRO
}
fl |= flag(typ.Kind()) << flagKindShift
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
var val unsafe.Pointer
switch {
case fl&flagIndir != 0:
// Indirect. Just bump pointer.
val = unsafe.Pointer(uintptr(v.val) + field.offset)
case bigEndian:
// Direct. Discard leading bytes.
val = unsafe.Pointer(uintptr(v.val) << (field.offset * 8))
default:
// Direct. Discard leading bytes.
val = unsafe.Pointer(uintptr(v.val) >> (field.offset * 8))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return Value{typ, val, fl}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// FieldByIndex returns the nested field corresponding to index.
// It panics if v's Kind is not struct.
func (v Value) FieldByIndex(index []int) Value {
v.mustBe(Struct)
for i, x := range index {
if i > 0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
if v.Kind() == Ptr && v.Elem().Kind() == Struct {
v = v.Elem()
}
}
v = v.Field(x)
}
return v
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// FieldByName returns the struct field with the given name.
// It returns the zero Value if no field was found.
// It panics if v's Kind is not struct.
func (v Value) FieldByName(name string) Value {
v.mustBe(Struct)
if f, ok := v.typ.FieldByName(name); ok {
return v.FieldByIndex(f.Index)
}
return Value{}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// FieldByNameFunc returns the struct field with a name
// that satisfies the match function.
// It panics if v's Kind is not struct.
// It returns the zero Value if no field was found.
func (v Value) FieldByNameFunc(match func(string) bool) Value {
v.mustBe(Struct)
if f, ok := v.typ.FieldByNameFunc(match); ok {
return v.FieldByIndex(f.Index)
}
return Value{}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Float returns v's underlying value, as a float64.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// It panics if v's Kind is not Float32 or Float64
func (v Value) Float() float64 {
k := v.kind()
switch k {
case Float32:
if v.flag&flagIndir != 0 {
return float64(*(*float32)(v.val))
}
return float64(*(*float32)(unsafe.Pointer(&v.val)))
case Float64:
if v.flag&flagIndir != 0 {
return *(*float64)(v.val)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return *(*float64)(unsafe.Pointer(&v.val))
}
panic(&ValueError{"reflect.Value.Float", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
var uint8Type = TypeOf(uint8(0)).(*rtype)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// Index returns v's i'th element.
// It panics if v's Kind is not Array, Slice, or String or i is out of range.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Index(i int) Value {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Array:
tt := (*arrayType)(unsafe.Pointer(v.typ))
if i < 0 || i > int(tt.len) {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect: array index out of range")
}
typ := tt.elem
fl := v.flag & (flagRO | flagIndir | flagAddr) // bits same as overall array
fl |= flag(typ.Kind()) << flagKindShift
offset := uintptr(i) * typ.size
var val unsafe.Pointer
switch {
case fl&flagIndir != 0:
// Indirect. Just bump pointer.
val = unsafe.Pointer(uintptr(v.val) + offset)
case bigEndian:
// Direct. Discard leading bytes.
val = unsafe.Pointer(uintptr(v.val) << (offset * 8))
default:
// Direct. Discard leading bytes.
val = unsafe.Pointer(uintptr(v.val) >> (offset * 8))
}
return Value{typ, val, fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Slice:
// Element flag same as Elem of Ptr.
// Addressable, indirect, possibly read-only.
fl := flagAddr | flagIndir | v.flag&flagRO
s := (*SliceHeader)(v.val)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
if i < 0 || i >= s.Len {
panic("reflect: slice index out of range")
}
tt := (*sliceType)(unsafe.Pointer(v.typ))
typ := tt.elem
fl |= flag(typ.Kind()) << flagKindShift
val := unsafe.Pointer(s.Data + uintptr(i)*typ.size)
return Value{typ, val, fl}
case String:
fl := v.flag&flagRO | flag(Uint8<<flagKindShift)
s := (*StringHeader)(v.val)
if i < 0 || i >= s.Len {
panic("reflect: string index out of range")
}
val := *(*byte)(unsafe.Pointer(s.Data + uintptr(i)))
return Value{uint8Type, unsafe.Pointer(uintptr(val)), fl}
}
panic(&ValueError{"reflect.Value.Index", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Int returns v's underlying value, as an int64.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Int, Int8, Int16, Int32, or Int64.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Int() int64 {
k := v.kind()
var p unsafe.Pointer
if v.flag&flagIndir != 0 {
p = v.val
} else {
// The escape analysis is good enough that &v.val
// does not trigger a heap allocation.
p = unsafe.Pointer(&v.val)
}
switch k {
case Int:
return int64(*(*int)(p))
case Int8:
return int64(*(*int8)(p))
case Int16:
return int64(*(*int16)(p))
case Int32:
return int64(*(*int32)(p))
case Int64:
return int64(*(*int64)(p))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
panic(&ValueError{"reflect.Value.Int", k})
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
// CanInterface returns true if Interface can be used without panicking.
func (v Value) CanInterface() bool {
if v.flag == 0 {
panic(&ValueError{"reflect.Value.CanInterface", Invalid})
}
return v.flag&flagRO == 0
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Interface returns v's current value as an interface{}.
// It is equivalent to:
// var i interface{} = (v's underlying value)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// If v is a method obtained by invoking Value.Method
// (as opposed to Type.Method), Interface cannot return an
// interface value, so it panics.
// It also panics if the Value was obtained by accessing
// unexported struct fields.
func (v Value) Interface() (i interface{}) {
return valueInterface(v, true)
}
func valueInterface(v Value, safe bool) interface{} {
if v.flag == 0 {
panic(&ValueError{"reflect.Value.Interface", 0})
}
if safe && v.flag&flagRO != 0 {
// Do not allow access to unexported values via Interface,
// because they might be pointers that should not be
// writable or methods or function that should not be callable.
panic("reflect.Value.Interface: cannot return value obtained from unexported field or method")
}
if v.flag&flagMethod != 0 {
v = makeMethodValue("Interface", v)
}
k := v.kind()
if k == Interface {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Special case: return the element inside the interface.
// Empty interface has one layout, all interfaces with
// methods have a second layout.
if v.NumMethod() == 0 {
return *(*interface{})(v.val)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return *(*interface {
M()
})(v.val)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
// Non-interface value.
var eface emptyInterface
eface.typ = v.typ
eface.word = v.iword()
// Don't need to allocate if v is not addressable or fits in one word.
if v.flag&flagAddr != 0 && v.typ.size > ptrSize {
// eface.word is a pointer to the actual data,
// which might be changed. We need to return
// a pointer to unchanging data, so make a copy.
ptr := unsafe_New(v.typ)
memmove(ptr, unsafe.Pointer(eface.word), v.typ.size)
eface.word = iword(ptr)
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
return *(*interface{})(unsafe.Pointer(&eface))
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// InterfaceData returns the interface v's value as a uintptr pair.
// It panics if v's Kind is not Interface.
func (v Value) InterfaceData() [2]uintptr {
v.mustBe(Interface)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// We treat this as a read operation, so we allow
// it even for unexported data, because the caller
// has to import "unsafe" to turn it into something
// that can be abused.
// Interface value is always bigger than a word; assume flagIndir.
return *(*[2]uintptr)(v.val)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// IsNil returns true if v is a nil value.
// It panics if v's Kind is not Chan, Func, Interface, Map, Ptr, or Slice.
func (v Value) IsNil() bool {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Chan, Func, Map, Ptr:
if v.flag&flagMethod != 0 {
return false
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
ptr := v.val
if v.flag&flagIndir != 0 {
ptr = *(*unsafe.Pointer)(ptr)
}
return ptr == nil
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Interface, Slice:
// Both interface and slice are nil if first word is 0.
// Both are always bigger than a word; assume flagIndir.
return *(*unsafe.Pointer)(v.val) == nil
}
panic(&ValueError{"reflect.Value.IsNil", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// IsValid returns true if v represents a value.
// It returns false if v is the zero Value.
// If IsValid returns false, all other methods except String panic.
// Most functions and methods never return an invalid value.
// If one does, its documentation states the conditions explicitly.
func (v Value) IsValid() bool {
return v.flag != 0
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Kind returns v's Kind.
// If v is the zero Value (IsValid returns false), Kind returns Invalid.
func (v Value) Kind() Kind {
return v.kind()
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Len returns v's length.
// It panics if v's Kind is not Array, Chan, Map, Slice, or String.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Len() int {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Array:
tt := (*arrayType)(unsafe.Pointer(v.typ))
return int(tt.len)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Chan:
return chanlen(v.iword())
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Map:
return maplen(v.iword())
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Slice:
// Slice is bigger than a word; assume flagIndir.
return (*SliceHeader)(v.val).Len
case String:
// String is bigger than a word; assume flagIndir.
return (*StringHeader)(v.val).Len
}
panic(&ValueError{"reflect.Value.Len", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// MapIndex returns the value associated with key in the map v.
// It panics if v's Kind is not Map.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It returns the zero Value if key is not found in the map or if v represents a nil map.
// As in Go, the key's value must be assignable to the map's key type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) MapIndex(key Value) Value {
v.mustBe(Map)
tt := (*mapType)(unsafe.Pointer(v.typ))
// Do not require key to be exported, so that DeepEqual
// and other programs can use all the keys returned by
// MapKeys as arguments to MapIndex. If either the map
// or the key is unexported, though, the result will be
// considered unexported. This is consistent with the
// behavior for structs, which allow read but not write
// of unexported fields.
key = key.assignTo("reflect.Value.MapIndex", tt.key, nil)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
word, ok := mapaccess(v.typ, v.iword(), key.iword())
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
if !ok {
return Value{}
}
typ := tt.elem
fl := (v.flag | key.flag) & flagRO
if typ.size > ptrSize {
fl |= flagIndir
}
fl |= flag(typ.Kind()) << flagKindShift
return Value{typ, unsafe.Pointer(word), fl}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// MapKeys returns a slice containing all the keys present in the map,
// in unspecified order.
// It panics if v's Kind is not Map.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It returns an empty slice if v represents a nil map.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) MapKeys() []Value {
v.mustBe(Map)
tt := (*mapType)(unsafe.Pointer(v.typ))
keyType := tt.key
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
fl := v.flag & flagRO
fl |= flag(keyType.Kind()) << flagKindShift
if keyType.size > ptrSize {
fl |= flagIndir
}
m := v.iword()
mlen := int(0)
if m != nil {
mlen = maplen(m)
}
it := mapiterinit(v.typ, m)
a := make([]Value, mlen)
var i int
for i = 0; i < len(a); i++ {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
keyWord, ok := mapiterkey(it)
if !ok {
break
}
a[i] = Value{keyType, unsafe.Pointer(keyWord), fl}
mapiternext(it)
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
return a[:i]
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Method returns a function value corresponding to v's i'th method.
// The arguments to a Call on the returned function should not include
// a receiver; the returned function will always use v as the receiver.
// Method panics if i is out of range or if v is a nil interface value.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Method(i int) Value {
if v.typ == nil {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic(&ValueError{"reflect.Value.Method", Invalid})
}
if v.flag&flagMethod != 0 || i < 0 || i >= v.typ.NumMethod() {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect: Method index out of range")
}
if v.typ.Kind() == Interface && v.IsNil() {
panic("reflect: Method on nil interface value")
}
fl := v.flag & (flagRO | flagIndir)
fl |= flag(Func) << flagKindShift
fl |= flag(i)<<flagMethodShift | flagMethod
return Value{v.typ, v.val, fl}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// NumMethod returns the number of methods in the value's method set.
func (v Value) NumMethod() int {
if v.typ == nil {
panic(&ValueError{"reflect.Value.NumMethod", Invalid})
}
if v.flag&flagMethod != 0 {
return 0
}
return v.typ.NumMethod()
}
// MethodByName returns a function value corresponding to the method
// of v with the given name.
// The arguments to a Call on the returned function should not include
// a receiver; the returned function will always use v as the receiver.
// It returns the zero Value if no method was found.
func (v Value) MethodByName(name string) Value {
if v.typ == nil {
panic(&ValueError{"reflect.Value.MethodByName", Invalid})
}
if v.flag&flagMethod != 0 {
return Value{}
}
m, ok := v.typ.MethodByName(name)
if !ok {
return Value{}
}
return v.Method(m.Index)
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// NumField returns the number of fields in the struct v.
// It panics if v's Kind is not Struct.
func (v Value) NumField() int {
v.mustBe(Struct)
tt := (*structType)(unsafe.Pointer(v.typ))
return len(tt.fields)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// OverflowComplex returns true if the complex128 x cannot be represented by v's type.
// It panics if v's Kind is not Complex64 or Complex128.
func (v Value) OverflowComplex(x complex128) bool {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Complex64:
return overflowFloat32(real(x)) || overflowFloat32(imag(x))
case Complex128:
return false
}
panic(&ValueError{"reflect.Value.OverflowComplex", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// OverflowFloat returns true if the float64 x cannot be represented by v's type.
// It panics if v's Kind is not Float32 or Float64.
func (v Value) OverflowFloat(x float64) bool {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Float32:
return overflowFloat32(x)
case Float64:
return false
}
panic(&ValueError{"reflect.Value.OverflowFloat", k})
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
func overflowFloat32(x float64) bool {
if x < 0 {
x = -x
}
return math.MaxFloat32 < x && x <= math.MaxFloat64
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// OverflowInt returns true if the int64 x cannot be represented by v's type.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Int, Int8, int16, Int32, or Int64.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) OverflowInt(x int64) bool {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Int, Int8, Int16, Int32, Int64:
bitSize := v.typ.size * 8
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
trunc := (x << (64 - bitSize)) >> (64 - bitSize)
return x != trunc
}
panic(&ValueError{"reflect.Value.OverflowInt", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// OverflowUint returns true if the uint64 x cannot be represented by v's type.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Uint, Uintptr, Uint8, Uint16, Uint32, or Uint64.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) OverflowUint(x uint64) bool {
k := v.kind()
switch k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Uint, Uintptr, Uint8, Uint16, Uint32, Uint64:
bitSize := v.typ.size * 8
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
trunc := (x << (64 - bitSize)) >> (64 - bitSize)
return x != trunc
}
panic(&ValueError{"reflect.Value.OverflowUint", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Pointer returns v's value as a uintptr.
// It returns uintptr instead of unsafe.Pointer so that
// code using reflect cannot obtain unsafe.Pointers
// without importing the unsafe package explicitly.
// It panics if v's Kind is not Chan, Func, Map, Ptr, Slice, or UnsafePointer.
//
// If v's Kind is Func, the returned pointer is an underlying
// code pointer, but not necessarily enough to identify a
// single function uniquely. The only guarantee is that the
// result is zero if and only if v is a nil func Value.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Pointer() uintptr {
k := v.kind()
switch k {
case Chan, Map, Ptr, UnsafePointer:
p := v.val
if v.flag&flagIndir != 0 {
p = *(*unsafe.Pointer)(p)
}
return uintptr(p)
case Func:
if v.flag&flagMethod != 0 {
// As the doc comment says, the returned pointer is an
// underlying code pointer but not necessarily enough to
// identify a single function uniquely. All method expressions
// created via reflect have the same underlying code pointer,
// so their Pointers are equal. The function used here must
// match the one used in makeMethodValue.
f := methodValueCall
return **(**uintptr)(unsafe.Pointer(&f))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
p := v.val
if v.flag&flagIndir != 0 {
p = *(*unsafe.Pointer)(p)
}
// Non-nil func value points at data block.
// First word of data block is actual code.
if p != nil {
p = *(*unsafe.Pointer)(p)
}
return uintptr(p)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Slice:
return (*SliceHeader)(v.val).Data
}
panic(&ValueError{"reflect.Value.Pointer", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Recv receives and returns a value from the channel v.
// It panics if v's Kind is not Chan.
// The receive blocks until a value is ready.
// The boolean value ok is true if the value x corresponds to a send
// on the channel, false if it is a zero value received because the channel is closed.
func (v Value) Recv() (x Value, ok bool) {
v.mustBe(Chan)
v.mustBeExported()
return v.recv(false)
}
// internal recv, possibly non-blocking (nb).
// v is known to be a channel.
func (v Value) recv(nb bool) (val Value, ok bool) {
tt := (*chanType)(unsafe.Pointer(v.typ))
if ChanDir(tt.dir)&RecvDir == 0 {
panic("reflect: recv on send-only channel")
}
word, selected, ok := chanrecv(v.typ, v.iword(), nb)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
if selected {
typ := tt.elem
fl := flag(typ.Kind()) << flagKindShift
if typ.size > ptrSize {
fl |= flagIndir
}
val = Value{typ, unsafe.Pointer(word), fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Send sends x on the channel v.
// It panics if v's kind is not Chan or if x's type is not the same type as v's element type.
// As in Go, x's value must be assignable to the channel's element type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Send(x Value) {
v.mustBe(Chan)
v.mustBeExported()
v.send(x, false)
}
// internal send, possibly non-blocking.
// v is known to be a channel.
func (v Value) send(x Value, nb bool) (selected bool) {
tt := (*chanType)(unsafe.Pointer(v.typ))
if ChanDir(tt.dir)&SendDir == 0 {
panic("reflect: send on recv-only channel")
}
x.mustBeExported()
x = x.assignTo("reflect.Value.Send", tt.elem, nil)
return chansend(v.typ, v.iword(), x.iword(), nb)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Set assigns x to the value v.
// It panics if CanSet returns false.
// As in Go, x's value must be assignable to v's type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Set(x Value) {
v.mustBeAssignable()
x.mustBeExported() // do not let unexported x leak
var target *interface{}
if v.kind() == Interface {
target = (*interface{})(v.val)
}
x = x.assignTo("reflect.Set", v.typ, target)
if x.flag&flagIndir != 0 {
memmove(v.val, x.val, v.typ.size)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
} else {
storeIword(v.val, iword(x.val), v.typ.size)
}
}
// SetBool sets v's underlying value.
// It panics if v's Kind is not Bool or if CanSet() is false.
func (v Value) SetBool(x bool) {
v.mustBeAssignable()
v.mustBe(Bool)
*(*bool)(v.val) = x
}
// SetBytes sets v's underlying value.
// It panics if v's underlying value is not a slice of bytes.
func (v Value) SetBytes(x []byte) {
v.mustBeAssignable()
v.mustBe(Slice)
if v.typ.Elem().Kind() != Uint8 {
panic("reflect.Value.SetBytes of non-byte slice")
}
*(*[]byte)(v.val) = x
}
// setRunes sets v's underlying value.
// It panics if v's underlying value is not a slice of runes (int32s).
func (v Value) setRunes(x []rune) {
v.mustBeAssignable()
v.mustBe(Slice)
if v.typ.Elem().Kind() != Int32 {
panic("reflect.Value.setRunes of non-rune slice")
}
*(*[]rune)(v.val) = x
}
// SetComplex sets v's underlying value to x.
// It panics if v's Kind is not Complex64 or Complex128, or if CanSet() is false.
func (v Value) SetComplex(x complex128) {
v.mustBeAssignable()
switch k := v.kind(); k {
default:
panic(&ValueError{"reflect.Value.SetComplex", k})
case Complex64:
*(*complex64)(v.val) = complex64(x)
case Complex128:
*(*complex128)(v.val) = x
}
}
// SetFloat sets v's underlying value to x.
// It panics if v's Kind is not Float32 or Float64, or if CanSet() is false.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) SetFloat(x float64) {
v.mustBeAssignable()
switch k := v.kind(); k {
default:
panic(&ValueError{"reflect.Value.SetFloat", k})
case Float32:
*(*float32)(v.val) = float32(x)
case Float64:
*(*float64)(v.val) = x
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetInt sets v's underlying value to x.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Int, Int8, Int16, Int32, or Int64, or if CanSet() is false.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) SetInt(x int64) {
v.mustBeAssignable()
switch k := v.kind(); k {
default:
panic(&ValueError{"reflect.Value.SetInt", k})
case Int:
*(*int)(v.val) = int(x)
case Int8:
*(*int8)(v.val) = int8(x)
case Int16:
*(*int16)(v.val) = int16(x)
case Int32:
*(*int32)(v.val) = int32(x)
case Int64:
*(*int64)(v.val) = x
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetLen sets v's length to n.
// It panics if v's Kind is not Slice or if n is negative or
// greater than the capacity of the slice.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) SetLen(n int) {
v.mustBeAssignable()
v.mustBe(Slice)
s := (*SliceHeader)(v.val)
if n < 0 || n > int(s.Cap) {
panic("reflect: slice length out of range in SetLen")
}
s.Len = n
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetCap sets v's capacity to n.
// It panics if v's Kind is not Slice or if n is smaller than the length or
// greater than the capacity of the slice.
func (v Value) SetCap(n int) {
v.mustBeAssignable()
v.mustBe(Slice)
s := (*SliceHeader)(v.val)
if n < int(s.Len) || n > int(s.Cap) {
panic("reflect: slice capacity out of range in SetCap")
}
s.Cap = n
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// SetMapIndex sets the value associated with key in the map v to val.
// It panics if v's Kind is not Map.
// If val is the zero Value, SetMapIndex deletes the key from the map.
// As in Go, key's value must be assignable to the map's key type,
// and val's value must be assignable to the map's value type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) SetMapIndex(key, val Value) {
v.mustBe(Map)
v.mustBeExported()
key.mustBeExported()
tt := (*mapType)(unsafe.Pointer(v.typ))
key = key.assignTo("reflect.Value.SetMapIndex", tt.key, nil)
if val.typ != nil {
val.mustBeExported()
val = val.assignTo("reflect.Value.SetMapIndex", tt.elem, nil)
}
mapassign(v.typ, v.iword(), key.iword(), val.iword(), val.typ != nil)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetUint sets v's underlying value to x.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Uint, Uintptr, Uint8, Uint16, Uint32, or Uint64, or if CanSet() is false.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) SetUint(x uint64) {
v.mustBeAssignable()
switch k := v.kind(); k {
default:
panic(&ValueError{"reflect.Value.SetUint", k})
case Uint:
*(*uint)(v.val) = uint(x)
case Uint8:
*(*uint8)(v.val) = uint8(x)
case Uint16:
*(*uint16)(v.val) = uint16(x)
case Uint32:
*(*uint32)(v.val) = uint32(x)
case Uint64:
*(*uint64)(v.val) = x
case Uintptr:
*(*uintptr)(v.val) = uintptr(x)
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetPointer sets the unsafe.Pointer value v to x.
// It panics if v's Kind is not UnsafePointer.
func (v Value) SetPointer(x unsafe.Pointer) {
v.mustBeAssignable()
v.mustBe(UnsafePointer)
*(*unsafe.Pointer)(v.val) = x
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// SetString sets v's underlying value to x.
// It panics if v's Kind is not String or if CanSet() is false.
func (v Value) SetString(x string) {
v.mustBeAssignable()
v.mustBe(String)
*(*string)(v.val) = x
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Slice returns v[i:j].
// It panics if v's Kind is not Array, Slice or String, or if v is an unaddressable array,
// or if the indexes are out of bounds.
func (v Value) Slice(i, j int) Value {
var (
cap int
typ *sliceType
base unsafe.Pointer
)
switch kind := v.kind(); kind {
default:
panic(&ValueError{"reflect.Value.Slice", kind})
case Array:
if v.flag&flagAddr == 0 {
panic("reflect.Value.Slice: slice of unaddressable array")
}
tt := (*arrayType)(unsafe.Pointer(v.typ))
cap = int(tt.len)
typ = (*sliceType)(unsafe.Pointer(tt.slice))
base = v.val
case Slice:
typ = (*sliceType)(unsafe.Pointer(v.typ))
s := (*SliceHeader)(v.val)
base = unsafe.Pointer(s.Data)
cap = s.Cap
case String:
s := (*StringHeader)(v.val)
if i < 0 || j < i || j > s.Len {
panic("reflect.Value.Slice: string slice index out of bounds")
}
var x string
val := (*StringHeader)(unsafe.Pointer(&x))
val.Data = s.Data + uintptr(i)
val.Len = j - i
return Value{v.typ, unsafe.Pointer(&x), v.flag}
}
if i < 0 || j < i || j > cap {
panic("reflect.Value.Slice: slice index out of bounds")
}
// Declare slice so that gc can see the base pointer in it.
var x []unsafe.Pointer
// Reinterpret as *SliceHeader to edit.
s := (*SliceHeader)(unsafe.Pointer(&x))
s.Data = uintptr(base) + uintptr(i)*typ.elem.Size()
s.Len = j - i
s.Cap = cap - i
fl := v.flag&flagRO | flagIndir | flag(Slice)<<flagKindShift
return Value{typ.common(), unsafe.Pointer(&x), fl}
}
// Slice3 is the 3-index form of the slice operation: it returns v[i:j:k].
// It panics if v's Kind is not Array or Slice, or if v is an unaddressable array,
// or if the indexes are out of bounds.
func (v Value) Slice3(i, j, k int) Value {
var (
cap int
typ *sliceType
base unsafe.Pointer
)
switch kind := v.kind(); kind {
default:
panic(&ValueError{"reflect.Value.Slice3", kind})
case Array:
if v.flag&flagAddr == 0 {
panic("reflect.Value.Slice: slice of unaddressable array")
}
tt := (*arrayType)(unsafe.Pointer(v.typ))
cap = int(tt.len)
typ = (*sliceType)(unsafe.Pointer(tt.slice))
base = v.val
case Slice:
typ = (*sliceType)(unsafe.Pointer(v.typ))
s := (*SliceHeader)(v.val)
base = unsafe.Pointer(s.Data)
cap = s.Cap
}
if i < 0 || j < i || k < j || k > cap {
panic("reflect.Value.Slice3: slice index out of bounds")
}
// Declare slice so that the garbage collector
// can see the base pointer in it.
var x []unsafe.Pointer
// Reinterpret as *SliceHeader to edit.
s := (*SliceHeader)(unsafe.Pointer(&x))
s.Data = uintptr(base) + uintptr(i)*typ.elem.Size()
s.Len = j - i
s.Cap = k - i
fl := v.flag&flagRO | flagIndir | flag(Slice)<<flagKindShift
return Value{typ.common(), unsafe.Pointer(&x), fl}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// String returns the string v's underlying value, as a string.
// String is a special case because of Go's String method convention.
// Unlike the other getters, it does not panic if v's Kind is not String.
// Instead, it returns a string of the form "<T value>" where T is v's type.
func (v Value) String() string {
switch k := v.kind(); k {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Invalid:
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
return "<invalid Value>"
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case String:
return *(*string)(v.val)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// If you call String on a reflect.Value of other type, it's better to
// print something than to panic. Useful in debugging.
return "<" + v.typ.String() + " Value>"
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// TryRecv attempts to receive a value from the channel v but will not block.
// It panics if v's Kind is not Chan.
// If the receive cannot finish without blocking, x is the zero Value.
// The boolean ok is true if the value x corresponds to a send
// on the channel, false if it is a zero value received because the channel is closed.
func (v Value) TryRecv() (x Value, ok bool) {
v.mustBe(Chan)
v.mustBeExported()
return v.recv(true)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// TrySend attempts to send x on the channel v but will not block.
// It panics if v's Kind is not Chan.
// It returns true if the value was sent, false otherwise.
// As in Go, x's value must be assignable to the channel's element type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) TrySend(x Value) bool {
v.mustBe(Chan)
v.mustBeExported()
return v.send(x, true)
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Type returns v's type.
func (v Value) Type() Type {
f := v.flag
if f == 0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic(&ValueError{"reflect.Value.Type", Invalid})
}
if f&flagMethod == 0 {
// Easy case
return v.typ
}
// Method value.
// v.typ describes the receiver, not the method type.
i := int(v.flag) >> flagMethodShift
if v.typ.Kind() == Interface {
// Method on interface.
tt := (*interfaceType)(unsafe.Pointer(v.typ))
if i < 0 || i >= len(tt.methods) {
panic("reflect: internal error: invalid method index")
}
m := &tt.methods[i]
return m.typ
}
// Method on concrete type.
ut := v.typ.uncommon()
if ut == nil || i < 0 || i >= len(ut.methods) {
panic("reflect: internal error: invalid method index")
}
m := &ut.methods[i]
return m.mtyp
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// Uint returns v's underlying value, as a uint64.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v's Kind is not Uint, Uintptr, Uint8, Uint16, Uint32, or Uint64.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) Uint() uint64 {
k := v.kind()
var p unsafe.Pointer
if v.flag&flagIndir != 0 {
p = v.val
} else {
// The escape analysis is good enough that &v.val
// does not trigger a heap allocation.
p = unsafe.Pointer(&v.val)
}
switch k {
case Uint:
return uint64(*(*uint)(p))
case Uint8:
return uint64(*(*uint8)(p))
case Uint16:
return uint64(*(*uint16)(p))
case Uint32:
return uint64(*(*uint32)(p))
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
case Uint64:
return uint64(*(*uint64)(p))
case Uintptr:
return uint64(*(*uintptr)(p))
}
panic(&ValueError{"reflect.Value.Uint", k})
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
// UnsafeAddr returns a pointer to v's data.
// It is for advanced clients that also import the "unsafe" package.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// It panics if v is not addressable.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func (v Value) UnsafeAddr() uintptr {
if v.typ == nil {
panic(&ValueError{"reflect.Value.UnsafeAddr", Invalid})
}
if v.flag&flagAddr == 0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect.Value.UnsafeAddr of unaddressable value")
}
return uintptr(v.val)
}
// StringHeader is the runtime representation of a string.
// It cannot be used safely or portably and its representation may
// change in a later release.
// Moreover, the Data field is not sufficient to guarantee the data
// it references will not be garbage collected, so programs must keep
// a separate, correctly typed pointer to the underlying data.
type StringHeader struct {
Data uintptr
Len int
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// SliceHeader is the runtime representation of a slice.
// It cannot be used safely or portably and its representation may
// change in a later release.
// Moreover, the Data field is not sufficient to guarantee the data
// it references will not be garbage collected, so programs must keep
// a separate, correctly typed pointer to the underlying data.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
type SliceHeader struct {
Data uintptr
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
Len int
Cap int
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
func typesMustMatch(what string, t1, t2 Type) {
if t1 != t2 {
panic(what + ": " + t1.String() + " != " + t2.String())
}
}
// grow grows the slice s so that it can hold extra more values, allocating
// more capacity if needed. It also returns the old and new slice lengths.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func grow(s Value, extra int) (Value, int, int) {
i0 := s.Len()
i1 := i0 + extra
if i1 < i0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
panic("reflect.Append: slice overflow")
}
m := s.Cap()
if i1 <= m {
return s.Slice(0, i1), i0, i1
}
if m == 0 {
m = extra
} else {
for m < i1 {
if i0 < 1024 {
m += m
} else {
m += m / 4
}
}
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
t := MakeSlice(s.Type(), i1, m)
Copy(t, s)
return t, i0, i1
}
// Append appends the values x to a slice s and returns the resulting slice.
// As in Go, each x's value must be assignable to the slice's element type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func Append(s Value, x ...Value) Value {
s.mustBe(Slice)
s, i0, i1 := grow(s, len(x))
for i, j := i0, 0; i < i1; i, j = i+1, j+1 {
s.Index(i).Set(x[j])
}
return s
}
// AppendSlice appends a slice t to a slice s and returns the resulting slice.
// The slices s and t must have the same element type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func AppendSlice(s, t Value) Value {
s.mustBe(Slice)
t.mustBe(Slice)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
typesMustMatch("reflect.AppendSlice", s.Type().Elem(), t.Type().Elem())
s, i0, i1 := grow(s, t.Len())
Copy(s.Slice(i0, i1), t)
return s
}
// Copy copies the contents of src into dst until either
// dst has been filled or src has been exhausted.
// It returns the number of elements copied.
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Dst and src each must have kind Slice or Array, and
// dst and src must have the same element type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func Copy(dst, src Value) int {
dk := dst.kind()
if dk != Array && dk != Slice {
panic(&ValueError{"reflect.Copy", dk})
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
if dk == Array {
dst.mustBeAssignable()
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
dst.mustBeExported()
sk := src.kind()
if sk != Array && sk != Slice {
panic(&ValueError{"reflect.Copy", sk})
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
src.mustBeExported()
de := dst.typ.Elem()
se := src.typ.Elem()
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
typesMustMatch("reflect.Copy", de, se)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
n := dst.Len()
if sn := src.Len(); n > sn {
n = sn
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// If sk is an in-line array, cannot take its address.
// Instead, copy element by element.
if src.flag&flagIndir == 0 {
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
for i := 0; i < n; i++ {
dst.Index(i).Set(src.Index(i))
}
return n
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// Copy via memmove.
var da, sa unsafe.Pointer
if dk == Array {
da = dst.val
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
} else {
da = unsafe.Pointer((*SliceHeader)(dst.val).Data)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
if sk == Array {
sa = src.val
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
} else {
sa = unsafe.Pointer((*SliceHeader)(src.val).Data)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
memmove(da, sa, uintptr(n)*de.Size())
return n
}
// A runtimeSelect is a single case passed to rselect.
// This must match ../runtime/chan.c:/runtimeSelect
type runtimeSelect struct {
dir uintptr // 0, SendDir, or RecvDir
typ *rtype // channel type
ch iword // interface word for channel
val iword // interface word for value (for SendDir)
}
// rselect runs a select. It returns the index of the chosen case,
// and if the case was a receive, the interface word of the received
// value and the conventional OK bool to indicate whether the receive
// corresponds to a sent value.
func rselect([]runtimeSelect) (chosen int, recv iword, recvOK bool)
// A SelectDir describes the communication direction of a select case.
type SelectDir int
// NOTE: These values must match ../runtime/chan.c:/SelectDir.
const (
_ SelectDir = iota
SelectSend // case Chan <- Send
SelectRecv // case <-Chan:
SelectDefault // default
)
// A SelectCase describes a single case in a select operation.
// The kind of case depends on Dir, the communication direction.
//
// If Dir is SelectDefault, the case represents a default case.
// Chan and Send must be zero Values.
//
// If Dir is SelectSend, the case represents a send operation.
// Normally Chan's underlying value must be a channel, and Send's underlying value must be
// assignable to the channel's element type. As a special case, if Chan is a zero Value,
// then the case is ignored, and the field Send will also be ignored and may be either zero
// or non-zero.
//
// If Dir is SelectRecv, the case represents a receive operation.
// Normally Chan's underlying value must be a channel and Send must be a zero Value.
// If Chan is a zero Value, then the case is ignored, but Send must still be a zero Value.
// When a receive operation is selected, the received Value is returned by Select.
//
type SelectCase struct {
Dir SelectDir // direction of case
Chan Value // channel to use (for send or receive)
Send Value // value to send (for send)
}
// Select executes a select operation described by the list of cases.
// Like the Go select statement, it blocks until at least one of the cases
// can proceed, makes a uniform pseudo-random choice,
// and then executes that case. It returns the index of the chosen case
// and, if that case was a receive operation, the value received and a
// boolean indicating whether the value corresponds to a send on the channel
// (as opposed to a zero value received because the channel is closed).
func Select(cases []SelectCase) (chosen int, recv Value, recvOK bool) {
// NOTE: Do not trust that caller is not modifying cases data underfoot.
// The range is safe because the caller cannot modify our copy of the len
// and each iteration makes its own copy of the value c.
runcases := make([]runtimeSelect, len(cases))
haveDefault := false
for i, c := range cases {
rc := &runcases[i]
rc.dir = uintptr(c.Dir)
switch c.Dir {
default:
panic("reflect.Select: invalid Dir")
case SelectDefault: // default
if haveDefault {
panic("reflect.Select: multiple default cases")
}
haveDefault = true
if c.Chan.IsValid() {
panic("reflect.Select: default case has Chan value")
}
if c.Send.IsValid() {
panic("reflect.Select: default case has Send value")
}
case SelectSend:
ch := c.Chan
if !ch.IsValid() {
break
}
ch.mustBe(Chan)
ch.mustBeExported()
tt := (*chanType)(unsafe.Pointer(ch.typ))
if ChanDir(tt.dir)&SendDir == 0 {
panic("reflect.Select: SendDir case using recv-only channel")
}
rc.ch = ch.iword()
rc.typ = &tt.rtype
v := c.Send
if !v.IsValid() {
panic("reflect.Select: SendDir case missing Send value")
}
v.mustBeExported()
v = v.assignTo("reflect.Select", tt.elem, nil)
rc.val = v.iword()
case SelectRecv:
if c.Send.IsValid() {
panic("reflect.Select: RecvDir case has Send value")
}
ch := c.Chan
if !ch.IsValid() {
break
}
ch.mustBe(Chan)
ch.mustBeExported()
tt := (*chanType)(unsafe.Pointer(ch.typ))
rc.typ = &tt.rtype
if ChanDir(tt.dir)&RecvDir == 0 {
panic("reflect.Select: RecvDir case using send-only channel")
}
rc.ch = ch.iword()
}
}
chosen, word, recvOK := rselect(runcases)
if runcases[chosen].dir == uintptr(SelectRecv) {
tt := (*chanType)(unsafe.Pointer(runcases[chosen].typ))
typ := tt.elem
fl := flag(typ.Kind()) << flagKindShift
if typ.size > ptrSize {
fl |= flagIndir
}
recv = Value{typ, unsafe.Pointer(word), fl}
}
return chosen, recv, recvOK
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
/*
* constructors
*/
// implemented in package runtime
func unsafe_New(*rtype) unsafe.Pointer
func unsafe_NewArray(*rtype, int) unsafe.Pointer
// MakeSlice creates a new zero-initialized slice value
// for the specified slice type, length, and capacity.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func MakeSlice(typ Type, len, cap int) Value {
if typ.Kind() != Slice {
panic("reflect.MakeSlice of non-slice type")
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
if len < 0 {
panic("reflect.MakeSlice: negative len")
}
if cap < 0 {
panic("reflect.MakeSlice: negative cap")
}
if len > cap {
panic("reflect.MakeSlice: len > cap")
}
// Declare slice so that gc can see the base pointer in it.
var x []unsafe.Pointer
// Reinterpret as *SliceHeader to edit.
s := (*SliceHeader)(unsafe.Pointer(&x))
s.Data = uintptr(unsafe_NewArray(typ.Elem().(*rtype), cap))
s.Len = len
s.Cap = cap
return Value{typ.common(), unsafe.Pointer(&x), flagIndir | flag(Slice)<<flagKindShift}
}
// MakeChan creates a new channel with the specified type and buffer size.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func MakeChan(typ Type, buffer int) Value {
if typ.Kind() != Chan {
panic("reflect.MakeChan of non-chan type")
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
if buffer < 0 {
panic("reflect.MakeChan: negative buffer size")
}
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
if typ.ChanDir() != BothDir {
panic("reflect.MakeChan: unidirectional channel type")
}
ch := makechan(typ.(*rtype), uint64(buffer))
return Value{typ.common(), unsafe.Pointer(ch), flag(Chan) << flagKindShift}
}
// MakeMap creates a new map of the specified type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func MakeMap(typ Type) Value {
if typ.Kind() != Map {
panic("reflect.MakeMap of non-map type")
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
}
m := makemap(typ.(*rtype))
return Value{typ.common(), unsafe.Pointer(m), flag(Map) << flagKindShift}
}
// Indirect returns the value that v points to.
// If v is a nil pointer, Indirect returns a zero Value.
// If v is not a pointer, Indirect returns v.
func Indirect(v Value) Value {
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
if v.Kind() != Ptr {
return v
}
return v.Elem()
}
// ValueOf returns a new Value initialized to the concrete value
// stored in the interface i. ValueOf(nil) returns the zero Value.
func ValueOf(i interface{}) Value {
if i == nil {
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
return Value{}
}
// TODO(rsc): Eliminate this terrible hack.
// In the call to packValue, eface.typ doesn't escape,
// and eface.word is an integer. So it looks like
// i (= eface) doesn't escape. But really it does,
// because eface.word is actually a pointer.
escapes(i)
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// For an interface value with the noAddr bit set,
// the representation is identical to an empty interface.
eface := *(*emptyInterface)(unsafe.Pointer(&i))
typ := eface.typ
fl := flag(typ.Kind()) << flagKindShift
if typ.size > ptrSize {
fl |= flagIndir
}
return Value{typ, unsafe.Pointer(eface.word), fl}
}
// Zero returns a Value representing the zero value for the specified type.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
// The result is different from the zero value of the Value struct,
// which represents no value at all.
// For example, Zero(TypeOf(42)) returns a Value with Kind Int and value 0.
// The returned value is neither addressable nor settable.
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
func Zero(typ Type) Value {
if typ == nil {
reflect: new Type and Value definitions Type is now an interface that implements all the possible type methods. Instead of a type switch on a reflect.Type t, switch on t.Kind(). If a method is invoked on the wrong kind of type (for example, calling t.Field(0) when t.Kind() != Struct), the call panics. There is one method renaming: t.(*ChanType).Dir() is now t.ChanDir(). Value is now a struct value that implements all the possible value methods. Instead of a type switch on a reflect.Value v, switch on v.Kind(). If a method is invoked on the wrong kind of value (for example, calling t.Recv() when t.Kind() != Chan), the call panics. Since Value is now a struct, not an interface, its zero value cannot be compared to nil. Instead of v != nil, use v.IsValid(). Instead of other uses of nil as a Value, use Value{}, the zero value. Many methods have been renamed, most due to signature conflicts: OLD NEW v.(*ArrayValue).Elem v.Index v.(*BoolValue).Get v.Bool v.(*BoolValue).Set v.SetBool v.(*ChanType).Dir v.ChanDir v.(*ChanValue).Get v.Pointer v.(*ComplexValue).Get v.Complex v.(*ComplexValue).Overflow v.OverflowComplex v.(*ComplexValue).Set v.SetComplex v.(*FloatValue).Get v.Float v.(*FloatValue).Overflow v.OverflowFloat v.(*FloatValue).Set v.SetFloat v.(*FuncValue).Get v.Pointer v.(*InterfaceValue).Get v.InterfaceData v.(*IntValue).Get v.Int v.(*IntValue).Overflow v.OverflowInt v.(*IntValue).Set v.SetInt v.(*MapValue).Elem v.MapIndex v.(*MapValue).Get v.Pointer v.(*MapValue).Keys v.MapKeys v.(*MapValue).SetElem v.SetMapIndex v.(*PtrValue).Get v.Pointer v.(*SliceValue).Elem v.Index v.(*SliceValue).Get v.Pointer v.(*StringValue).Get v.String v.(*StringValue).Set v.SetString v.(*UintValue).Get v.Uint v.(*UintValue).Overflow v.OverflowUint v.(*UintValue).Set v.SetUint v.(*UnsafePointerValue).Get v.Pointer v.(*UnsafePointerValue).Set v.SetPointer Part of the motivation for this change is to enable a more efficient implementation of Value, one that does not allocate memory during most operations. To reduce the size of the CL, this CL's implementation is a wrapper around the old API. Later CLs will make the implementation more efficient without changing the API. Other CLs to be submitted at the same time as this one add support for this change to gofix (4343047) and update the Go source tree (4353043). R=gri, iant, niemeyer, r, rog, gustavo, r2 CC=golang-dev https://golang.org/cl/4281055
2011-04-08 10:26:51 -06:00
panic("reflect: Zero(nil)")
}
t := typ.common()
fl := flag(t.Kind()) << flagKindShift
if t.size <= ptrSize {
return Value{t, nil, fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
return Value{t, unsafe_New(typ.(*rtype)), fl | flagIndir}
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// New returns a Value representing a pointer to a new zero value
// for the specified type. That is, the returned Value's Type is PtrTo(t).
func New(typ Type) Value {
if typ == nil {
panic("reflect: New(nil)")
}
ptr := unsafe_New(typ.(*rtype))
fl := flag(Ptr) << flagKindShift
return Value{typ.common().ptrTo(), ptr, fl}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
}
// NewAt returns a Value representing a pointer to a value of the
// specified type, using p as that pointer.
func NewAt(typ Type, p unsafe.Pointer) Value {
fl := flag(Ptr) << flagKindShift
return Value{typ.common().ptrTo(), p, fl}
}
// assignTo returns a value v that can be assigned directly to typ.
// It panics if v is not assignable to typ.
// For a conversion to an interface type, target is a suggested scratch space to use.
func (v Value) assignTo(context string, dst *rtype, target *interface{}) Value {
if v.flag&flagMethod != 0 {
v = makeMethodValue(context, v)
}
switch {
case directlyAssignable(dst, v.typ):
// Overwrite type so that they match.
// Same memory layout, so no harm done.
v.typ = dst
fl := v.flag & (flagRO | flagAddr | flagIndir)
fl |= flag(dst.Kind()) << flagKindShift
return Value{dst, v.val, fl}
case implements(dst, v.typ):
if target == nil {
target = new(interface{})
}
x := valueInterface(v, false)
if dst.NumMethod() == 0 {
*target = x
} else {
ifaceE2I(dst, x, unsafe.Pointer(target))
}
return Value{dst, unsafe.Pointer(target), flagIndir | flag(Interface)<<flagKindShift}
}
// Failed.
panic(context + ": value of type " + v.typ.String() + " is not assignable to type " + dst.String())
}
// Convert returns the value v converted to type t.
// If the usual Go conversion rules do not allow conversion
// of the value v to type t, Convert panics.
func (v Value) Convert(t Type) Value {
if v.flag&flagMethod != 0 {
v = makeMethodValue("Convert", v)
}
op := convertOp(t.common(), v.typ)
if op == nil {
panic("reflect.Value.Convert: value of type " + v.typ.String() + " cannot be converted to type " + t.String())
}
return op(v, t)
}
// convertOp returns the function to convert a value of type src
// to a value of type dst. If the conversion is illegal, convertOp returns nil.
func convertOp(dst, src *rtype) func(Value, Type) Value {
switch src.Kind() {
case Int, Int8, Int16, Int32, Int64:
switch dst.Kind() {
case Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
return cvtInt
case Float32, Float64:
return cvtIntFloat
case String:
return cvtIntString
}
case Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
switch dst.Kind() {
case Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
return cvtUint
case Float32, Float64:
return cvtUintFloat
case String:
return cvtUintString
}
case Float32, Float64:
switch dst.Kind() {
case Int, Int8, Int16, Int32, Int64:
return cvtFloatInt
case Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
return cvtFloatUint
case Float32, Float64:
return cvtFloat
}
case Complex64, Complex128:
switch dst.Kind() {
case Complex64, Complex128:
return cvtComplex
}
case String:
if dst.Kind() == Slice && dst.Elem().PkgPath() == "" {
switch dst.Elem().Kind() {
case Uint8:
return cvtStringBytes
case Int32:
return cvtStringRunes
}
}
case Slice:
if dst.Kind() == String && src.Elem().PkgPath() == "" {
switch src.Elem().Kind() {
case Uint8:
return cvtBytesString
case Int32:
return cvtRunesString
}
}
}
// dst and src have same underlying type.
if haveIdenticalUnderlyingType(dst, src) {
return cvtDirect
}
// dst and src are unnamed pointer types with same underlying base type.
if dst.Kind() == Ptr && dst.Name() == "" &&
src.Kind() == Ptr && src.Name() == "" &&
haveIdenticalUnderlyingType(dst.Elem().common(), src.Elem().common()) {
return cvtDirect
}
if implements(dst, src) {
if src.Kind() == Interface {
return cvtI2I
}
return cvtT2I
}
return nil
}
// makeInt returns a Value of type t equal to bits (possibly truncated),
// where t is a signed or unsigned int type.
func makeInt(f flag, bits uint64, t Type) Value {
typ := t.common()
if typ.size > ptrSize {
// Assume ptrSize >= 4, so this must be uint64.
ptr := unsafe_New(typ)
*(*uint64)(unsafe.Pointer(ptr)) = bits
return Value{typ, ptr, f | flagIndir | flag(typ.Kind())<<flagKindShift}
}
var w iword
switch typ.size {
case 1:
*(*uint8)(unsafe.Pointer(&w)) = uint8(bits)
case 2:
*(*uint16)(unsafe.Pointer(&w)) = uint16(bits)
case 4:
*(*uint32)(unsafe.Pointer(&w)) = uint32(bits)
case 8:
*(*uint64)(unsafe.Pointer(&w)) = uint64(bits)
}
return Value{typ, unsafe.Pointer(w), f | flag(typ.Kind())<<flagKindShift}
}
// makeFloat returns a Value of type t equal to v (possibly truncated to float32),
// where t is a float32 or float64 type.
func makeFloat(f flag, v float64, t Type) Value {
typ := t.common()
if typ.size > ptrSize {
// Assume ptrSize >= 4, so this must be float64.
ptr := unsafe_New(typ)
*(*float64)(unsafe.Pointer(ptr)) = v
return Value{typ, ptr, f | flagIndir | flag(typ.Kind())<<flagKindShift}
}
var w iword
switch typ.size {
case 4:
*(*float32)(unsafe.Pointer(&w)) = float32(v)
case 8:
*(*float64)(unsafe.Pointer(&w)) = v
}
return Value{typ, unsafe.Pointer(w), f | flag(typ.Kind())<<flagKindShift}
}
// makeComplex returns a Value of type t equal to v (possibly truncated to complex64),
// where t is a complex64 or complex128 type.
func makeComplex(f flag, v complex128, t Type) Value {
typ := t.common()
if typ.size > ptrSize {
ptr := unsafe_New(typ)
switch typ.size {
case 8:
*(*complex64)(unsafe.Pointer(ptr)) = complex64(v)
case 16:
*(*complex128)(unsafe.Pointer(ptr)) = v
}
return Value{typ, ptr, f | flagIndir | flag(typ.Kind())<<flagKindShift}
}
// Assume ptrSize <= 8 so this must be complex64.
var w iword
*(*complex64)(unsafe.Pointer(&w)) = complex64(v)
return Value{typ, unsafe.Pointer(w), f | flag(typ.Kind())<<flagKindShift}
}
func makeString(f flag, v string, t Type) Value {
ret := New(t).Elem()
ret.SetString(v)
ret.flag = ret.flag&^flagAddr | f
return ret
}
func makeBytes(f flag, v []byte, t Type) Value {
ret := New(t).Elem()
ret.SetBytes(v)
ret.flag = ret.flag&^flagAddr | f
return ret
}
func makeRunes(f flag, v []rune, t Type) Value {
ret := New(t).Elem()
ret.setRunes(v)
ret.flag = ret.flag&^flagAddr | f
return ret
}
// These conversion functions are returned by convertOp
// for classes of conversions. For example, the first function, cvtInt,
// takes any value v of signed int type and returns the value converted
// to type t, where t is any signed or unsigned int type.
// convertOp: intXX -> [u]intXX
func cvtInt(v Value, t Type) Value {
return makeInt(v.flag&flagRO, uint64(v.Int()), t)
}
// convertOp: uintXX -> [u]intXX
func cvtUint(v Value, t Type) Value {
return makeInt(v.flag&flagRO, v.Uint(), t)
}
// convertOp: floatXX -> intXX
func cvtFloatInt(v Value, t Type) Value {
return makeInt(v.flag&flagRO, uint64(int64(v.Float())), t)
}
// convertOp: floatXX -> uintXX
func cvtFloatUint(v Value, t Type) Value {
return makeInt(v.flag&flagRO, uint64(v.Float()), t)
}
// convertOp: intXX -> floatXX
func cvtIntFloat(v Value, t Type) Value {
return makeFloat(v.flag&flagRO, float64(v.Int()), t)
}
// convertOp: uintXX -> floatXX
func cvtUintFloat(v Value, t Type) Value {
return makeFloat(v.flag&flagRO, float64(v.Uint()), t)
}
// convertOp: floatXX -> floatXX
func cvtFloat(v Value, t Type) Value {
return makeFloat(v.flag&flagRO, v.Float(), t)
}
// convertOp: complexXX -> complexXX
func cvtComplex(v Value, t Type) Value {
return makeComplex(v.flag&flagRO, v.Complex(), t)
}
// convertOp: intXX -> string
func cvtIntString(v Value, t Type) Value {
return makeString(v.flag&flagRO, string(v.Int()), t)
}
// convertOp: uintXX -> string
func cvtUintString(v Value, t Type) Value {
return makeString(v.flag&flagRO, string(v.Uint()), t)
}
// convertOp: []byte -> string
func cvtBytesString(v Value, t Type) Value {
return makeString(v.flag&flagRO, string(v.Bytes()), t)
}
// convertOp: string -> []byte
func cvtStringBytes(v Value, t Type) Value {
return makeBytes(v.flag&flagRO, []byte(v.String()), t)
}
// convertOp: []rune -> string
func cvtRunesString(v Value, t Type) Value {
return makeString(v.flag&flagRO, string(v.runes()), t)
}
// convertOp: string -> []rune
func cvtStringRunes(v Value, t Type) Value {
return makeRunes(v.flag&flagRO, []rune(v.String()), t)
}
// convertOp: direct copy
func cvtDirect(v Value, typ Type) Value {
f := v.flag
t := typ.common()
val := v.val
if f&flagAddr != 0 {
// indirect, mutable word - make a copy
ptr := unsafe_New(t)
memmove(ptr, val, t.size)
val = ptr
f &^= flagAddr
}
return Value{t, val, v.flag&flagRO | f}
}
// convertOp: concrete -> interface
func cvtT2I(v Value, typ Type) Value {
target := new(interface{})
x := valueInterface(v, false)
if typ.NumMethod() == 0 {
*target = x
} else {
ifaceE2I(typ.(*rtype), x, unsafe.Pointer(target))
}
return Value{typ.common(), unsafe.Pointer(target), v.flag&flagRO | flagIndir | flag(Interface)<<flagKindShift}
}
// convertOp: interface -> interface
func cvtI2I(v Value, typ Type) Value {
if v.IsNil() {
ret := Zero(typ)
ret.flag |= v.flag & flagRO
return ret
}
return cvtT2I(v.Elem(), typ)
}
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
// implemented in ../pkg/runtime
func chancap(ch iword) int
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
func chanclose(ch iword)
func chanlen(ch iword) int
func chanrecv(t *rtype, ch iword, nb bool) (val iword, selected, received bool)
func chansend(t *rtype, ch iword, val iword, nb bool) bool
func makechan(typ *rtype, size uint64) (ch iword)
func makemap(t *rtype) (m iword)
func mapaccess(t *rtype, m iword, key iword) (val iword, ok bool)
func mapassign(t *rtype, m iword, key, val iword, ok bool)
func mapiterinit(t *rtype, m iword) *byte
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
func mapiterkey(it *byte) (key iword, ok bool)
func mapiternext(it *byte)
func maplen(m iword) int
reflect: more efficient; cannot Set result of NewValue anymore * Reduces malloc counts during gob encoder/decoder test from 6/6 to 3/5. The current reflect uses Set to mean two subtly different things. (1) If you have a reflect.Value v, it might just represent itself (as in v = reflect.NewValue(42)), in which case calling v.Set only changed v, not any other data in the program. (2) If you have a reflect Value v derived from a pointer or a slice (as in x := []int{42}; v = reflect.NewValue(x).Index(0)), v represents the value held there. Changing x[0] affects the value returned by v.Int(), and calling v.Set affects x[0]. This was not really by design; it just happened that way. The motivation for the new reflect implementation was to remove mallocs. The use case (1) has an implicit malloc inside it. If you can do: v := reflect.NewValue(0) v.Set(42) i := v.Int() // i = 42 then that implies that v is referring to some underlying chunk of memory in order to remember the 42; that is, NewValue must have allocated some memory. Almost all the time you are using reflect the goal is to inspect or to change other data, not to manipulate data stored solely inside a reflect.Value. This CL removes use case (1), so that an assignable reflect.Value must always refer to some other piece of data in the program. Put another way, removing this case would make v := reflect.NewValue(0) v.Set(42) as illegal as 0 = 42. It would also make this illegal: x := 0 v := reflect.NewValue(x) v.Set(42) for the same reason. (Note that right now, v.Set(42) "succeeds" but does not change the value of x.) If you really wanted to make v refer to x, you'd start with &x and dereference it: x := 0 v := reflect.NewValue(&x).Elem() // v = *&x v.Set(42) It's pretty rare, except in tests, to want to use NewValue and then call Set to change the Value itself instead of some other piece of data in the program. I haven't seen it happen once yet while making the tree build with this change. For the same reasons, reflect.Zero (formerly reflect.MakeZero) would also return an unassignable, unaddressable value. This invalidates the (awkward) idiom: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.PointTo(v) which, when the API changed, turned into: pv := ... some Ptr Value we have ... v := reflect.Zero(pv.Type().Elem()) pv.Set(v.Addr()) In both, it is far from clear what the code is trying to do. Now that it is possible, this CL adds reflect.New(Type) Value that does the obvious thing (same as Go's new), so this code would be replaced by: pv := ... some Ptr Value we have ... pv.Set(reflect.New(pv.Type().Elem())) The changes just described can be confusing to think about, but I believe it is because the old API was confusing - it was conflating two different kinds of Values - and that the new API by itself is pretty simple: you can only Set (or call Addr on) a Value if it actually addresses some real piece of data; that is, only if it is the result of dereferencing a Ptr or indexing a Slice. If you really want the old behavior, you'd get it by translating: v := reflect.NewValue(x) into v := reflect.New(reflect.Typeof(x)).Elem() v.Set(reflect.NewValue(x)) Gofix will not be able to help with this, because whether and how to change the code depends on whether the original code meant use (1) or use (2), so the developer has to read and think about the code. You can see the effect on packages in the tree in https://golang.org/cl/4423043/. R=r CC=golang-dev https://golang.org/cl/4435042
2011-04-18 12:35:33 -06:00
func call(fn, arg unsafe.Pointer, n uint32)
func ifaceE2I(t *rtype, src interface{}, dst unsafe.Pointer)
// Dummy annotation marking that the value x escapes,
// for use in cases where the reflect code is so clever that
// the compiler cannot follow.
func escapes(x interface{}) {
if dummy.b {
dummy.x = x
}
}
var dummy struct {
b bool
x interface{}
}