1
0
mirror of https://github.com/golang/go synced 2024-11-18 20:24:41 -07:00
go/internal/lsp/source/identifier.go

409 lines
11 KiB
Go
Raw Normal View History

// Copyright 2018 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 source
import (
"context"
"fmt"
"go/ast"
"go/token"
"go/types"
"strconv"
"golang.org/x/tools/internal/event"
"golang.org/x/tools/internal/lsp/protocol"
errors "golang.org/x/xerrors"
)
// IdentifierInfo holds information about an identifier in Go source.
type IdentifierInfo struct {
Name string
Snapshot Snapshot
mappedRange
Type struct {
mappedRange
Object types.Object
}
Declaration Declaration
ident *ast.Ident
// enclosing is an expression used to determine the link anchor for an
// identifier. If it's a named type, it should be exported.
enclosing types.Type
pkg Package
qf types.Qualifier
}
type Declaration struct {
MappedRange []mappedRange
node ast.Node
obj types.Object
// typeSwitchImplicit indicates that the declaration is in an implicit
// type switch. Its type is the type of the variable on the right-hand
// side of the type switch.
typeSwitchImplicit types.Type
}
// Identifier returns identifier information for a position
// in a file, accounting for a potentially incomplete selector.
func Identifier(ctx context.Context, snapshot Snapshot, fh FileHandle, pos protocol.Position) (*IdentifierInfo, error) {
ctx, done := event.Start(ctx, "source.Identifier")
defer done()
pkg, pgf, err := getParsedFile(ctx, snapshot, fh, NarrowestPackage)
if err != nil {
return nil, fmt.Errorf("getting file for Identifier: %w", err)
}
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
spn, err := pgf.Mapper.PointSpan(pos)
if err != nil {
return nil, err
}
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
rng, err := spn.Range(pgf.Mapper.Converter)
if err != nil {
return nil, err
}
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
return findIdentifier(ctx, snapshot, pkg, pgf.File, rng.Start)
}
var ErrNoIdentFound = errors.New("no identifier found")
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
func findIdentifier(ctx context.Context, s Snapshot, pkg Package, file *ast.File, pos token.Pos) (*IdentifierInfo, error) {
// Handle import specs separately, as there is no formal position for a
// package declaration.
if result, err := importSpec(s, pkg, file, pos); result != nil || err != nil {
if s.View().Options().ImportShortcut.ShowDefinition() {
return result, err
}
return nil, nil
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
path := pathEnclosingObjNode(file, pos)
if path == nil {
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
return nil, ErrNoIdentFound
}
view := s.View()
qf := qualifier(file, pkg.GetTypes(), pkg.GetTypesInfo())
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
ident, _ := path[0].(*ast.Ident)
if ident == nil {
return nil, ErrNoIdentFound
}
// Special case for package declarations, since they have no
// corresponding types.Object.
if ident == file.Name {
rng, err := posToMappedRange(view, pkg, file.Name.Pos(), file.Name.End())
if err != nil {
return nil, err
}
var declAST *ast.File
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
for _, pgf := range pkg.CompiledGoFiles() {
if pgf.File.Doc != nil {
declAST = pgf.File
}
}
// If there's no package documentation, just use current file.
if declAST == nil {
declAST = file
}
declRng, err := posToMappedRange(view, pkg, declAST.Name.Pos(), declAST.Name.End())
if err != nil {
return nil, err
}
return &IdentifierInfo{
Name: file.Name.Name,
ident: file.Name,
mappedRange: rng,
pkg: pkg,
qf: qf,
Snapshot: s,
Declaration: Declaration{
node: declAST.Name,
MappedRange: []mappedRange{declRng},
},
}, nil
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
result := &IdentifierInfo{
Snapshot: s,
qf: qf,
pkg: pkg,
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
ident: ident,
enclosing: searchForEnclosing(pkg.GetTypesInfo(), path),
}
var wasEmbeddedField bool
for _, n := range path[1:] {
if field, ok := n.(*ast.Field); ok {
wasEmbeddedField = len(field.Names) == 0
break
}
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
result.Name = result.ident.Name
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
var err error
if result.mappedRange, err = posToMappedRange(view, pkg, result.ident.Pos(), result.ident.End()); err != nil {
return nil, err
}
result.Declaration.obj = pkg.GetTypesInfo().ObjectOf(result.ident)
if result.Declaration.obj == nil {
// If there was no types.Object for the declaration, there might be an
// implicit local variable declaration in a type switch.
if objs, typ := typeSwitchImplicits(pkg, path); len(objs) > 0 {
// There is no types.Object for the declaration of an implicit local variable,
// but all of the types.Objects associated with the usages of this variable can be
// used to connect it back to the declaration.
// Preserve the first of these objects and treat it as if it were the declaring object.
result.Declaration.obj = objs[0]
result.Declaration.typeSwitchImplicit = typ
} else {
// Probably a type error.
return nil, errors.Errorf("no object for ident %v", result.Name)
}
}
// Handle builtins separately.
if result.Declaration.obj.Parent() == types.Universe {
builtin, err := view.BuiltinPackage(ctx)
if err != nil {
return nil, err
}
decl, ok := builtin.Package().Scope.Lookup(result.Name).Decl.(ast.Node)
if !ok {
return nil, errors.Errorf("no declaration for %s", result.Name)
}
result.Declaration.node = decl
// The builtin package isn't in the dependency graph, so the usual utilities
// won't work here.
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
rng := newMappedRange(view.Session().Cache().FileSet(), builtin.ParsedFile().Mapper, decl.Pos(), decl.Pos()+token.Pos(len(result.Name)))
result.Declaration.MappedRange = append(result.Declaration.MappedRange, rng)
return result, nil
}
if wasEmbeddedField {
// The original position was on the embedded field declaration, so we
// try to dig out the type and jump to that instead.
if v, ok := result.Declaration.obj.(*types.Var); ok {
if typObj := typeToObject(v.Type()); typObj != nil {
result.Declaration.obj = typObj
}
}
}
rng, err := objToMappedRange(view, pkg, result.Declaration.obj)
if err != nil {
return nil, err
}
result.Declaration.MappedRange = append(result.Declaration.MappedRange, rng)
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
if result.Declaration.node, err = objToDecl(ctx, s, pkg, result.Declaration.obj); err != nil {
return nil, err
}
typ := pkg.GetTypesInfo().TypeOf(result.ident)
if typ == nil {
return result, nil
}
result.Type.Object = typeToObject(typ)
if result.Type.Object != nil {
// Identifiers with the type "error" are a special case with no position.
if hasErrorType(result.Type.Object) {
return result, nil
}
if result.Type.mappedRange, err = objToMappedRange(view, pkg, result.Type.Object); err != nil {
return nil, err
}
}
return result, nil
}
func searchForEnclosing(info *types.Info, path []ast.Node) types.Type {
for _, n := range path {
switch n := n.(type) {
case *ast.SelectorExpr:
if sel, ok := info.Selections[n]; ok {
recv := deref(sel.Recv())
// Keep track of the last exported type seen.
var exported types.Type
if named, ok := recv.(*types.Named); ok && named.Obj().Exported() {
exported = named
}
// We don't want the last element, as that's the field or
// method itself.
for _, index := range sel.Index()[:len(sel.Index())-1] {
if r, ok := recv.Underlying().(*types.Struct); ok {
recv = deref(r.Field(index).Type())
if named, ok := recv.(*types.Named); ok && named.Obj().Exported() {
exported = named
}
}
}
return exported
}
case *ast.CompositeLit:
if t, ok := info.Types[n]; ok {
return t.Type
}
case *ast.TypeSpec:
if _, ok := n.Type.(*ast.StructType); ok {
if t, ok := info.Defs[n.Name]; ok {
return t.Type()
}
}
}
}
return nil
}
func typeToObject(typ types.Type) types.Object {
switch typ := typ.(type) {
case *types.Named:
return typ.Obj()
case *types.Pointer:
return typeToObject(typ.Elem())
default:
return nil
}
}
func hasErrorType(obj types.Object) bool {
return types.IsInterface(obj.Type()) && obj.Pkg() == nil && obj.Name() == "error"
}
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
func objToDecl(ctx context.Context, snapshot Snapshot, srcPkg Package, obj types.Object) (ast.Decl, error) {
pgf, _, err := findPosInPackage(snapshot.View(), srcPkg, obj.Pos())
if err != nil {
return nil, err
}
internal/lsp: replace ParseGoHandle with concrete data ParseGoHandles serve two purposes: they pin cache entries so that redundant calculations are cached, and they allow users to obtain the actual parsed AST. The former is an implementation detail, and the latter turns out to just be an annoyance. Parsed Go files are obtained from two places. By far the most common is from a type checked package. But a type checked package must by definition have already parsed all the files it contains, so the PGH is already computed and cannot have failed. Type checked packages can simply return the parsed file without requiring a separate Check operation. We do want to pin the cache entries in this case, which I've done by holding on to the PGH in cache.pkg. There are some cases where we directly parse a file, such as for the FoldingRange LSP call, which doesn't need type information. Those parses can actually fail, so we do need an error check. But we don't need the PGH; in all cases we are immediately using and discarding it. So it turns out we don't actually need the PGH type at all, at least not in the public API. Instead, we can pass around a concrete struct that has the various pieces of data directly available. This uncovered a bug in typeCheck: it should fail if it encounters any real errors. Change-Id: I203bf2dd79d5d65c01392d69c2cf4f7744fde7fc Reviewed-on: https://go-review.googlesource.com/c/tools/+/244021 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-07-21 13:15:06 -06:00
posToDecl, err := snapshot.PosToDecl(ctx, pgf)
if err != nil {
return nil, err
}
return posToDecl[obj.Pos()], nil
}
// importSpec handles positions inside of an *ast.ImportSpec.
func importSpec(s Snapshot, pkg Package, file *ast.File, pos token.Pos) (*IdentifierInfo, error) {
var imp *ast.ImportSpec
for _, spec := range file.Imports {
if spec.Path.Pos() <= pos && pos < spec.Path.End() {
imp = spec
}
}
if imp == nil {
return nil, nil
}
importPath, err := strconv.Unquote(imp.Path.Value)
if err != nil {
return nil, errors.Errorf("import path not quoted: %s (%v)", imp.Path.Value, err)
}
result := &IdentifierInfo{
Snapshot: s,
Name: importPath,
pkg: pkg,
}
if result.mappedRange, err = posToMappedRange(s.View(), pkg, imp.Path.Pos(), imp.Path.End()); err != nil {
return nil, err
}
// Consider the "declaration" of an import spec to be the imported package.
importedPkg, err := pkg.GetImport(importPath)
if err != nil {
return nil, err
}
// Return all of the files in the package as the definition of the import spec.
for _, dst := range importedPkg.GetSyntax() {
rng, err := posToMappedRange(s.View(), pkg, dst.Pos(), dst.End())
if err != nil {
return nil, err
}
result.Declaration.MappedRange = append(result.Declaration.MappedRange, rng)
}
result.Declaration.node = imp
return result, nil
}
// typeSwitchImplicits returns all the implicit type switch objects that
// correspond to the leaf *ast.Ident. It also returns the original type
// associated with the identifier (outside of a case clause).
func typeSwitchImplicits(pkg Package, path []ast.Node) ([]types.Object, types.Type) {
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
ident, _ := path[0].(*ast.Ident)
if ident == nil {
return nil, nil
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
var (
ts *ast.TypeSwitchStmt
assign *ast.AssignStmt
cc *ast.CaseClause
obj = pkg.GetTypesInfo().ObjectOf(ident)
)
// Walk our ancestors to determine if our leaf ident refers to a
// type switch variable, e.g. the "a" from "switch a := b.(type)".
Outer:
for i := 1; i < len(path); i++ {
switch n := path[i].(type) {
case *ast.AssignStmt:
// Check if ident is the "a" in "a := foo.(type)". The "a" in
// this case has no types.Object, so check for ident equality.
if len(n.Lhs) == 1 && n.Lhs[0] == ident {
assign = n
}
case *ast.CaseClause:
// Check if ident is a use of "a" within a case clause. Each
// case clause implicitly maps "a" to a different types.Object,
// so check if ident's object is the case clause's implicit
// object.
if obj != nil && pkg.GetTypesInfo().Implicits[n] == obj {
cc = n
}
case *ast.TypeSwitchStmt:
// Look for the type switch that owns our previously found
// *ast.AssignStmt or *ast.CaseClause.
if n.Assign == assign {
ts = n
break Outer
}
for _, stmt := range n.Body.List {
if stmt == cc {
ts = n
break Outer
}
}
}
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
if ts == nil {
return nil, nil
}
internal/lsp: refactor find-references and rename The main goal is to push the package variant logic from internal/lsp into internal/lsp/source so all users of internal/lsp/source benefit. "references" and "rename" now have top-level source.References() and source.Rename() entry points (as opposed to hanging off source.Identifier()). I expanded objectsAtProtocolPos() to know about implicit objects (type switch and import spec), and to handle *ast.ImportSpec generically. This gets rid of special case handling of *types.PkgName in various places. The biggest practical benefit, though, is that "references" no longer needs to compute the objectpath for every types.Object comparison it does, instead using direct types.Object equality. This speeds up "references" and "rename" a lot. Two other notable improvements that fell out of not using source.Identifier()'s logic: - Finding references on an embedded field now shows references to the field, not the type being embedded. - Finding references on an imported object now works correctly (previously it searched the importing package's dependents rather than the imported package's dependents). Finally, I refactored findIdentifier() to use pathEnclosingObjNode() instead of astutil.PathEnclosingInterval. Now we only need a single call to get the path because pathEnclosingObjNode() has the "try pos || try pos-1" logic built in. Change-Id: I667be9bed6ad83912404b90257c5c1485b3a7025 Reviewed-on: https://go-review.googlesource.com/c/tools/+/211999 Run-TryBot: Muir Manders <muir@mnd.rs> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2019-12-17 22:06:31 -07:00
// Our leaf ident refers to a type switch variable. Fan out to the
// type switch's implicit case clause objects.
var objs []types.Object
for _, cc := range ts.Body.List {
if ccObj := pkg.GetTypesInfo().Implicits[cc]; ccObj != nil {
objs = append(objs, ccObj)
}
}
// The right-hand side of a type switch should only have one
// element, and we need to track its type in order to generate
// hover information for implicit type switch variables.
var typ types.Type
if assign, ok := ts.Assign.(*ast.AssignStmt); ok && len(assign.Rhs) == 1 {
if rhs := assign.Rhs[0].(*ast.TypeAssertExpr); ok {
typ = pkg.GetTypesInfo().TypeOf(rhs.X)
}
}
return objs, typ
}