1
0
mirror of https://github.com/golang/go synced 2024-11-18 23:44:43 -07:00
go/astutil/enclosing.go

626 lines
16 KiB
Go
Raw Normal View History

// Copyright 2013 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 astutil
// This file defines utilities for working with source positions.
import (
"fmt"
"go/ast"
"go/token"
"sort"
)
// PathEnclosingInterval returns the node that encloses the source
// interval [start, end), and all its ancestors up to the AST root.
//
// The definition of "enclosing" used by this function considers
// additional whitespace abutting a node to be enclosed by it.
// In this example:
//
// z := x + y // add them
// <-A->
// <----B----->
//
// the ast.BinaryExpr(+) node is considered to enclose interval B
// even though its [Pos()..End()) is actually only interval A.
// This behaviour makes user interfaces more tolerant of imperfect
// input.
//
// This function treats tokens as nodes, though they are not included
// in the result. e.g. PathEnclosingInterval("+") returns the
// enclosing ast.BinaryExpr("x + y").
//
// If start==end, the 1-char interval following start is used instead.
//
// The 'exact' result is true if the interval contains only path[0]
// and perhaps some adjacent whitespace. It is false if the interval
// overlaps multiple children of path[0], or if it contains only
// interior whitespace of path[0].
// In this example:
//
// z := x + y // add them
// <--C--> <---E-->
// ^
// D
//
// intervals C, D and E are inexact. C is contained by the
// z-assignment statement, because it spans three of its children (:=,
// x, +). So too is the 1-char interval D, because it contains only
// interior whitespace of the assignment. E is considered interior
// whitespace of the BlockStmt containing the assignment.
//
// Precondition: [start, end) both lie within the same file as root.
// TODO(adonovan): return (nil, false) in this case and remove precond.
// Requires FileSet; see loader.tokenFileContainsPos.
//
// Postcondition: path is never nil; it always contains at least 'root'.
//
func PathEnclosingInterval(root *ast.File, start, end token.Pos) (path []ast.Node, exact bool) {
// fmt.Printf("EnclosingInterval %d %d\n", start, end) // debugging
// Precondition: node.[Pos..End) and adjoining whitespace contain [start, end).
var visit func(node ast.Node) bool
visit = func(node ast.Node) bool {
path = append(path, node)
nodePos := node.Pos()
nodeEnd := node.End()
// fmt.Printf("visit(%T, %d, %d)\n", node, nodePos, nodeEnd) // debugging
// Intersect [start, end) with interval of node.
if start < nodePos {
start = nodePos
}
if end > nodeEnd {
end = nodeEnd
}
// Find sole child that contains [start, end).
children := childrenOf(node)
l := len(children)
for i, child := range children {
// [childPos, childEnd) is unaugmented interval of child.
childPos := child.Pos()
childEnd := child.End()
// [augPos, augEnd) is whitespace-augmented interval of child.
augPos := childPos
augEnd := childEnd
if i > 0 {
augPos = children[i-1].End() // start of preceding whitespace
}
if i < l-1 {
nextChildPos := children[i+1].Pos()
// Does [start, end) lie between child and next child?
if start >= augEnd && end <= nextChildPos {
return false // inexact match
}
augEnd = nextChildPos // end of following whitespace
}
// fmt.Printf("\tchild %d: [%d..%d)\tcontains interval [%d..%d)?\n",
// i, augPos, augEnd, start, end) // debugging
// Does augmented child strictly contain [start, end)?
if augPos <= start && end <= augEnd {
_, isToken := child.(tokenNode)
return isToken || visit(child)
}
// Does [start, end) overlap multiple children?
// i.e. left-augmented child contains start
// but LR-augmented child does not contain end.
if start < childEnd && end > augEnd {
break
}
}
// No single child contained [start, end),
// so node is the result. Is it exact?
// (It's tempting to put this condition before the
// child loop, but it gives the wrong result in the
// case where a node (e.g. ExprStmt) and its sole
// child have equal intervals.)
if start == nodePos && end == nodeEnd {
return true // exact match
}
return false // inexact: overlaps multiple children
}
if start > end {
start, end = end, start
}
if start < root.End() && end > root.Pos() {
if start == end {
end = start + 1 // empty interval => interval of size 1
}
exact = visit(root)
// Reverse the path:
for i, l := 0, len(path); i < l/2; i++ {
path[i], path[l-1-i] = path[l-1-i], path[i]
}
} else {
// Selection lies within whitespace preceding the
// first (or following the last) declaration in the file.
// The result nonetheless always includes the ast.File.
path = append(path, root)
}
return
}
// tokenNode is a dummy implementation of ast.Node for a single token.
// They are used transiently by PathEnclosingInterval but never escape
// this package.
//
type tokenNode struct {
pos token.Pos
end token.Pos
}
func (n tokenNode) Pos() token.Pos {
return n.pos
}
func (n tokenNode) End() token.Pos {
return n.end
}
func tok(pos token.Pos, len int) ast.Node {
return tokenNode{pos, pos + token.Pos(len)}
}
// childrenOf returns the direct non-nil children of ast.Node n.
// It may include fake ast.Node implementations for bare tokens.
// it is not safe to call (e.g.) ast.Walk on such nodes.
//
func childrenOf(n ast.Node) []ast.Node {
var children []ast.Node
// First add nodes for all true subtrees.
ast.Inspect(n, func(node ast.Node) bool {
if node == n { // push n
return true // recur
}
if node != nil { // push child
children = append(children, node)
}
return false // no recursion
})
// Then add fake Nodes for bare tokens.
switch n := n.(type) {
case *ast.ArrayType:
children = append(children,
tok(n.Lbrack, len("[")),
tok(n.Elt.End(), len("]")))
case *ast.AssignStmt:
children = append(children,
tok(n.TokPos, len(n.Tok.String())))
case *ast.BasicLit:
children = append(children,
tok(n.ValuePos, len(n.Value)))
case *ast.BinaryExpr:
children = append(children, tok(n.OpPos, len(n.Op.String())))
case *ast.BlockStmt:
children = append(children,
tok(n.Lbrace, len("{")),
tok(n.Rbrace, len("}")))
case *ast.BranchStmt:
children = append(children,
tok(n.TokPos, len(n.Tok.String())))
case *ast.CallExpr:
children = append(children,
tok(n.Lparen, len("(")),
tok(n.Rparen, len(")")))
if n.Ellipsis != 0 {
children = append(children, tok(n.Ellipsis, len("...")))
}
case *ast.CaseClause:
if n.List == nil {
children = append(children,
tok(n.Case, len("default")))
} else {
children = append(children,
tok(n.Case, len("case")))
}
children = append(children, tok(n.Colon, len(":")))
case *ast.ChanType:
switch n.Dir {
case ast.RECV:
children = append(children, tok(n.Begin, len("<-chan")))
case ast.SEND:
children = append(children, tok(n.Begin, len("chan<-")))
case ast.RECV | ast.SEND:
children = append(children, tok(n.Begin, len("chan")))
}
case *ast.CommClause:
if n.Comm == nil {
children = append(children,
tok(n.Case, len("default")))
} else {
children = append(children,
tok(n.Case, len("case")))
}
children = append(children, tok(n.Colon, len(":")))
case *ast.Comment:
// nop
case *ast.CommentGroup:
// nop
case *ast.CompositeLit:
children = append(children,
tok(n.Lbrace, len("{")),
tok(n.Rbrace, len("{")))
case *ast.DeclStmt:
// nop
case *ast.DeferStmt:
children = append(children,
tok(n.Defer, len("defer")))
case *ast.Ellipsis:
children = append(children,
tok(n.Ellipsis, len("...")))
case *ast.EmptyStmt:
// nop
case *ast.ExprStmt:
// nop
case *ast.Field:
// TODO(adonovan): Field.{Doc,Comment,Tag}?
case *ast.FieldList:
children = append(children,
tok(n.Opening, len("(")),
tok(n.Closing, len(")")))
case *ast.File:
// TODO test: Doc
children = append(children,
tok(n.Package, len("package")))
case *ast.ForStmt:
children = append(children,
tok(n.For, len("for")))
case *ast.FuncDecl:
// TODO(adonovan): FuncDecl.Comment?
// Uniquely, FuncDecl breaks the invariant that
// preorder traversal yields tokens in lexical order:
// in fact, FuncDecl.Recv precedes FuncDecl.Type.Func.
//
// As a workaround, we inline the case for FuncType
// here and order things correctly.
//
children = nil // discard ast.Walk(FuncDecl) info subtrees
children = append(children, tok(n.Type.Func, len("func")))
if n.Recv != nil {
children = append(children, n.Recv)
}
children = append(children, n.Name)
if n.Type.Params != nil {
children = append(children, n.Type.Params)
}
if n.Type.Results != nil {
children = append(children, n.Type.Results)
}
if n.Body != nil {
children = append(children, n.Body)
}
case *ast.FuncLit:
// nop
case *ast.FuncType:
if n.Func != 0 {
children = append(children,
tok(n.Func, len("func")))
}
case *ast.GenDecl:
children = append(children,
tok(n.TokPos, len(n.Tok.String())))
if n.Lparen != 0 {
children = append(children,
tok(n.Lparen, len("(")),
tok(n.Rparen, len(")")))
}
case *ast.GoStmt:
children = append(children,
tok(n.Go, len("go")))
case *ast.Ident:
children = append(children,
tok(n.NamePos, len(n.Name)))
case *ast.IfStmt:
children = append(children,
tok(n.If, len("if")))
case *ast.ImportSpec:
// TODO(adonovan): ImportSpec.{Doc,EndPos}?
case *ast.IncDecStmt:
children = append(children,
tok(n.TokPos, len(n.Tok.String())))
case *ast.IndexExpr:
children = append(children,
tok(n.Lbrack, len("{")),
tok(n.Rbrack, len("}")))
case *ast.InterfaceType:
children = append(children,
tok(n.Interface, len("interface")))
case *ast.KeyValueExpr:
children = append(children,
tok(n.Colon, len(":")))
case *ast.LabeledStmt:
children = append(children,
tok(n.Colon, len(":")))
case *ast.MapType:
children = append(children,
tok(n.Map, len("map")))
case *ast.ParenExpr:
children = append(children,
tok(n.Lparen, len("(")),
tok(n.Rparen, len(")")))
case *ast.RangeStmt:
children = append(children,
tok(n.For, len("for")),
tok(n.TokPos, len(n.Tok.String())))
case *ast.ReturnStmt:
children = append(children,
tok(n.Return, len("return")))
case *ast.SelectStmt:
children = append(children,
tok(n.Select, len("select")))
case *ast.SelectorExpr:
// nop
case *ast.SendStmt:
children = append(children,
tok(n.Arrow, len("<-")))
case *ast.SliceExpr:
children = append(children,
tok(n.Lbrack, len("[")),
tok(n.Rbrack, len("]")))
case *ast.StarExpr:
children = append(children, tok(n.Star, len("*")))
case *ast.StructType:
children = append(children, tok(n.Struct, len("struct")))
case *ast.SwitchStmt:
children = append(children, tok(n.Switch, len("switch")))
case *ast.TypeAssertExpr:
children = append(children,
tok(n.Lparen-1, len(".")),
tok(n.Lparen, len("(")),
tok(n.Rparen, len(")")))
case *ast.TypeSpec:
// TODO(adonovan): TypeSpec.{Doc,Comment}?
case *ast.TypeSwitchStmt:
children = append(children, tok(n.Switch, len("switch")))
case *ast.UnaryExpr:
children = append(children, tok(n.OpPos, len(n.Op.String())))
case *ast.ValueSpec:
// TODO(adonovan): ValueSpec.{Doc,Comment}?
default:
// Includes *ast.BadDecl, *ast.BadExpr, *ast.BadStmt.
panic(fmt.Sprintf("unexpected node type %T", n))
}
// TODO(adonovan): opt: merge the logic of ast.Inspect() into
// the switch above so we can make interleaved callbacks for
// both Nodes and Tokens in the right order and avoid the need
// to sort.
sort.Sort(byPos(children))
return children
}
type byPos []ast.Node
func (sl byPos) Len() int {
return len(sl)
}
func (sl byPos) Less(i, j int) bool {
return sl[i].Pos() < sl[j].Pos()
}
func (sl byPos) Swap(i, j int) {
sl[i], sl[j] = sl[j], sl[i]
}
// NodeDescription returns a description of the concrete type of n suitable
// for a user interface.
//
// TODO(adonovan): in some cases (e.g. Field, FieldList, Ident,
// StarExpr) we could be much more specific given the path to the AST
// root. Perhaps we should do that.
//
func NodeDescription(n ast.Node) string {
switch n := n.(type) {
case *ast.ArrayType:
return "array type"
case *ast.AssignStmt:
return "assignment"
case *ast.BadDecl:
return "bad declaration"
case *ast.BadExpr:
return "bad expression"
case *ast.BadStmt:
return "bad statement"
case *ast.BasicLit:
return "basic literal"
case *ast.BinaryExpr:
return fmt.Sprintf("binary %s operation", n.Op)
case *ast.BlockStmt:
return "block"
case *ast.BranchStmt:
switch n.Tok {
case token.BREAK:
return "break statement"
case token.CONTINUE:
return "continue statement"
case token.GOTO:
return "goto statement"
case token.FALLTHROUGH:
return "fall-through statement"
}
case *ast.CallExpr:
return "function call (or conversion)"
case *ast.CaseClause:
return "case clause"
case *ast.ChanType:
return "channel type"
case *ast.CommClause:
return "communication clause"
case *ast.Comment:
return "comment"
case *ast.CommentGroup:
return "comment group"
case *ast.CompositeLit:
return "composite literal"
case *ast.DeclStmt:
return NodeDescription(n.Decl) + " statement"
case *ast.DeferStmt:
return "defer statement"
case *ast.Ellipsis:
return "ellipsis"
case *ast.EmptyStmt:
return "empty statement"
case *ast.ExprStmt:
return "expression statement"
case *ast.Field:
// Can be any of these:
// struct {x, y int} -- struct field(s)
// struct {T} -- anon struct field
// interface {I} -- interface embedding
// interface {f()} -- interface method
// func (A) func(B) C -- receiver, param(s), result(s)
return "field/method/parameter"
case *ast.FieldList:
return "field/method/parameter list"
case *ast.File:
return "source file"
case *ast.ForStmt:
return "for loop"
case *ast.FuncDecl:
return "function declaration"
case *ast.FuncLit:
return "function literal"
case *ast.FuncType:
return "function type"
case *ast.GenDecl:
switch n.Tok {
case token.IMPORT:
return "import declaration"
case token.CONST:
return "constant declaration"
case token.TYPE:
return "type declaration"
case token.VAR:
return "variable declaration"
}
case *ast.GoStmt:
return "go statement"
case *ast.Ident:
return "identifier"
case *ast.IfStmt:
return "if statement"
case *ast.ImportSpec:
return "import specification"
case *ast.IncDecStmt:
if n.Tok == token.INC {
return "increment statement"
}
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
return "decrement statement"
case *ast.IndexExpr:
return "index expression"
case *ast.InterfaceType:
return "interface type"
case *ast.KeyValueExpr:
return "key/value association"
case *ast.LabeledStmt:
return "statement label"
case *ast.MapType:
return "map type"
case *ast.Package:
return "package"
case *ast.ParenExpr:
return "parenthesized " + NodeDescription(n.X)
case *ast.RangeStmt:
return "range loop"
case *ast.ReturnStmt:
return "return statement"
case *ast.SelectStmt:
return "select statement"
case *ast.SelectorExpr:
return "selector"
case *ast.SendStmt:
return "channel send"
case *ast.SliceExpr:
return "slice expression"
case *ast.StarExpr:
return "*-operation" // load/store expr or pointer type
case *ast.StructType:
return "struct type"
case *ast.SwitchStmt:
return "switch statement"
case *ast.TypeAssertExpr:
return "type assertion"
case *ast.TypeSpec:
return "type specification"
case *ast.TypeSwitchStmt:
return "type switch"
case *ast.UnaryExpr:
return fmt.Sprintf("unary %s operation", n.Op)
case *ast.ValueSpec:
return "value specification"
}
panic(fmt.Sprintf("unexpected node type: %T", n))
}