blob: 2de739efad440c06658a7d9dddf6d1cd1cd2a8ca [file] [log] [blame] [edit]
// 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"
}
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))
}