summaryrefslogtreecommitdiff
path: root/src/pkg/go/parser/parser.go
diff options
context:
space:
mode:
Diffstat (limited to 'src/pkg/go/parser/parser.go')
-rw-r--r--src/pkg/go/parser/parser.go249
1 files changed, 155 insertions, 94 deletions
diff --git a/src/pkg/go/parser/parser.go b/src/pkg/go/parser/parser.go
index 20e505d97..a021a5abe 100644
--- a/src/pkg/go/parser/parser.go
+++ b/src/pkg/go/parser/parser.go
@@ -28,7 +28,7 @@ type parser struct {
// Tracing/debugging
mode Mode // parsing mode
trace bool // == (mode & Trace != 0)
- indent uint // indentation used for tracing output
+ indent int // indentation used for tracing output
// Comments
comments []*ast.CommentGroup
@@ -56,7 +56,7 @@ type parser struct {
unresolved []*ast.Ident // unresolved identifiers
imports []*ast.ImportSpec // list of imports
- // Label scope
+ // Label scopes
// (maintained by open/close LabelScope)
labelScope *ast.Scope // label scope for current function
targetStack [][]*ast.Ident // stack of unresolved labels
@@ -75,14 +75,6 @@ func (p *parser) init(fset *token.FileSet, filename string, src []byte, mode Mod
p.trace = mode&Trace != 0 // for convenience (p.trace is used frequently)
p.next()
-
- // set up the pkgScope here (as opposed to in parseFile) because
- // there are other parser entry points (ParseExpr, etc.)
- p.openScope()
- p.pkgScope = p.topScope
-
- // for the same reason, set up a label scope
- p.openLabelScope()
}
// ----------------------------------------------------------------------------
@@ -157,7 +149,7 @@ func (p *parser) shortVarDecl(decl *ast.AssignStmt, list []ast.Expr) {
}
}
} else {
- p.errorExpected(x.Pos(), "identifier")
+ p.errorExpected(x.Pos(), "identifier on left side of :=")
}
}
if n == 0 && p.mode&DeclarationErrors != 0 {
@@ -170,7 +162,12 @@ func (p *parser) shortVarDecl(decl *ast.AssignStmt, list []ast.Expr) {
// internal consistency.
var unresolved = new(ast.Object)
-func (p *parser) resolve(x ast.Expr) {
+// If x is an identifier, tryResolve attempts to resolve x by looking up
+// the object it denotes. If no object is found and collectUnresolved is
+// set, x is marked as unresolved and collected in the list of unresolved
+// identifiers.
+//
+func (p *parser) tryResolve(x ast.Expr, collectUnresolved bool) {
// nothing to do if x is not an identifier or the blank identifier
ident, _ := x.(*ast.Ident)
if ident == nil {
@@ -191,23 +188,30 @@ func (p *parser) resolve(x ast.Expr) {
// must be found either in the file scope, package scope
// (perhaps in another file), or universe scope --- collect
// them so that they can be resolved later
- ident.Obj = unresolved
- p.unresolved = append(p.unresolved, ident)
+ if collectUnresolved {
+ ident.Obj = unresolved
+ p.unresolved = append(p.unresolved, ident)
+ }
+}
+
+func (p *parser) resolve(x ast.Expr) {
+ p.tryResolve(x, true)
}
// ----------------------------------------------------------------------------
// Parsing support
func (p *parser) printTrace(a ...interface{}) {
- const dots = ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " +
- ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . "
- const n = uint(len(dots))
+ const dots = ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . "
+ const n = len(dots)
pos := p.file.Position(p.pos)
fmt.Printf("%5d:%3d: ", pos.Line, pos.Column)
i := 2 * p.indent
- for ; i > n; i -= n {
+ for i > n {
fmt.Print(dots)
+ i -= n
}
+ // i <= n
fmt.Print(dots[0:i])
fmt.Println(a...)
}
@@ -218,7 +222,7 @@ func trace(p *parser, msg string) *parser {
return p
}
-// Usage pattern: defer un(trace(p, "..."));
+// Usage pattern: defer un(trace(p, "..."))
func un(p *parser) {
p.indent--
p.printTrace(")")
@@ -304,14 +308,14 @@ func (p *parser) consumeCommentGroup(n int) (comments *ast.CommentGroup, endline
func (p *parser) next() {
p.leadComment = nil
p.lineComment = nil
- line := p.file.Line(p.pos) // current line
+ prev := p.pos
p.next0()
if p.tok == token.COMMENT {
var comment *ast.CommentGroup
var endline int
- if p.file.Line(p.pos) == line {
+ if p.file.Line(p.pos) == p.file.Line(prev) {
// The comment is on same line as the previous token; it
// cannot be a lead comment but may be a line comment.
comment, endline = p.consumeCommentGroup(0)
@@ -336,8 +340,26 @@ func (p *parser) next() {
}
}
+// A bailout panic is raised to indicate early termination.
+type bailout struct{}
+
func (p *parser) error(pos token.Pos, msg string) {
- p.errors.Add(p.file.Position(pos), msg)
+ epos := p.file.Position(pos)
+
+ // If AllErrors is not set, discard errors reported on the same line
+ // as the last recorded error and stop parsing if there are more than
+ // 10 errors.
+ if p.mode&AllErrors == 0 {
+ n := len(p.errors)
+ if n > 0 && p.errors[n-1].Pos.Line == epos.Line {
+ return // discard - likely a spurious error
+ }
+ if n > 10 {
+ panic(bailout{})
+ }
+ }
+
+ p.errors.Add(epos, msg)
}
func (p *parser) errorExpected(pos token.Pos, msg string) {
@@ -585,14 +607,15 @@ func (p *parser) parseTypeName() ast.Expr {
return ident
}
-func (p *parser) parseArrayType(ellipsisOk bool) ast.Expr {
+func (p *parser) parseArrayType() ast.Expr {
if p.trace {
defer un(trace(p, "ArrayType"))
}
lbrack := p.expect(token.LBRACK)
var len ast.Expr
- if ellipsisOk && p.tok == token.ELLIPSIS {
+ // always permit ellipsis for more fault-tolerant parsing
+ if p.tok == token.ELLIPSIS {
len = &ast.Ellipsis{Ellipsis: p.pos}
p.next()
} else if p.tok != token.RBRACK {
@@ -704,7 +727,7 @@ func (p *parser) tryVarType(isParam bool) ast.Expr {
if isParam && p.tok == token.ELLIPSIS {
pos := p.pos
p.next()
- typ := p.tryIdentOrType(isParam) // don't use parseType so we can provide better error message
+ typ := p.tryIdentOrType() // don't use parseType so we can provide better error message
if typ != nil {
p.resolve(typ)
} else {
@@ -713,7 +736,7 @@ func (p *parser) tryVarType(isParam bool) ast.Expr {
}
return &ast.Ellipsis{Ellipsis: pos, Elt: typ}
}
- return p.tryIdentOrType(false)
+ return p.tryIdentOrType()
}
// If the result is an identifier, it is not resolved.
@@ -931,29 +954,31 @@ func (p *parser) parseChanType() *ast.ChanType {
pos := p.pos
dir := ast.SEND | ast.RECV
+ var arrow token.Pos
if p.tok == token.CHAN {
p.next()
if p.tok == token.ARROW {
+ arrow = p.pos
p.next()
dir = ast.SEND
}
} else {
- p.expect(token.ARROW)
+ arrow = p.expect(token.ARROW)
p.expect(token.CHAN)
dir = ast.RECV
}
value := p.parseType()
- return &ast.ChanType{Begin: pos, Dir: dir, Value: value}
+ return &ast.ChanType{Begin: pos, Arrow: arrow, Dir: dir, Value: value}
}
// If the result is an identifier, it is not resolved.
-func (p *parser) tryIdentOrType(ellipsisOk bool) ast.Expr {
+func (p *parser) tryIdentOrType() ast.Expr {
switch p.tok {
case token.IDENT:
return p.parseTypeName()
case token.LBRACK:
- return p.parseArrayType(ellipsisOk)
+ return p.parseArrayType()
case token.STRUCT:
return p.parseStructType()
case token.MUL:
@@ -980,7 +1005,7 @@ func (p *parser) tryIdentOrType(ellipsisOk bool) ast.Expr {
}
func (p *parser) tryType() ast.Expr {
- typ := p.tryIdentOrType(false)
+ typ := p.tryIdentOrType()
if typ != nil {
p.resolve(typ)
}
@@ -1088,7 +1113,7 @@ func (p *parser) parseOperand(lhs bool) ast.Expr {
return p.parseFuncTypeOrLit()
}
- if typ := p.tryIdentOrType(true); typ != nil {
+ if typ := p.tryIdentOrType(); typ != nil {
// could be type for composite literal or conversion
_, isIdent := typ.(*ast.Ident)
assert(!isIdent, "type cannot be identifier")
@@ -1193,14 +1218,35 @@ func (p *parser) parseElement(keyOk bool) ast.Expr {
return p.parseLiteralValue(nil)
}
- x := p.checkExpr(p.parseExpr(keyOk)) // don't resolve if map key
+ // Because the parser doesn't know the composite literal type, it cannot
+ // know if a key that's an identifier is a struct field name or a name
+ // denoting a value. The former is not resolved by the parser or the
+ // resolver.
+ //
+ // Instead, _try_ to resolve such a key if possible. If it resolves,
+ // it a) has correctly resolved, or b) incorrectly resolved because
+ // the key is a struct field with a name matching another identifier.
+ // In the former case we are done, and in the latter case we don't
+ // care because the type checker will do a separate field lookup.
+ //
+ // If the key does not resolve, it a) must be defined at the top
+ // level in another file of the same package, the universe scope, or be
+ // undeclared; or b) it is a struct field. In the former case, the type
+ // checker can do a top-level lookup, and in the latter case it will do
+ // a separate field lookup.
+ x := p.checkExpr(p.parseExpr(keyOk))
if keyOk {
if p.tok == token.COLON {
colon := p.pos
p.next()
+ // Try to resolve the key but don't collect it
+ // as unresolved identifier if it fails so that
+ // we don't get (possibly false) errors about
+ // undeclared names.
+ p.tryResolve(x, false)
return &ast.KeyValueExpr{Key: x, Colon: colon, Value: p.parseElement(false)}
}
- p.resolve(x) // not a map key
+ p.resolve(x) // not a key
}
return x
@@ -1404,16 +1450,49 @@ func (p *parser) parseUnaryExpr(lhs bool) ast.Expr {
case token.ARROW:
// channel type or receive expression
- pos := p.pos
+ arrow := p.pos
p.next()
- if p.tok == token.CHAN {
- p.next()
- value := p.parseType()
- return &ast.ChanType{Begin: pos, Dir: ast.RECV, Value: value}
- }
+
+ // If the next token is token.CHAN we still don't know if it
+ // is a channel type or a receive operation - we only know
+ // once we have found the end of the unary expression. There
+ // are two cases:
+ //
+ // <- type => (<-type) must be channel type
+ // <- expr => <-(expr) is a receive from an expression
+ //
+ // In the first case, the arrow must be re-associated with
+ // the channel type parsed already:
+ //
+ // <- (chan type) => (<-chan type)
+ // <- (chan<- type) => (<-chan (<-type))
x := p.parseUnaryExpr(false)
- return &ast.UnaryExpr{OpPos: pos, Op: token.ARROW, X: p.checkExpr(x)}
+
+ // determine which case we have
+ if typ, ok := x.(*ast.ChanType); ok {
+ // (<-type)
+
+ // re-associate position info and <-
+ dir := ast.SEND
+ for ok && dir == ast.SEND {
+ if typ.Dir == ast.RECV {
+ // error: (<-type) is (<-(<-chan T))
+ p.errorExpected(typ.Arrow, "'chan'")
+ }
+ arrow, typ.Begin, typ.Arrow = typ.Arrow, arrow, arrow
+ dir, typ.Dir = typ.Dir, ast.RECV
+ typ, ok = typ.Value.(*ast.ChanType)
+ }
+ if dir == ast.SEND {
+ p.errorExpected(arrow, "channel type")
+ }
+
+ return x
+ }
+
+ // <-(expr)
+ return &ast.UnaryExpr{OpPos: arrow, Op: token.ARROW, X: p.checkExpr(x)}
case token.MUL:
// pointer type or unary "*" expression
@@ -1774,7 +1853,7 @@ func (p *parser) parseSwitchStmt() ast.Stmt {
//
// switch t := 0; t := x.(T) { ... }
//
- // (this code is not valid Go because the first t will
+ // (this code is not valid Go because the first t
// cannot be accessed and thus is never used, the extra
// scope is needed for the correct error message).
//
@@ -2012,7 +2091,7 @@ func (p *parser) parseStmt() (s ast.Stmt) {
// ----------------------------------------------------------------------------
// Declarations
-type parseSpecFunction func(p *parser, doc *ast.CommentGroup, iota int) ast.Spec
+type parseSpecFunction func(p *parser, doc *ast.CommentGroup, keyword token.Token, iota int) ast.Spec
func isValidImport(lit string) bool {
const illegalChars = `!"#$%&'()*,:;<=>?[\]^{|}` + "`\uFFFD"
@@ -2025,7 +2104,7 @@ func isValidImport(lit string) bool {
return s != ""
}
-func parseImportSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
+func (p *parser) parseImportSpec(doc *ast.CommentGroup, _ token.Token, _ int) ast.Spec {
if p.trace {
defer un(trace(p, "ImportSpec"))
}
@@ -2063,15 +2142,15 @@ func parseImportSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
return spec
}
-func parseConstSpec(p *parser, doc *ast.CommentGroup, iota int) ast.Spec {
+func (p *parser) parseValueSpec(doc *ast.CommentGroup, keyword token.Token, iota int) ast.Spec {
if p.trace {
- defer un(trace(p, "ConstSpec"))
+ defer un(trace(p, keyword.String()+"Spec"))
}
idents := p.parseIdentList()
typ := p.tryType()
var values []ast.Expr
- if typ != nil || p.tok == token.ASSIGN || iota == 0 {
+ if p.tok == token.ASSIGN || keyword == token.CONST && (typ != nil || iota == 0) || keyword == token.VAR && typ == nil {
p.expect(token.ASSIGN)
values = p.parseRhsList()
}
@@ -2088,12 +2167,16 @@ func parseConstSpec(p *parser, doc *ast.CommentGroup, iota int) ast.Spec {
Values: values,
Comment: p.lineComment,
}
- p.declare(spec, iota, p.topScope, ast.Con, idents...)
+ kind := ast.Con
+ if keyword == token.VAR {
+ kind = ast.Var
+ }
+ p.declare(spec, iota, p.topScope, kind, idents...)
return spec
}
-func parseTypeSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
+func (p *parser) parseTypeSpec(doc *ast.CommentGroup, _ token.Token, _ int) ast.Spec {
if p.trace {
defer un(trace(p, "TypeSpec"))
}
@@ -2114,36 +2197,6 @@ func parseTypeSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
return spec
}
-func parseVarSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
- if p.trace {
- defer un(trace(p, "VarSpec"))
- }
-
- idents := p.parseIdentList()
- typ := p.tryType()
- var values []ast.Expr
- if typ == nil || p.tok == token.ASSIGN {
- p.expect(token.ASSIGN)
- values = p.parseRhsList()
- }
- p.expectSemi() // call before accessing p.linecomment
-
- // Go spec: The scope of a constant or variable identifier declared inside
- // a function begins at the end of the ConstSpec or VarSpec and ends at
- // the end of the innermost containing block.
- // (Global identifiers are resolved in a separate phase after parsing.)
- spec := &ast.ValueSpec{
- Doc: doc,
- Names: idents,
- Type: typ,
- Values: values,
- Comment: p.lineComment,
- }
- p.declare(spec, nil, p.topScope, ast.Var, idents...)
-
- return spec
-}
-
func (p *parser) parseGenDecl(keyword token.Token, f parseSpecFunction) *ast.GenDecl {
if p.trace {
defer un(trace(p, "GenDecl("+keyword.String()+")"))
@@ -2157,12 +2210,12 @@ func (p *parser) parseGenDecl(keyword token.Token, f parseSpecFunction) *ast.Gen
lparen = p.pos
p.next()
for iota := 0; p.tok != token.RPAREN && p.tok != token.EOF; iota++ {
- list = append(list, f(p, p.leadComment, iota))
+ list = append(list, f(p, p.leadComment, keyword, iota))
}
rparen = p.expect(token.RPAREN)
p.expectSemi()
} else {
- list = append(list, f(p, nil, 0))
+ list = append(list, f(p, nil, keyword, 0))
}
return &ast.GenDecl{
@@ -2262,14 +2315,11 @@ func (p *parser) parseDecl(sync func(*parser)) ast.Decl {
var f parseSpecFunction
switch p.tok {
- case token.CONST:
- f = parseConstSpec
+ case token.CONST, token.VAR:
+ f = (*parser).parseValueSpec
case token.TYPE:
- f = parseTypeSpec
-
- case token.VAR:
- f = parseVarSpec
+ f = (*parser).parseTypeSpec
case token.FUNC:
return p.parseFuncDecl()
@@ -2292,6 +2342,12 @@ func (p *parser) parseFile() *ast.File {
defer un(trace(p, "File"))
}
+ // Don't bother parsing the rest if we had errors scanning the first token.
+ // Likely not a Go source file at all.
+ if p.errors.Len() != 0 {
+ return nil
+ }
+
// package clause
doc := p.leadComment
pos := p.expect(token.PACKAGE)
@@ -2303,15 +2359,19 @@ func (p *parser) parseFile() *ast.File {
}
p.expectSemi()
- var decls []ast.Decl
-
- // Don't bother parsing the rest if we had errors already.
+ // Don't bother parsing the rest if we had errors parsing the package clause.
// Likely not a Go source file at all.
+ if p.errors.Len() != 0 {
+ return nil
+ }
- if p.errors.Len() == 0 && p.mode&PackageClauseOnly == 0 {
+ p.openScope()
+ p.pkgScope = p.topScope
+ var decls []ast.Decl
+ if p.mode&PackageClauseOnly == 0 {
// import decls
for p.tok == token.IMPORT {
- decls = append(decls, p.parseGenDecl(token.IMPORT, parseImportSpec))
+ decls = append(decls, p.parseGenDecl(token.IMPORT, (*parser).parseImportSpec))
}
if p.mode&ImportsOnly == 0 {
@@ -2321,8 +2381,9 @@ func (p *parser) parseFile() *ast.File {
}
}
}
-
- assert(p.topScope == p.pkgScope, "imbalanced scopes")
+ p.closeScope()
+ assert(p.topScope == nil, "unbalanced scopes")
+ assert(p.labelScope == nil, "unbalanced label scopes")
// resolve global identifiers within the same file
i := 0
generated by cgit v1.2.3 (git 2.46.0) at 2025-10-31 00:17:13 +0000