// 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 parser implements a parser for Go source files. Input may be // provided in a variety of forms (see the various Parse* functions); the // output is an abstract syntax tree (AST) representing the Go source. The // parser is invoked through one of the Parse* functions. // package parser import ( "fmt" "go/ast" "go/scanner" "go/token" "strconv" "strings" "unicode" ) // The parser structure holds the parser's internal state. type parser struct { file *token.File errors scanner.ErrorList scanner scanner.Scanner // Tracing/debugging mode Mode // parsing mode trace bool // == (mode & Trace != 0) indent int // indentation used for tracing output // Comments comments []*ast.CommentGroup leadComment *ast.CommentGroup // last lead comment lineComment *ast.CommentGroup // last line comment // Next token pos token.Pos // token position tok token.Token // one token look-ahead lit string // token literal // Error recovery // (used to limit the number of calls to syncXXX functions // w/o making scanning progress - avoids potential endless // loops across multiple parser functions during error recovery) syncPos token.Pos // last synchronization position syncCnt int // number of calls to syncXXX without progress // Non-syntactic parser control exprLev int // < 0: in control clause, >= 0: in expression inRhs bool // if set, the parser is parsing a rhs expression // Ordinary identifier scopes pkgScope *ast.Scope // pkgScope.Outer == nil topScope *ast.Scope // top-most scope; may be pkgScope unresolved []*ast.Ident // unresolved identifiers imports []*ast.ImportSpec // list of imports // Label scopes // (maintained by open/close LabelScope) labelScope *ast.Scope // label scope for current function targetStack [][]*ast.Ident // stack of unresolved labels } func (p *parser) init(fset *token.FileSet, filename string, src []byte, mode Mode) { p.file = fset.AddFile(filename, -1, len(src)) var m scanner.Mode if mode&ParseComments != 0 { m = scanner.ScanComments } eh := func(pos token.Position, msg string) { p.errors.Add(pos, msg) } p.scanner.Init(p.file, src, eh, m) p.mode = mode p.trace = mode&Trace != 0 // for convenience (p.trace is used frequently) p.next() } // ---------------------------------------------------------------------------- // Scoping support func (p *parser) openScope() { p.topScope = ast.NewScope(p.topScope) } func (p *parser) closeScope() { p.topScope = p.topScope.Outer } func (p *parser) openLabelScope() { p.labelScope = ast.NewScope(p.labelScope) p.targetStack = append(p.targetStack, nil) } func (p *parser) closeLabelScope() { // resolve labels n := len(p.targetStack) - 1 scope := p.labelScope for _, ident := range p.targetStack[n] { ident.Obj = scope.Lookup(ident.Name) if ident.Obj == nil && p.mode&DeclarationErrors != 0 { p.error(ident.Pos(), fmt.Sprintf("label %s undefined", ident.Name)) } } // pop label scope p.targetStack = p.targetStack[0:n] p.labelScope = p.labelScope.Outer } func (p *parser) declare(decl, data interface{}, scope *ast.Scope, kind ast.ObjKind, idents ...*ast.Ident) { for _, ident := range idents { assert(ident.Obj == nil, "identifier already declared or resolved") obj := ast.NewObj(kind, ident.Name) // remember the corresponding declaration for redeclaration // errors and global variable resolution/typechecking phase obj.Decl = decl obj.Data = data ident.Obj = obj if ident.Name != "_" { if alt := scope.Insert(obj); alt != nil && p.mode&DeclarationErrors != 0 { prevDecl := "" if pos := alt.Pos(); pos.IsValid() { prevDecl = fmt.Sprintf("\n\tprevious declaration at %s", p.file.Position(pos)) } p.error(ident.Pos(), fmt.Sprintf("%s redeclared in this block%s", ident.Name, prevDecl)) } } } } func (p *parser) shortVarDecl(decl *ast.AssignStmt, list []ast.Expr) { // Go spec: A short variable declaration may redeclare variables // provided they were originally declared in the same block with // the same type, and at least one of the non-blank variables is new. n := 0 // number of new variables for _, x := range list { if ident, isIdent := x.(*ast.Ident); isIdent { assert(ident.Obj == nil, "identifier already declared or resolved") obj := ast.NewObj(ast.Var, ident.Name) // remember corresponding assignment for other tools obj.Decl = decl ident.Obj = obj if ident.Name != "_" { if alt := p.topScope.Insert(obj); alt != nil { ident.Obj = alt // redeclaration } else { n++ // new declaration } } } else { p.errorExpected(x.Pos(), "identifier on left side of :=") } } if n == 0 && p.mode&DeclarationErrors != 0 { p.error(list[0].Pos(), "no new variables on left side of :=") } } // The unresolved object is a sentinel to mark identifiers that have been added // to the list of unresolved identifiers. The sentinel is only used for verifying // internal consistency. var unresolved = new(ast.Object) // 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 { return } assert(ident.Obj == nil, "identifier already declared or resolved") if ident.Name == "_" { return } // try to resolve the identifier for s := p.topScope; s != nil; s = s.Outer { if obj := s.Lookup(ident.Name); obj != nil { ident.Obj = obj return } } // all local scopes are known, so any unresolved identifier // 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 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 = len(dots) pos := p.file.Position(p.pos) fmt.Printf("%5d:%3d: ", pos.Line, pos.Column) i := 2 * p.indent for i > n { fmt.Print(dots) i -= n } // i <= n fmt.Print(dots[0:i]) fmt.Println(a...) } func trace(p *parser, msg string) *parser { p.printTrace(msg, "(") p.indent++ return p } // Usage pattern: defer un(trace(p, "...")) func un(p *parser) { p.indent-- p.printTrace(")") } // Advance to the next token. func (p *parser) next0() { // Because of one-token look-ahead, print the previous token // when tracing as it provides a more readable output. The // very first token (!p.pos.IsValid()) is not initialized // (it is token.ILLEGAL), so don't print it . if p.trace && p.pos.IsValid() { s := p.tok.String() switch { case p.tok.IsLiteral(): p.printTrace(s, p.lit) case p.tok.IsOperator(), p.tok.IsKeyword(): p.printTrace("\"" + s + "\"") default: p.printTrace(s) } } p.pos, p.tok, p.lit = p.scanner.Scan() } // Consume a comment and return it and the line on which it ends. func (p *parser) consumeComment() (comment *ast.Comment, endline int) { // /*-style comments may end on a different line than where they start. // Scan the comment for '\n' chars and adjust endline accordingly. endline = p.file.Line(p.pos) if p.lit[1] == '*' { // don't use range here - no need to decode Unicode code points for i := 0; i < len(p.lit); i++ { if p.lit[i] == '\n' { endline++ } } } comment = &ast.Comment{Slash: p.pos, Text: p.lit} p.next0() return } // Consume a group of adjacent comments, add it to the parser's // comments list, and return it together with the line at which // the last comment in the group ends. A non-comment token or n // empty lines terminate a comment group. // func (p *parser) consumeCommentGroup(n int) (comments *ast.CommentGroup, endline int) { var list []*ast.Comment endline = p.file.Line(p.pos) for p.tok == token.COMMENT && p.file.Line(p.pos) <= endline+n { var comment *ast.Comment comment, endline = p.consumeComment() list = append(list, comment) } // add comment group to the comments list comments = &ast.CommentGroup{List: list} p.comments = append(p.comments, comments) return } // Advance to the next non-comment token. In the process, collect // any comment groups encountered, and remember the last lead and // and line comments. // // A lead comment is a comment group that starts and ends in a // line without any other tokens and that is followed by a non-comment // token on the line immediately after the comment group. // // A line comment is a comment group that follows a non-comment // token on the same line, and that has no tokens after it on the line // where it ends. // // Lead and line comments may be considered documentation that is // stored in the AST. // func (p *parser) next() { p.leadComment = nil p.lineComment = nil prev := p.pos p.next0() if p.tok == token.COMMENT { var comment *ast.CommentGroup var endline int 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) if p.file.Line(p.pos) != endline { // The next token is on a different line, thus // the last comment group is a line comment. p.lineComment = comment } } // consume successor comments, if any endline = -1 for p.tok == token.COMMENT { comment, endline = p.consumeCommentGroup(1) } if endline+1 == p.file.Line(p.pos) { // The next token is following on the line immediately after the // comment group, thus the last comment group is a lead comment. p.leadComment = comment } } } // A bailout panic is raised to indicate early termination. type bailout struct{} func (p *parser) error(pos token.Pos, msg string) { 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) { msg = "expected " + msg if pos == p.pos { // the error happened at the current position; // make the error message more specific if p.tok == token.SEMICOLON && p.lit == "\n" { msg += ", found newline" } else { msg += ", found '" + p.tok.String() + "'" if p.tok.IsLiteral() { msg += " " + p.lit } } } p.error(pos, msg) } func (p *parser) expect(tok token.Token) token.Pos { pos := p.pos if p.tok != tok { p.errorExpected(pos, "'"+tok.String()+"'") } p.next() // make progress return pos } // expectClosing is like expect but provides a better error message // for the common case of a missing comma before a newline. // func (p *parser) expectClosing(tok token.Token, context string) token.Pos { if p.tok != tok && p.tok == token.SEMICOLON && p.lit == "\n" { p.error(p.pos, "missing ',' before newline in "+context) p.next() } return p.expect(tok) } func (p *parser) expectSemi() { // semicolon is optional before a closing ')' or '}' if p.tok != token.RPAREN && p.tok != token.RBRACE { if p.tok == token.SEMICOLON { p.next() } else { p.errorExpected(p.pos, "';'") syncStmt(p) } } } func (p *parser) atComma(context string) bool { if p.tok == token.COMMA { return true } if p.tok == token.SEMICOLON && p.lit == "\n" { p.error(p.pos, "missing ',' before newline in "+context) return true // "insert" the comma and continue } return false } func assert(cond bool, msg string) { if !cond { panic("go/parser internal error: " + msg) } } // syncStmt advances to the next statement. // Used for synchronization after an error. // func syncStmt(p *parser) { for { switch p.tok { case token.BREAK, token.CONST, token.CONTINUE, token.DEFER, token.FALLTHROUGH, token.FOR, token.GO, token.GOTO, token.IF, token.RETURN, token.SELECT, token.SWITCH, token.TYPE, token.VAR: // Return only if parser made some progress since last // sync or if it has not reached 10 sync calls without // progress. Otherwise consume at least one token to // avoid an endless parser loop (it is possible that // both parseOperand and parseStmt call syncStmt and // correctly do not advance, thus the need for the // invocation limit p.syncCnt). if p.pos == p.syncPos && p.syncCnt < 10 { p.syncCnt++ return } if p.pos > p.syncPos { p.syncPos = p.pos p.syncCnt = 0 return } // Reaching here indicates a parser bug, likely an // incorrect token list in this function, but it only // leads to skipping of possibly correct code if a // previous error is present, and thus is preferred // over a non-terminating parse. case token.EOF: return } p.next() } } // syncDecl advances to the next declaration. // Used for synchronization after an error. // func syncDecl(p *parser) { for { switch p.tok { case token.CONST, token.TYPE, token.VAR: // see comments in syncStmt if p.pos == p.syncPos && p.syncCnt < 10 { p.syncCnt++ return } if p.pos > p.syncPos { p.syncPos = p.pos p.syncCnt = 0 return } case token.EOF: return } p.next() } } // safePos returns a valid file position for a given position: If pos // is valid to begin with, safePos returns pos. If pos is out-of-range, // safePos returns the EOF position. // // This is hack to work around "artificial" end positions in the AST which // are computed by adding 1 to (presumably valid) token positions. If the // token positions are invalid due to parse errors, the resulting end position // may be past the file's EOF position, which would lead to panics if used // later on. // func (p *parser) safePos(pos token.Pos) (res token.Pos) { defer func() { if recover() != nil { res = token.Pos(p.file.Base() + p.file.Size()) // EOF position } }() _ = p.file.Offset(pos) // trigger a panic if position is out-of-range return pos } // ---------------------------------------------------------------------------- // Identifiers func (p *parser) parseIdent() *ast.Ident { pos := p.pos name := "_" if p.tok == token.IDENT { name = p.lit p.next() } else { p.expect(token.IDENT) // use expect() error handling } return &ast.Ident{NamePos: pos, Name: name} } func (p *parser) parseIdentList() (list []*ast.Ident) { if p.trace { defer un(trace(p, "IdentList")) } list = append(list, p.parseIdent()) for p.tok == token.COMMA { p.next() list = append(list, p.parseIdent()) } return } // ---------------------------------------------------------------------------- // Common productions // If lhs is set, result list elements which are identifiers are not resolved. func (p *parser) parseExprList(lhs bool) (list []ast.Expr) { if p.trace { defer un(trace(p, "ExpressionList")) } list = append(list, p.checkExpr(p.parseExpr(lhs))) for p.tok == token.COMMA { p.next() list = append(list, p.checkExpr(p.parseExpr(lhs))) } return } func (p *parser) parseLhsList() []ast.Expr { old := p.inRhs p.inRhs = false list := p.parseExprList(true) switch p.tok { case token.DEFINE: // lhs of a short variable declaration // but doesn't enter scope until later: // caller must call p.shortVarDecl(p.makeIdentList(list)) // at appropriate time. case token.COLON: // lhs of a label declaration or a communication clause of a select // statement (parseLhsList is not called when parsing the case clause // of a switch statement): // - labels are declared by the caller of parseLhsList // - for communication clauses, if there is a stand-alone identifier // followed by a colon, we have a syntax error; there is no need // to resolve the identifier in that case default: // identifiers must be declared elsewhere for _, x := range list { p.resolve(x) } } p.inRhs = old return list } func (p *parser) parseRhsList() []ast.Expr { old := p.inRhs p.inRhs = true list := p.parseExprList(false) p.inRhs = old return list } // ---------------------------------------------------------------------------- // Types func (p *parser) parseType() ast.Expr { if p.trace { defer un(trace(p, "Type")) } typ := p.tryType() if typ == nil { pos := p.pos p.errorExpected(pos, "type") p.next() // make progress return &ast.BadExpr{From: pos, To: p.pos} } return typ } // If the result is an identifier, it is not resolved. func (p *parser) parseTypeName() ast.Expr { if p.trace { defer un(trace(p, "TypeName")) } ident := p.parseIdent() // don't resolve ident yet - it may be a parameter or field name if p.tok == token.PERIOD { // ident is a package name p.next() p.resolve(ident) sel := p.parseIdent() return &ast.SelectorExpr{X: ident, Sel: sel} } return ident } func (p *parser) parseArrayType() ast.Expr { if p.trace { defer un(trace(p, "ArrayType")) } lbrack := p.expect(token.LBRACK) p.exprLev++ var len ast.Expr // 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 { len = p.parseRhs() } p.exprLev-- p.expect(token.RBRACK) elt := p.parseType() return &ast.ArrayType{Lbrack: lbrack, Len: len, Elt: elt} } func (p *parser) makeIdentList(list []ast.Expr) []*ast.Ident { idents := make([]*ast.Ident, len(list)) for i, x := range list { ident, isIdent := x.(*ast.Ident) if !isIdent { if _, isBad := x.(*ast.BadExpr); !isBad { // only report error if it's a new one p.errorExpected(x.Pos(), "identifier") } ident = &ast.Ident{NamePos: x.Pos(), Name: "_"} } idents[i] = ident } return idents } func (p *parser) parseFieldDecl(scope *ast.Scope) *ast.Field { if p.trace { defer un(trace(p, "FieldDecl")) } doc := p.leadComment // FieldDecl list, typ := p.parseVarList(false) // Tag var tag *ast.BasicLit if p.tok == token.STRING { tag = &ast.BasicLit{ValuePos: p.pos, Kind: p.tok, Value: p.lit} p.next() } // analyze case var idents []*ast.Ident if typ != nil { // IdentifierList Type idents = p.makeIdentList(list) } else { // ["*"] TypeName (AnonymousField) typ = list[0] // we always have at least one element if n := len(list); n > 1 || !isTypeName(deref(typ)) { pos := typ.Pos() p.errorExpected(pos, "anonymous field") typ = &ast.BadExpr{From: pos, To: p.safePos(list[n-1].End())} } } p.expectSemi() // call before accessing p.linecomment field := &ast.Field{Doc: doc, Names: idents, Type: typ, Tag: tag, Comment: p.lineComment} p.declare(field, nil, scope, ast.Var, idents...) p.resolve(typ) return field } func (p *parser) parseStructType() *ast.StructType { if p.trace { defer un(trace(p, "StructType")) } pos := p.expect(token.STRUCT) lbrace := p.expect(token.LBRACE) scope := ast.NewScope(nil) // struct scope var list []*ast.Field for p.tok == token.IDENT || p.tok == token.MUL || p.tok == token.LPAREN { // a field declaration cannot start with a '(' but we accept // it here for more robust parsing and better error messages // (parseFieldDecl will check and complain if necessary) list = append(list, p.parseFieldDecl(scope)) } rbrace := p.expect(token.RBRACE) return &ast.StructType{ Struct: pos, Fields: &ast.FieldList{ Opening: lbrace, List: list, Closing: rbrace, }, } } func (p *parser) parsePointerType() *ast.StarExpr { if p.trace { defer un(trace(p, "PointerType")) } star := p.expect(token.MUL) base := p.parseType() return &ast.StarExpr{Star: star, X: base} } // If the result is an identifier, it is not resolved. func (p *parser) tryVarType(isParam bool) ast.Expr { if isParam && p.tok == token.ELLIPSIS { pos := p.pos p.next() typ := p.tryIdentOrType() // don't use parseType so we can provide better error message if typ != nil { p.resolve(typ) } else { p.error(pos, "'...' parameter is missing type") typ = &ast.BadExpr{From: pos, To: p.pos} } return &ast.Ellipsis{Ellipsis: pos, Elt: typ} } return p.tryIdentOrType() } // If the result is an identifier, it is not resolved. func (p *parser) parseVarType(isParam bool) ast.Expr { typ := p.tryVarType(isParam) if typ == nil { pos := p.pos p.errorExpected(pos, "type") p.next() // make progress typ = &ast.BadExpr{From: pos, To: p.pos} } return typ } // If any of the results are identifiers, they are not resolved. func (p *parser) parseVarList(isParam bool) (list []ast.Expr, typ ast.Expr) { if p.trace { defer un(trace(p, "VarList")) } // a list of identifiers looks like a list of type names // // parse/tryVarType accepts any type (including parenthesized // ones) even though the syntax does not permit them here: we // accept them all for more robust parsing and complain later for typ := p.parseVarType(isParam); typ != nil; { list = append(list, typ) if p.tok != token.COMMA { break } p.next() typ = p.tryVarType(isParam) // maybe nil as in: func f(int,) {} } // if we had a list of identifiers, it must be followed by a type typ = p.tryVarType(isParam) return } func (p *parser) parseParameterList(scope *ast.Scope, ellipsisOk bool) (params []*ast.Field) { if p.trace { defer un(trace(p, "ParameterList")) } // ParameterDecl list, typ := p.parseVarList(ellipsisOk) // analyze case if typ != nil { // IdentifierList Type idents := p.makeIdentList(list) field := &ast.Field{Names: idents, Type: typ} params = append(params, field) // Go spec: The scope of an identifier denoting a function // parameter or result variable is the function body. p.declare(field, nil, scope, ast.Var, idents...) p.resolve(typ) if !p.atComma("parameter list") { return } p.next() for p.tok != token.RPAREN && p.tok != token.EOF { idents := p.parseIdentList() typ := p.parseVarType(ellipsisOk) field := &ast.Field{Names: idents, Type: typ} params = append(params, field) // Go spec: The scope of an identifier denoting a function // parameter or result variable is the function body. p.declare(field, nil, scope, ast.Var, idents...) p.resolve(typ) if !p.atComma("parameter list") { break } p.next() } return } // Type { "," Type } (anonymous parameters) params = make([]*ast.Field, len(list)) for i, typ := range list { p.resolve(typ) params[i] = &ast.Field{Type: typ} } return } func (p *parser) parseParameters(scope *ast.Scope, ellipsisOk bool) *ast.FieldList { if p.trace { defer un(trace(p, "Parameters")) } var params []*ast.Field lparen := p.expect(token.LPAREN) if p.tok != token.RPAREN { params = p.parseParameterList(scope, ellipsisOk) } rparen := p.expect(token.RPAREN) return &ast.FieldList{Opening: lparen, List: params, Closing: rparen} } func (p *parser) parseResult(scope *ast.Scope) *ast.FieldList { if p.trace { defer un(trace(p, "Result")) } if p.tok == token.LPAREN { return p.parseParameters(scope, false) } typ := p.tryType() if typ != nil { list := make([]*ast.Field, 1) list[0] = &ast.Field{Type: typ} return &ast.FieldList{List: list} } return nil } func (p *parser) parseSignature(scope *ast.Scope) (params, results *ast.FieldList) { if p.trace { defer un(trace(p, "Signature")) } params = p.parseParameters(scope, true) results = p.parseResult(scope) return } func (p *parser) parseFuncType() (*ast.FuncType, *ast.Scope) { if p.trace { defer un(trace(p, "FuncType")) } pos := p.expect(token.FUNC) scope := ast.NewScope(p.topScope) // function scope params, results := p.parseSignature(scope) return &ast.FuncType{Func: pos, Params: params, Results: results}, scope } func (p *parser) parseMethodSpec(scope *ast.Scope) *ast.Field { if p.trace { defer un(trace(p, "MethodSpec")) } doc := p.leadComment var idents []*ast.Ident var typ ast.Expr x := p.parseTypeName() if ident, isIdent := x.(*ast.Ident); isIdent && p.tok == token.LPAREN { // method idents = []*ast.Ident{ident} scope := ast.NewScope(nil) // method scope params, results := p.parseSignature(scope) typ = &ast.FuncType{Func: token.NoPos, Params: params, Results: results} } else { // embedded interface typ = x p.resolve(typ) } p.expectSemi() // call before accessing p.linecomment spec := &ast.Field{Doc: doc, Names: idents, Type: typ, Comment: p.lineComment} p.declare(spec, nil, scope, ast.Fun, idents...) return spec } func (p *parser) parseInterfaceType() *ast.InterfaceType { if p.trace { defer un(trace(p, "InterfaceType")) } pos := p.expect(token.INTERFACE) lbrace := p.expect(token.LBRACE) scope := ast.NewScope(nil) // interface scope var list []*ast.Field for p.tok == token.IDENT { list = append(list, p.parseMethodSpec(scope)) } rbrace := p.expect(token.RBRACE) return &ast.InterfaceType{ Interface: pos, Methods: &ast.FieldList{ Opening: lbrace, List: list, Closing: rbrace, }, } } func (p *parser) parseMapType() *ast.MapType { if p.trace { defer un(trace(p, "MapType")) } pos := p.expect(token.MAP) p.expect(token.LBRACK) key := p.parseType() p.expect(token.RBRACK) value := p.parseType() return &ast.MapType{Map: pos, Key: key, Value: value} } func (p *parser) parseChanType() *ast.ChanType { if p.trace { defer un(trace(p, "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 { arrow = p.expect(token.ARROW) p.expect(token.CHAN) dir = ast.RECV } value := p.parseType() 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() ast.Expr { switch p.tok { case token.IDENT: return p.parseTypeName() case token.LBRACK: return p.parseArrayType() case token.STRUCT: return p.parseStructType() case token.MUL: return p.parsePointerType() case token.FUNC: typ, _ := p.parseFuncType() return typ case token.INTERFACE: return p.parseInterfaceType() case token.MAP: return p.parseMapType() case token.CHAN, token.ARROW: return p.parseChanType() case token.LPAREN: lparen := p.pos p.next() typ := p.parseType() rparen := p.expect(token.RPAREN) return &ast.ParenExpr{Lparen: lparen, X: typ, Rparen: rparen} } // no type found return nil } func (p *parser) tryType() ast.Expr { typ := p.tryIdentOrType() if typ != nil { p.resolve(typ) } return typ } // ---------------------------------------------------------------------------- // Blocks func (p *parser) parseStmtList() (list []ast.Stmt) { if p.trace { defer un(trace(p, "StatementList")) } for p.tok != token.CASE && p.tok != token.DEFAULT && p.tok != token.RBRACE && p.tok != token.EOF { list = append(list, p.parseStmt()) } return } func (p *parser) parseBody(scope *ast.Scope) *ast.BlockStmt { if p.trace { defer un(trace(p, "Body")) } lbrace := p.expect(token.LBRACE) p.topScope = scope // open function scope p.openLabelScope() list := p.parseStmtList() p.closeLabelScope() p.closeScope() rbrace := p.expect(token.RBRACE) return &ast.BlockStmt{Lbrace: lbrace, List: list, Rbrace: rbrace} } func (p *parser) parseBlockStmt() *ast.BlockStmt { if p.trace { defer un(trace(p, "BlockStmt")) } lbrace := p.expect(token.LBRACE) p.openScope() list := p.parseStmtList() p.closeScope() rbrace := p.expect(token.RBRACE) return &ast.BlockStmt{Lbrace: lbrace, List: list, Rbrace: rbrace} } // ---------------------------------------------------------------------------- // Expressions func (p *parser) parseFuncTypeOrLit() ast.Expr { if p.trace { defer un(trace(p, "FuncTypeOrLit")) } typ, scope := p.parseFuncType() if p.tok != token.LBRACE { // function type only return typ } p.exprLev++ body := p.parseBody(scope) p.exprLev-- return &ast.FuncLit{Type: typ, Body: body} } // parseOperand may return an expression or a raw type (incl. array // types of the form [...]T. Callers must verify the result. // If lhs is set and the result is an identifier, it is not resolved. // func (p *parser) parseOperand(lhs bool) ast.Expr { if p.trace { defer un(trace(p, "Operand")) } switch p.tok { case token.IDENT: x := p.parseIdent() if !lhs { p.resolve(x) } return x case token.INT, token.FLOAT, token.IMAG, token.CHAR, token.STRING: x := &ast.BasicLit{ValuePos: p.pos, Kind: p.tok, Value: p.lit} p.next() return x case token.LPAREN: lparen := p.pos p.next() p.exprLev++ x := p.parseRhsOrType() // types may be parenthesized: (some type) p.exprLev-- rparen := p.expect(token.RPAREN) return &ast.ParenExpr{Lparen: lparen, X: x, Rparen: rparen} case token.FUNC: return p.parseFuncTypeOrLit() } if typ := p.tryIdentOrType(); typ != nil { // could be type for composite literal or conversion _, isIdent := typ.(*ast.Ident) assert(!isIdent, "type cannot be identifier") return typ } // we have an error pos := p.pos p.errorExpected(pos, "operand") syncStmt(p) return &ast.BadExpr{From: pos, To: p.pos} } func (p *parser) parseSelector(x ast.Expr) ast.Expr { if p.trace { defer un(trace(p, "Selector")) } sel := p.parseIdent() return &ast.SelectorExpr{X: x, Sel: sel} } func (p *parser) parseTypeAssertion(x ast.Expr) ast.Expr { if p.trace { defer un(trace(p, "TypeAssertion")) } lparen := p.expect(token.LPAREN) var typ ast.Expr if p.tok == token.TYPE { // type switch: typ == nil p.next() } else { typ = p.parseType() } rparen := p.expect(token.RPAREN) return &ast.TypeAssertExpr{X: x, Type: typ, Lparen: lparen, Rparen: rparen} } func (p *parser) parseIndexOrSlice(x ast.Expr) ast.Expr { if p.trace { defer un(trace(p, "IndexOrSlice")) } const N = 3 // change the 3 to 2 to disable 3-index slices lbrack := p.expect(token.LBRACK) p.exprLev++ var index [N]ast.Expr var colons [N - 1]token.Pos if p.tok != token.COLON { index[0] = p.parseRhs() } ncolons := 0 for p.tok == token.COLON && ncolons < len(colons) { colons[ncolons] = p.pos ncolons++ p.next() if p.tok != token.COLON && p.tok != token.RBRACK && p.tok != token.EOF { index[ncolons] = p.parseRhs() } } p.exprLev-- rbrack := p.expect(token.RBRACK) if ncolons > 0 { // slice expression slice3 := false if ncolons == 2 { slice3 = true // Check presence of 2nd and 3rd index here rather than during type-checking // to prevent erroneous programs from passing through gofmt (was issue 7305). if index[1] == nil { p.error(colons[0], "2nd index required in 3-index slice") index[1] = &ast.BadExpr{From: colons[0] + 1, To: colons[1]} } if index[2] == nil { p.error(colons[1], "3rd index required in 3-index slice") index[2] = &ast.BadExpr{From: colons[1] + 1, To: rbrack} } } return &ast.SliceExpr{X: x, Lbrack: lbrack, Low: index[0], High: index[1], Max: index[2], Slice3: slice3, Rbrack: rbrack} } return &ast.IndexExpr{X: x, Lbrack: lbrack, Index: index[0], Rbrack: rbrack} } func (p *parser) parseCallOrConversion(fun ast.Expr) *ast.CallExpr { if p.trace { defer un(trace(p, "CallOrConversion")) } lparen := p.expect(token.LPAREN) p.exprLev++ var list []ast.Expr var ellipsis token.Pos for p.tok != token.RPAREN && p.tok != token.EOF && !ellipsis.IsValid() { list = append(list, p.parseRhsOrType()) // builtins may expect a type: make(some type, ...) if p.tok == token.ELLIPSIS { ellipsis = p.pos p.next() } if !p.atComma("argument list") { break } p.next() } p.exprLev-- rparen := p.expectClosing(token.RPAREN, "argument list") return &ast.CallExpr{Fun: fun, Lparen: lparen, Args: list, Ellipsis: ellipsis, Rparen: rparen} } func (p *parser) parseElement(keyOk bool) ast.Expr { if p.trace { defer un(trace(p, "Element")) } if p.tok == token.LBRACE { return p.parseLiteralValue(nil) } // 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 key } return x } func (p *parser) parseElementList() (list []ast.Expr) { if p.trace { defer un(trace(p, "ElementList")) } for p.tok != token.RBRACE && p.tok != token.EOF { list = append(list, p.parseElement(true)) if !p.atComma("composite literal") { break } p.next() } return } func (p *parser) parseLiteralValue(typ ast.Expr) ast.Expr { if p.trace { defer un(trace(p, "LiteralValue")) } lbrace := p.expect(token.LBRACE) var elts []ast.Expr p.exprLev++ if p.tok != token.RBRACE { elts = p.parseElementList() } p.exprLev-- rbrace := p.expectClosing(token.RBRACE, "composite literal") return &ast.CompositeLit{Type: typ, Lbrace: lbrace, Elts: elts, Rbrace: rbrace} } // checkExpr checks that x is an expression (and not a type). func (p *parser) checkExpr(x ast.Expr) ast.Expr { switch unparen(x).(type) { case *ast.BadExpr: case *ast.Ident: case *ast.BasicLit: case *ast.FuncLit: case *ast.CompositeLit: case *ast.ParenExpr: panic("unreachable") case *ast.SelectorExpr: case *ast.IndexExpr: case *ast.SliceExpr: case *ast.TypeAssertExpr: // If t.Type == nil we have a type assertion of the form // y.(type), which is only allowed in type switch expressions. // It's hard to exclude those but for the case where we are in // a type switch. Instead be lenient and test this in the type // checker. case *ast.CallExpr: case *ast.StarExpr: case *ast.UnaryExpr: case *ast.BinaryExpr: default: // all other nodes are not proper expressions p.errorExpected(x.Pos(), "expression") x = &ast.BadExpr{From: x.Pos(), To: p.safePos(x.End())} } return x } // isTypeName returns true iff x is a (qualified) TypeName. func isTypeName(x ast.Expr) bool { switch t := x.(type) { case *ast.BadExpr: case *ast.Ident: case *ast.SelectorExpr: _, isIdent := t.X.(*ast.Ident) return isIdent default: return false // all other nodes are not type names } return true } // isLiteralType returns true iff x is a legal composite literal type. func isLiteralType(x ast.Expr) bool { switch t := x.(type) { case *ast.BadExpr: case *ast.Ident: case *ast.SelectorExpr: _, isIdent := t.X.(*ast.Ident) return isIdent case *ast.ArrayType: case *ast.StructType: case *ast.MapType: default: return false // all other nodes are not legal composite literal types } return true } // If x is of the form *T, deref returns T, otherwise it returns x. func deref(x ast.Expr) ast.Expr { if p, isPtr := x.(*ast.StarExpr); isPtr { x = p.X } return x } // If x is of the form (T), unparen returns unparen(T), otherwise it returns x. func unparen(x ast.Expr) ast.Expr { if p, isParen := x.(*ast.ParenExpr); isParen { x = unparen(p.X) } return x } // checkExprOrType checks that x is an expression or a type // (and not a raw type such as [...]T). // func (p *parser) checkExprOrType(x ast.Expr) ast.Expr { switch t := unparen(x).(type) { case *ast.ParenExpr: panic("unreachable") case *ast.UnaryExpr: case *ast.ArrayType: if len, isEllipsis := t.Len.(*ast.Ellipsis); isEllipsis { p.error(len.Pos(), "expected array length, found '...'") x = &ast.BadExpr{From: x.Pos(), To: p.safePos(x.End())} } } // all other nodes are expressions or types return x } // If lhs is set and the result is an identifier, it is not resolved. func (p *parser) parsePrimaryExpr(lhs bool) ast.Expr { if p.trace { defer un(trace(p, "PrimaryExpr")) } x := p.parseOperand(lhs) L: for { switch p.tok { case token.PERIOD: p.next() if lhs { p.resolve(x) } switch p.tok { case token.IDENT: x = p.parseSelector(p.checkExprOrType(x)) case token.LPAREN: x = p.parseTypeAssertion(p.checkExpr(x)) default: pos := p.pos p.errorExpected(pos, "selector or type assertion") p.next() // make progress x = &ast.BadExpr{From: pos, To: p.pos} } case token.LBRACK: if lhs { p.resolve(x) } x = p.parseIndexOrSlice(p.checkExpr(x)) case token.LPAREN: if lhs { p.resolve(x) } x = p.parseCallOrConversion(p.checkExprOrType(x)) case token.LBRACE: if isLiteralType(x) && (p.exprLev >= 0 || !isTypeName(x)) { if lhs { p.resolve(x) } x = p.parseLiteralValue(x) } else { break L } default: break L } lhs = false // no need to try to resolve again } return x } // If lhs is set and the result is an identifier, it is not resolved. func (p *parser) parseUnaryExpr(lhs bool) ast.Expr { if p.trace { defer un(trace(p, "UnaryExpr")) } switch p.tok { case token.ADD, token.SUB, token.NOT, token.XOR, token.AND: pos, op := p.pos, p.tok p.next() x := p.parseUnaryExpr(false) return &ast.UnaryExpr{OpPos: pos, Op: op, X: p.checkExpr(x)} case token.ARROW: // channel type or receive expression arrow := p.pos p.next() // 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) // 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 pos := p.pos p.next() x := p.parseUnaryExpr(false) return &ast.StarExpr{Star: pos, X: p.checkExprOrType(x)} } return p.parsePrimaryExpr(lhs) } func (p *parser) tokPrec() (token.Token, int) { tok := p.tok if p.inRhs && tok == token.ASSIGN { tok = token.EQL } return tok, tok.Precedence() } // If lhs is set and the result is an identifier, it is not resolved. func (p *parser) parseBinaryExpr(lhs bool, prec1 int) ast.Expr { if p.trace { defer un(trace(p, "BinaryExpr")) } x := p.parseUnaryExpr(lhs) for _, prec := p.tokPrec(); prec >= prec1; prec-- { for { op, oprec := p.tokPrec() if oprec != prec { break } pos := p.expect(op) if lhs { p.resolve(x) lhs = false } y := p.parseBinaryExpr(false, prec+1) x = &ast.BinaryExpr{X: p.checkExpr(x), OpPos: pos, Op: op, Y: p.checkExpr(y)} } } return x } // If lhs is set and the result is an identifier, it is not resolved. // The result may be a type or even a raw type ([...]int). Callers must // check the result (using checkExpr or checkExprOrType), depending on // context. func (p *parser) parseExpr(lhs bool) ast.Expr { if p.trace { defer un(trace(p, "Expression")) } return p.parseBinaryExpr(lhs, token.LowestPrec+1) } func (p *parser) parseRhs() ast.Expr { old := p.inRhs p.inRhs = true x := p.checkExpr(p.parseExpr(false)) p.inRhs = old return x } func (p *parser) parseRhsOrType() ast.Expr { old := p.inRhs p.inRhs = true x := p.checkExprOrType(p.parseExpr(false)) p.inRhs = old return x } // ---------------------------------------------------------------------------- // Statements // Parsing modes for parseSimpleStmt. const ( basic = iota labelOk rangeOk ) // parseSimpleStmt returns true as 2nd result if it parsed the assignment // of a range clause (with mode == rangeOk). The returned statement is an // assignment with a right-hand side that is a single unary expression of // the form "range x". No guarantees are given for the left-hand side. func (p *parser) parseSimpleStmt(mode int) (ast.Stmt, bool) { if p.trace { defer un(trace(p, "SimpleStmt")) } x := p.parseLhsList() switch p.tok { case token.DEFINE, token.ASSIGN, token.ADD_ASSIGN, token.SUB_ASSIGN, token.MUL_ASSIGN, token.QUO_ASSIGN, token.REM_ASSIGN, token.AND_ASSIGN, token.OR_ASSIGN, token.XOR_ASSIGN, token.SHL_ASSIGN, token.SHR_ASSIGN, token.AND_NOT_ASSIGN: // assignment statement, possibly part of a range clause pos, tok := p.pos, p.tok p.next() var y []ast.Expr isRange := false if mode == rangeOk && p.tok == token.RANGE && (tok == token.DEFINE || tok == token.ASSIGN) { pos := p.pos p.next() y = []ast.Expr{&ast.UnaryExpr{OpPos: pos, Op: token.RANGE, X: p.parseRhs()}} isRange = true } else { y = p.parseRhsList() } as := &ast.AssignStmt{Lhs: x, TokPos: pos, Tok: tok, Rhs: y} if tok == token.DEFINE { p.shortVarDecl(as, x) } return as, isRange } if len(x) > 1 { p.errorExpected(x[0].Pos(), "1 expression") // continue with first expression } switch p.tok { case token.COLON: // labeled statement colon := p.pos p.next() if label, isIdent := x[0].(*ast.Ident); mode == labelOk && isIdent { // Go spec: The scope of a label is the body of the function // in which it is declared and excludes the body of any nested // function. stmt := &ast.LabeledStmt{Label: label, Colon: colon, Stmt: p.parseStmt()} p.declare(stmt, nil, p.labelScope, ast.Lbl, label) return stmt, false } // The label declaration typically starts at x[0].Pos(), but the label // declaration may be erroneous due to a token after that position (and // before the ':'). If SpuriousErrors is not set, the (only) error re- // ported for the line is the illegal label error instead of the token // before the ':' that caused the problem. Thus, use the (latest) colon // position for error reporting. p.error(colon, "illegal label declaration") return &ast.BadStmt{From: x[0].Pos(), To: colon + 1}, false case token.ARROW: // send statement arrow := p.pos p.next() y := p.parseRhs() return &ast.SendStmt{Chan: x[0], Arrow: arrow, Value: y}, false case token.INC, token.DEC: // increment or decrement s := &ast.IncDecStmt{X: x[0], TokPos: p.pos, Tok: p.tok} p.next() return s, false } // expression return &ast.ExprStmt{X: x[0]}, false } func (p *parser) parseCallExpr(callType string) *ast.CallExpr { x := p.parseRhsOrType() // could be a conversion: (some type)(x) if call, isCall := x.(*ast.CallExpr); isCall { return call } if _, isBad := x.(*ast.BadExpr); !isBad { // only report error if it's a new one p.error(p.safePos(x.End()), fmt.Sprintf("function must be invoked in %s statement", callType)) } return nil } func (p *parser) parseGoStmt() ast.Stmt { if p.trace { defer un(trace(p, "GoStmt")) } pos := p.expect(token.GO) call := p.parseCallExpr("go") p.expectSemi() if call == nil { return &ast.BadStmt{From: pos, To: pos + 2} // len("go") } return &ast.GoStmt{Go: pos, Call: call} } func (p *parser) parseDeferStmt() ast.Stmt { if p.trace { defer un(trace(p, "DeferStmt")) } pos := p.expect(token.DEFER) call := p.parseCallExpr("defer") p.expectSemi() if call == nil { return &ast.BadStmt{From: pos, To: pos + 5} // len("defer") } return &ast.DeferStmt{Defer: pos, Call: call} } func (p *parser) parseReturnStmt() *ast.ReturnStmt { if p.trace { defer un(trace(p, "ReturnStmt")) } pos := p.pos p.expect(token.RETURN) var x []ast.Expr if p.tok != token.SEMICOLON && p.tok != token.RBRACE { x = p.parseRhsList() } p.expectSemi() return &ast.ReturnStmt{Return: pos, Results: x} } func (p *parser) parseBranchStmt(tok token.Token) *ast.BranchStmt { if p.trace { defer un(trace(p, "BranchStmt")) } pos := p.expect(tok) var label *ast.Ident if tok != token.FALLTHROUGH && p.tok == token.IDENT { label = p.parseIdent() // add to list of unresolved targets n := len(p.targetStack) - 1 p.targetStack[n] = append(p.targetStack[n], label) } p.expectSemi() return &ast.BranchStmt{TokPos: pos, Tok: tok, Label: label} } func (p *parser) makeExpr(s ast.Stmt, kind string) ast.Expr { if s == nil { return nil } if es, isExpr := s.(*ast.ExprStmt); isExpr { return p.checkExpr(es.X) } p.error(s.Pos(), fmt.Sprintf("expected %s, found simple statement (missing parentheses around composite literal?)", kind)) return &ast.BadExpr{From: s.Pos(), To: p.safePos(s.End())} } func (p *parser) parseIfStmt() *ast.IfStmt { if p.trace { defer un(trace(p, "IfStmt")) } pos := p.expect(token.IF) p.openScope() defer p.closeScope() var s ast.Stmt var x ast.Expr { prevLev := p.exprLev p.exprLev = -1 if p.tok == token.SEMICOLON { p.next() x = p.parseRhs() } else { s, _ = p.parseSimpleStmt(basic) if p.tok == token.SEMICOLON { p.next() x = p.parseRhs() } else { x = p.makeExpr(s, "boolean expression") s = nil } } p.exprLev = prevLev } body := p.parseBlockStmt() var else_ ast.Stmt if p.tok == token.ELSE { p.next() else_ = p.parseStmt() } else { p.expectSemi() } return &ast.IfStmt{If: pos, Init: s, Cond: x, Body: body, Else: else_} } func (p *parser) parseTypeList() (list []ast.Expr) { if p.trace { defer un(trace(p, "TypeList")) } list = append(list, p.parseType()) for p.tok == token.COMMA { p.next() list = append(list, p.parseType()) } return } func (p *parser) parseCaseClause(typeSwitch bool) *ast.CaseClause { if p.trace { defer un(trace(p, "CaseClause")) } pos := p.pos var list []ast.Expr if p.tok == token.CASE { p.next() if typeSwitch { list = p.parseTypeList() } else { list = p.parseRhsList() } } else { p.expect(token.DEFAULT) } colon := p.expect(token.COLON) p.openScope() body := p.parseStmtList() p.closeScope() return &ast.CaseClause{Case: pos, List: list, Colon: colon, Body: body} } func isTypeSwitchAssert(x ast.Expr) bool { a, ok := x.(*ast.TypeAssertExpr) return ok && a.Type == nil } func isTypeSwitchGuard(s ast.Stmt) bool { switch t := s.(type) { case *ast.ExprStmt: // x.(nil) return isTypeSwitchAssert(t.X) case *ast.AssignStmt: // v := x.(nil) return len(t.Lhs) == 1 && t.Tok == token.DEFINE && len(t.Rhs) == 1 && isTypeSwitchAssert(t.Rhs[0]) } return false } func (p *parser) parseSwitchStmt() ast.Stmt { if p.trace { defer un(trace(p, "SwitchStmt")) } pos := p.expect(token.SWITCH) p.openScope() defer p.closeScope() var s1, s2 ast.Stmt if p.tok != token.LBRACE { prevLev := p.exprLev p.exprLev = -1 if p.tok != token.SEMICOLON { s2, _ = p.parseSimpleStmt(basic) } if p.tok == token.SEMICOLON { p.next() s1 = s2 s2 = nil if p.tok != token.LBRACE { // A TypeSwitchGuard may declare a variable in addition // to the variable declared in the initial SimpleStmt. // Introduce extra scope to avoid redeclaration errors: // // switch t := 0; t := x.(T) { ... } // // (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). // // If we don't have a type switch, s2 must be an expression. // Having the extra nested but empty scope won't affect it. p.openScope() defer p.closeScope() s2, _ = p.parseSimpleStmt(basic) } } p.exprLev = prevLev } typeSwitch := isTypeSwitchGuard(s2) lbrace := p.expect(token.LBRACE) var list []ast.Stmt for p.tok == token.CASE || p.tok == token.DEFAULT { list = append(list, p.parseCaseClause(typeSwitch)) } rbrace := p.expect(token.RBRACE) p.expectSemi() body := &ast.BlockStmt{Lbrace: lbrace, List: list, Rbrace: rbrace} if typeSwitch { return &ast.TypeSwitchStmt{Switch: pos, Init: s1, Assign: s2, Body: body} } return &ast.SwitchStmt{Switch: pos, Init: s1, Tag: p.makeExpr(s2, "switch expression"), Body: body} } func (p *parser) parseCommClause() *ast.CommClause { if p.trace { defer un(trace(p, "CommClause")) } p.openScope() pos := p.pos var comm ast.Stmt if p.tok == token.CASE { p.next() lhs := p.parseLhsList() if p.tok == token.ARROW { // SendStmt if len(lhs) > 1 { p.errorExpected(lhs[0].Pos(), "1 expression") // continue with first expression } arrow := p.pos p.next() rhs := p.parseRhs() comm = &ast.SendStmt{Chan: lhs[0], Arrow: arrow, Value: rhs} } else { // RecvStmt if tok := p.tok; tok == token.ASSIGN || tok == token.DEFINE { // RecvStmt with assignment if len(lhs) > 2 { p.errorExpected(lhs[0].Pos(), "1 or 2 expressions") // continue with first two expressions lhs = lhs[0:2] } pos := p.pos p.next() rhs := p.parseRhs() as := &ast.AssignStmt{Lhs: lhs, TokPos: pos, Tok: tok, Rhs: []ast.Expr{rhs}} if tok == token.DEFINE { p.shortVarDecl(as, lhs) } comm = as } else { // lhs must be single receive operation if len(lhs) > 1 { p.errorExpected(lhs[0].Pos(), "1 expression") // continue with first expression } comm = &ast.ExprStmt{X: lhs[0]} } } } else { p.expect(token.DEFAULT) } colon := p.expect(token.COLON) body := p.parseStmtList() p.closeScope() return &ast.CommClause{Case: pos, Comm: comm, Colon: colon, Body: body} } func (p *parser) parseSelectStmt() *ast.SelectStmt { if p.trace { defer un(trace(p, "SelectStmt")) } pos := p.expect(token.SELECT) lbrace := p.expect(token.LBRACE) var list []ast.Stmt for p.tok == token.CASE || p.tok == token.DEFAULT { list = append(list, p.parseCommClause()) } rbrace := p.expect(token.RBRACE) p.expectSemi() body := &ast.BlockStmt{Lbrace: lbrace, List: list, Rbrace: rbrace} return &ast.SelectStmt{Select: pos, Body: body} } func (p *parser) parseForStmt() ast.Stmt { if p.trace { defer un(trace(p, "ForStmt")) } pos := p.expect(token.FOR) p.openScope() defer p.closeScope() var s1, s2, s3 ast.Stmt var isRange bool if p.tok != token.LBRACE { prevLev := p.exprLev p.exprLev = -1 if p.tok != token.SEMICOLON { if p.tok == token.RANGE { // "for range x" (nil lhs in assignment) pos := p.pos p.next() y := []ast.Expr{&ast.UnaryExpr{OpPos: pos, Op: token.RANGE, X: p.parseRhs()}} s2 = &ast.AssignStmt{Rhs: y} isRange = true } else { s2, isRange = p.parseSimpleStmt(rangeOk) } } if !isRange && p.tok == token.SEMICOLON { p.next() s1 = s2 s2 = nil if p.tok != token.SEMICOLON { s2, _ = p.parseSimpleStmt(basic) } p.expectSemi() if p.tok != token.LBRACE { s3, _ = p.parseSimpleStmt(basic) } } p.exprLev = prevLev } body := p.parseBlockStmt() p.expectSemi() if isRange { as := s2.(*ast.AssignStmt) // check lhs var key, value ast.Expr switch len(as.Lhs) { case 0: // nothing to do case 1: key = as.Lhs[0] case 2: key, value = as.Lhs[0], as.Lhs[1] default: p.errorExpected(as.Lhs[len(as.Lhs)-1].Pos(), "at most 2 expressions") return &ast.BadStmt{From: pos, To: p.safePos(body.End())} } // parseSimpleStmt returned a right-hand side that // is a single unary expression of the form "range x" x := as.Rhs[0].(*ast.UnaryExpr).X return &ast.RangeStmt{ For: pos, Key: key, Value: value, TokPos: as.TokPos, Tok: as.Tok, X: x, Body: body, } } // regular for statement return &ast.ForStmt{ For: pos, Init: s1, Cond: p.makeExpr(s2, "boolean or range expression"), Post: s3, Body: body, } } func (p *parser) parseStmt() (s ast.Stmt) { if p.trace { defer un(trace(p, "Statement")) } switch p.tok { case token.CONST, token.TYPE, token.VAR: s = &ast.DeclStmt{Decl: p.parseDecl(syncStmt)} case // tokens that may start an expression token.IDENT, token.INT, token.FLOAT, token.IMAG, token.CHAR, token.STRING, token.FUNC, token.LPAREN, // operands token.LBRACK, token.STRUCT, // composite types token.ADD, token.SUB, token.MUL, token.AND, token.XOR, token.ARROW, token.NOT: // unary operators s, _ = p.parseSimpleStmt(labelOk) // because of the required look-ahead, labeled statements are // parsed by parseSimpleStmt - don't expect a semicolon after // them if _, isLabeledStmt := s.(*ast.LabeledStmt); !isLabeledStmt { p.expectSemi() } case token.GO: s = p.parseGoStmt() case token.DEFER: s = p.parseDeferStmt() case token.RETURN: s = p.parseReturnStmt() case token.BREAK, token.CONTINUE, token.GOTO, token.FALLTHROUGH: s = p.parseBranchStmt(p.tok) case token.LBRACE: s = p.parseBlockStmt() p.expectSemi() case token.IF: s = p.parseIfStmt() case token.SWITCH: s = p.parseSwitchStmt() case token.SELECT: s = p.parseSelectStmt() case token.FOR: s = p.parseForStmt() case token.SEMICOLON: s = &ast.EmptyStmt{Semicolon: p.pos} p.next() case token.RBRACE: // a semicolon may be omitted before a closing "}" s = &ast.EmptyStmt{Semicolon: p.pos} default: // no statement found pos := p.pos p.errorExpected(pos, "statement") syncStmt(p) s = &ast.BadStmt{From: pos, To: p.pos} } return } // ---------------------------------------------------------------------------- // Declarations type parseSpecFunction func(doc *ast.CommentGroup, keyword token.Token, iota int) ast.Spec func isValidImport(lit string) bool { const illegalChars = `!"#$%&'()*,:;<=>?[\]^{|}` + "`\uFFFD" s, _ := strconv.Unquote(lit) // go/scanner returns a legal string literal for _, r := range s { if !unicode.IsGraphic(r) || unicode.IsSpace(r) || strings.ContainsRune(illegalChars, r) { return false } } return s != "" } func (p *parser) parseImportSpec(doc *ast.CommentGroup, _ token.Token, _ int) ast.Spec { if p.trace { defer un(trace(p, "ImportSpec")) } var ident *ast.Ident switch p.tok { case token.PERIOD: ident = &ast.Ident{NamePos: p.pos, Name: "."} p.next() case token.IDENT: ident = p.parseIdent() } pos := p.pos var path string if p.tok == token.STRING { path = p.lit if !isValidImport(path) { p.error(pos, "invalid import path: "+path) } p.next() } else { p.expect(token.STRING) // use expect() error handling } p.expectSemi() // call before accessing p.linecomment // collect imports spec := &ast.ImportSpec{ Doc: doc, Name: ident, Path: &ast.BasicLit{ValuePos: pos, Kind: token.STRING, Value: path}, Comment: p.lineComment, } p.imports = append(p.imports, spec) return spec } func (p *parser) parseValueSpec(doc *ast.CommentGroup, keyword token.Token, iota int) ast.Spec { if p.trace { defer un(trace(p, keyword.String()+"Spec")) } idents := p.parseIdentList() typ := p.tryType() var values []ast.Expr // always permit optional initialization for more tolerant parsing if p.tok == token.ASSIGN { p.next() 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, } kind := ast.Con if keyword == token.VAR { kind = ast.Var } p.declare(spec, iota, p.topScope, kind, idents...) return spec } func (p *parser) parseTypeSpec(doc *ast.CommentGroup, _ token.Token, _ int) ast.Spec { if p.trace { defer un(trace(p, "TypeSpec")) } ident := p.parseIdent() // Go spec: The scope of a type identifier declared inside a function begins // at the identifier in the TypeSpec and ends at the end of the innermost // containing block. // (Global identifiers are resolved in a separate phase after parsing.) spec := &ast.TypeSpec{Doc: doc, Name: ident} p.declare(spec, nil, p.topScope, ast.Typ, ident) spec.Type = p.parseType() p.expectSemi() // call before accessing p.linecomment spec.Comment = p.lineComment return spec } func (p *parser) parseGenDecl(keyword token.Token, f parseSpecFunction) *ast.GenDecl { if p.trace { defer un(trace(p, "GenDecl("+keyword.String()+")")) } doc := p.leadComment pos := p.expect(keyword) var lparen, rparen token.Pos var list []ast.Spec if p.tok == token.LPAREN { lparen = p.pos p.next() for iota := 0; p.tok != token.RPAREN && p.tok != token.EOF; iota++ { list = append(list, f(p.leadComment, keyword, iota)) } rparen = p.expect(token.RPAREN) p.expectSemi() } else { list = append(list, f(nil, keyword, 0)) } return &ast.GenDecl{ Doc: doc, TokPos: pos, Tok: keyword, Lparen: lparen, Specs: list, Rparen: rparen, } } func (p *parser) parseFuncDecl() *ast.FuncDecl { if p.trace { defer un(trace(p, "FunctionDecl")) } doc := p.leadComment pos := p.expect(token.FUNC) scope := ast.NewScope(p.topScope) // function scope var recv *ast.FieldList if p.tok == token.LPAREN { recv = p.parseParameters(scope, false) } ident := p.parseIdent() params, results := p.parseSignature(scope) var body *ast.BlockStmt if p.tok == token.LBRACE { body = p.parseBody(scope) } p.expectSemi() decl := &ast.FuncDecl{ Doc: doc, Recv: recv, Name: ident, Type: &ast.FuncType{ Func: pos, Params: params, Results: results, }, Body: body, } if recv == nil { // Go spec: The scope of an identifier denoting a constant, type, // variable, or function (but not method) declared at top level // (outside any function) is the package block. // // init() functions cannot be referred to and there may // be more than one - don't put them in the pkgScope if ident.Name != "init" { p.declare(decl, nil, p.pkgScope, ast.Fun, ident) } } return decl } func (p *parser) parseDecl(sync func(*parser)) ast.Decl { if p.trace { defer un(trace(p, "Declaration")) } var f parseSpecFunction switch p.tok { case token.CONST, token.VAR: f = p.parseValueSpec case token.TYPE: f = p.parseTypeSpec case token.FUNC: return p.parseFuncDecl() default: pos := p.pos p.errorExpected(pos, "declaration") sync(p) return &ast.BadDecl{From: pos, To: p.pos} } return p.parseGenDecl(p.tok, f) } // ---------------------------------------------------------------------------- // Source files func (p *parser) parseFile() *ast.File { if p.trace { 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) // Go spec: The package clause is not a declaration; // the package name does not appear in any scope. ident := p.parseIdent() if ident.Name == "_" && p.mode&DeclarationErrors != 0 { p.error(p.pos, "invalid package name _") } p.expectSemi() // 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 } 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, p.parseImportSpec)) } if p.mode&ImportsOnly == 0 { // rest of package body for p.tok != token.EOF { decls = append(decls, p.parseDecl(syncDecl)) } } } p.closeScope() assert(p.topScope == nil, "unbalanced scopes") assert(p.labelScope == nil, "unbalanced label scopes") // resolve global identifiers within the same file i := 0 for _, ident := range p.unresolved { // i <= index for current ident assert(ident.Obj == unresolved, "object already resolved") ident.Obj = p.pkgScope.Lookup(ident.Name) // also removes unresolved sentinel if ident.Obj == nil { p.unresolved[i] = ident i++ } } return &ast.File{ Doc: doc, Package: pos, Name: ident, Decls: decls, Scope: p.pkgScope, Imports: p.imports, Unresolved: p.unresolved[0:i], Comments: p.comments, } }