diff options
Diffstat (limited to 'src/go/parser')
-rw-r--r-- | src/go/parser/error_test.go | 187 | ||||
-rw-r--r-- | src/go/parser/example_test.go | 43 | ||||
-rw-r--r-- | src/go/parser/interface.go | 241 | ||||
-rw-r--r-- | src/go/parser/parser.go | 2591 | ||||
-rw-r--r-- | src/go/parser/parser_test.go | 579 | ||||
-rw-r--r-- | src/go/parser/performance_test.go | 30 | ||||
-rw-r--r-- | src/go/parser/short_test.go | 136 | ||||
-rw-r--r-- | src/go/parser/testdata/commas.src | 19 | ||||
-rw-r--r-- | src/go/parser/testdata/issue11377.src | 27 | ||||
-rw-r--r-- | src/go/parser/testdata/issue23434.src | 25 | ||||
-rw-r--r-- | src/go/parser/testdata/issue3106.src | 46 | ||||
-rw-r--r-- | src/go/parser/testdata/issue34946.src | 22 | ||||
-rw-r--r-- | src/go/parser/testdata/issue42951/not_a_file.go/invalid.go | 1 |
13 files changed, 3947 insertions, 0 deletions
diff --git a/src/go/parser/error_test.go b/src/go/parser/error_test.go new file mode 100644 index 0000000..358a844 --- /dev/null +++ b/src/go/parser/error_test.go @@ -0,0 +1,187 @@ +// Copyright 2012 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. + +// This file implements a parser test harness. The files in the testdata +// directory are parsed and the errors reported are compared against the +// error messages expected in the test files. The test files must end in +// .src rather than .go so that they are not disturbed by gofmt runs. +// +// Expected errors are indicated in the test files by putting a comment +// of the form /* ERROR "rx" */ immediately following an offending token. +// The harness will verify that an error matching the regular expression +// rx is reported at that source position. +// +// For instance, the following test file indicates that a "not declared" +// error should be reported for the undeclared variable x: +// +// package p +// func f() { +// _ = x /* ERROR "not declared" */ + 1 +// } + +package parser + +import ( + "go/scanner" + "go/token" + "os" + "path/filepath" + "regexp" + "strings" + "testing" +) + +const testdata = "testdata" + +// getFile assumes that each filename occurs at most once +func getFile(fset *token.FileSet, filename string) (file *token.File) { + fset.Iterate(func(f *token.File) bool { + if f.Name() == filename { + if file != nil { + panic(filename + " used multiple times") + } + file = f + } + return true + }) + return file +} + +func getPos(fset *token.FileSet, filename string, offset int) token.Pos { + if f := getFile(fset, filename); f != nil { + return f.Pos(offset) + } + return token.NoPos +} + +// ERROR comments must be of the form /* ERROR "rx" */ and rx is +// a regular expression that matches the expected error message. +// The special form /* ERROR HERE "rx" */ must be used for error +// messages that appear immediately after a token, rather than at +// a token's position. +// +var errRx = regexp.MustCompile(`^/\* *ERROR *(HERE)? *"([^"]*)" *\*/$`) + +// expectedErrors collects the regular expressions of ERROR comments found +// in files and returns them as a map of error positions to error messages. +// +func expectedErrors(fset *token.FileSet, filename string, src []byte) map[token.Pos]string { + errors := make(map[token.Pos]string) + + var s scanner.Scanner + // file was parsed already - do not add it again to the file + // set otherwise the position information returned here will + // not match the position information collected by the parser + s.Init(getFile(fset, filename), src, nil, scanner.ScanComments) + var prev token.Pos // position of last non-comment, non-semicolon token + var here token.Pos // position immediately after the token at position prev + + for { + pos, tok, lit := s.Scan() + switch tok { + case token.EOF: + return errors + case token.COMMENT: + s := errRx.FindStringSubmatch(lit) + if len(s) == 3 { + pos := prev + if s[1] == "HERE" { + pos = here + } + errors[pos] = string(s[2]) + } + case token.SEMICOLON: + // don't use the position of auto-inserted (invisible) semicolons + if lit != ";" { + break + } + fallthrough + default: + prev = pos + var l int // token length + if tok.IsLiteral() { + l = len(lit) + } else { + l = len(tok.String()) + } + here = prev + token.Pos(l) + } + } +} + +// compareErrors compares the map of expected error messages with the list +// of found errors and reports discrepancies. +// +func compareErrors(t *testing.T, fset *token.FileSet, expected map[token.Pos]string, found scanner.ErrorList) { + for _, error := range found { + // error.Pos is a token.Position, but we want + // a token.Pos so we can do a map lookup + pos := getPos(fset, error.Pos.Filename, error.Pos.Offset) + if msg, found := expected[pos]; found { + // we expect a message at pos; check if it matches + rx, err := regexp.Compile(msg) + if err != nil { + t.Errorf("%s: %v", error.Pos, err) + continue + } + if match := rx.MatchString(error.Msg); !match { + t.Errorf("%s: %q does not match %q", error.Pos, error.Msg, msg) + continue + } + // we have a match - eliminate this error + delete(expected, pos) + } else { + // To keep in mind when analyzing failed test output: + // If the same error position occurs multiple times in errors, + // this message will be triggered (because the first error at + // the position removes this position from the expected errors). + t.Errorf("%s: unexpected error: %s", error.Pos, error.Msg) + } + } + + // there should be no expected errors left + if len(expected) > 0 { + t.Errorf("%d errors not reported:", len(expected)) + for pos, msg := range expected { + t.Errorf("%s: %s\n", fset.Position(pos), msg) + } + } +} + +func checkErrors(t *testing.T, filename string, input interface{}) { + src, err := readSource(filename, input) + if err != nil { + t.Error(err) + return + } + + fset := token.NewFileSet() + _, err = ParseFile(fset, filename, src, DeclarationErrors|AllErrors) + found, ok := err.(scanner.ErrorList) + if err != nil && !ok { + t.Error(err) + return + } + found.RemoveMultiples() + + // we are expecting the following errors + // (collect these after parsing a file so that it is found in the file set) + expected := expectedErrors(fset, filename, src) + + // verify errors returned by the parser + compareErrors(t, fset, expected, found) +} + +func TestErrors(t *testing.T) { + list, err := os.ReadDir(testdata) + if err != nil { + t.Fatal(err) + } + for _, d := range list { + name := d.Name() + if !d.IsDir() && !strings.HasPrefix(name, ".") && strings.HasSuffix(name, ".src") { + checkErrors(t, filepath.Join(testdata, name), nil) + } + } +} diff --git a/src/go/parser/example_test.go b/src/go/parser/example_test.go new file mode 100644 index 0000000..c2f7f29 --- /dev/null +++ b/src/go/parser/example_test.go @@ -0,0 +1,43 @@ +// Copyright 2012 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_test + +import ( + "fmt" + "go/parser" + "go/token" +) + +func ExampleParseFile() { + fset := token.NewFileSet() // positions are relative to fset + + src := `package foo + +import ( + "fmt" + "time" +) + +func bar() { + fmt.Println(time.Now()) +}` + + // Parse src but stop after processing the imports. + f, err := parser.ParseFile(fset, "", src, parser.ImportsOnly) + if err != nil { + fmt.Println(err) + return + } + + // Print the imports from the file's AST. + for _, s := range f.Imports { + fmt.Println(s.Path.Value) + } + + // output: + // + // "fmt" + // "time" +} diff --git a/src/go/parser/interface.go b/src/go/parser/interface.go new file mode 100644 index 0000000..56ff5fe --- /dev/null +++ b/src/go/parser/interface.go @@ -0,0 +1,241 @@ +// 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. + +// This file contains the exported entry points for invoking the parser. + +package parser + +import ( + "bytes" + "errors" + "go/ast" + "go/token" + "io" + "io/fs" + "os" + "path/filepath" + "strings" +) + +// If src != nil, readSource converts src to a []byte if possible; +// otherwise it returns an error. If src == nil, readSource returns +// the result of reading the file specified by filename. +// +func readSource(filename string, src interface{}) ([]byte, error) { + if src != nil { + switch s := src.(type) { + case string: + return []byte(s), nil + case []byte: + return s, nil + case *bytes.Buffer: + // is io.Reader, but src is already available in []byte form + if s != nil { + return s.Bytes(), nil + } + case io.Reader: + return io.ReadAll(s) + } + return nil, errors.New("invalid source") + } + return os.ReadFile(filename) +} + +// A Mode value is a set of flags (or 0). +// They control the amount of source code parsed and other optional +// parser functionality. +// +type Mode uint + +const ( + PackageClauseOnly Mode = 1 << iota // stop parsing after package clause + ImportsOnly // stop parsing after import declarations + ParseComments // parse comments and add them to AST + Trace // print a trace of parsed productions + DeclarationErrors // report declaration errors + SpuriousErrors // same as AllErrors, for backward-compatibility + AllErrors = SpuriousErrors // report all errors (not just the first 10 on different lines) +) + +// ParseFile parses the source code of a single Go source file and returns +// the corresponding ast.File node. The source code may be provided via +// the filename of the source file, or via the src parameter. +// +// If src != nil, ParseFile parses the source from src and the filename is +// only used when recording position information. The type of the argument +// for the src parameter must be string, []byte, or io.Reader. +// If src == nil, ParseFile parses the file specified by filename. +// +// The mode parameter controls the amount of source text parsed and other +// optional parser functionality. Position information is recorded in the +// file set fset, which must not be nil. +// +// If the source couldn't be read, the returned AST is nil and the error +// indicates the specific failure. If the source was read but syntax +// errors were found, the result is a partial AST (with ast.Bad* nodes +// representing the fragments of erroneous source code). Multiple errors +// are returned via a scanner.ErrorList which is sorted by source position. +// +func ParseFile(fset *token.FileSet, filename string, src interface{}, mode Mode) (f *ast.File, err error) { + if fset == nil { + panic("parser.ParseFile: no token.FileSet provided (fset == nil)") + } + + // get source + text, err := readSource(filename, src) + if err != nil { + return nil, err + } + + var p parser + defer func() { + if e := recover(); e != nil { + // resume same panic if it's not a bailout + if _, ok := e.(bailout); !ok { + panic(e) + } + } + + // set result values + if f == nil { + // source is not a valid Go source file - satisfy + // ParseFile API and return a valid (but) empty + // *ast.File + f = &ast.File{ + Name: new(ast.Ident), + Scope: ast.NewScope(nil), + } + } + + p.errors.Sort() + err = p.errors.Err() + }() + + // parse source + p.init(fset, filename, text, mode) + f = p.parseFile() + + return +} + +// ParseDir calls ParseFile for all files with names ending in ".go" in the +// directory specified by path and returns a map of package name -> package +// AST with all the packages found. +// +// If filter != nil, only the files with fs.FileInfo entries passing through +// the filter (and ending in ".go") are considered. The mode bits are passed +// to ParseFile unchanged. Position information is recorded in fset, which +// must not be nil. +// +// If the directory couldn't be read, a nil map and the respective error are +// returned. If a parse error occurred, a non-nil but incomplete map and the +// first error encountered are returned. +// +func ParseDir(fset *token.FileSet, path string, filter func(fs.FileInfo) bool, mode Mode) (pkgs map[string]*ast.Package, first error) { + list, err := os.ReadDir(path) + if err != nil { + return nil, err + } + + pkgs = make(map[string]*ast.Package) + for _, d := range list { + if d.IsDir() || !strings.HasSuffix(d.Name(), ".go") { + continue + } + if filter != nil { + info, err := d.Info() + if err != nil { + return nil, err + } + if !filter(info) { + continue + } + } + filename := filepath.Join(path, d.Name()) + if src, err := ParseFile(fset, filename, nil, mode); err == nil { + name := src.Name.Name + pkg, found := pkgs[name] + if !found { + pkg = &ast.Package{ + Name: name, + Files: make(map[string]*ast.File), + } + pkgs[name] = pkg + } + pkg.Files[filename] = src + } else if first == nil { + first = err + } + } + + return +} + +// ParseExprFrom is a convenience function for parsing an expression. +// The arguments have the same meaning as for ParseFile, but the source must +// be a valid Go (type or value) expression. Specifically, fset must not +// be nil. +// +// If the source couldn't be read, the returned AST is nil and the error +// indicates the specific failure. If the source was read but syntax +// errors were found, the result is a partial AST (with ast.Bad* nodes +// representing the fragments of erroneous source code). Multiple errors +// are returned via a scanner.ErrorList which is sorted by source position. +// +func ParseExprFrom(fset *token.FileSet, filename string, src interface{}, mode Mode) (expr ast.Expr, err error) { + if fset == nil { + panic("parser.ParseExprFrom: no token.FileSet provided (fset == nil)") + } + + // get source + text, err := readSource(filename, src) + if err != nil { + return nil, err + } + + var p parser + defer func() { + if e := recover(); e != nil { + // resume same panic if it's not a bailout + if _, ok := e.(bailout); !ok { + panic(e) + } + } + p.errors.Sort() + err = p.errors.Err() + }() + + // parse expr + p.init(fset, filename, text, mode) + // Set up pkg-level scopes to avoid nil-pointer errors. + // This is not needed for a correct expression x as the + // parser will be ok with a nil topScope, but be cautious + // in case of an erroneous x. + p.openScope() + p.pkgScope = p.topScope + expr = p.parseRhsOrType() + p.closeScope() + assert(p.topScope == nil, "unbalanced scopes") + + // If a semicolon was inserted, consume it; + // report an error if there's more tokens. + if p.tok == token.SEMICOLON && p.lit == "\n" { + p.next() + } + p.expect(token.EOF) + + return +} + +// ParseExpr is a convenience function for obtaining the AST of an expression x. +// The position information recorded in the AST is undefined. The filename used +// in error messages is the empty string. +// +// If syntax errors were found, the result is a partial AST (with ast.Bad* nodes +// representing the fragments of erroneous source code). Multiple errors are +// returned via a scanner.ErrorList which is sorted by source position. +// +func ParseExpr(x string) (ast.Expr, error) { + return ParseExprFrom(token.NewFileSet(), "", []byte(x), 0) +} diff --git a/src/go/parser/parser.go b/src/go/parser/parser.go new file mode 100644 index 0000000..31a7398 --- /dev/null +++ b/src/go/parser/parser.go @@ -0,0 +1,2591 @@ +// 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. +// +// The parser accepts a larger language than is syntactically permitted by +// the Go spec, for simplicity, and for improved robustness in the presence +// of syntax errors. For instance, in method declarations, the receiver is +// treated like an ordinary parameter list and thus may contain multiple +// entries where the spec permits exactly one. Consequently, the corresponding +// field in the AST (ast.FuncDecl.Recv) field is not restricted to one entry. +// +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 parser.advance + // 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 parser.advance calls 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 +// 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 || p.tok == token.EOF { + // 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 + switch { + case p.tok == token.SEMICOLON && p.lit == "\n": + msg += ", found newline" + case p.tok.IsLiteral(): + // print 123 rather than 'INT', etc. + msg += ", found " + p.lit + default: + msg += ", found '" + p.tok.String() + "'" + } + } + 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 +} + +// expect2 is like expect, but it returns an invalid position +// if the expected token is not found. +func (p *parser) expect2(tok token.Token) (pos token.Pos) { + if p.tok == tok { + pos = p.pos + } else { + p.errorExpected(p.pos, "'"+tok.String()+"'") + } + p.next() // make progress + return +} + +// 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 { + switch p.tok { + case token.COMMA: + // permit a ',' instead of a ';' but complain + p.errorExpected(p.pos, "';'") + fallthrough + case token.SEMICOLON: + p.next() + default: + p.errorExpected(p.pos, "';'") + p.advance(stmtStart) + } + } +} + +func (p *parser) atComma(context string, follow token.Token) bool { + if p.tok == token.COMMA { + return true + } + if p.tok != follow { + msg := "missing ','" + if p.tok == token.SEMICOLON && p.lit == "\n" { + msg += " before newline" + } + p.error(p.pos, msg+" in "+context) + return true // "insert" comma and continue + } + return false +} + +func assert(cond bool, msg string) { + if !cond { + panic("go/parser internal error: " + msg) + } +} + +// advance consumes tokens until the current token p.tok +// is in the 'to' set, or token.EOF. For error recovery. +func (p *parser) advance(to map[token.Token]bool) { + for ; p.tok != token.EOF; p.next() { + if to[p.tok] { + // Return only if parser made some progress since last + // sync or if it has not reached 10 advance calls without + // progress. Otherwise consume at least one token to + // avoid an endless parser loop (it is possible that + // both parseOperand and parseStmt call advance 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. + } + } +} + +var stmtStart = map[token.Token]bool{ + token.BREAK: true, + token.CONST: true, + token.CONTINUE: true, + token.DEFER: true, + token.FALLTHROUGH: true, + token.FOR: true, + token.GO: true, + token.GOTO: true, + token.IF: true, + token.RETURN: true, + token.SELECT: true, + token.SWITCH: true, + token.TYPE: true, + token.VAR: true, +} + +var declStart = map[token.Token]bool{ + token.CONST: true, + token.TYPE: true, + token.VAR: true, +} + +var exprEnd = map[token.Token]bool{ + token.COMMA: true, + token.COLON: true, + token.SEMICOLON: true, + token.RPAREN: true, + token.RBRACK: true, + token.RBRACE: true, +} + +// 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.advance(exprEnd) + 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 + + // 1st FieldDecl + // A type name used as an anonymous field looks like a field identifier. + var list []ast.Expr + for { + list = append(list, p.parseVarType(false)) + if p.tok != token.COMMA { + break + } + p.next() + } + + typ := p.tryVarType(false) + + // 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 { + p.errorExpected(p.pos, "type") + typ = &ast.BadExpr{From: p.pos, To: p.pos} + } else if !isTypeName(deref(typ)) { + p.errorExpected(typ.Pos(), "anonymous field") + typ = &ast.BadExpr{From: typ.Pos(), To: p.safePos(typ.End())} + } + } + + // Tag + var tag *ast.BasicLit + if p.tok == token.STRING { + tag = &ast.BasicLit{ValuePos: p.pos, Kind: p.tok, Value: p.lit} + p.next() + } + + 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 +} + +func (p *parser) parseParameterList(scope *ast.Scope, ellipsisOk bool) (params []*ast.Field) { + if p.trace { + defer un(trace(p, "ParameterList")) + } + + // 1st ParameterDecl + // A list of identifiers looks like a list of type names. + var list []ast.Expr + for { + list = append(list, p.parseVarType(ellipsisOk)) + if p.tok != token.COMMA { + break + } + p.next() + if p.tok == token.RPAREN { + break + } + } + + // analyze case + if typ := p.tryVarType(ellipsisOk); 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", token.RPAREN) { + 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", token.RPAREN) { + 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.expect2(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.expect2(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") + p.advance(stmtStart) + 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", token.RPAREN) { + 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) parseValue(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 { + // 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) + } else { + // not a key + p.resolve(x) + } + } + + return x +} + +func (p *parser) parseElement() ast.Expr { + if p.trace { + defer un(trace(p, "Element")) + } + + x := p.parseValue(true) + if p.tok == token.COLON { + colon := p.pos + p.next() + x = &ast.KeyValueExpr{Key: x, Colon: colon, Value: p.parseValue(false)} + } + + 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()) + if !p.atComma("composite literal", token.RBRACE) { + 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 reports whether 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 reports whether 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.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 + sel := &ast.Ident{NamePos: pos, Name: "_"} + x = &ast.SelectorExpr{X: x, Sel: sel} + } + 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 { + op, oprec := p.tokPrec() + if oprec < prec1 { + return x + } + pos := p.expect(op) + if lhs { + p.resolve(x) + lhs = false + } + y := p.parseBinaryExpr(false, oprec+1) + x = &ast.BinaryExpr{X: p.checkExpr(x), OpPos: pos, Op: op, Y: p.checkExpr(y)} + } +} + +// 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 + // reported 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, want string) ast.Expr { + if s == nil { + return nil + } + if es, isExpr := s.(*ast.ExprStmt); isExpr { + return p.checkExpr(es.X) + } + found := "simple statement" + if _, isAss := s.(*ast.AssignStmt); isAss { + found = "assignment" + } + p.error(s.Pos(), fmt.Sprintf("expected %s, found %s (missing parentheses around composite literal?)", want, found)) + return &ast.BadExpr{From: s.Pos(), To: p.safePos(s.End())} +} + +// parseIfHeader is an adjusted version of parser.header +// in cmd/compile/internal/syntax/parser.go, which has +// been tuned for better error handling. +func (p *parser) parseIfHeader() (init ast.Stmt, cond ast.Expr) { + if p.tok == token.LBRACE { + p.error(p.pos, "missing condition in if statement") + cond = &ast.BadExpr{From: p.pos, To: p.pos} + return + } + // p.tok != token.LBRACE + + outer := p.exprLev + p.exprLev = -1 + + if p.tok != token.SEMICOLON { + // accept potential variable declaration but complain + if p.tok == token.VAR { + p.next() + p.error(p.pos, fmt.Sprintf("var declaration not allowed in 'IF' initializer")) + } + init, _ = p.parseSimpleStmt(basic) + } + + var condStmt ast.Stmt + var semi struct { + pos token.Pos + lit string // ";" or "\n"; valid if pos.IsValid() + } + if p.tok != token.LBRACE { + if p.tok == token.SEMICOLON { + semi.pos = p.pos + semi.lit = p.lit + p.next() + } else { + p.expect(token.SEMICOLON) + } + if p.tok != token.LBRACE { + condStmt, _ = p.parseSimpleStmt(basic) + } + } else { + condStmt = init + init = nil + } + + if condStmt != nil { + cond = p.makeExpr(condStmt, "boolean expression") + } else if semi.pos.IsValid() { + if semi.lit == "\n" { + p.error(semi.pos, "unexpected newline, expecting { after if clause") + } else { + p.error(semi.pos, "missing condition in if statement") + } + } + + // make sure we have a valid AST + if cond == nil { + cond = &ast.BadExpr{From: p.pos, To: p.pos} + } + + p.exprLev = outer + return +} + +func (p *parser) parseIfStmt() *ast.IfStmt { + if p.trace { + defer un(trace(p, "IfStmt")) + } + + pos := p.expect(token.IF) + p.openScope() + defer p.closeScope() + + init, cond := p.parseIfHeader() + body := p.parseBlockStmt() + + var else_ ast.Stmt + if p.tok == token.ELSE { + p.next() + switch p.tok { + case token.IF: + else_ = p.parseIfStmt() + case token.LBRACE: + else_ = p.parseBlockStmt() + p.expectSemi() + default: + p.errorExpected(p.pos, "if statement or block") + else_ = &ast.BadStmt{From: p.pos, To: p.pos} + } + } else { + p.expectSemi() + } + + return &ast.IfStmt{If: pos, Init: init, Cond: cond, 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 (p *parser) isTypeSwitchGuard(s ast.Stmt) bool { + switch t := s.(type) { + case *ast.ExprStmt: + // x.(type) + return isTypeSwitchAssert(t.X) + case *ast.AssignStmt: + // v := x.(type) + if len(t.Lhs) == 1 && len(t.Rhs) == 1 && isTypeSwitchAssert(t.Rhs[0]) { + switch t.Tok { + case token.ASSIGN: + // permit v = x.(type) but complain + p.error(t.TokPos, "expected ':=', found '='") + fallthrough + case token.DEFINE: + return true + } + } + } + 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 := p.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(stmtStart)} + 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, token.MAP, token.CHAN, token.INTERFACE, // 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: + // Is it ever possible to have an implicit semicolon + // producing an empty statement in a valid program? + // (handle correctly anyway) + s = &ast.EmptyStmt{Semicolon: p.pos, Implicit: p.lit == "\n"} + p.next() + case token.RBRACE: + // a semicolon may be omitted before a closing "}" + s = &ast.EmptyStmt{Semicolon: p.pos, Implicit: true} + default: + // no statement found + pos := p.pos + p.errorExpected(pos, "statement") + p.advance(stmtStart) + 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")) + } + + pos := p.pos + 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 + + switch keyword { + case token.VAR: + if typ == nil && values == nil { + p.error(pos, "missing variable type or initialization") + } + case token.CONST: + if values == nil && (iota == 0 || typ != nil) { + p.error(pos, "missing constant value") + } + } + + // 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) + if p.tok == token.ASSIGN { + spec.Assign = p.pos + p.next() + } + 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() + } else if p.tok == token.SEMICOLON { + p.next() + if p.tok == token.LBRACE { + // opening { of function declaration on next line + p.error(p.pos, "unexpected semicolon or newline before {") + body = p.parseBody(scope) + p.expectSemi() + } + } else { + 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 map[token.Token]bool) 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") + p.advance(sync) + 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(declStart)) + } + } + } + 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, + } +} diff --git a/src/go/parser/parser_test.go b/src/go/parser/parser_test.go new file mode 100644 index 0000000..a4f882d --- /dev/null +++ b/src/go/parser/parser_test.go @@ -0,0 +1,579 @@ +// 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 + +import ( + "bytes" + "fmt" + "go/ast" + "go/token" + "io/fs" + "strings" + "testing" +) + +var validFiles = []string{ + "parser.go", + "parser_test.go", + "error_test.go", + "short_test.go", +} + +func TestParse(t *testing.T) { + for _, filename := range validFiles { + _, err := ParseFile(token.NewFileSet(), filename, nil, DeclarationErrors) + if err != nil { + t.Fatalf("ParseFile(%s): %v", filename, err) + } + } +} + +func nameFilter(filename string) bool { + switch filename { + case "parser.go", "interface.go", "parser_test.go": + return true + case "parser.go.orig": + return true // permit but should be ignored by ParseDir + } + return false +} + +func dirFilter(f fs.FileInfo) bool { return nameFilter(f.Name()) } + +func TestParseFile(t *testing.T) { + src := "package p\nvar _=s[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]" + _, err := ParseFile(token.NewFileSet(), "", src, 0) + if err == nil { + t.Errorf("ParseFile(%s) succeeded unexpectedly", src) + } +} + +func TestParseExprFrom(t *testing.T) { + src := "s[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]+\ns[::]" + _, err := ParseExprFrom(token.NewFileSet(), "", src, 0) + if err == nil { + t.Errorf("ParseExprFrom(%s) succeeded unexpectedly", src) + } +} + +func TestParseDir(t *testing.T) { + path := "." + pkgs, err := ParseDir(token.NewFileSet(), path, dirFilter, 0) + if err != nil { + t.Fatalf("ParseDir(%s): %v", path, err) + } + if n := len(pkgs); n != 1 { + t.Errorf("got %d packages; want 1", n) + } + pkg := pkgs["parser"] + if pkg == nil { + t.Errorf(`package "parser" not found`) + return + } + if n := len(pkg.Files); n != 3 { + t.Errorf("got %d package files; want 3", n) + } + for filename := range pkg.Files { + if !nameFilter(filename) { + t.Errorf("unexpected package file: %s", filename) + } + } +} + +func TestIssue42951(t *testing.T) { + path := "./testdata/issue42951" + _, err := ParseDir(token.NewFileSet(), path, nil, 0) + if err != nil { + t.Errorf("ParseDir(%s): %v", path, err) + } +} + +func TestParseExpr(t *testing.T) { + // just kicking the tires: + // a valid arithmetic expression + src := "a + b" + x, err := ParseExpr(src) + if err != nil { + t.Errorf("ParseExpr(%q): %v", src, err) + } + // sanity check + if _, ok := x.(*ast.BinaryExpr); !ok { + t.Errorf("ParseExpr(%q): got %T, want *ast.BinaryExpr", src, x) + } + + // a valid type expression + src = "struct{x *int}" + x, err = ParseExpr(src) + if err != nil { + t.Errorf("ParseExpr(%q): %v", src, err) + } + // sanity check + if _, ok := x.(*ast.StructType); !ok { + t.Errorf("ParseExpr(%q): got %T, want *ast.StructType", src, x) + } + + // an invalid expression + src = "a + *" + x, err = ParseExpr(src) + if err == nil { + t.Errorf("ParseExpr(%q): got no error", src) + } + if x == nil { + t.Errorf("ParseExpr(%q): got no (partial) result", src) + } + if _, ok := x.(*ast.BinaryExpr); !ok { + t.Errorf("ParseExpr(%q): got %T, want *ast.BinaryExpr", src, x) + } + + // a valid expression followed by extra tokens is invalid + src = "a[i] := x" + if _, err := ParseExpr(src); err == nil { + t.Errorf("ParseExpr(%q): got no error", src) + } + + // a semicolon is not permitted unless automatically inserted + src = "a + b\n" + if _, err := ParseExpr(src); err != nil { + t.Errorf("ParseExpr(%q): got error %s", src, err) + } + src = "a + b;" + if _, err := ParseExpr(src); err == nil { + t.Errorf("ParseExpr(%q): got no error", src) + } + + // various other stuff following a valid expression + const validExpr = "a + b" + const anything = "dh3*#D)#_" + for _, c := range "!)]};," { + src := validExpr + string(c) + anything + if _, err := ParseExpr(src); err == nil { + t.Errorf("ParseExpr(%q): got no error", src) + } + } + + // ParseExpr must not crash + for _, src := range valids { + ParseExpr(src) + } +} + +func TestColonEqualsScope(t *testing.T) { + f, err := ParseFile(token.NewFileSet(), "", `package p; func f() { x, y, z := x, y, z }`, 0) + if err != nil { + t.Fatal(err) + } + + // RHS refers to undefined globals; LHS does not. + as := f.Decls[0].(*ast.FuncDecl).Body.List[0].(*ast.AssignStmt) + for _, v := range as.Rhs { + id := v.(*ast.Ident) + if id.Obj != nil { + t.Errorf("rhs %s has Obj, should not", id.Name) + } + } + for _, v := range as.Lhs { + id := v.(*ast.Ident) + if id.Obj == nil { + t.Errorf("lhs %s does not have Obj, should", id.Name) + } + } +} + +func TestVarScope(t *testing.T) { + f, err := ParseFile(token.NewFileSet(), "", `package p; func f() { var x, y, z = x, y, z }`, 0) + if err != nil { + t.Fatal(err) + } + + // RHS refers to undefined globals; LHS does not. + as := f.Decls[0].(*ast.FuncDecl).Body.List[0].(*ast.DeclStmt).Decl.(*ast.GenDecl).Specs[0].(*ast.ValueSpec) + for _, v := range as.Values { + id := v.(*ast.Ident) + if id.Obj != nil { + t.Errorf("rhs %s has Obj, should not", id.Name) + } + } + for _, id := range as.Names { + if id.Obj == nil { + t.Errorf("lhs %s does not have Obj, should", id.Name) + } + } +} + +func TestObjects(t *testing.T) { + const src = ` +package p +import fmt "fmt" +const pi = 3.14 +type T struct{} +var x int +func f() { L: } +` + + f, err := ParseFile(token.NewFileSet(), "", src, 0) + if err != nil { + t.Fatal(err) + } + + objects := map[string]ast.ObjKind{ + "p": ast.Bad, // not in a scope + "fmt": ast.Bad, // not resolved yet + "pi": ast.Con, + "T": ast.Typ, + "x": ast.Var, + "int": ast.Bad, // not resolved yet + "f": ast.Fun, + "L": ast.Lbl, + } + + ast.Inspect(f, func(n ast.Node) bool { + if ident, ok := n.(*ast.Ident); ok { + obj := ident.Obj + if obj == nil { + if objects[ident.Name] != ast.Bad { + t.Errorf("no object for %s", ident.Name) + } + return true + } + if obj.Name != ident.Name { + t.Errorf("names don't match: obj.Name = %s, ident.Name = %s", obj.Name, ident.Name) + } + kind := objects[ident.Name] + if obj.Kind != kind { + t.Errorf("%s: obj.Kind = %s; want %s", ident.Name, obj.Kind, kind) + } + } + return true + }) +} + +func TestUnresolved(t *testing.T) { + f, err := ParseFile(token.NewFileSet(), "", ` +package p +// +func f1a(int) +func f2a(byte, int, float) +func f3a(a, b int, c float) +func f4a(...complex) +func f5a(a s1a, b ...complex) +// +func f1b(*int) +func f2b([]byte, (int), *float) +func f3b(a, b *int, c []float) +func f4b(...*complex) +func f5b(a s1a, b ...[]complex) +// +type s1a struct { int } +type s2a struct { byte; int; s1a } +type s3a struct { a, b int; c float } +// +type s1b struct { *int } +type s2b struct { byte; int; *float } +type s3b struct { a, b *s3b; c []float } +`, 0) + if err != nil { + t.Fatal(err) + } + + want := "int " + // f1a + "byte int float " + // f2a + "int float " + // f3a + "complex " + // f4a + "complex " + // f5a + // + "int " + // f1b + "byte int float " + // f2b + "int float " + // f3b + "complex " + // f4b + "complex " + // f5b + // + "int " + // s1a + "byte int " + // s2a + "int float " + // s3a + // + "int " + // s1a + "byte int float " + // s2a + "float " // s3a + + // collect unresolved identifiers + var buf bytes.Buffer + for _, u := range f.Unresolved { + buf.WriteString(u.Name) + buf.WriteByte(' ') + } + got := buf.String() + + if got != want { + t.Errorf("\ngot: %s\nwant: %s", got, want) + } +} + +var imports = map[string]bool{ + `"a"`: true, + "`a`": true, + `"a/b"`: true, + `"a.b"`: true, + `"m\x61th"`: true, + `"greek/αβ"`: true, + `""`: false, + + // Each of these pairs tests both `` vs "" strings + // and also use of invalid characters spelled out as + // escape sequences and written directly. + // For example `"\x00"` tests import "\x00" + // while "`\x00`" tests import `<actual-NUL-byte>`. + `"\x00"`: false, + "`\x00`": false, + `"\x7f"`: false, + "`\x7f`": false, + `"a!"`: false, + "`a!`": false, + `"a b"`: false, + "`a b`": false, + `"a\\b"`: false, + "`a\\b`": false, + "\"`a`\"": false, + "`\"a\"`": false, + `"\x80\x80"`: false, + "`\x80\x80`": false, + `"\xFFFD"`: false, + "`\xFFFD`": false, +} + +func TestImports(t *testing.T) { + for path, isValid := range imports { + src := fmt.Sprintf("package p; import %s", path) + _, err := ParseFile(token.NewFileSet(), "", src, 0) + switch { + case err != nil && isValid: + t.Errorf("ParseFile(%s): got %v; expected no error", src, err) + case err == nil && !isValid: + t.Errorf("ParseFile(%s): got no error; expected one", src) + } + } +} + +func TestCommentGroups(t *testing.T) { + f, err := ParseFile(token.NewFileSet(), "", ` +package p /* 1a */ /* 1b */ /* 1c */ // 1d +/* 2a +*/ +// 2b +const pi = 3.1415 +/* 3a */ // 3b +/* 3c */ const e = 2.7182 + +// Example from issue 3139 +func ExampleCount() { + fmt.Println(strings.Count("cheese", "e")) + fmt.Println(strings.Count("five", "")) // before & after each rune + // Output: + // 3 + // 5 +} +`, ParseComments) + if err != nil { + t.Fatal(err) + } + expected := [][]string{ + {"/* 1a */", "/* 1b */", "/* 1c */", "// 1d"}, + {"/* 2a\n*/", "// 2b"}, + {"/* 3a */", "// 3b", "/* 3c */"}, + {"// Example from issue 3139"}, + {"// before & after each rune"}, + {"// Output:", "// 3", "// 5"}, + } + if len(f.Comments) != len(expected) { + t.Fatalf("got %d comment groups; expected %d", len(f.Comments), len(expected)) + } + for i, exp := range expected { + got := f.Comments[i].List + if len(got) != len(exp) { + t.Errorf("got %d comments in group %d; expected %d", len(got), i, len(exp)) + continue + } + for j, exp := range exp { + got := got[j].Text + if got != exp { + t.Errorf("got %q in group %d; expected %q", got, i, exp) + } + } + } +} + +func getField(file *ast.File, fieldname string) *ast.Field { + parts := strings.Split(fieldname, ".") + for _, d := range file.Decls { + if d, ok := d.(*ast.GenDecl); ok && d.Tok == token.TYPE { + for _, s := range d.Specs { + if s, ok := s.(*ast.TypeSpec); ok && s.Name.Name == parts[0] { + if s, ok := s.Type.(*ast.StructType); ok { + for _, f := range s.Fields.List { + for _, name := range f.Names { + if name.Name == parts[1] { + return f + } + } + } + } + } + } + } + } + return nil +} + +// Don't use ast.CommentGroup.Text() - we want to see exact comment text. +func commentText(c *ast.CommentGroup) string { + var buf bytes.Buffer + if c != nil { + for _, c := range c.List { + buf.WriteString(c.Text) + } + } + return buf.String() +} + +func checkFieldComments(t *testing.T, file *ast.File, fieldname, lead, line string) { + f := getField(file, fieldname) + if f == nil { + t.Fatalf("field not found: %s", fieldname) + } + if got := commentText(f.Doc); got != lead { + t.Errorf("got lead comment %q; expected %q", got, lead) + } + if got := commentText(f.Comment); got != line { + t.Errorf("got line comment %q; expected %q", got, line) + } +} + +func TestLeadAndLineComments(t *testing.T) { + f, err := ParseFile(token.NewFileSet(), "", ` +package p +type T struct { + /* F1 lead comment */ + // + F1 int /* F1 */ // line comment + // F2 lead + // comment + F2 int // F2 line comment + // f3 lead comment + f3 int // f3 line comment +} +`, ParseComments) + if err != nil { + t.Fatal(err) + } + checkFieldComments(t, f, "T.F1", "/* F1 lead comment *///", "/* F1 */// line comment") + checkFieldComments(t, f, "T.F2", "// F2 lead// comment", "// F2 line comment") + checkFieldComments(t, f, "T.f3", "// f3 lead comment", "// f3 line comment") + ast.FileExports(f) + checkFieldComments(t, f, "T.F1", "/* F1 lead comment *///", "/* F1 */// line comment") + checkFieldComments(t, f, "T.F2", "// F2 lead// comment", "// F2 line comment") + if getField(f, "T.f3") != nil { + t.Error("not expected to find T.f3") + } +} + +// TestIssue9979 verifies that empty statements are contained within their enclosing blocks. +func TestIssue9979(t *testing.T) { + for _, src := range []string{ + "package p; func f() {;}", + "package p; func f() {L:}", + "package p; func f() {L:;}", + "package p; func f() {L:\n}", + "package p; func f() {L:\n;}", + "package p; func f() { ; }", + "package p; func f() { L: }", + "package p; func f() { L: ; }", + "package p; func f() { L: \n}", + "package p; func f() { L: \n; }", + } { + fset := token.NewFileSet() + f, err := ParseFile(fset, "", src, 0) + if err != nil { + t.Fatal(err) + } + + var pos, end token.Pos + ast.Inspect(f, func(x ast.Node) bool { + switch s := x.(type) { + case *ast.BlockStmt: + pos, end = s.Pos()+1, s.End()-1 // exclude "{", "}" + case *ast.LabeledStmt: + pos, end = s.Pos()+2, s.End() // exclude "L:" + case *ast.EmptyStmt: + // check containment + if s.Pos() < pos || s.End() > end { + t.Errorf("%s: %T[%d, %d] not inside [%d, %d]", src, s, s.Pos(), s.End(), pos, end) + } + // check semicolon + offs := fset.Position(s.Pos()).Offset + if ch := src[offs]; ch != ';' != s.Implicit { + want := "want ';'" + if s.Implicit { + want = "but ';' is implicit" + } + t.Errorf("%s: found %q at offset %d; %s", src, ch, offs, want) + } + } + return true + }) + } +} + +// TestIncompleteSelection ensures that an incomplete selector +// expression is parsed as a (blank) *ast.SelectorExpr, not a +// *ast.BadExpr. +func TestIncompleteSelection(t *testing.T) { + for _, src := range []string{ + "package p; var _ = fmt.", // at EOF + "package p; var _ = fmt.\ntype X int", // not at EOF + } { + fset := token.NewFileSet() + f, err := ParseFile(fset, "", src, 0) + if err == nil { + t.Errorf("ParseFile(%s) succeeded unexpectedly", src) + continue + } + + const wantErr = "expected selector or type assertion" + if !strings.Contains(err.Error(), wantErr) { + t.Errorf("ParseFile returned wrong error %q, want %q", err, wantErr) + } + + var sel *ast.SelectorExpr + ast.Inspect(f, func(n ast.Node) bool { + if n, ok := n.(*ast.SelectorExpr); ok { + sel = n + } + return true + }) + if sel == nil { + t.Error("found no *ast.SelectorExpr") + continue + } + const wantSel = "&{fmt _}" + if fmt.Sprint(sel) != wantSel { + t.Errorf("found selector %s, want %s", sel, wantSel) + continue + } + } +} + +func TestLastLineComment(t *testing.T) { + const src = `package main +type x int // comment +` + fset := token.NewFileSet() + f, err := ParseFile(fset, "", src, ParseComments) + if err != nil { + t.Fatal(err) + } + comment := f.Decls[0].(*ast.GenDecl).Specs[0].(*ast.TypeSpec).Comment.List[0].Text + if comment != "// comment" { + t.Errorf("got %q, want %q", comment, "// comment") + } +} diff --git a/src/go/parser/performance_test.go b/src/go/parser/performance_test.go new file mode 100644 index 0000000..f81bcee --- /dev/null +++ b/src/go/parser/performance_test.go @@ -0,0 +1,30 @@ +// Copyright 2012 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 + +import ( + "go/token" + "os" + "testing" +) + +var src = readFile("parser.go") + +func readFile(filename string) []byte { + data, err := os.ReadFile(filename) + if err != nil { + panic(err) + } + return data +} + +func BenchmarkParse(b *testing.B) { + b.SetBytes(int64(len(src))) + for i := 0; i < b.N; i++ { + if _, err := ParseFile(token.NewFileSet(), "", src, ParseComments); err != nil { + b.Fatalf("benchmark failed due to parse error: %s", err) + } + } +} diff --git a/src/go/parser/short_test.go b/src/go/parser/short_test.go new file mode 100644 index 0000000..49bb681 --- /dev/null +++ b/src/go/parser/short_test.go @@ -0,0 +1,136 @@ +// 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. + +// This file contains test cases for short valid and invalid programs. + +package parser + +import "testing" + +var valids = []string{ + "package p\n", + `package p;`, + `package p; import "fmt"; func f() { fmt.Println("Hello, World!") };`, + `package p; func f() { if f(T{}) {} };`, + `package p; func f() { _ = <-chan int(nil) };`, + `package p; func f() { _ = (<-chan int)(nil) };`, + `package p; func f() { _ = (<-chan <-chan int)(nil) };`, + `package p; func f() { _ = <-chan <-chan <-chan <-chan <-int(nil) };`, + `package p; func f(func() func() func());`, + `package p; func f(...T);`, + `package p; func f(float, ...int);`, + `package p; func f(x int, a ...int) { f(0, a...); f(1, a...,) };`, + `package p; func f(int,) {};`, + `package p; func f(...int,) {};`, + `package p; func f(x ...int,) {};`, + `package p; type T []int; var a []bool; func f() { if a[T{42}[0]] {} };`, + `package p; type T []int; func g(int) bool { return true }; func f() { if g(T{42}[0]) {} };`, + `package p; type T []int; func f() { for _ = range []int{T{42}[0]} {} };`, + `package p; var a = T{{1, 2}, {3, 4}}`, + `package p; func f() { select { case <- c: case c <- d: case c <- <- d: case <-c <- d: } };`, + `package p; func f() { select { case x := (<-c): } };`, + `package p; func f() { if ; true {} };`, + `package p; func f() { switch ; {} };`, + `package p; func f() { for _ = range "foo" + "bar" {} };`, + `package p; func f() { var s []int; g(s[:], s[i:], s[:j], s[i:j], s[i:j:k], s[:j:k]) };`, + `package p; var ( _ = (struct {*T}).m; _ = (interface {T}).m )`, + `package p; func ((T),) m() {}`, + `package p; func ((*T),) m() {}`, + `package p; func (*(T),) m() {}`, + `package p; func _(x []int) { for range x {} }`, + `package p; func _() { if [T{}.n]int{} {} }`, + `package p; func _() { map[int]int{}[0]++; map[int]int{}[0] += 1 }`, + `package p; func _(x interface{f()}) { interface{f()}(x).f() }`, + `package p; func _(x chan int) { chan int(x) <- 0 }`, + `package p; const (x = 0; y; z)`, // issue 9639 + `package p; var _ = map[P]int{P{}:0, {}:1}`, + `package p; var _ = map[*P]int{&P{}:0, {}:1}`, + `package p; type T = int`, + `package p; type (T = p.T; _ = struct{}; x = *T)`, +} + +func TestValid(t *testing.T) { + for _, src := range valids { + checkErrors(t, src, src) + } +} + +var invalids = []string{ + `foo /* ERROR "expected 'package'" */ !`, + `package p; func f() { if { /* ERROR "missing condition" */ } };`, + `package p; func f() { if ; /* ERROR "missing condition" */ {} };`, + `package p; func f() { if f(); /* ERROR "missing condition" */ {} };`, + `package p; func f() { if _ = range /* ERROR "expected operand" */ x; true {} };`, + `package p; func f() { switch _ /* ERROR "expected switch expression" */ = range x; true {} };`, + `package p; func f() { for _ = range x ; /* ERROR "expected '{'" */ ; {} };`, + `package p; func f() { for ; ; _ = range /* ERROR "expected operand" */ x {} };`, + `package p; func f() { for ; _ /* ERROR "expected boolean or range expression" */ = range x ; {} };`, + `package p; func f() { switch t = /* ERROR "expected ':=', found '='" */ t.(type) {} };`, + `package p; func f() { switch t /* ERROR "expected switch expression" */ , t = t.(type) {} };`, + `package p; func f() { switch t /* ERROR "expected switch expression" */ = t.(type), t {} };`, + `package p; var a = [ /* ERROR "expected expression" */ 1]int;`, + `package p; var a = [ /* ERROR "expected expression" */ ...]int;`, + `package p; var a = struct /* ERROR "expected expression" */ {}`, + `package p; var a = func /* ERROR "expected expression" */ ();`, + `package p; var a = interface /* ERROR "expected expression" */ {}`, + `package p; var a = [ /* ERROR "expected expression" */ ]int`, + `package p; var a = map /* ERROR "expected expression" */ [int]int`, + `package p; var a = chan /* ERROR "expected expression" */ int;`, + `package p; var a = []int{[ /* ERROR "expected expression" */ ]int};`, + `package p; var a = ( /* ERROR "expected expression" */ []int);`, + `package p; var a = a[[ /* ERROR "expected expression" */ ]int:[]int];`, + `package p; var a = <- /* ERROR "expected expression" */ chan int;`, + `package p; func f() { select { case _ <- chan /* ERROR "expected expression" */ int: } };`, + `package p; func f() { _ = (<-<- /* ERROR "expected 'chan'" */ chan int)(nil) };`, + `package p; func f() { _ = (<-chan<-chan<-chan<-chan<-chan<- /* ERROR "expected channel type" */ int)(nil) };`, + `package p; func f() { var t []int; t /* ERROR "expected identifier on left side of :=" */ [0] := 0 };`, + `package p; func f() { if x := g(); x /* ERROR "expected boolean expression" */ = 0 {}};`, + `package p; func f() { _ = x = /* ERROR "expected '=='" */ 0 {}};`, + `package p; func f() { _ = 1 == func()int { var x bool; x = x = /* ERROR "expected '=='" */ true; return x }() };`, + `package p; func f() { var s []int; _ = s[] /* ERROR "expected operand" */ };`, + `package p; func f() { var s []int; _ = s[i:j: /* ERROR "3rd index required" */ ] };`, + `package p; func f() { var s []int; _ = s[i: /* ERROR "2nd index required" */ :k] };`, + `package p; func f() { var s []int; _ = s[i: /* ERROR "2nd index required" */ :] };`, + `package p; func f() { var s []int; _ = s[: /* ERROR "2nd index required" */ :] };`, + `package p; func f() { var s []int; _ = s[: /* ERROR "2nd index required" */ ::] };`, + `package p; func f() { var s []int; _ = s[i:j:k: /* ERROR "expected ']'" */ l] };`, + `package p; func f() { for x /* ERROR "boolean or range expression" */ = []string {} }`, + `package p; func f() { for x /* ERROR "boolean or range expression" */ := []string {} }`, + `package p; func f() { for i /* ERROR "boolean or range expression" */ , x = []string {} }`, + `package p; func f() { for i /* ERROR "boolean or range expression" */ , x := []string {} }`, + `package p; func f() { go f /* ERROR HERE "function must be invoked" */ }`, + `package p; func f() { defer func() {} /* ERROR HERE "function must be invoked" */ }`, + `package p; func f() { go func() { func() { f(x func /* ERROR "missing ','" */ (){}) } } }`, + `package p; func f(x func(), u v func /* ERROR "missing ','" */ ()){}`, + + // issue 8656 + `package p; func f() (a b string /* ERROR "missing ','" */ , ok bool)`, + + // issue 9639 + `package p; var x /* ERROR "missing variable type or initialization" */ , y, z;`, + `package p; const x /* ERROR "missing constant value" */ ;`, + `package p; const x /* ERROR "missing constant value" */ int;`, + `package p; const (x = 0; y; z /* ERROR "missing constant value" */ int);`, + + // issue 12437 + `package p; var _ = struct { x int, /* ERROR "expected ';', found ','" */ }{};`, + `package p; var _ = struct { x int, /* ERROR "expected ';', found ','" */ y float }{};`, + + // issue 11611 + `package p; type _ struct { int, } /* ERROR "expected type, found '}'" */ ;`, + `package p; type _ struct { int, float } /* ERROR "expected type, found '}'" */ ;`, + `package p; type _ struct { ( /* ERROR "expected anonymous field" */ int) };`, + `package p; func _()(x, y, z ... /* ERROR "expected '\)', found '...'" */ int){}`, + `package p; func _()(... /* ERROR "expected type, found '...'" */ int){}`, + + // issue 13475 + `package p; func f() { if true {} else ; /* ERROR "expected if statement or block" */ }`, + `package p; func f() { if true {} else defer /* ERROR "expected if statement or block" */ f() }`, +} + +func TestInvalid(t *testing.T) { + for _, src := range invalids { + checkErrors(t, src, src) + } +} diff --git a/src/go/parser/testdata/commas.src b/src/go/parser/testdata/commas.src new file mode 100644 index 0000000..e0603cf --- /dev/null +++ b/src/go/parser/testdata/commas.src @@ -0,0 +1,19 @@ +// Copyright 2012 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. + +// Test case for error messages/parser synchronization +// after missing commas. + +package p + +var _ = []int{ + 0/* ERROR HERE "missing ','" */ +} + +var _ = []int{ + 0, + 1, + 2, + 3/* ERROR HERE "missing ','" */ +} diff --git a/src/go/parser/testdata/issue11377.src b/src/go/parser/testdata/issue11377.src new file mode 100644 index 0000000..1c43800 --- /dev/null +++ b/src/go/parser/testdata/issue11377.src @@ -0,0 +1,27 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Test case for issue 11377: Better synchronization of +// parser after certain syntax errors. + +package p + +func bad1() { + if f()) /* ERROR "expected ';', found '\)'" */ { + return + } +} + +// There shouldn't be any errors down below. + +func F1() {} +func F2() {} +func F3() {} +func F4() {} +func F5() {} +func F6() {} +func F7() {} +func F8() {} +func F9() {} +func F10() {} diff --git a/src/go/parser/testdata/issue23434.src b/src/go/parser/testdata/issue23434.src new file mode 100644 index 0000000..24a0832 --- /dev/null +++ b/src/go/parser/testdata/issue23434.src @@ -0,0 +1,25 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Test case for issue 23434: Better synchronization of +// parser after missing type. There should be exactly +// one error each time, with now follow errors. + +package p + +func g() { + m := make(map[string]! /* ERROR "expected type, found '!'" */ ) + for { + x := 1 + print(x) + } +} + +func f() { + m := make(map[string]) /* ERROR "expected type, found '\)'" */ + for { + x := 1 + print(x) + } +} diff --git a/src/go/parser/testdata/issue3106.src b/src/go/parser/testdata/issue3106.src new file mode 100644 index 0000000..2db10be --- /dev/null +++ b/src/go/parser/testdata/issue3106.src @@ -0,0 +1,46 @@ +// Copyright 2012 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. + +// Test case for issue 3106: Better synchronization of +// parser after certain syntax errors. + +package main + +func f() { + var m Mutex + c := MakeCond(&m) + percent := 0 + const step = 10 + for i := 0; i < 5; i++ { + go func() { + for { + // Emulates some useful work. + time.Sleep(1e8) + m.Lock() + defer + if /* ERROR "expected ';', found 'if'" */ percent == 100 { + m.Unlock() + break + } + percent++ + if percent % step == 0 { + //c.Signal() + } + m.Unlock() + } + }() + } + for { + m.Lock() + if percent == 0 || percent % step != 0 { + c.Wait() + } + fmt.Print(",") + if percent == 100 { + m.Unlock() + break + } + m.Unlock() + } +} diff --git a/src/go/parser/testdata/issue34946.src b/src/go/parser/testdata/issue34946.src new file mode 100644 index 0000000..6bb15e1 --- /dev/null +++ b/src/go/parser/testdata/issue34946.src @@ -0,0 +1,22 @@ +// Copyright 2019 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. + +// Test case for issue 34946: Better synchronization of +// parser for function declarations that start their +// body's opening { on a new line. + +package p + +// accept Allman/BSD-style declaration but complain +// (implicit semicolon between signature and body) +func _() int +{ /* ERROR "unexpected semicolon or newline before {" */ + { return 0 } +} + +func _() {} + +func _(); { /* ERROR "unexpected semicolon or newline before {" */ } + +func _() {} diff --git a/src/go/parser/testdata/issue42951/not_a_file.go/invalid.go b/src/go/parser/testdata/issue42951/not_a_file.go/invalid.go new file mode 100644 index 0000000..bb698be --- /dev/null +++ b/src/go/parser/testdata/issue42951/not_a_file.go/invalid.go @@ -0,0 +1 @@ +This file should not be parsed by ParseDir. |