go/scanner: return literal as string instead of []byte

Removed many string conversions in dependent code.
Runs all tests. No change to gofmt output.

R=r
CC=golang-dev
https://golang.org/cl/4291070

src/cmd/cgo/gcc.go

@@ -1156,7 +1156,7 @@ func (c *typeConv) Opaque(n int64) ast.Expr {
 func (c *typeConv) intExpr(n int64) ast.Expr {
 	return &ast.BasicLit{
 		Kind:  token.INT,
-		Value: []byte(strconv.Itoa64(n)),
+		Value: strconv.Itoa64(n),
 	}
 }
 

src/cmd/godoc/spec.go

@@ -27,7 +27,7 @@ type ebnfParser struct {
 	prev    int         // offset of previous token
 	pos     token.Pos   // token position
 	tok     token.Token // one token look-ahead
-	lit     []byte      // token literal
+	lit     string      // token literal
 }
 
 
@@ -63,7 +63,7 @@ func (p *ebnfParser) errorExpected(pos token.Pos, msg string) {
 		// make the error message more specific
 		msg += ", found '" + p.tok.String() + "'"
 		if p.tok.IsLiteral() {
-			msg += " " + string(p.lit)
+			msg += " " + p.lit
 		}
 	}
 	p.Error(p.file.Position(pos), msg)
@@ -81,7 +81,7 @@ func (p *ebnfParser) expect(tok token.Token) token.Pos {
 
 
 func (p *ebnfParser) parseIdentifier(def bool) {
-	name := string(p.lit)
+	name := p.lit
 	p.expect(token.IDENT)
 	if def {
 		fmt.Fprintf(p.out, `<a id="%s">%s</a>`, name, name)

src/cmd/govet/govet.go

@@ -7,7 +7,6 @@
 package main
 
 import (
-	"bytes"
 	"flag"
 	"fmt"
 	"io"
@@ -257,7 +256,7 @@ func (f *File) checkPrintf(call *ast.CallExpr, name string, skip int) {
 		return
 	}
 	if lit.Kind == token.STRING {
-		if bytes.IndexByte(lit.Value, '%') < 0 {
+		if strings.Index(lit.Value, "%") < 0 {
 			if len(call.Args) > skip+1 {
 				f.Badf(call.Pos(), "no formatting directive in %s call", name)
 			}
@@ -284,11 +283,11 @@ func (f *File) checkPrintf(call *ast.CallExpr, name string, skip int) {
 // parsePrintfVerb returns the number of bytes and number of arguments
 // consumed by the Printf directive that begins s, including its percent sign
 // and verb.
-func parsePrintfVerb(s []byte) (nbytes, nargs int) {
+func parsePrintfVerb(s string) (nbytes, nargs int) {
 	// There's guaranteed a percent sign.
 	nbytes = 1
 	end := len(s)
-	// There may be flags
+	// There may be flags.
 FlagLoop:
 	for nbytes < end {
 		switch s[nbytes] {
@@ -308,7 +307,7 @@ FlagLoop:
 			}
 		}
 	}
-	// There may be a width
+	// There may be a width.
 	getNum()
 	// If there's a period, there may be a precision.
 	if nbytes < end && s[nbytes] == '.' {
@@ -316,7 +315,7 @@ FlagLoop:
 		getNum()
 	}
 	// Now a verb.
-	c, w := utf8.DecodeRune(s[nbytes:])
+	c, w := utf8.DecodeRuneInString(s[nbytes:])
 	nbytes += w
 	if c != '%' {
 		nargs++
@@ -325,8 +324,6 @@ FlagLoop:
 }
 
 
-var terminalNewline = []byte(`\n"`) // \n at end of interpreted string
-
 // checkPrint checks a call to an unformatted print routine such as Println.
 // The skip argument records how many arguments to ignore; that is,
 // call.Args[skip] is the first argument to be printed.
@@ -341,7 +338,7 @@ func (f *File) checkPrint(call *ast.CallExpr, name string, skip int) {
 	}
 	arg := args[skip]
 	if lit, ok := arg.(*ast.BasicLit); ok && lit.Kind == token.STRING {
-		if bytes.IndexByte(lit.Value, '%') >= 0 {
+		if strings.Index(lit.Value, "%") >= 0 {
 			f.Badf(call.Pos(), "possible formatting directive in %s call", name)
 		}
 	}
@@ -349,7 +346,7 @@ func (f *File) checkPrint(call *ast.CallExpr, name string, skip int) {
 		// The last item, if a string, should not have a newline.
 		arg = args[len(call.Args)-1]
 		if lit, ok := arg.(*ast.BasicLit); ok && lit.Kind == token.STRING {
-			if bytes.HasSuffix(lit.Value, terminalNewline) {
+			if strings.HasSuffix(lit.Value, `\n"`) {
 				f.Badf(call.Pos(), "%s call ends with newline", name)
 			}
 		}

src/pkg/ebnf/parser.go

@@ -18,7 +18,7 @@ type parser struct {
 	scanner scanner.Scanner
 	pos     token.Pos   // token position
 	tok     token.Token // one token look-ahead
-	lit     []byte      // token literal
+	lit     string      // token literal
 }
 
 
@@ -44,7 +44,7 @@ func (p *parser) errorExpected(pos token.Pos, msg string) {
 		// make the error message more specific
 		msg += ", found '" + p.tok.String() + "'"
 		if p.tok.IsLiteral() {
-			msg += " " + string(p.lit)
+			msg += " " + p.lit
 		}
 	}
 	p.error(pos, msg)
@@ -63,7 +63,7 @@ func (p *parser) expect(tok token.Token) token.Pos {
 
 func (p *parser) parseIdentifier() *Name {
 	pos := p.pos
-	name := string(p.lit)
+	name := p.lit
 	p.expect(token.IDENT)
 	return &Name{pos, name}
 }
@@ -73,7 +73,7 @@ func (p *parser) parseToken() *Token {
 	pos := p.pos
 	value := ""
 	if p.tok == token.STRING {
-		value, _ = strconv.Unquote(string(p.lit))
+		value, _ = strconv.Unquote(p.lit)
 		// Unquote may fail with an error, but only if the scanner found
 		// an illegal string in the first place. In this case the error
 		// has already been reported.

src/pkg/exp/datafmt/parser.go

@@ -22,7 +22,7 @@ type parser struct {
 	file    *token.File
 	pos     token.Pos   // token position
 	tok     token.Token // one token look-ahead
-	lit     []byte      // token literal
+	lit     string      // token literal
 
 	packs map[string]string // PackageName -> ImportPath
 	rules map[string]expr   // RuleName -> Expression
@@ -62,7 +62,7 @@ func (p *parser) errorExpected(pos token.Pos, msg string) {
 		// make the error message more specific
 		msg += ", found '" + p.tok.String() + "'"
 		if p.tok.IsLiteral() {
-			msg += " " + string(p.lit)
+			msg += " " + p.lit
 		}
 	}
 	p.error(pos, msg)
@@ -80,7 +80,7 @@ func (p *parser) expect(tok token.Token) token.Pos {
 
 
 func (p *parser) parseIdentifier() string {
-	name := string(p.lit)
+	name := p.lit
 	p.expect(token.IDENT)
 	return name
 }
@@ -130,7 +130,7 @@ func (p *parser) parseRuleName() (string, bool) {
 func (p *parser) parseString() string {
 	s := ""
 	if p.tok == token.STRING {
-		s, _ = strconv.Unquote(string(p.lit))
+		s, _ = strconv.Unquote(p.lit)
 		// Unquote may fail with an error, but only if the scanner found
 		// an illegal string in the first place. In this case the error
 		// has already been reported.
@@ -181,7 +181,7 @@ func (p *parser) parseField() expr {
 	var fname string
 	switch p.tok {
 	case token.ILLEGAL:
-		if string(p.lit) != "@" {
+		if p.lit != "@" {
 			return nil
 		}
 		fname = "@"

src/pkg/exp/ogle/cmd.go

@@ -205,7 +205,7 @@ func parseLoad(args []byte) (ident string, path string, err os.Error) {
 	sc, ev := newScanner(args)
 
 	var toks [4]token.Token
-	var lits [4][]byte
+	var lits [4]string
 	for i := range toks {
 		_, toks[i], lits[i] = sc.Scan()
 	}

src/pkg/go/ast/ast.go

@@ -66,7 +66,7 @@ type Decl interface {
 // A Comment node represents a single //-style or /*-style comment.
 type Comment struct {
 	Slash token.Pos // position of "/" starting the comment
-	Text  []byte    // comment text (excluding '\n' for //-style comments)
+	Text  string    // comment text (excluding '\n' for //-style comments)
 }
 
 
@@ -199,7 +199,7 @@ type (
 	BasicLit struct {
 		ValuePos token.Pos   // literal position
 		Kind     token.Token // token.INT, token.FLOAT, token.IMAG, token.CHAR, or token.STRING
-		Value    []byte      // literal string; e.g. 42, 0x7f, 3.14, 1e-9, 2.4i, 'a', '\x7f', "foo" or `\m\n\o`
+		Value    string      // literal string; e.g. 42, 0x7f, 3.14, 1e-9, 2.4i, 'a', '\x7f', "foo" or `\m\n\o`
 	}
 
 	// A FuncLit node represents a function literal.

src/pkg/go/ast/filter.go

@@ -304,7 +304,7 @@ const (
 // separator is an empty //-style comment that is interspersed between
 // different comment groups when they are concatenated into a single group
 //
-var separator = &Comment{noPos, []byte("//")}
+var separator = &Comment{noPos, "//"}
 
 
 // MergePackageFiles creates a file AST by merging the ASTs of the

src/pkg/go/doc/comment.go

@@ -286,7 +286,7 @@ func unindent(block [][]byte) {
 // nor to have trailing spaces at the end of lines.
 // The comment markers have already been removed.
 //
-// Turn each run of multiple \n into </p><p>
+// Turn each run of multiple \n into </p><p>.
 // Turn each run of indented lines into a <pre> block without indent.
 //
 // URLs in the comment text are converted into links; if the URL also appears

src/pkg/go/doc/doc.go

@@ -66,7 +66,7 @@ func (doc *docReader) addDoc(comments *ast.CommentGroup) {
 	n2 := len(comments.List)
 	list := make([]*ast.Comment, n1+1+n2) // + 1 for separator line
 	copy(list, doc.doc.List)
-	list[n1] = &ast.Comment{token.NoPos, []byte("//")} // separator line
+	list[n1] = &ast.Comment{token.NoPos, "//"} // separator line
 	copy(list[n1+1:], comments.List)
 	doc.doc = &ast.CommentGroup{list}
 }
@@ -105,7 +105,7 @@ func baseTypeName(typ ast.Expr) string {
 		// if the type is not exported, the effect to
 		// a client is as if there were no type name
 		if t.IsExported() {
-			return string(t.Name)
+			return t.Name
 		}
 	case *ast.StarExpr:
 		return baseTypeName(t.X)
@@ -300,9 +300,9 @@ func (doc *docReader) addFile(src *ast.File) {
 	// collect BUG(...) comments
 	for _, c := range src.Comments {
 		text := c.List[0].Text
-		if m := bug_markers.FindIndex(text); m != nil {
+		if m := bug_markers.FindStringIndex(text); m != nil {
 			// found a BUG comment; maybe empty
-			if btxt := text[m[1]:]; bug_content.Match(btxt) {
+			if btxt := text[m[1]:]; bug_content.MatchString(btxt) {
 				// non-empty BUG comment; collect comment without BUG prefix
 				list := copyCommentList(c.List)
 				list[0].Text = text[m[1]:]

src/pkg/go/parser/parser.go

@@ -47,9 +47,9 @@ type parser struct {
 	lineComment *ast.CommentGroup // last line comment
 
 	// Next token
-	pos  token.Pos   // token position
-	tok  token.Token // one token look-ahead
-	lit_ []byte      // token literal (slice into original source, don't hold on to it)
+	pos token.Pos   // token position
+	tok token.Token // one token look-ahead
+	lit string      // token literal
 
 	// Non-syntactic parser control
 	exprLev int // < 0: in control clause, >= 0: in expression
@@ -96,15 +96,6 @@ func (p *parser) init(fset *token.FileSet, filename string, src []byte, mode uin
 }
 
 
-func (p *parser) lit() []byte {
-	// make a copy of p.lit_ so that we don't hold on to
-	// a copy of the entire source indirectly in the AST
-	t := make([]byte, len(p.lit_))
-	copy(t, p.lit_)
-	return t
-}
-
-
 // ----------------------------------------------------------------------------
 // Scoping support
 
@@ -261,7 +252,7 @@ func (p *parser) next0() {
 		s := p.tok.String()
 		switch {
 		case p.tok.IsLiteral():
-			p.printTrace(s, string(p.lit_))
+			p.printTrace(s, p.lit)
 		case p.tok.IsOperator(), p.tok.IsKeyword():
 			p.printTrace("\"" + s + "\"")
 		default:
@@ -269,7 +260,7 @@ func (p *parser) next0() {
 		}
 	}
 
-	p.pos, p.tok, p.lit_ = p.scanner.Scan()
+	p.pos, p.tok, p.lit = p.scanner.Scan()
 }
 
 // Consume a comment and return it and the line on which it ends.
@@ -277,15 +268,16 @@ 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] == '*' {
-		for _, b := range p.lit_ {
-			if b == '\n' {
+	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{p.pos, p.lit()}
+	comment = &ast.Comment{p.pos, p.lit}
 	p.next0()
 
 	return
@@ -375,12 +367,12 @@ func (p *parser) errorExpected(pos token.Pos, msg string) {
 	if pos == p.pos {
 		// the error happened at the current position;
 		// make the error message more specific
-		if p.tok == token.SEMICOLON && p.lit_[0] == '\n' {
+		if p.tok == token.SEMICOLON && p.lit[0] == '\n' {
 			msg += ", found newline"
 		} else {
 			msg += ", found '" + p.tok.String() + "'"
 			if p.tok.IsLiteral() {
-				msg += " " + string(p.lit_)
+				msg += " " + p.lit
 			}
 		}
 	}
@@ -419,7 +411,7 @@ func (p *parser) parseIdent() *ast.Ident {
 	pos := p.pos
 	name := "_"
 	if p.tok == token.IDENT {
-		name = string(p.lit_)
+		name = p.lit
 		p.next()
 	} else {
 		p.expect(token.IDENT) // use expect() error handling
@@ -581,7 +573,7 @@ func (p *parser) parseFieldDecl(scope *ast.Scope) *ast.Field {
 	// optional tag
 	var tag *ast.BasicLit
 	if p.tok == token.STRING {
-		tag = &ast.BasicLit{p.pos, p.tok, p.lit()}
+		tag = &ast.BasicLit{p.pos, p.tok, p.lit}
 		p.next()
 	}
 
@@ -1024,7 +1016,7 @@ func (p *parser) parseOperand(lhs bool) ast.Expr {
 		return x
 
 	case token.INT, token.FLOAT, token.IMAG, token.CHAR, token.STRING:
-		x := &ast.BasicLit{p.pos, p.tok, p.lit()}
+		x := &ast.BasicLit{p.pos, p.tok, p.lit}
 		p.next()
 		return x
 
@@ -1978,7 +1970,7 @@ func parseImportSpec(p *parser, doc *ast.CommentGroup, _ int) ast.Spec {
 
 	var path *ast.BasicLit
 	if p.tok == token.STRING {
-		path = &ast.BasicLit{p.pos, p.tok, p.lit()}
+		path = &ast.BasicLit{p.pos, p.tok, p.lit}
 		p.next()
 	} else {
 		p.expect(token.STRING) // use expect() error handling

src/pkg/go/printer/nodes.go

@@ -363,7 +363,7 @@ func (p *printer) isOneLineFieldList(list []*ast.Field) bool {
 
 
 func (p *printer) setLineComment(text string) {
-	p.setComment(&ast.CommentGroup{[]*ast.Comment{&ast.Comment{token.NoPos, []byte(text)}}})
+	p.setComment(&ast.CommentGroup{[]*ast.Comment{&ast.Comment{token.NoPos, text}}})
 }
 
 
@@ -527,7 +527,7 @@ func walkBinary(e *ast.BinaryExpr) (has4, has5 bool, maxProblem int) {
 		}
 
 	case *ast.StarExpr:
-		if e.Op.String() == "/" {
+		if e.Op == token.QUO {
 			maxProblem = 5
 		}
 

src/pkg/go/scanner/scanner.go

@@ -538,14 +538,12 @@ func (S *Scanner) switch4(tok0, tok1 token.Token, ch2 int, tok2, tok3 token.Toke
 }
 
 
-var newline = []byte{'\n'}
-
-// Scan scans the next token and returns the token position pos,
-// the token tok, and the literal text lit corresponding to the
+// Scan scans the next token and returns the token position,
+// the token, and the literal string corresponding to the
 // token. The source end is indicated by token.EOF.
 //
 // If the returned token is token.SEMICOLON, the corresponding
-// literal value is ";" if the semicolon was present in the source,
+// literal string is ";" if the semicolon was present in the source,
 // and "\n" if the semicolon was inserted because of a newline or
 // at EOF.
 //
@@ -560,7 +558,7 @@ var newline = []byte{'\n'}
 // set with Init. Token positions are relative to that file
 // and thus relative to the file set.
 //
-func (S *Scanner) Scan() (token.Pos, token.Token, []byte) {
+func (S *Scanner) Scan() (token.Pos, token.Token, string) {
 scanAgain:
 	S.skipWhitespace()
 
@@ -586,7 +584,7 @@ scanAgain:
 		case -1:
 			if S.insertSemi {
 				S.insertSemi = false // EOF consumed
-				return S.file.Pos(offs), token.SEMICOLON, newline
+				return S.file.Pos(offs), token.SEMICOLON, "\n"
 			}
 			tok = token.EOF
 		case '\n':
@@ -594,7 +592,7 @@ scanAgain:
 			// set in the first place and exited early
 			// from S.skipWhitespace()
 			S.insertSemi = false // newline consumed
-			return S.file.Pos(offs), token.SEMICOLON, newline
+			return S.file.Pos(offs), token.SEMICOLON, "\n"
 		case '"':
 			insertSemi = true
 			tok = token.STRING
@@ -662,7 +660,7 @@ scanAgain:
 					S.offset = offs
 					S.rdOffset = offs + 1
 					S.insertSemi = false // newline consumed
-					return S.file.Pos(offs), token.SEMICOLON, newline
+					return S.file.Pos(offs), token.SEMICOLON, "\n"
 				}
 				S.scanComment()
 				if S.mode&ScanComments == 0 {
@@ -711,5 +709,9 @@ scanAgain:
 	if S.mode&InsertSemis != 0 {
 		S.insertSemi = insertSemi
 	}
-	return S.file.Pos(offs), tok, S.src[offs:S.offset]
+
+	// TODO(gri): The scanner API should change such that the literal string
+	//            is only valid if an actual literal was scanned. This will
+	//            permit a more efficient implementation.
+	return S.file.Pos(offs), tok, string(S.src[offs:S.offset])
 }

src/pkg/go/scanner/scanner_test.go

@@ -234,12 +234,11 @@ func TestScan(t *testing.T) {
 	index := 0
 	epos := token.Position{"", 0, 1, 1} // expected position
 	for {
-		pos, tok, litb := s.Scan()
+		pos, tok, lit := s.Scan()
 		e := elt{token.EOF, "", special}
 		if index < len(tokens) {
 			e = tokens[index]
 		}
-		lit := string(litb)
 		if tok == token.EOF {
 			lit = "<EOF>"
 			epos.Line = src_linecount
@@ -257,7 +256,7 @@ func TestScan(t *testing.T) {
 		}
 		epos.Offset += len(lit) + len(whitespace)
 		epos.Line += newlineCount(lit) + whitespace_linecount
-		if tok == token.COMMENT && litb[1] == '/' {
+		if tok == token.COMMENT && lit[1] == '/' {
 			// correct for unaccounted '/n' in //-style comment
 			epos.Offset++
 			epos.Line++
@@ -292,7 +291,7 @@ func checkSemi(t *testing.T, line string, mode uint) {
 			semiPos.Column++
 			pos, tok, lit = S.Scan()
 			if tok == token.SEMICOLON {
-				if string(lit) != semiLit {
+				if lit != semiLit {
 					t.Errorf(`bad literal for %q: got %q, expected %q`, line, lit, semiLit)
 				}
 				checkPos(t, line, pos, semiPos)
@@ -493,7 +492,7 @@ func TestLineComments(t *testing.T) {
 	for _, s := range segments {
 		p, _, lit := S.Scan()
 		pos := file.Position(p)
-		checkPos(t, string(lit), p, token.Position{s.filename, pos.Offset, s.line, pos.Column})
+		checkPos(t, lit, p, token.Position{s.filename, pos.Offset, s.line, pos.Column})
 	}
 
 	if S.ErrorCount != 0 {
@@ -547,10 +546,10 @@ func TestIllegalChars(t *testing.T) {
 	for offs, ch := range src {
 		pos, tok, lit := s.Scan()
 		if poffs := file.Offset(pos); poffs != offs {
-			t.Errorf("bad position for %s: got %d, expected %d", string(lit), poffs, offs)
+			t.Errorf("bad position for %s: got %d, expected %d", lit, poffs, offs)
 		}
-		if tok == token.ILLEGAL && string(lit) != string(ch) {
-			t.Errorf("bad token: got %s, expected %s", string(lit), string(ch))
+		if tok == token.ILLEGAL && lit != string(ch) {
+			t.Errorf("bad token: got %s, expected %s", lit, string(ch))
 		}
 	}
 

src/pkg/go/typechecker/typechecker_test.go

@@ -78,7 +78,7 @@ func expectedErrors(t *testing.T, pkg *ast.Package) (list scanner.ErrorList) {
 			case token.EOF:
 				break loop
 			case token.COMMENT:
-				s := errRx.FindSubmatch(lit)
+				s := errRx.FindStringSubmatch(lit)
 				if len(s) == 2 {
 					list = append(list, &scanner.Error{fset.Position(prev), string(s[1])})
 				}