- better comment classification

- related cleanups

R=r
OCL=19227
CL=19227

usr/gri/pretty/ast.go

@@ -258,14 +258,14 @@ export var BadDecl = NewDecl(0, Scanner.ILLEGAL, false);
 // Program
 
 export type Comment struct {
-	pos int;
+	pos, tok int;
 	text string;
 }
 
 
-export func NewComment(pos int, text string) *Comment {
+export func NewComment(pos, tok int, text string) *Comment {
 	c := new(Comment);
-	c.pos, c.text = pos, text;
+	c.pos, c.tok, c.text = pos, tok, text;
 	return c;
 }
 

usr/gri/pretty/parser.go

@@ -77,8 +77,15 @@ func (P *Parser) Next0() {
 
 
 func (P *Parser) Next() {
-	for P.Next0(); P.tok == Scanner.COMMENT; P.Next0() {
-		P.comments.Add(AST.NewComment(P.pos, P.val));
+	// TODO This is too expensive for every token - fix
+	for P.Next0();
+		P.tok == Scanner.COMMENT_WW ||
+		P.tok == Scanner.COMMENT_WB ||
+		P.tok == Scanner.COMMENT_BW ||
+		P.tok == Scanner.COMMENT_BB ;
+		P.Next0() 
+	{
+		P.comments.Add(AST.NewComment(P.pos, P.tok, P.val));
 	}
 }
 

usr/gri/pretty/printer.go

@@ -234,45 +234,36 @@ func (P *Printer) String(pos int, s string) {
 		assert(len(text) >= 3);  // classification char + "//" or "/*"
 		
 		// classify comment
-		switch text[0] {
-		case ' ':
-			// not only white space before comment on the same line
-			// - put into next cell if //-style comment
-			// - preceed with a space if /*-style comment
-			//print("[case a][", text[1 : len(text)], "]");
-			if text[2] == '/' {
-				P.buf.Tab();
-			} else {
-				P.buf.Print(" ");
-			}
+		switch comment.tok {
+		case Scanner.COMMENT_BB:
+			// black space before and after comment on the same line
+			// - print surrounded by blanks
+			P.buf.Print(" ");
+			P.buf.Print(text);
+			P.buf.Print(" ");
+
+		case Scanner.COMMENT_BW:
+			// only white space after comment on the same line
+			// - put into next cell
+			P.buf.Tab();
+			P.buf.Print(text);
 			
-			/*
-		case '\n':
-			// comment starts at beginning of line
-			// - reproduce exactly
-			//print("[case b][", text[1 : len(text)], "]");
-			if !P.buf.AtLineBegin() {
-				P.buf.Newline();
-			}
-			*/
-			
-		case '\n', '\t':
+		case Scanner.COMMENT_WW, Scanner.COMMENT_WB:
 			// only white space before comment on the same line
 			// - indent
-			//print("[case c][", text[1 : len(text)], "]");
 			if !P.buf.EmptyLine() {
 				P.buf.Newline();
 			}
 			for i := P.indent; i > 0; i-- {
 				P.buf.Tab();
 			}
+			P.buf.Print(text);
 
 		default:
 			panic("UNREACHABLE");
 		}
 		
-		P.buf.Print(text[1 : len(text)]);
-		if text[2] == '/' {
+		if text[1] == '/' {
 			// line comments must end in newline
 			// TODO should we set P.newl instead?
 			P.buf.Newline();

usr/gri/pretty/scanner.go

@@ -13,9 +13,13 @@ export const (
 	INT;
 	FLOAT;
 	STRING;
-	COMMENT;
 	EOF;
 
+	COMMENT_BB;
+	COMMENT_BW;
+	COMMENT_WB;
+	COMMENT_WW;
+
 	ADD;
 	SUB;
 	MUL;
@@ -116,9 +120,13 @@ export func TokenString(tok int) string {
 	case INT: return "INT";
 	case FLOAT: return "FLOAT";
 	case STRING: return "STRING";
-	case COMMENT: return "COMMENT";
 	case EOF: return "EOF";
 
+	case COMMENT_BB: return "COMMENT_BB";
+	case COMMENT_BW: return "COMMENT_BW";
+	case COMMENT_WB: return "COMMENT_WB";
+	case COMMENT_WW: return "COMMENT_WW";
+
 	case ADD: return "+";
 	case SUB: return "-";
 	case MUL: return "*";
@@ -518,29 +526,23 @@ func (S *Scanner) Expect(ch int) {
 }
 
 
-func (S *Scanner) SkipWhitespace() int {
-	pos := -1;  // no new line position yet
-	
-	if S.chpos == 0 {
-		// file beginning is always start of a new line
-		pos = 0;
-	}
-	
+// Returns true if a newline was seen, returns false otherwise.
+func (S *Scanner) SkipWhitespace() bool {
+	sawnl := S.chpos == 0;  // file beginning is always start of a new line
 	for {
 		switch S.ch {
 		case '\t', '\r', ' ':  // nothing to do
-		case '\n': pos = S.pos;  // remember start of new line
-		default: goto exit;
+		case '\n': sawnl = true;
+		default: return sawnl;
 		}
 		S.Next();
 	}
-
-exit:
-	return pos;
+	panic("UNREACHABLE");
+	return false;
 }
 
 
-func (S *Scanner) ScanComment(nlpos int) string {
+func (S *Scanner) ScanComment(leading_ws bool) (tok int, val string) {
 	// first '/' already consumed
 	pos := S.chpos - 1;
 	
@@ -575,6 +577,12 @@ func (S *Scanner) ScanComment(nlpos int) string {
 exit:
 	comment := S.src[pos : S.chpos];
 
+	// skip whitespace but stop at line end
+	for S.ch == '\t' || S.ch == '\r' || S.ch == ' ' {
+		S.Next();
+	}
+	trailing_ws := S.ch == '\n';
+
 	if S.testmode {
 		// interpret ERROR and SYNC comments
 		oldpos := -1;
@@ -595,18 +603,22 @@ exit:
 			S.ErrorMsg(oldpos, "ERROR not found");
 		}
 	}
-	
-	if nlpos < 0 {
-		// not only whitespace before comment on this line
-		comment = " " + comment;
-	} else if nlpos == pos {
-		// comment starts at the beginning of the line
-		comment = "\n" + comment;
+
+	if leading_ws {
+		if trailing_ws {
+			tok = COMMENT_WW;
+		} else {
+			tok = COMMENT_WB;
+		}
 	} else {
-		// only whitespace before comment on this line
-		comment = "\t" + comment;
+		if trailing_ws {
+			tok = COMMENT_BW;
+		} else {
+			tok = COMMENT_BB;
+		}
 	}
-	return comment;
+
+	return tok, comment;
 }
 
 
@@ -835,7 +847,7 @@ func (S *Scanner) Select4(tok0, tok1, ch2, tok2, tok3 int) int {
 
 
 func (S *Scanner) Scan() (pos, tok int, val string) {
-	nlpos := S.SkipWhitespace();
+	sawnl := S.SkipWhitespace();
 	
 	pos, tok = S.chpos, ILLEGAL;
 	
@@ -875,7 +887,7 @@ func (S *Scanner) Scan() (pos, tok int, val string) {
 		case '*': tok = S.Select2(MUL, MUL_ASSIGN);
 		case '/':
 			if S.ch == '/' || S.ch == '*' {
-				tok, val = COMMENT, S.ScanComment(nlpos);
+				tok, val = S.ScanComment(sawnl);
 			} else {
 				tok = S.Select2(QUO, QUO_ASSIGN);
 			}