[dev.inline] cmd/compile/internal/syntax: report byte offset rather then rune count for column value

This will only become user-visible if error messages show column information.
Per the discussion in #10324.

For #10324.

Change-Id: I5959c1655aba74bb1a22fdc261cd728ffcfa6912
Reviewed-on: https://go-review.googlesource.com/34244
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

4 files changed, 78 insertions, 69 deletions

diff --git a/src/cmd/compile/internal/syntax/pos.go b/src/cmd/compile/internal/syntax/pos.go
index 6601df9ec7..01a03ff30c 100644
--- a/src/cmd/compile/internal/syntax/pos.go
+++ b/src/cmd/compile/internal/syntax/pos.go
@@ -60,7 +60,8 @@ func (p Pos) String() string {
 // posString formats a (filename, line, col) tuple as a printable position.
 func posString(filename string, line, col uint) string {
 	s := filename + ":" + strconv.FormatUint(uint64(line), 10)
-	if col != 0 {
+	// col == colMax is interpreted as unknown column value
+	if col < colMax {
 		s += ":" + strconv.FormatUint(uint64(col), 10)
 	}
 	return s
@@ -147,8 +148,8 @@ func makeLico(line, col uint) lico {
 		line = lineMax
 	}
 	if col > colMax {
-		// cannot represent column, use 0 to indicate unknown column
-		col = 0
+		// cannot represent column, use max. column so we have some information
+		col = colMax
 	}
 	return lico(line<<colBits | col)
 }
diff --git a/src/cmd/compile/internal/syntax/pos_test.go b/src/cmd/compile/internal/syntax/pos_test.go
index ff29e7b0a3..c9ecd3de68 100644
--- a/src/cmd/compile/internal/syntax/pos_test.go
+++ b/src/cmd/compile/internal/syntax/pos_test.go
@@ -28,11 +28,11 @@ func TestPos(t *testing.T) {
 		relFilename string
 		relLine     uint
 	}{
-		{Pos{}, ":0", "", 0, 0, "", 0},
+		{Pos{}, ":0:0", "", 0, 0, "", 0},
 		{MakePos(nil, 2, 3), ":2:3", "", 2, 3, "", 2},
 		{MakePos(f0, 2, 3), ":2:3", "", 2, 3, "", 2},
 		{MakePos(f1, 1, 1), "f1:1:1", "f1", 1, 1, "f1", 1},
-		{MakePos(f2, 7, 10), "f2:16:10[:0]", "", 7, 10, "f2", 16},
+		{MakePos(f2, 7, 10), "f2:16:10[:0:0]", "", 7, 10, "f2", 16},
 		{MakePos(f3, 12, 7), "f3:101:7[f1:10:1]", "f1", 12, 7, "f3", 101},
 		{MakePos(f4, 25, 1), "f4:114:1[f3:99:1[f1:10:1]]", "f3", 25, 1, "f4", 114}, // doesn't occur in Go code
 	} {
@@ -68,15 +68,15 @@ func TestLico(t *testing.T) {
 		string    string
 		line, col uint
 	}{
-		{0, ":0", 0, 0},
-		{makeLico(0, 0), ":0", 0, 0},
+		{0, ":0:0", 0, 0},
+		{makeLico(0, 0), ":0:0", 0, 0},
 		{makeLico(0, 1), ":0:1", 0, 1},
-		{makeLico(1, 0), ":1", 1, 0},
+		{makeLico(1, 0), ":1:0", 1, 0},
 		{makeLico(1, 1), ":1:1", 1, 1},
 		{makeLico(2, 3), ":2:3", 2, 3},
 		{makeLico(lineMax, 1), fmt.Sprintf(":%d:1", lineMax), lineMax, 1},
 		{makeLico(lineMax+1, 1), fmt.Sprintf(":%d:1", lineMax), lineMax, 1}, // line too large, stick with max. line
-		{makeLico(1, colMax), fmt.Sprintf(":1:%d", colMax), 1, colMax},
+		{makeLico(1, colMax), ":1", 1, colMax},
 		{makeLico(1, colMax+1), ":1", 1, 0}, // column too large
 		{makeLico(lineMax+1, colMax+1), fmt.Sprintf(":%d", lineMax), lineMax, 0},
 	} {
diff --git a/src/cmd/compile/internal/syntax/scanner_test.go b/src/cmd/compile/internal/syntax/scanner_test.go
index 5532780399..c0943e7bce 100644
--- a/src/cmd/compile/internal/syntax/scanner_test.go
+++ b/src/cmd/compile/internal/syntax/scanner_test.go
@@ -263,71 +263,76 @@ func TestScanErrors(t *testing.T) {
 		// token.
 
 		// rune-level errors
-		{"fo\x00o", "invalid NUL character", 1, 3},
-		{"foo\n\ufeff bar", "invalid BOM in the middle of the file", 2, 1},
-		{"foo\n\n\xff    ", "invalid UTF-8 encoding", 3, 1},
+		{"fo\x00o", "invalid NUL character", 1, 2},
+		{"foo\n\ufeff bar", "invalid BOM in the middle of the file", 2, 0},
+		{"foo\n\n\xff    ", "invalid UTF-8 encoding", 3, 0},
 
 		// token-level errors
-		{"x + ~y", "bitwise complement operator is ^", 1, 5},
-		{"foo$bar = 0", "invalid character U+0024 '$'", 1, 4},
-		{"const x = 0xyz", "malformed hex constant", 1, 13},
-		{"0123456789", "malformed octal constant", 1, 11},
-		{"0123456789. /* foobar", "comment not terminated", 1, 13},   // valid float constant
-		{"0123456789e0 /*\nfoobar", "comment not terminated", 1, 14}, // valid float constant
-		{"var a, b = 08, 07\n", "malformed octal constant", 1, 14},
-		{"(x + 1.0e+x)", "malformed floating-point constant exponent", 1, 11},
+		{"\u00BD" /* ½ */, "invalid identifier character U+00BD '½'", 1, 0},
+		{"\U0001d736\U0001d737\U0001d738_½" /* 𝜶𝜷𝜸_½ */, "invalid identifier character U+00BD '½'", 1, 13 /* byte offset */},
+		{"\U0001d7d8" /* 𝟘 */, "identifier cannot begin with digit U+1D7D8 '𝟘'", 1, 0},
+		{"foo\U0001d7d8_½" /* foo𝟘_½ */, "invalid identifier character U+00BD '½'", 1, 8 /* byte offset */},
 
-		{`''`, "empty character literal or unescaped ' in character literal", 1, 2},
-		{"'\n", "newline in character literal", 1, 2},
-		{`'\`, "missing '", 1, 3},
-		{`'\'`, "missing '", 1, 4},
-		{`'\x`, "missing '", 1, 4},
-		{`'\x'`, "non-hex character in escape sequence: '", 1, 4},
-		{`'\y'`, "unknown escape sequence", 1, 3},
-		{`'\x0'`, "non-hex character in escape sequence: '", 1, 5},
-		{`'\00'`, "non-octal character in escape sequence: '", 1, 5},
-		{`'\377' /*`, "comment not terminated", 1, 8}, // valid octal escape
-		{`'\378`, "non-octal character in escape sequence: 8", 1, 5},
-		{`'\400'`, "octal escape value > 255: 256", 1, 6},
-		{`'xx`, "missing '", 1, 3},
+		{"x + ~y", "bitwise complement operator is ^", 1, 4},
+		{"foo$bar = 0", "invalid character U+0024 '$'", 1, 3},
+		{"const x = 0xyz", "malformed hex constant", 1, 12},
+		{"0123456789", "malformed octal constant", 1, 10},
+		{"0123456789. /* foobar", "comment not terminated", 1, 12},   // valid float constant
+		{"0123456789e0 /*\nfoobar", "comment not terminated", 1, 13}, // valid float constant
+		{"var a, b = 08, 07\n", "malformed octal constant", 1, 13},
+		{"(x + 1.0e+x)", "malformed floating-point constant exponent", 1, 10},
 
-		{"\"\n", "newline in string", 1, 2},
-		{`"`, "string not terminated", 1, 1},
-		{`"foo`, "string not terminated", 1, 1},
-		{"`", "string not terminated", 1, 1},
-		{"`foo", "string not terminated", 1, 1},
-		{"/*/", "comment not terminated", 1, 1},
-		{"/*\n\nfoo", "comment not terminated", 1, 1},
-		{"/*\n\nfoo", "comment not terminated", 1, 1},
-		{`"\`, "string not terminated", 1, 1},
-		{`"\"`, "string not terminated", 1, 1},
-		{`"\x`, "string not terminated", 1, 1},
-		{`"\x"`, "non-hex character in escape sequence: \"", 1, 4},
-		{`"\y"`, "unknown escape sequence", 1, 3},
-		{`"\x0"`, "non-hex character in escape sequence: \"", 1, 5},
-		{`"\00"`, "non-octal character in escape sequence: \"", 1, 5},
-		{`"\377" /*`, "comment not terminated", 1, 8}, // valid octal escape
-		{`"\378"`, "non-octal character in escape sequence: 8", 1, 5},
-		{`"\400"`, "octal escape value > 255: 256", 1, 6},
+		{`''`, "empty character literal or unescaped ' in character literal", 1, 1},
+		{"'\n", "newline in character literal", 1, 1},
+		{`'\`, "missing '", 1, 2},
+		{`'\'`, "missing '", 1, 3},
+		{`'\x`, "missing '", 1, 3},
+		{`'\x'`, "non-hex character in escape sequence: '", 1, 3},
+		{`'\y'`, "unknown escape sequence", 1, 2},
+		{`'\x0'`, "non-hex character in escape sequence: '", 1, 4},
+		{`'\00'`, "non-octal character in escape sequence: '", 1, 4},
+		{`'\377' /*`, "comment not terminated", 1, 7}, // valid octal escape
+		{`'\378`, "non-octal character in escape sequence: 8", 1, 4},
+		{`'\400'`, "octal escape value > 255: 256", 1, 5},
+		{`'xx`, "missing '", 1, 2},
 
-		{`s := "foo\z"`, "unknown escape sequence", 1, 11},
-		{`s := "foo\z00\nbar"`, "unknown escape sequence", 1, 11},
-		{`"\x`, "string not terminated", 1, 1},
-		{`"\x"`, "non-hex character in escape sequence: \"", 1, 4},
-		{`var s string = "\x"`, "non-hex character in escape sequence: \"", 1, 19},
-		{`return "\Uffffffff"`, "escape sequence is invalid Unicode code point", 1, 19},
+		{"\"\n", "newline in string", 1, 1},
+		{`"`, "string not terminated", 1, 0},
+		{`"foo`, "string not terminated", 1, 0},
+		{"`", "string not terminated", 1, 0},
+		{"`foo", "string not terminated", 1, 0},
+		{"/*/", "comment not terminated", 1, 0},
+		{"/*\n\nfoo", "comment not terminated", 1, 0},
+		{"/*\n\nfoo", "comment not terminated", 1, 0},
+		{`"\`, "string not terminated", 1, 0},
+		{`"\"`, "string not terminated", 1, 0},
+		{`"\x`, "string not terminated", 1, 0},
+		{`"\x"`, "non-hex character in escape sequence: \"", 1, 3},
+		{`"\y"`, "unknown escape sequence", 1, 2},
+		{`"\x0"`, "non-hex character in escape sequence: \"", 1, 4},
+		{`"\00"`, "non-octal character in escape sequence: \"", 1, 4},
+		{`"\377" /*`, "comment not terminated", 1, 7}, // valid octal escape
+		{`"\378"`, "non-octal character in escape sequence: 8", 1, 4},
+		{`"\400"`, "octal escape value > 255: 256", 1, 5},
+
+		{`s := "foo\z"`, "unknown escape sequence", 1, 10},
+		{`s := "foo\z00\nbar"`, "unknown escape sequence", 1, 10},
+		{`"\x`, "string not terminated", 1, 0},
+		{`"\x"`, "non-hex character in escape sequence: \"", 1, 3},
+		{`var s string = "\x"`, "non-hex character in escape sequence: \"", 1, 18},
+		{`return "\Uffffffff"`, "escape sequence is invalid Unicode code point", 1, 18},
 
 		// TODO(gri) move these test cases into an appropriate parser test
-		// {`//line :`, "invalid line number: ", 1, 9},
-		// {`//line :x`, "invalid line number: x", 1, 9},
-		// {`//line foo :`, "invalid line number: ", 1, 13},
-		// {`//line foo:123abc`, "invalid line number: 123abc", 1, 12},
-		// {`/**///line foo:x`, "invalid line number: x", 1, 16},
-		// {`//line foo:0`, "invalid line number: 0", 1, 12},
-		// {fmt.Sprintf(`//line foo:%d`, lineMax+1), fmt.Sprintf("invalid line number: %d", lineMax+1), 1, 12},
+		// {`//line :`, "invalid line number: ", 1, 8},
+		// {`//line :x`, "invalid line number: x", 1, 8},
+		// {`//line foo :`, "invalid line number: ", 1, 12},
+		// {`//line foo:123abc`, "invalid line number: 123abc", 1, 11},
+		// {`/**///line foo:x`, "invalid line number: x", 1, 15},
+		// {`//line foo:0`, "invalid line number: 0", 1, 11},
+		// {fmt.Sprintf(`//line foo:%d`, lineMax+1), fmt.Sprintf("invalid line number: %d", lineMax+1), 1, 11},
 
 		// former problem cases
-		{"package p\n\n\xef", "invalid UTF-8 encoding", 3, 1},
+		{"package p\n\n\xef", "invalid UTF-8 encoding", 3, 0},
 	} {
 		var s scanner
 		nerrors := 0
diff --git a/src/cmd/compile/internal/syntax/source.go b/src/cmd/compile/internal/syntax/source.go
index 037742d73c..4ce35a3615 100644
--- a/src/cmd/compile/internal/syntax/source.go
+++ b/src/cmd/compile/internal/syntax/source.go
@@ -32,7 +32,7 @@ type source struct {
 	offs        int   // source offset of buf
 	r0, r, w    int   // previous/current read and write buf positions, excluding sentinel
 	line0, line uint  // previous/current line
-	col0, col   uint  // previous/current column
+	col0, col   uint  // previous/current column (byte offsets from line start)
 	ioerr       error // pending io error
 
 	// literal buffer
@@ -50,7 +50,7 @@ func (s *source) init(src io.Reader, errh func(line, pos uint, msg string)) {
 	s.offs = 0
 	s.r0, s.r, s.w = 0, 0, 0
 	s.line0, s.line = 1, 1
-	s.col0, s.col = 1, 1
+	s.col0, s.col = 0, 0
 	s.ioerr = nil
 
 	s.lit = s.lit[:0]
@@ -102,6 +102,9 @@ redo:
 	// (invariant: s.buf[s.w] == utf8.RuneSelf)
 	if b := s.buf[s.r]; b < utf8.RuneSelf {
 		s.r++
+		// TODO(gri) Optimization: Instead of adjusting s.col for each character,
+		// remember the line offset instead and then compute the offset as needed
+		// (which is less often).
 		s.col++
 		if b == 0 {
 			s.error("invalid NUL character")
@@ -109,7 +112,7 @@ redo:
 		}
 		if b == '\n' {
 			s.line++
-			s.col = 1
+			s.col = 0
 		}
 		return rune(b)
 	}
@@ -125,7 +128,7 @@ redo:
 	// uncommon case: not ASCII
 	r, w := utf8.DecodeRune(s.buf[s.r:s.w])
 	s.r += w
-	s.col++
+	s.col += uint(w)
 
 	if r == utf8.RuneError && w == 1 {
 		s.error("invalid UTF-8 encoding")