regexp: document and implement that invalid UTF-8 bytes are the same …

…as U+FFFD

What should it mean to run a regexp match on invalid UTF-8 bytes?
The coherent behavior options are:

1. Invalid UTF-8 does not match any character classes,
   nor a U+FFFD literal (nor \x{fffd}).
2. Each byte of invalid UTF-8 is treated identically to a U+FFFD in the input,
   as a utf8.DecodeRune loop might.

RE2 uses Rule 1.
Because it works byte at a time, it can also provide \C to match any
single byte of input, which matches invalid UTF-8 as well.
This provides the nice property that a match for a regexp without \C
is guaranteed to be valid UTF-8.

Unfortunately, today Go has an incoherent mix of these two, although
mostly Rule 2. This is a deviation from RE2, and it gives up the nice
property, but we probably can't correct that at this point.
In particular .* already matches entire inputs today, valid UTF-8 or
not, and I doubt we can break that.

This CL adopts Rule 2 officially, fixing the few places that deviate from it.

Fixes #48749.

Change-Id: I96402527c5dfb1146212f568ffa09dde91d71244
Reviewed-on: https://go-review.googlesource.com/c/go/+/354569
Trust: Russ Cox <[email protected]>
Run-TryBot: Russ Cox <[email protected]>
TryBot-Result: Go Bot <[email protected]>
Reviewed-by: Rob Pike <[email protected]>

src/regexp/all_test.go

@@ -372,6 +372,9 @@ var literalPrefixTests = []MetaTest{
 	{`^^0$$`, ``, ``, false},
 	{`^$^$`, ``, ``, false},
 	{`$$0^^`, ``, ``, false},
+	{`a\x{fffd}b`, ``, `a`, false},
+	{`\x{fffd}b`, ``, ``, false},
+	{"\ufffd", ``, ``, false},
 }
 
 func TestQuoteMeta(t *testing.T) {

src/regexp/find_test.go

@@ -116,6 +116,13 @@ var findTests = []FindTest{
 	{"\\`", "`", build(1, 0, 1)},
 	{"[\\`]+", "`", build(1, 0, 1)},
 
+	{"\ufffd", "\xff", build(1, 0, 1)},
+	{"\ufffd", "hello\xffworld", build(1, 5, 6)},
+	{`.*`, "hello\xffworld", build(1, 0, 11)},
+	{`\x{fffd}`, "\xc2\x00", build(1, 0, 1)},
+	{"[\ufffd]", "\xff", build(1, 0, 1)},
+	{`[\x{fffd}]`, "\xc2\x00", build(1, 0, 1)},
+
 	// long set of matches (longer than startSize)
 	{
 		".",

src/regexp/onepass.go

@@ -9,6 +9,7 @@ import (
 	"sort"
 	"strings"
 	"unicode"
+	"unicode/utf8"
 )
 
 // "One-pass" regexp execution.
@@ -55,7 +56,7 @@ func onePassPrefix(p *syntax.Prog) (prefix string, complete bool, pc uint32) {
 
 	// Have prefix; gather characters.
 	var buf strings.Builder
-	for iop(i) == syntax.InstRune && len(i.Rune) == 1 && syntax.Flags(i.Arg)&syntax.FoldCase == 0 {
+	for iop(i) == syntax.InstRune && len(i.Rune) == 1 && syntax.Flags(i.Arg)&syntax.FoldCase == 0 && i.Rune[0] != utf8.RuneError {
 		buf.WriteRune(i.Rune[0])
 		pc, i = i.Out, &p.Inst[i.Out]
 	}

src/regexp/regexp.go

@@ -20,6 +20,8 @@
 // or any book about automata theory.
 //
 // All characters are UTF-8-encoded code points.
+// Following utf8.DecodeRune, each byte of an invalid UTF-8 sequence
+// is treated as if it encoded utf8.RuneError (U+FFFD).
 //
 // There are 16 methods of Regexp that match a regular expression and identify
 // the matched text. Their names are matched by this regular expression:
@@ -276,7 +278,11 @@ func minInputLen(re *syntax.Regexp) int {
 	case syntax.OpLiteral:
 		l := 0
 		for _, r := range re.Rune {
-			l += utf8.RuneLen(r)
+			if r == utf8.RuneError {
+				l++
+			} else {
+				l += utf8.RuneLen(r)
+			}
 		}
 		return l
 	case syntax.OpCapture, syntax.OpPlus:

src/regexp/syntax/prog.go

@@ -8,6 +8,7 @@ import (
 	"strconv"
 	"strings"
 	"unicode"
+	"unicode/utf8"
 )
 
 // Compiled program.
@@ -154,7 +155,7 @@ func (p *Prog) Prefix() (prefix string, complete bool) {
 
 	// Have prefix; gather characters.
 	var buf strings.Builder
-	for i.op() == InstRune && len(i.Rune) == 1 && Flags(i.Arg)&FoldCase == 0 {
+	for i.op() == InstRune && len(i.Rune) == 1 && Flags(i.Arg)&FoldCase == 0 && i.Rune[0] != utf8.RuneError {
 		buf.WriteRune(i.Rune[0])
 		i = p.skipNop(i.Out)
 	}