all: make Unicode surrogate halves illegal as UTF-8

Surrogate halves are part of UTF-16 and should never appear in UTF-8.
(The rune that two combined halves represent in UTF-16 should
be encoded directly.)

Encoding: encode as RuneError.
Decoding: convert to RuneError, consume one byte.

This requires changing:
        package unicode/utf8
        runtime for range over string
Also added utf8.ValidRune and fixed bug in utf.RuneLen.

Fixes #3927.

R=golang-dev, rsc, bsiegert
CC=golang-dev
https://golang.org/cl/6458099
This commit is contained in:
Rob Pike 2012-08-08 14:01:23 -07:00
parent 4939b7b065
commit c48b77b1b5
4 changed files with 136 additions and 16 deletions

View file

@ -47,6 +47,9 @@ enum
Runeerror = 0xFFFD,
Runeself = 0x80,
SurrogateMin = 0xD800,
SurrogateMax = 0xDFFF,
Bad = Runeerror,
Runemax = 0x10FFFF, /* maximum rune value */
@ -128,6 +131,8 @@ runtime·charntorune(int32 *rune, uint8 *str, int32 length)
l = ((((c << Bitx) | c1) << Bitx) | c2) & Rune3;
if(l <= Rune2)
goto bad;
if (SurrogateMin <= l && l <= SurrogateMax)
goto bad;
*rune = l;
return 3;
}
@ -193,13 +198,15 @@ runtime·runetochar(byte *str, int32 rune) /* note: in original, arg2 was point
}
/*
* If the Rune is out of range, convert it to the error rune.
* If the Rune is out of range or a surrogate half, convert it to the error rune.
* Do this test here because the error rune encodes to three bytes.
* Doing it earlier would duplicate work, since an out of range
* Rune wouldn't have fit in one or two bytes.
*/
if (c > Runemax)
c = Runeerror;
if (SurrogateMin <= c && c <= SurrogateMax)
c = Runeerror;
/*
* three character sequence

View file

@ -18,6 +18,12 @@ const (
UTFMax = 4 // maximum number of bytes of a UTF-8 encoded Unicode character.
)
// Code points in the surrogate range are not valid for UTF-8.
const (
surrogateMin = 0xD800
surrogateMax = 0xDFFF
)
const (
t1 = 0x00 // 0000 0000
tx = 0x80 // 1000 0000
@ -34,7 +40,6 @@ const (
rune1Max = 1<<7 - 1
rune2Max = 1<<11 - 1
rune3Max = 1<<16 - 1
rune4Max = 1<<21 - 1
)
func decodeRuneInternal(p []byte) (r rune, size int, short bool) {
@ -87,6 +92,9 @@ func decodeRuneInternal(p []byte) (r rune, size int, short bool) {
if r <= rune2Max {
return RuneError, 1, false
}
if surrogateMin <= r && r <= surrogateMax {
return RuneError, 1, false
}
return r, 3, false
}
@ -162,6 +170,9 @@ func decodeRuneInStringInternal(s string) (r rune, size int, short bool) {
if r <= rune2Max {
return RuneError, 1, false
}
if surrogateMin <= r && r <= surrogateMax {
return RuneError, 1, false
}
return r, 3, false
}
@ -295,15 +306,20 @@ func DecodeLastRuneInString(s string) (r rune, size int) {
}
// RuneLen returns the number of bytes required to encode the rune.
// It returns -1 if the rune is not a valid value to encode in UTF-8.
func RuneLen(r rune) int {
switch {
case r < 0:
return -1
case r <= rune1Max:
return 1
case r <= rune2Max:
return 2
case surrogateMin <= r && r <= surrogateMax:
return -1
case r <= rune3Max:
return 3
case r <= rune4Max:
case r <= MaxRune:
return 4
}
return -1
@ -328,6 +344,10 @@ func EncodeRune(p []byte, r rune) int {
r = RuneError
}
if surrogateMin <= r && r <= surrogateMax {
r = RuneError
}
if uint32(r) <= rune3Max {
p[0] = t3 | byte(r>>12)
p[1] = tx | byte(r>>6)&maskx
@ -407,3 +427,17 @@ func ValidString(s string) bool {
}
return true
}
// ValidRune reports whether r can be legally encoded as UTF-8.
// Code points that are out of range or a surrogate half are illegal.
func ValidRune(r rune) bool {
switch {
case r < 0:
return false
case surrogateMin <= r && r <= surrogateMax:
return false
case r > MaxRune:
return false
}
return true
}

View file

@ -56,6 +56,8 @@ var utf8map = []Utf8Map{
{0x07ff, "\xdf\xbf"},
{0x0800, "\xe0\xa0\x80"},
{0x0801, "\xe0\xa0\x81"},
{0xd7ff, "\xed\x9f\xbf"}, // last code point before surrogate half.
{0xe000, "\xee\x80\x80"}, // first code point after surrogate half.
{0xfffe, "\xef\xbf\xbe"},
{0xffff, "\xef\xbf\xbf"},
{0x10000, "\xf0\x90\x80\x80"},
@ -65,6 +67,11 @@ var utf8map = []Utf8Map{
{0xFFFD, "\xef\xbf\xbd"},
}
var surrogateMap = []Utf8Map{
{0xd800, "\xed\xa0\x80"}, // surrogate min decodes to (RuneError, 1)
{0xdfff, "\xed bf bf"}, // surrogate max decodes to (RuneError, 1)
}
var testStrings = []string{
"",
"abcd",
@ -75,8 +82,7 @@ var testStrings = []string{
}
func TestFullRune(t *testing.T) {
for i := 0; i < len(utf8map); i++ {
m := utf8map[i]
for _, m := range utf8map {
b := []byte(m.str)
if !FullRune(b) {
t.Errorf("FullRune(%q) (%U) = false, want true", b, m.r)
@ -97,8 +103,7 @@ func TestFullRune(t *testing.T) {
}
func TestEncodeRune(t *testing.T) {
for i := 0; i < len(utf8map); i++ {
m := utf8map[i]
for _, m := range utf8map {
b := []byte(m.str)
var buf [10]byte
n := EncodeRune(buf[0:], m.r)
@ -110,8 +115,7 @@ func TestEncodeRune(t *testing.T) {
}
func TestDecodeRune(t *testing.T) {
for i := 0; i < len(utf8map); i++ {
m := utf8map[i]
for _, m := range utf8map {
b := []byte(m.str)
r, size := DecodeRune(b)
if r != m.r || size != len(b) {
@ -168,6 +172,21 @@ func TestDecodeRune(t *testing.T) {
}
}
func TestDecodeSurrogateRune(t *testing.T) {
for _, m := range surrogateMap {
b := []byte(m.str)
r, size := DecodeRune(b)
if r != RuneError || size != 1 {
t.Errorf("DecodeRune(%q) = %x, %d want %x, %d", b, r, size, RuneError, 1)
}
s := m.str
r, size = DecodeRuneInString(s)
if r != RuneError || size != 1 {
t.Errorf("DecodeRune(%q) = %x, %d want %x, %d", b, r, size, RuneError, 1)
}
}
}
// Check that DecodeRune and DecodeLastRune correspond to
// the equivalent range loop.
func TestSequencing(t *testing.T) {
@ -284,8 +303,7 @@ var runecounttests = []RuneCountTest{
}
func TestRuneCount(t *testing.T) {
for i := 0; i < len(runecounttests); i++ {
tt := runecounttests[i]
for _, tt := range runecounttests {
if out := RuneCountInString(tt.in); out != tt.out {
t.Errorf("RuneCountInString(%q) = %d, want %d", tt.in, out, tt.out)
}
@ -295,6 +313,32 @@ func TestRuneCount(t *testing.T) {
}
}
type RuneLenTest struct {
r rune
size int
}
var runelentests = []RuneLenTest{
{0, 1},
{'e', 1},
{'é', 2},
{'☺', 3},
{RuneError, 3},
{MaxRune, 4},
{0xD800, -1},
{0xDFFF, -1},
{MaxRune + 1, -1},
{-1, -1},
}
func TestRuneLen(t *testing.T) {
for _, tt := range runelentests {
if size := RuneLen(tt.r); size != tt.size {
t.Errorf("RuneLen(%#U) = %d, want %d", tt.r, size, tt.size)
}
}
}
type ValidTest struct {
in string
out bool
@ -314,17 +358,45 @@ var validTests = []ValidTest{
{string("\xF7\xBF\xBF\xBF"), true}, // U+1FFFFF
{string("\xFB\xBF\xBF\xBF\xBF"), false}, // 0x3FFFFFF; out of range
{string("\xc0\x80"), false}, // U+0000 encoded in two bytes: incorrect
// TODO {string("\xed\xa0\x80"), false }, // U+D800 high surrogate (sic)
// TODO {string("\xed\xbf\xbf"), false }, // U+DFFF low surrogate (sic)
{string("\xed\xa0\x80"), false}, // U+D800 high surrogate (sic)
{string("\xed\xbf\xbf"), false}, // U+DFFF low surrogate (sic)
}
func TestValid(t *testing.T) {
for i, tt := range validTests {
for _, tt := range validTests {
if Valid([]byte(tt.in)) != tt.out {
t.Errorf("%d. Valid(%q) = %v; want %v", i, tt.in, !tt.out, tt.out)
t.Errorf("Valid(%q) = %v; want %v", tt.in, !tt.out, tt.out)
}
if ValidString(tt.in) != tt.out {
t.Errorf("%d. ValidString(%q) = %v; want %v", i, tt.in, !tt.out, tt.out)
t.Errorf("ValidString(%q) = %v; want %v", tt.in, !tt.out, tt.out)
}
}
}
type ValidRuneTest struct {
r rune
ok bool
}
var validrunetests = []ValidRuneTest{
{0, true},
{'e', true},
{'é', true},
{'☺', true},
{RuneError, true},
{MaxRune, true},
{0xD7FF, true},
{0xD800, false},
{0xDFFF, false},
{0xE000, true},
{MaxRune + 1, false},
{-1, false},
}
func TestValidRune(t *testing.T) {
for _, tt := range validrunetests {
if ok := ValidRune(tt.r); ok != tt.ok {
t.Errorf("ValidRune(%#U) = %t, want %t", tt.r, ok, tt.ok)
}
}
}

View file

@ -57,6 +57,13 @@ func main() {
ok = false
}
for _, c := range "a\xed\xa0\x80a" {
if c != 'a' && c != utf8.RuneError {
fmt.Printf("surrogate UTF-8 does not error: %U\n", c)
ok = false
}
}
if !ok {
fmt.Println("BUG: stringrange")
os.Exit(1)