tinygo-org/tinygo
1
0
Fork 0
mirror of https://github.com/tinygo-org/tinygo.git synced 2026-08-16 02:33:28 +00:00
Code Issues Packages Projects Releases 6 Wiki Activity

loader, unicode/utf8: fix hiBits overflow on 16-bit AVR targets

Browse Source 
Go 1.26 added SWAR optimizations to unicode/utf8 that use:

  const ptrSize = 4 << (^uintptr(0) >> 63)
  const hiBits = 0x8080808080808080 >> (64 - 8*ptrSize)

This formula only distinguishes 32-bit and 64-bit architectures.
On AVR (16-bit uintptr), ptrSize computes as 4, and hiBits becomes
0x80808080 (2155905152) which overflows the 16-bit uintptr type.

Fix by providing a patched unicode/utf8 overlay for Go 1.26+ that
uses a ptrSize formula handling all three sizes (16/32/64-bit):

  const ptrSize = 1 << (^uintptr(0)>>15&1 + ^uintptr(0)>>31&1 + ^uintptr(0)>>63&1)

This evaluates to 2 on AVR, 4 on 32-bit, and 8 on 64-bit. The
word() helper also gains a ptrSize==2 case for 16-bit loads.
This commit is contained in:
deadprogram
2026-04-10 13:17:58 +02:00
committed by Ron Evans
parent 98b3c27c76
commit 3c07a36e95
3 changed files with 1396 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
loader/goroot.go
+10
View File
@@ -273,6 +273,16 @@ func pathsToOverride(goMinor int, needsSyscallPackage bool) map[string]bool {
paths["internal/syscall/"] = true paths["internal/syscall/"] = true
paths["internal/syscall/unix/"] = false paths["internal/syscall/unix/"] = false
} }
if goMinor >= 26 {
// Go 1.26 added SWAR optimizations to unicode/utf8 that use
// constants assuming at least 32-bit uintptr. TinyGo supports
// 16-bit targets (AVR) where these constants overflow, so we
// provide a patched version.
paths["unicode/"] = true
paths["unicode/utf8/"] = false
}
return paths return paths
} }
src/unicode/utf8/utf8.go
+585
View File
@@ -0,0 +1,585 @@
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package utf8 implements functions and constants to support text encoded in
// UTF-8. It includes functions to translate between runes and UTF-8 byte sequences.
// See https://en.wikipedia.org/wiki/UTF-8
package utf8
// The conditions RuneError==unicode.ReplacementChar and
// MaxRune==unicode.MaxRune are verified in the tests.
// Defining them locally avoids this package depending on package unicode.
// Numbers fundamental to the encoding.
const (
RuneError = '\uFFFD' // the "error" Rune or "Unicode replacement character"
RuneSelf = 0x80 // characters below RuneSelf are represented as themselves in a single byte.
MaxRune = '\U0010FFFF' // Maximum valid Unicode code point.
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 = 0b00000000
tx = 0b10000000
t2 = 0b11000000
t3 = 0b11100000
t4 = 0b11110000
t5 = 0b11111000
maskx = 0b00111111
mask2 = 0b00011111
mask3 = 0b00001111
mask4 = 0b00000111
rune1Max = 1<<7 - 1
rune2Max = 1<<11 - 1
rune3Max = 1<<16 - 1
// The default lowest and highest continuation byte.
locb = 0b10000000
hicb = 0b10111111
// These names of these constants are chosen to give nice alignment in the
// table below. The first nibble is an index into acceptRanges or F for
// special one-byte cases. The second nibble is the Rune length or the
// Status for the special one-byte case.
xx = 0xF1 // invalid: size 1
as = 0xF0 // ASCII: size 1
s1 = 0x02 // accept 0, size 2
s2 = 0x13 // accept 1, size 3
s3 = 0x03 // accept 0, size 3
s4 = 0x23 // accept 2, size 3
s5 = 0x34 // accept 3, size 4
s6 = 0x04 // accept 0, size 4
s7 = 0x44 // accept 4, size 4
)
const (
runeErrorByte0 = t3 | (RuneError >> 12)
runeErrorByte1 = tx | (RuneError>>6)&maskx
runeErrorByte2 = tx | RuneError&maskx
)
// first is information about the first byte in a UTF-8 sequence.
var first = [256]uint8{
// 1 2 3 4 5 6 7 8 9 A B C D E F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x00-0x0F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x10-0x1F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x20-0x2F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x30-0x3F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x40-0x4F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x50-0x5F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x60-0x6F
as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x70-0x7F
// 1 2 3 4 5 6 7 8 9 A B C D E F
xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x80-0x8F
xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x90-0x9F
xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xA0-0xAF
xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xB0-0xBF
xx, xx, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xC0-0xCF
s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xD0-0xDF
s2, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s4, s3, s3, // 0xE0-0xEF
s5, s6, s6, s6, s7, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xF0-0xFF
}
// acceptRange gives the range of valid values for the second byte in a UTF-8
// sequence.
type acceptRange struct {
lo uint8 // lowest value for second byte.
hi uint8 // highest value for second byte.
}
// acceptRanges has size 16 to avoid bounds checks in the code that uses it.
var acceptRanges = [16]acceptRange{
0: {locb, hicb},
1: {0xA0, hicb},
2: {locb, 0x9F},
3: {0x90, hicb},
4: {locb, 0x8F},
}
// FullRune reports whether the bytes in p begin with a full UTF-8 encoding of a rune.
// An invalid encoding is considered a full Rune since it will convert as a width-1 error rune.
func FullRune(p []byte) bool {
n := len(p)
if n == 0 {
return false
}
x := first[p[0]]
if n >= int(x&7) {
return true // ASCII, invalid or valid.
}
// Must be short or invalid.
accept := acceptRanges[x>>4]
if n > 1 && (p[1] < accept.lo || accept.hi < p[1]) {
return true
} else if n > 2 && (p[2] < locb || hicb < p[2]) {
return true
}
return false
}
// FullRuneInString is like FullRune but its input is a string.
func FullRuneInString(s string) bool {
n := len(s)
if n == 0 {
return false
}
x := first[s[0]]
if n >= int(x&7) {
return true // ASCII, invalid, or valid.
}
// Must be short or invalid.
accept := acceptRanges[x>>4]
if n > 1 && (s[1] < accept.lo || accept.hi < s[1]) {
return true
} else if n > 2 && (s[2] < locb || hicb < s[2]) {
return true
}
return false
}
// DecodeRune unpacks the first UTF-8 encoding in p and returns the rune and
// its width in bytes. If p is empty it returns ([RuneError], 0). Otherwise, if
// the encoding is invalid, it returns (RuneError, 1). Both are impossible
// results for correct, non-empty UTF-8.
//
// An encoding is invalid if it is incorrect UTF-8, encodes a rune that is
// out of range, or is not the shortest possible UTF-8 encoding for the
// value. No other validation is performed.
func DecodeRune(p []byte) (r rune, size int) {
// Inlineable fast path for ASCII characters; see #48195.
// This implementation is weird but effective at rendering the
// function inlineable.
for _, b := range p {
if b < RuneSelf {
return rune(b), 1
}
break
}
r, size = decodeRuneSlow(p)
return
}
func decodeRuneSlow(p []byte) (r rune, size int) {
n := len(p)
if n < 1 {
return RuneError, 0
}
p0 := p[0]
x := first[p0]
if x >= as {
// The following code simulates an additional check for x == xx and
// handling the ASCII and invalid cases accordingly. This mask-and-or
// approach prevents an additional branch.
mask := rune(x) << 31 >> 31 // Create 0x0000 or 0xFFFF.
return rune(p[0])&^mask | RuneError&mask, 1
}
sz := int(x & 7)
accept := acceptRanges[x>>4]
if n < sz {
return RuneError, 1
}
b1 := p[1]
if b1 < accept.lo || accept.hi < b1 {
return RuneError, 1
}
if sz <= 2 { // <= instead of == to help the compiler eliminate some bounds checks
return rune(p0&mask2)<<6 | rune(b1&maskx), 2
}
b2 := p[2]
if b2 < locb || hicb < b2 {
return RuneError, 1
}
if sz <= 3 {
return rune(p0&mask3)<<12 | rune(b1&maskx)<<6 | rune(b2&maskx), 3
}
b3 := p[3]
if b3 < locb || hicb < b3 {
return RuneError, 1
}
return rune(p0&mask4)<<18 | rune(b1&maskx)<<12 | rune(b2&maskx)<<6 | rune(b3&maskx), 4
}
// DecodeRuneInString is like [DecodeRune] but its input is a string. If s is
// empty it returns ([RuneError], 0). Otherwise, if the encoding is invalid, it
// returns (RuneError, 1). Both are impossible results for correct, non-empty
// UTF-8.
//
// An encoding is invalid if it is incorrect UTF-8, encodes a rune that is
// out of range, or is not the shortest possible UTF-8 encoding for the
// value. No other validation is performed.
func DecodeRuneInString(s string) (r rune, size int) {
// Inlineable fast path for ASCII characters; see #48195.
// This implementation is a bit weird but effective at rendering the
// function inlineable.
if s != "" && s[0] < RuneSelf {
return rune(s[0]), 1
} else {
r, size = decodeRuneInStringSlow(s)
}
return
}
func decodeRuneInStringSlow(s string) (rune, int) {
n := len(s)
if n < 1 {
return RuneError, 0
}
s0 := s[0]
x := first[s0]
if x >= as {
// The following code simulates an additional check for x == xx and
// handling the ASCII and invalid cases accordingly. This mask-and-or
// approach prevents an additional branch.
mask := rune(x) << 31 >> 31 // Create 0x0000 or 0xFFFF.
return rune(s[0])&^mask | RuneError&mask, 1
}
sz := int(x & 7)
accept := acceptRanges[x>>4]
if n < sz {
return RuneError, 1
}
s1 := s[1]
if s1 < accept.lo || accept.hi < s1 {
return RuneError, 1
}
if sz <= 2 { // <= instead of == to help the compiler eliminate some bounds checks
return rune(s0&mask2)<<6 | rune(s1&maskx), 2
}
s2 := s[2]
if s2 < locb || hicb < s2 {
return RuneError, 1
}
if sz <= 3 {
return rune(s0&mask3)<<12 | rune(s1&maskx)<<6 | rune(s2&maskx), 3
}
s3 := s[3]
if s3 < locb || hicb < s3 {
return RuneError, 1
}
return rune(s0&mask4)<<18 | rune(s1&maskx)<<12 | rune(s2&maskx)<<6 | rune(s3&maskx), 4
}
// DecodeLastRune unpacks the last UTF-8 encoding in p and returns the rune and
// its width in bytes. If p is empty it returns ([RuneError], 0). Otherwise, if
// the encoding is invalid, it returns (RuneError, 1). Both are impossible
// results for correct, non-empty UTF-8.
//
// An encoding is invalid if it is incorrect UTF-8, encodes a rune that is
// out of range, or is not the shortest possible UTF-8 encoding for the
// value. No other validation is performed.
func DecodeLastRune(p []byte) (r rune, size int) {
end := len(p)
if end == 0 {
return RuneError, 0
}
start := end - 1
r = rune(p[start])
if r < RuneSelf {
return r, 1
}
// guard against O(n^2) behavior when traversing
// backwards through strings with long sequences of
// invalid UTF-8.
lim := max(end-UTFMax, 0)
for start--; start >= lim; start-- {
if RuneStart(p[start]) {
break
}
}
if start < 0 {
start = 0
}
r, size = DecodeRune(p[start:end])
if start+size != end {
return RuneError, 1
}
return r, size
}
// DecodeLastRuneInString is like [DecodeLastRune] but its input is a string. If
// s is empty it returns ([RuneError], 0). Otherwise, if the encoding is invalid,
// it returns (RuneError, 1). Both are impossible results for correct,
// non-empty UTF-8.
//
// An encoding is invalid if it is incorrect UTF-8, encodes a rune that is
// out of range, or is not the shortest possible UTF-8 encoding for the
// value. No other validation is performed.
func DecodeLastRuneInString(s string) (r rune, size int) {
end := len(s)
if end == 0 {
return RuneError, 0
}
start := end - 1
r = rune(s[start])
if r < RuneSelf {
return r, 1
}
// guard against O(n^2) behavior when traversing
// backwards through strings with long sequences of
// invalid UTF-8.
lim := max(end-UTFMax, 0)
for start--; start >= lim; start-- {
if RuneStart(s[start]) {
break
}
}
if start < 0 {
start = 0
}
r, size = DecodeRuneInString(s[start:end])
if start+size != end {
return RuneError, 1
}
return r, size
}
// RuneLen returns the number of bytes in the UTF-8 encoding of 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 <= MaxRune:
return 4
}
return -1
}
// EncodeRune writes into p (which must be large enough) the UTF-8 encoding of the rune.
// If the rune is out of range, it writes the encoding of [RuneError].
// It returns the number of bytes written.
func EncodeRune(p []byte, r rune) int {
// This function is inlineable for fast handling of ASCII.
if uint32(r) <= rune1Max {
p[0] = byte(r)
return 1
}
return encodeRuneNonASCII(p, r)
}
func encodeRuneNonASCII(p []byte, r rune) int {
// Negative values are erroneous. Making it unsigned addresses the problem.
switch i := uint32(r); {
case i <= rune2Max:
_ = p[1] // eliminate bounds checks
p[0] = t2 | byte(r>>6)
p[1] = tx | byte(r)&maskx
return 2
case i < surrogateMin, surrogateMax < i && i <= rune3Max:
_ = p[2] // eliminate bounds checks
p[0] = t3 | byte(r>>12)
p[1] = tx | byte(r>>6)&maskx
p[2] = tx | byte(r)&maskx
return 3
case i > rune3Max && i <= MaxRune:
_ = p[3] // eliminate bounds checks
p[0] = t4 | byte(r>>18)
p[1] = tx | byte(r>>12)&maskx
p[2] = tx | byte(r>>6)&maskx
p[3] = tx | byte(r)&maskx
return 4
default:
_ = p[2] // eliminate bounds checks
p[0] = runeErrorByte0
p[1] = runeErrorByte1
p[2] = runeErrorByte2
return 3
}
}
// AppendRune appends the UTF-8 encoding of r to the end of p and
// returns the extended buffer. If the rune is out of range,
// it appends the encoding of [RuneError].
func AppendRune(p []byte, r rune) []byte {
// This function is inlineable for fast handling of ASCII.
if uint32(r) <= rune1Max {
return append(p, byte(r))
}
return appendRuneNonASCII(p, r)
}
func appendRuneNonASCII(p []byte, r rune) []byte {
// Negative values are erroneous. Making it unsigned addresses the problem.
switch i := uint32(r); {
case i <= rune2Max:
return append(p, t2|byte(r>>6), tx|byte(r)&maskx)
case i < surrogateMin, surrogateMax < i && i <= rune3Max:
return append(p, t3|byte(r>>12), tx|byte(r>>6)&maskx, tx|byte(r)&maskx)
case i > rune3Max && i <= MaxRune:
return append(p, t4|byte(r>>18), tx|byte(r>>12)&maskx, tx|byte(r>>6)&maskx, tx|byte(r)&maskx)
default:
return append(p, runeErrorByte0, runeErrorByte1, runeErrorByte2)
}
}
// RuneCount returns the number of runes in p. Erroneous and short
// encodings are treated as single runes of width 1 byte.
func RuneCount(p []byte) int {
np := len(p)
var n int
for ; n < np; n++ {
if c := p[n]; c >= RuneSelf {
// non-ASCII slow path
return n + RuneCountInString(string(p[n:]))
}
}
return n
}
// RuneCountInString is like [RuneCount] but its input is a string.
func RuneCountInString(s string) (n int) {
for range s {
n++
}
return n
}
// RuneStart reports whether the byte could be the first byte of an encoded,
// possibly invalid rune. Second and subsequent bytes always have the top two
// bits set to 10.
func RuneStart(b byte) bool { return b&0xC0 != 0x80 }
// ptrSize is the size of a uintptr in bytes.
// The upstream Go formula (4 << (^uintptr(0) >> 63)) only distinguishes 32-bit
// and 64-bit architectures. TinyGo also supports 16-bit targets (e.g. AVR),
// where uintptr is 2 bytes, so we use a formula that handles all three sizes.
const ptrSize = 1 << (^uintptr(0)>>15&1 + ^uintptr(0)>>31&1 + ^uintptr(0)>>63&1)
const hiBits = 0x8080808080808080 >> (64 - 8*ptrSize)
func word[T string | []byte](s T) uintptr {
if ptrSize == 2 {
return uintptr(s[0]) | uintptr(s[1])<<8
}
if ptrSize == 4 {
return uintptr(s[0]) | uintptr(s[1])<<8 | uintptr(s[2])<<16 | uintptr(s[3])<<24
}
return uintptr(uint64(s[0]) | uint64(s[1])<<8 | uint64(s[2])<<16 | uint64(s[3])<<24 | uint64(s[4])<<32 | uint64(s[5])<<40 | uint64(s[6])<<48 | uint64(s[7])<<56)
}
// Valid reports whether p consists entirely of valid UTF-8-encoded runes.
func Valid(p []byte) bool {
// This optimization avoids the need to recompute the capacity
// when generating code for slicing p, bringing it to parity with
// ValidString, which was 20% faster on long ASCII strings.
p = p[:len(p):len(p)]
for len(p) > 0 {
p0 := p[0]
if p0 < RuneSelf {
p = p[1:]
// If there's one ASCII byte, there are probably more.
// Advance quickly through ASCII-only data.
// Note: using > instead of >= here is intentional. That avoids
// needing pointing-past-the-end fixup on the slice operations.
if len(p) > ptrSize && word(p)&hiBits == 0 {
p = p[ptrSize:]
if len(p) > 2*ptrSize && (word(p)|word(p[ptrSize:]))&hiBits == 0 {
p = p[2*ptrSize:]
for len(p) > 4*ptrSize && ((word(p)|word(p[ptrSize:]))|(word(p[2*ptrSize:])|word(p[3*ptrSize:])))&hiBits == 0 {
p = p[4*ptrSize:]
}
}
}
continue
}
x := first[p0]
size := int(x & 7)
accept := acceptRanges[x>>4]
switch size {
case 2:
if len(p) < 2 || p[1] < accept.lo || accept.hi < p[1] {
return false
}
p = p[2:]
case 3:
if len(p) < 3 || p[1] < accept.lo || accept.hi < p[1] || p[2] < locb || hicb < p[2] {
return false
}
p = p[3:]
case 4:
if len(p) < 4 || p[1] < accept.lo || accept.hi < p[1] || p[2] < locb || hicb < p[2] || p[3] < locb || hicb < p[3] {
return false
}
p = p[4:]
default:
return false // illegal starter byte
}
}
return true
}
// ValidString reports whether s consists entirely of valid UTF-8-encoded runes.
func ValidString(s string) bool {
for len(s) > 0 {
s0 := s[0]
if s0 < RuneSelf {
s = s[1:]
// If there's one ASCII byte, there are probably more.
// Advance quickly through ASCII-only data.
// Note: using > instead of >= here is intentional. That avoids
// needing pointing-past-the-end fixup on the slice operations.
if len(s) > ptrSize && word(s)&hiBits == 0 {
s = s[ptrSize:]
if len(s) > 2*ptrSize && (word(s)|word(s[ptrSize:]))&hiBits == 0 {
s = s[2*ptrSize:]
for len(s) > 4*ptrSize && ((word(s)|word(s[ptrSize:]))|(word(s[2*ptrSize:])|word(s[3*ptrSize:])))&hiBits == 0 {
s = s[4*ptrSize:]
}
}
}
continue
}
x := first[s0]
size := int(x & 7)
accept := acceptRanges[x>>4]
switch size {
case 2:
if len(s) < 2 || s[1] < accept.lo || accept.hi < s[1] {
return false
}
s = s[2:]
case 3:
if len(s) < 3 || s[1] < accept.lo || accept.hi < s[1] || s[2] < locb || hicb < s[2] {
return false
}
s = s[3:]
case 4:
if len(s) < 4 || s[1] < accept.lo || accept.hi < s[1] || s[2] < locb || hicb < s[2] || s[3] < locb || hicb < s[3] {
return false
}
s = s[4:]
default:
return false // illegal starter byte
}
}
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 0 <= r && r < surrogateMin:
return true
case surrogateMax < r && r <= MaxRune:
return true
}
return false
}
src/unicode/utf8/utf8_test.go
+801
View File
@@ -0,0 +1,801 @@
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package utf8_test
import (
"bytes"
"strings"
"testing"
"unicode"
. "unicode/utf8"
)
// Validate the constants redefined from unicode.
func TestConstants(t *testing.T) {
if MaxRune != unicode.MaxRune {
t.Errorf("utf8.MaxRune is wrong: %x should be %x", MaxRune, unicode.MaxRune)
}
if RuneError != unicode.ReplacementChar {
t.Errorf("utf8.RuneError is wrong: %x should be %x", RuneError, unicode.ReplacementChar)
}
}
type Utf8Map struct {
r rune
str string
}
var utf8map = []Utf8Map{
{0x0000, "\x00"},
{0x0001, "\x01"},
{0x007e, "\x7e"},
{0x007f, "\x7f"},
{0x0080, "\xc2\x80"},
{0x0081, "\xc2\x81"},
{0x00bf, "\xc2\xbf"},
{0x00c0, "\xc3\x80"},
{0x00c1, "\xc3\x81"},
{0x00c8, "\xc3\x88"},
{0x00d0, "\xc3\x90"},
{0x00e0, "\xc3\xa0"},
{0x00f0, "\xc3\xb0"},
{0x00f8, "\xc3\xb8"},
{0x00ff, "\xc3\xbf"},
{0x0100, "\xc4\x80"},
{0x07ff, "\xdf\xbf"},
{0x0400, "\xd0\x80"},
{0x0800, "\xe0\xa0\x80"},
{0x0801, "\xe0\xa0\x81"},
{0x1000, "\xe1\x80\x80"},
{0xd000, "\xed\x80\x80"},
{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"},
{0x10001, "\xf0\x90\x80\x81"},
{0x40000, "\xf1\x80\x80\x80"},
{0x10fffe, "\xf4\x8f\xbf\xbe"},
{0x10ffff, "\xf4\x8f\xbf\xbf"},
{0xFFFD, "\xef\xbf\xbd"},
}
var surrogateMap = []Utf8Map{
{0xd800, "\xed\xa0\x80"}, // surrogate min decodes to (RuneError, 1)
{0xdfff, "\xed\xbf\xbf"}, // surrogate max decodes to (RuneError, 1)
}
var testStrings = []string{
"",
"abcd",
"☺☻☹",
"日a本b語ç日ð本Ê語þ日¥本¼語i日©",
"日a本b語ç日ð本Ê語þ日¥本¼語i日©日a本b語ç日ð本Ê語þ日¥本¼語i日©日a本b語ç日ð本Ê語þ日¥本¼語i日©",
"\x80\x80\x80\x80",
}
func TestFullRune(t *testing.T) {
for _, m := range utf8map {
b := []byte(m.str)
if !FullRune(b) {
t.Errorf("FullRune(%q) (%U) = false, want true", b, m.r)
}
s := m.str
if !FullRuneInString(s) {
t.Errorf("FullRuneInString(%q) (%U) = false, want true", s, m.r)
}
b1 := b[0 : len(b)-1]
if FullRune(b1) {
t.Errorf("FullRune(%q) = true, want false", b1)
}
s1 := string(b1)
if FullRuneInString(s1) {
t.Errorf("FullRune(%q) = true, want false", s1)
}
}
for _, s := range []string{"\xc0", "\xc1"} {
b := []byte(s)
if !FullRune(b) {
t.Errorf("FullRune(%q) = false, want true", s)
}
if !FullRuneInString(s) {
t.Errorf("FullRuneInString(%q) = false, want true", s)
}
}
}
func TestEncodeRune(t *testing.T) {
for _, m := range utf8map {
b := []byte(m.str)
var buf [10]byte
n := EncodeRune(buf[0:], m.r)
b1 := buf[0:n]
if !bytes.Equal(b, b1) {
t.Errorf("EncodeRune(%#04x) = %q want %q", m.r, b1, b)
}
}
}
func TestAppendRune(t *testing.T) {
for _, m := range utf8map {
if buf := AppendRune(nil, m.r); string(buf) != m.str {
t.Errorf("AppendRune(nil, %#04x) = %s, want %s", m.r, buf, m.str)
}
if buf := AppendRune([]byte("init"), m.r); string(buf) != "init"+m.str {
t.Errorf("AppendRune(init, %#04x) = %s, want %s", m.r, buf, "init"+m.str)
}
}
}
func TestDecodeRune(t *testing.T) {
for _, m := range utf8map {
b := []byte(m.str)
r, size := DecodeRune(b)
if r != m.r || size != len(b) {
t.Errorf("DecodeRune(%q) = %#04x, %d want %#04x, %d", b, r, size, m.r, len(b))
}
s := m.str
r, size = DecodeRuneInString(s)
if r != m.r || size != len(b) {
t.Errorf("DecodeRuneInString(%q) = %#04x, %d want %#04x, %d", s, r, size, m.r, len(b))
}
// there's an extra byte that bytes left behind - make sure trailing byte works
r, size = DecodeRune(b[0:cap(b)])
if r != m.r || size != len(b) {
t.Errorf("DecodeRune(%q) = %#04x, %d want %#04x, %d", b, r, size, m.r, len(b))
}
s = m.str + "\x00"
r, size = DecodeRuneInString(s)
if r != m.r || size != len(b) {
t.Errorf("DecodeRuneInString(%q) = %#04x, %d want %#04x, %d", s, r, size, m.r, len(b))
}
// make sure missing bytes fail
wantsize := 1
if wantsize >= len(b) {
wantsize = 0
}
r, size = DecodeRune(b[0 : len(b)-1])
if r != RuneError || size != wantsize {
t.Errorf("DecodeRune(%q) = %#04x, %d want %#04x, %d", b[:len(b)-1], r, size, RuneError, wantsize)
}
s = m.str[0 : len(m.str)-1]
r, size = DecodeRuneInString(s)
if r != RuneError || size != wantsize {
t.Errorf("DecodeRuneInString(%q) = %#04x, %d want %#04x, %d", s, r, size, RuneError, wantsize)
}
// make sure bad sequences fail
if len(b) == 1 {
b[0] = 0x80
} else {
b[len(b)-1] = 0x7F
}
r, size = DecodeRune(b)
if r != RuneError || size != 1 {
t.Errorf("DecodeRune(%q) = %#04x, %d want %#04x, %d", b, r, size, RuneError, 1)
}
s = string(b)
r, size = DecodeRuneInString(s)
if r != RuneError || size != 1 {
t.Errorf("DecodeRuneInString(%q) = %#04x, %d want %#04x, %d", s, r, size, RuneError, 1)
}
}
}
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("DecodeRuneInString(%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) {
for _, ts := range testStrings {
for _, m := range utf8map {
for _, s := range []string{ts + m.str, m.str + ts, ts + m.str + ts} {
testSequence(t, s)
}
}
}
}
func runtimeRuneCount(s string) int {
return len([]rune(s)) // Replaced by gc with call to runtime.countrunes(s).
}
// Check that a range loop, len([]rune(string)) optimization and
// []rune conversions visit the same runes.
// Not really a test of this package, but the assumption is used here and
// it's good to verify.
func TestRuntimeConversion(t *testing.T) {
for _, ts := range testStrings {
count := RuneCountInString(ts)
if n := runtimeRuneCount(ts); n != count {
t.Errorf("%q: len([]rune()) counted %d runes; got %d from RuneCountInString", ts, n, count)
break
}
runes := []rune(ts)
if n := len(runes); n != count {
t.Errorf("%q: []rune() has length %d; got %d from RuneCountInString", ts, n, count)
break
}
i := 0
for _, r := range ts {
if r != runes[i] {
t.Errorf("%q[%d]: expected %c (%U); got %c (%U)", ts, i, runes[i], runes[i], r, r)
}
i++
}
}
}
var invalidSequenceTests = []string{
"\xed\xa0\x80\x80", // surrogate min
"\xed\xbf\xbf\x80", // surrogate max
// xx
"\x91\x80\x80\x80",
// s1
"\xC2\x7F\x80\x80",
"\xC2\xC0\x80\x80",
"\xDF\x7F\x80\x80",
"\xDF\xC0\x80\x80",
// s2
"\xE0\x9F\xBF\x80",
"\xE0\xA0\x7F\x80",
"\xE0\xBF\xC0\x80",
"\xE0\xC0\x80\x80",
// s3
"\xE1\x7F\xBF\x80",
"\xE1\x80\x7F\x80",
"\xE1\xBF\xC0\x80",
"\xE1\xC0\x80\x80",
//s4
"\xED\x7F\xBF\x80",
"\xED\x80\x7F\x80",
"\xED\x9F\xC0\x80",
"\xED\xA0\x80\x80",
// s5
"\xF0\x8F\xBF\xBF",
"\xF0\x90\x7F\xBF",
"\xF0\x90\x80\x7F",
"\xF0\xBF\xBF\xC0",
"\xF0\xBF\xC0\x80",
"\xF0\xC0\x80\x80",
// s6
"\xF1\x7F\xBF\xBF",
"\xF1\x80\x7F\xBF",
"\xF1\x80\x80\x7F",
"\xF1\xBF\xBF\xC0",
"\xF1\xBF\xC0\x80",
"\xF1\xC0\x80\x80",
// s7
"\xF4\x7F\xBF\xBF",
"\xF4\x80\x7F\xBF",
"\xF4\x80\x80\x7F",
"\xF4\x8F\xBF\xC0",
"\xF4\x8F\xC0\x80",
"\xF4\x90\x80\x80",
}
func runtimeDecodeRune(s string) rune {
for _, r := range s {
return r
}
return -1
}
func TestDecodeInvalidSequence(t *testing.T) {
for _, s := range invalidSequenceTests {
r1, _ := DecodeRune([]byte(s))
if want := RuneError; r1 != want {
t.Errorf("DecodeRune(%#x) = %#04x, want %#04x", s, r1, want)
return
}
r2, _ := DecodeRuneInString(s)
if want := RuneError; r2 != want {
t.Errorf("DecodeRuneInString(%q) = %#04x, want %#04x", s, r2, want)
return
}
if r1 != r2 {
t.Errorf("DecodeRune(%#x) = %#04x mismatch with DecodeRuneInString(%q) = %#04x", s, r1, s, r2)
return
}
r3 := runtimeDecodeRune(s)
if r2 != r3 {
t.Errorf("DecodeRuneInString(%q) = %#04x mismatch with runtime.decoderune(%q) = %#04x", s, r2, s, r3)
return
}
}
}
func testSequence(t *testing.T, s string) {
type info struct {
index int
r rune
}
index := make([]info, len(s))
b := []byte(s)
si := 0
j := 0
for i, r := range s {
if si != i {
t.Errorf("Sequence(%q) mismatched index %d, want %d", s, si, i)
return
}
index[j] = info{i, r}
j++
r1, size1 := DecodeRune(b[i:])
if r != r1 {
t.Errorf("DecodeRune(%q) = %#04x, want %#04x", s[i:], r1, r)
return
}
r2, size2 := DecodeRuneInString(s[i:])
if r != r2 {
t.Errorf("DecodeRuneInString(%q) = %#04x, want %#04x", s[i:], r2, r)
return
}
if size1 != size2 {
t.Errorf("DecodeRune/DecodeRuneInString(%q) size mismatch %d/%d", s[i:], size1, size2)
return
}
si += size1
}
j--
for si = len(s); si > 0; {
r1, size1 := DecodeLastRune(b[0:si])
r2, size2 := DecodeLastRuneInString(s[0:si])
if size1 != size2 {
t.Errorf("DecodeLastRune/DecodeLastRuneInString(%q, %d) size mismatch %d/%d", s, si, size1, size2)
return
}
if r1 != index[j].r {
t.Errorf("DecodeLastRune(%q, %d) = %#04x, want %#04x", s, si, r1, index[j].r)
return
}
if r2 != index[j].r {
t.Errorf("DecodeLastRuneInString(%q, %d) = %#04x, want %#04x", s, si, r2, index[j].r)
return
}
si -= size1
if si != index[j].index {
t.Errorf("DecodeLastRune(%q) index mismatch at %d, want %d", s, si, index[j].index)
return
}
j--
}
if si != 0 {
t.Errorf("DecodeLastRune(%q) finished at %d, not 0", s, si)
}
}
// Check that negative runes encode as U+FFFD.
func TestNegativeRune(t *testing.T) {
errorbuf := make([]byte, UTFMax)
errorbuf = errorbuf[0:EncodeRune(errorbuf, RuneError)]
buf := make([]byte, UTFMax)
buf = buf[0:EncodeRune(buf, -1)]
if !bytes.Equal(buf, errorbuf) {
t.Errorf("incorrect encoding [% x] for -1; expected [% x]", buf, errorbuf)
}
}
type RuneCountTest struct {
in string
out int
}
var runecounttests = []RuneCountTest{
{"abcd", 4},
{"☺☻☹", 3},
{"1,2,3,4", 7},
{"\xe2\x00", 2},
{"\xe2\x80", 2},
{"a\xe2\x80", 3},
}
func TestRuneCount(t *testing.T) {
for _, tt := range runecounttests {
if out := RuneCountInString(tt.in); out != tt.out {
t.Errorf("RuneCountInString(%q) = %d, want %d", tt.in, out, tt.out)
}
if out := RuneCount([]byte(tt.in)); out != tt.out {
t.Errorf("RuneCount(%q) = %d, want %d", tt.in, out, tt.out)
}
}
}
func TestRuneCountNonASCIIAllocation(t *testing.T) {
if n := testing.AllocsPerRun(10, func() {
s := []byte("日本語日本語日本語日")
_ = RuneCount(s)
}); n > 0 {
t.Errorf("unexpected RuneCount allocation, got %v, want 0", n)
}
}
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
}
var validTests = []ValidTest{
{"", true},
{"a", true},
{"abc", true},
{"Ж", true},
{"ЖЖ", true},
{"брэд-ЛГТМ", true},
{"☺☻☹", true},
{"aa\xe2", false},
{string([]byte{66, 250}), false},
{string([]byte{66, 250, 67}), false},
{"a\uFFFDb", true},
{string("\xF4\x8F\xBF\xBF"), true}, // U+10FFFF
{string("\xF4\x90\x80\x80"), false}, // U+10FFFF+1; out of range
{string("\xF7\xBF\xBF\xBF"), false}, // 0x1FFFFF; out of range
{string("\xFB\xBF\xBF\xBF\xBF"), false}, // 0x3FFFFFF; out of range
{string("\xc0\x80"), false}, // U+0000 encoded in two bytes: incorrect
{string("\xed\xa0\x80"), false}, // U+D800 high surrogate (sic)
{string("\xed\xbf\xbf"), false}, // U+DFFF low surrogate (sic)
}
func init() {
for i := range 100 {
validTests = append(validTests, ValidTest{in: strings.Repeat("a", i), out: true})
validTests = append(validTests, ValidTest{in: strings.Repeat("a", i) + "Ж", out: true})
validTests = append(validTests, ValidTest{in: strings.Repeat("a", i) + "\xe2", out: false})
validTests = append(validTests, ValidTest{in: strings.Repeat("a", i) + "Ж" + strings.Repeat("b", i), out: true})
validTests = append(validTests, ValidTest{in: strings.Repeat("a", i) + "\xe2" + strings.Repeat("b", i), out: false})
}
}
func TestValid(t *testing.T) {
for _, tt := range validTests {
if Valid([]byte(tt.in)) != tt.out {
t.Errorf("Valid(%q) = %v; want %v", tt.in, !tt.out, tt.out)
}
if ValidString(tt.in) != 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)
}
}
}
func BenchmarkRuneCountTenASCIIChars(b *testing.B) {
s := []byte("0123456789")
for i := 0; i < b.N; i++ {
RuneCount(s)
}
}
func BenchmarkRuneCountTenJapaneseChars(b *testing.B) {
s := []byte("日本語日本語日本語日")
for i := 0; i < b.N; i++ {
RuneCount(s)
}
}
func BenchmarkRuneCountInStringTenASCIIChars(b *testing.B) {
for i := 0; i < b.N; i++ {
RuneCountInString("0123456789")
}
}
func BenchmarkRuneCountInStringTenJapaneseChars(b *testing.B) {
for i := 0; i < b.N; i++ {
RuneCountInString("日本語日本語日本語日")
}
}
var ascii100000 = strings.Repeat("0123456789", 10000)
func BenchmarkValidTenASCIIChars(b *testing.B) {
s := []byte("0123456789")
for i := 0; i < b.N; i++ {
Valid(s)
}
}
func BenchmarkValid100KASCIIChars(b *testing.B) {
s := []byte(ascii100000)
for i := 0; i < b.N; i++ {
Valid(s)
}
}
func BenchmarkValidTenJapaneseChars(b *testing.B) {
s := []byte("日本語日本語日本語日")
for i := 0; i < b.N; i++ {
Valid(s)
}
}
func BenchmarkValidLongMostlyASCII(b *testing.B) {
longMostlyASCII := []byte(longStringMostlyASCII)
for i := 0; i < b.N; i++ {
Valid(longMostlyASCII)
}
}
func BenchmarkValidLongJapanese(b *testing.B) {
longJapanese := []byte(longStringJapanese)
for i := 0; i < b.N; i++ {
Valid(longJapanese)
}
}
func BenchmarkValidStringTenASCIIChars(b *testing.B) {
for i := 0; i < b.N; i++ {
ValidString("0123456789")
}
}
func BenchmarkValidString100KASCIIChars(b *testing.B) {
for i := 0; i < b.N; i++ {
ValidString(ascii100000)
}
}
func BenchmarkValidStringTenJapaneseChars(b *testing.B) {
for i := 0; i < b.N; i++ {
ValidString("日本語日本語日本語日")
}
}
func BenchmarkValidStringLongMostlyASCII(b *testing.B) {
for i := 0; i < b.N; i++ {
ValidString(longStringMostlyASCII)
}
}
func BenchmarkValidStringLongJapanese(b *testing.B) {
for i := 0; i < b.N; i++ {
ValidString(longStringJapanese)
}
}
var longStringMostlyASCII string // ~100KB, ~97% ASCII
var longStringJapanese string // ~100KB, non-ASCII
func init() {
const japanese = "日本語日本語日本語日"
var b strings.Builder
for i := 0; b.Len() < 100_000; i++ {
if i%100 == 0 {
b.WriteString(japanese)
} else {
b.WriteString("0123456789")
}
}
longStringMostlyASCII = b.String()
longStringJapanese = strings.Repeat(japanese, 100_000/len(japanese))
}
func BenchmarkEncodeASCIIRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, 'a') // 1 byte
}
}
func BenchmarkEncodeSpanishRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, 'Ñ') // 2 bytes
}
}
func BenchmarkEncodeJapaneseRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, '本') // 3 bytes
}
}
func BenchmarkEncodeMaxRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, MaxRune) // 4 bytes
}
}
func BenchmarkEncodeInvalidRuneMaxPlusOne(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, MaxRune+1) // 3 bytes: RuneError
}
}
func BenchmarkEncodeInvalidRuneSurrogate(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, 0xD800) // 3 bytes: RuneError
}
}
func BenchmarkEncodeInvalidRuneNegative(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
EncodeRune(buf, -1) // 3 bytes: RuneError
}
}
func BenchmarkAppendASCIIRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], 'a') // 1 byte
}
}
func BenchmarkAppendSpanishRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], 'Ñ') // 2 bytes
}
}
func BenchmarkAppendJapaneseRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], '本') // 3 bytes
}
}
func BenchmarkAppendMaxRune(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], MaxRune) // 4 bytes
}
}
func BenchmarkAppendInvalidRuneMaxPlusOne(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], MaxRune+1) // 3 bytes: RuneError
}
}
func BenchmarkAppendInvalidRuneSurrogate(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], 0xD800) // 3 bytes: RuneError
}
}
func BenchmarkAppendInvalidRuneNegative(b *testing.B) {
buf := make([]byte, UTFMax)
for i := 0; i < b.N; i++ {
AppendRune(buf[:0], -1) // 3 bytes: RuneError
}
}
func BenchmarkDecodeASCIIRune(b *testing.B) {
a := []byte{'a'}
for range b.N {
runeSink, sizeSink = DecodeRune(a)
}
}
func BenchmarkDecodeJapaneseRune(b *testing.B) {
nihon := []byte("本")
for range b.N {
runeSink, sizeSink = DecodeRune(nihon)
}
}
func BenchmarkDecodeASCIIRuneInString(b *testing.B) {
a := "a"
for range b.N {
runeSink, sizeSink = DecodeRuneInString(a)
}
}
func BenchmarkDecodeJapaneseRuneInString(b *testing.B) {
nihon := "本"
for range b.N {
runeSink, sizeSink = DecodeRuneInString(nihon)
}
}
var (
runeSink rune
sizeSink int
)
// boolSink is used to reference the return value of benchmarked
// functions to avoid dead code elimination.
var boolSink bool
func BenchmarkFullRune(b *testing.B) {
benchmarks := []struct {
name string
data []byte
}{
{"ASCII", []byte("a")},
{"Incomplete", []byte("\xf0\x90\x80")},
{"Japanese", []byte("本")},
}
for _, bm := range benchmarks {
b.Run(bm.name, func(b *testing.B) {
for i := 0; i < b.N; i++ {
boolSink = FullRune(bm.data)
}
})
}
}