Implement DFA Unicode Decoder
This is a separation of the DFA Unicode Decoder from https://chromium-review.googlesource.com/c/v8/v8/+/789560 I attempted to make the DFA's table a bit more explicit in this CL. Still, the linter prevents me from letting me present the array as a "table" in source code. For a better representation, please refer to https://docs.google.com/spreadsheets/d/1L9STtkmWs-A7HdK5ZmZ-wPZ_VBjQ3-Jj_xN9c6_hLKA - - - - - Now for a big copy-paste from 789560: Essentially, reworks a standard FSM (imagine an array of structs) and flattens it out into a single-dimension array. Using Table 3-7 of the Unicode 10.0.0 standard (page 126 of http://www.unicode.org/versions/Unicode10.0.0/ch03.pdf), we can nicely map all bytes into one of 12 character classes: 00. 0x00-0x7F 01. 0x80-0x8F (split from general continuation because this range is not valid after a 0xF0 leading byte) 02. 0x90-0x9F (split from general continuation because this range is not valid after a 0xE0 nor a 0xF4 leading byte) 03. 0xA0-0xBF (the rest of the continuation range) 04. 0xC0-0xC1, 0xF5-0xFF (the joined range of invalid bytes, notice this includes 255 which we use as a known bad byte during hex-to-int decoding) 05. 0xC2-0xDF (leading bytes which require any continuation byte afterwards) 06. 0xE0 (leading byte which requires a 0xA0-0xBF afterwards then any continuation byte after that) 07. 0xE1-0xEC, 0xEE-0xEF (leading bytes which requires any continuation afterwards then any continuation byte after that) 08. 0xED (leading byte which requires a 0x80-0x9F afterwards then any continuation byte after that) 09. 0xF1-F3 (leading bytes which requires any continuation byte afterwards then any continuation byte then any continuation byte) 10. 0xF0 (leading bytes which requires a 0x90-0xBF afterwards then any continuation byte then any continuation byte) 11. 0xF4 (leading bytes which requires a 0x80-0x8F afterwards then any continuation byte then any continuation byte) Note that 0xF0 and 0xF1-0xF3 were swapped so that fewer bytes were needed to represent the transition state ("9, 10, 10, 10" vs. "10, 9, 9, 9"). Using these 12 classes as "transitions", we can map from one state to the next. Each state is defined as some multiple of 12, so that we're always starting at the 0th column of each row of the FSM. From each state, we add the transition and get a index of the new row the FSM is entering. If at any point we encounter a bad byte, the state + bad-byte-transition is guaranteed to map us into the first row of the FSM (which contains no valid exiting transitions). The key differences from Björn's original (or his self-modified) DFA is the "bad" state is now mapped to 0 (or the first row of the FSM) instead of 12 (the second row). This saves ~50 bytes when gzipping, and also speeds up determining if a string is properly encoded (see his sample code at http://bjoern.hoehrmann.de/utf-8/decoder/dfa/#performance). Finally, I've replace his ternary check with an array access, to make the algorithm branchless. This places a requirement on the caller to 0 out the code point between successful decodings, which it could always have done because it's already branching. R=marja@google.com Bug: Change-Id: I574f208a84dc5d06caba17127b0d41f7ce1a3395 Reviewed-on: https://chromium-review.googlesource.com/805357 Commit-Queue: Justin Ridgewell <jridgewell@google.com> Reviewed-by: Marja Hölttä <marja@chromium.org> Reviewed-by: Mathias Bynens <mathias@chromium.org> Cr-Commit-Position: refs/heads/master@{#50012}
This commit is contained in:
Justin Ridgewell
Commit Bot
parent
a91d043dd4
commit
cedec225c9
1
BUILD.gn
1
BUILD.gn
@ -2052,6 +2052,7 @@ v8_source_set("v8_base") {
|
||||
"src/string-stream.h",
|
||||
"src/strtod.cc",
|
||||
"src/strtod.h",
|
||||
"src/third_party/utf8-decoder/utf8-decoder.h",
|
||||
"src/tracing/trace-event.cc",
|
||||
"src/tracing/trace-event.h",
|
||||
"src/tracing/traced-value.cc",
|
||||
|
||||
@ -203,7 +203,7 @@ class Utf8ExternalStreamingStream : public BufferedUtf16CharacterStream {
|
||||
Utf8ExternalStreamingStream(
|
||||
ScriptCompiler::ExternalSourceStream* source_stream,
|
||||
RuntimeCallStats* stats)
|
||||
: current_({0, {0, 0, unibrow::Utf8::Utf8IncrementalBuffer(0)}}),
|
||||
: current_({0, {0, 0, 0, unibrow::Utf8::State::kAccept}}),
|
||||
source_stream_(source_stream),
|
||||
stats_(stats) {}
|
||||
~Utf8ExternalStreamingStream() override {
|
||||
@ -223,7 +223,8 @@ class Utf8ExternalStreamingStream : public BufferedUtf16CharacterStream {
|
||||
struct StreamPosition {
|
||||
size_t bytes;
|
||||
size_t chars;
|
||||
unibrow::Utf8::Utf8IncrementalBuffer incomplete_char;
|
||||
uint32_t incomplete_char;
|
||||
unibrow::Utf8::State state;
|
||||
};
|
||||
|
||||
// Position contains a StreamPosition and the index of the chunk the position
|
||||
@ -268,25 +269,25 @@ bool Utf8ExternalStreamingStream::SkipToPosition(size_t position) {
|
||||
const Chunk& chunk = chunks_[current_.chunk_no];
|
||||
DCHECK(current_.pos.bytes >= chunk.start.bytes);
|
||||
|
||||
unibrow::Utf8::Utf8IncrementalBuffer incomplete_char =
|
||||
chunk.start.incomplete_char;
|
||||
unibrow::Utf8::State state = chunk.start.state;
|
||||
uint32_t incomplete_char = chunk.start.incomplete_char;
|
||||
size_t it = current_.pos.bytes - chunk.start.bytes;
|
||||
size_t chars = chunk.start.chars;
|
||||
while (it < chunk.length && chars < position) {
|
||||
unibrow::uchar t =
|
||||
unibrow::Utf8::ValueOfIncremental(chunk.data[it], &incomplete_char);
|
||||
unibrow::uchar t = unibrow::Utf8::ValueOfIncremental(
|
||||
chunk.data[it], &it, &state, &incomplete_char);
|
||||
if (t == kUtf8Bom && current_.pos.chars == 0) {
|
||||
// BOM detected at beginning of the stream. Don't copy it.
|
||||
} else if (t != unibrow::Utf8::kIncomplete) {
|
||||
chars++;
|
||||
if (t > unibrow::Utf16::kMaxNonSurrogateCharCode) chars++;
|
||||
}
|
||||
it++;
|
||||
}
|
||||
|
||||
current_.pos.bytes += it;
|
||||
current_.pos.chars = chars;
|
||||
current_.pos.incomplete_char = incomplete_char;
|
||||
current_.pos.state = state;
|
||||
current_.chunk_no += (it == chunk.length);
|
||||
|
||||
return current_.pos.chars == position;
|
||||
@ -304,31 +305,33 @@ void Utf8ExternalStreamingStream::FillBufferFromCurrentChunk() {
|
||||
uint16_t* cursor = buffer_ + (buffer_end_ - buffer_start_);
|
||||
DCHECK_EQ(cursor, buffer_end_);
|
||||
|
||||
unibrow::Utf8::State state = current_.pos.state;
|
||||
uint32_t incomplete_char = current_.pos.incomplete_char;
|
||||
|
||||
// If the current chunk is the last (empty) chunk we'll have to process
|
||||
// any left-over, partial characters.
|
||||
if (chunk.length == 0) {
|
||||
unibrow::uchar t =
|
||||
unibrow::Utf8::ValueOfIncrementalFinish(¤t_.pos.incomplete_char);
|
||||
unibrow::uchar t = unibrow::Utf8::ValueOfIncrementalFinish(&state);
|
||||
if (t != unibrow::Utf8::kBufferEmpty) {
|
||||
DCHECK_LT(t, unibrow::Utf16::kMaxNonSurrogateCharCode);
|
||||
DCHECK_EQ(t, unibrow::Utf8::kBadChar);
|
||||
*cursor = static_cast<uc16>(t);
|
||||
buffer_end_++;
|
||||
current_.pos.chars++;
|
||||
current_.pos.incomplete_char = 0;
|
||||
current_.pos.state = state;
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
unibrow::Utf8::Utf8IncrementalBuffer incomplete_char =
|
||||
current_.pos.incomplete_char;
|
||||
size_t it;
|
||||
for (it = current_.pos.bytes - chunk.start.bytes;
|
||||
it < chunk.length && cursor + 1 < buffer_start_ + kBufferSize; it++) {
|
||||
unibrow::uchar t =
|
||||
unibrow::Utf8::ValueOfIncremental(chunk.data[it], &incomplete_char);
|
||||
if (t == unibrow::Utf8::kIncomplete) continue;
|
||||
size_t it = current_.pos.bytes - chunk.start.bytes;
|
||||
while (it < chunk.length && cursor + 1 < buffer_start_ + kBufferSize) {
|
||||
unibrow::uchar t = unibrow::Utf8::ValueOfIncremental(
|
||||
chunk.data[it], &it, &state, &incomplete_char);
|
||||
if (V8_LIKELY(t < kUtf8Bom)) {
|
||||
*(cursor++) = static_cast<uc16>(t); // The by most frequent case.
|
||||
} else if (t == kUtf8Bom && current_.pos.bytes + it == 2) {
|
||||
} else if (t == unibrow::Utf8::kIncomplete) {
|
||||
continue;
|
||||
} else if (t == kUtf8Bom && current_.pos.bytes + it == 3) {
|
||||
// BOM detected at beginning of the stream. Don't copy it.
|
||||
} else if (t <= unibrow::Utf16::kMaxNonSurrogateCharCode) {
|
||||
*(cursor++) = static_cast<uc16>(t);
|
||||
@ -341,6 +344,7 @@ void Utf8ExternalStreamingStream::FillBufferFromCurrentChunk() {
|
||||
current_.pos.bytes = chunk.start.bytes + it;
|
||||
current_.pos.chars += (cursor - buffer_end_);
|
||||
current_.pos.incomplete_char = incomplete_char;
|
||||
current_.pos.state = state;
|
||||
current_.chunk_no += (it == chunk.length);
|
||||
|
||||
buffer_end_ = cursor;
|
||||
@ -396,16 +400,15 @@ void Utf8ExternalStreamingStream::SearchPosition(size_t position) {
|
||||
// checking whether the # bytes in a chunk are equal to the # chars, and if
|
||||
// so avoid the expensive SkipToPosition.)
|
||||
bool ascii_only_chunk =
|
||||
chunks_[chunk_no].start.incomplete_char ==
|
||||
unibrow::Utf8::Utf8IncrementalBuffer(0) &&
|
||||
chunks_[chunk_no].start.incomplete_char == 0 &&
|
||||
(chunks_[chunk_no + 1].start.bytes - chunks_[chunk_no].start.bytes) ==
|
||||
(chunks_[chunk_no + 1].start.chars - chunks_[chunk_no].start.chars);
|
||||
if (ascii_only_chunk) {
|
||||
size_t skip = position - chunks_[chunk_no].start.chars;
|
||||
current_ = {chunk_no,
|
||||
{chunks_[chunk_no].start.bytes + skip,
|
||||
chunks_[chunk_no].start.chars + skip,
|
||||
unibrow::Utf8::Utf8IncrementalBuffer(0)}};
|
||||
chunks_[chunk_no].start.chars + skip, 0,
|
||||
unibrow::Utf8::State::kAccept}};
|
||||
} else {
|
||||
current_ = {chunk_no, chunks_[chunk_no].start};
|
||||
SkipToPosition(position);
|
||||
|
||||
19
src/third_party/utf8-decoder/LICENSE
vendored
Normal file
19
src/third_party/utf8-decoder/LICENSE
vendored
Normal file
@ -0,0 +1,19 @@
|
||||
Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de>
|
||||
|
||||
Permission is hereby granted, free of charge, to any person obtaining a copy of
|
||||
this software and associated documentation files (the "Software"), to deal in
|
||||
the Software without restriction, including without limitation the rights to
|
||||
use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
|
||||
of the Software, and to permit persons to whom the Software is furnished to do
|
||||
so, subject to the following conditions:
|
||||
|
||||
The above copyright notice and this permission notice shall be included in all
|
||||
copies or substantial portions of the Software.
|
||||
|
||||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
SOFTWARE.
|
||||
18
src/third_party/utf8-decoder/README.v8
vendored
Normal file
18
src/third_party/utf8-decoder/README.v8
vendored
Normal file
@ -0,0 +1,18 @@
|
||||
Name: DFA UTF-8 Decoder
|
||||
Short Name: utf8-decoder
|
||||
URL: http://bjoern.hoehrmann.de/utf-8/decoder/dfa/
|
||||
Version: 0
|
||||
License: MIT
|
||||
License File: NOT_SHIPPED
|
||||
Security Critical: no
|
||||
|
||||
Description:
|
||||
Decodes UTF-8 bytes using a fast and simple definite finite automata.
|
||||
|
||||
Local modifications:
|
||||
- Rejection state has been mapped to row 0 (instead of row 1) of the DFA,
|
||||
saving some 50 bytes and making the table easier to reason about.
|
||||
- The transitions have been remapped to represent both a state transition and a
|
||||
bit mask for the incoming byte.
|
||||
- The caller must now zero out the code point buffer after successful or
|
||||
unsuccessful state transitions.
|
||||
78
src/third_party/utf8-decoder/utf8-decoder.h
vendored
Normal file
78
src/third_party/utf8-decoder/utf8-decoder.h
vendored
Normal file
@ -0,0 +1,78 @@
|
||||
// See http://bjoern.hoehrmann.de/utf-8/decoder/dfa/ for details.
|
||||
// The remapped transition table is justified at
|
||||
// https://docs.google.com/spreadsheets/d/1AZcQwuEL93HmNCljJWUwFMGqf7JAQ0puawZaUgP0E14
|
||||
|
||||
#include <stdint.h>
|
||||
|
||||
#ifndef __UTF8_DFA_DECODER_H
|
||||
#define __UTF8_DFA_DECODER_H
|
||||
|
||||
namespace Utf8DfaDecoder {
|
||||
|
||||
enum State : uint8_t {
|
||||
kReject = 0,
|
||||
kAccept = 12,
|
||||
kTwoByte = 24,
|
||||
kThreeByte = 36,
|
||||
kThreeByteLowMid = 48,
|
||||
kFourByte = 60,
|
||||
kFourByteLow = 72,
|
||||
kThreeByteHigh = 84,
|
||||
kFourByteMidHigh = 96,
|
||||
};
|
||||
|
||||
static inline void Decode(uint8_t byte, State* state, uint32_t* buffer) {
|
||||
// This first table maps bytes to character to a transition.
|
||||
static constexpr uint8_t transitions[] = {
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 00-0F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 10-1F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 20-2F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 30-3F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 40-4F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 50-5F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 60-6F
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 70-7F
|
||||
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 80-8F
|
||||
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // 90-9F
|
||||
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // A0-AF
|
||||
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // B0-BF
|
||||
9, 9, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, // C0-CF
|
||||
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, // D0-DF
|
||||
10, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 5, 5, // E0-EF
|
||||
11, 7, 7, 7, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, // F0-FF
|
||||
};
|
||||
|
||||
// This second table maps a state to a new state when adding a transition.
|
||||
// 00-7F
|
||||
// | 80-8F
|
||||
// | | 90-9F
|
||||
// | | | A0-BF
|
||||
// | | | | C2-DF
|
||||
// | | | | | E1-EC, EE, EF
|
||||
// | | | | | | ED
|
||||
// | | | | | | | F1-F3
|
||||
// | | | | | | | | F4
|
||||
// | | | | | | | | | C0, C1, F5-FF
|
||||
// | | | | | | | | | | E0
|
||||
// | | | | | | | | | | | F0
|
||||
static constexpr uint8_t states[] = {
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // REJECT = 0
|
||||
12, 0, 0, 0, 24, 36, 48, 60, 72, 0, 84, 96, // ACCEPT = 12
|
||||
0, 12, 12, 12, 0, 0, 0, 0, 0, 0, 0, 0, // 2-byte = 24
|
||||
0, 24, 24, 24, 0, 0, 0, 0, 0, 0, 0, 0, // 3-byte = 36
|
||||
0, 24, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 3-byte low/mid = 48
|
||||
0, 36, 36, 36, 0, 0, 0, 0, 0, 0, 0, 0, // 4-byte = 60
|
||||
0, 36, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 4-byte low = 72
|
||||
0, 0, 0, 24, 0, 0, 0, 0, 0, 0, 0, 0, // 3-byte high = 84
|
||||
0, 0, 36, 36, 0, 0, 0, 0, 0, 0, 0, 0, // 4-byte mid/high = 96
|
||||
};
|
||||
|
||||
DCHECK_NE(*state, State::kReject);
|
||||
uint8_t type = transitions[byte];
|
||||
*state = static_cast<State>(states[*state + type]);
|
||||
*buffer = (*buffer << 6) | (byte & (0x7F >> (type >> 1)));
|
||||
}
|
||||
|
||||
} // namespace Utf8DfaDecoder
|
||||
|
||||
#endif /* __UTF8_DFA_DECODER_H */
|
||||
@ -113,8 +113,8 @@ unsigned Utf8::Encode(char* str,
|
||||
uchar Utf8::ValueOf(const byte* bytes, size_t length, size_t* cursor) {
|
||||
if (length <= 0) return kBadChar;
|
||||
byte first = bytes[0];
|
||||
// Characters between 0000 and 0007F are encoded as a single character
|
||||
if (first <= kMaxOneByteChar) {
|
||||
// Characters between 0000 and 007F are encoded as a single character
|
||||
if (V8_LIKELY(first <= kMaxOneByteChar)) {
|
||||
*cursor += 1;
|
||||
return first;
|
||||
}
|
||||
|
||||
325
src/unicode.cc
325
src/unicode.cc
@ -193,306 +193,91 @@ static int LookupMapping(const int32_t* table,
|
||||
}
|
||||
}
|
||||
|
||||
static inline uint8_t NonASCIISequenceLength(byte first) {
|
||||
// clang-format off
|
||||
static const uint8_t lengths[256] = {
|
||||
// The first 128 entries correspond to ASCII characters.
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 00 - 0f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 10 - 1f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 20 - 2f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 30 - 3f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 40 - 4f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 50 - 5f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 60 - 6f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 70 - 7f */
|
||||
// The following 64 entries correspond to continuation bytes.
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80 - 8f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 90 - 9f */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* a0 - af */
|
||||
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* b0 - bf */
|
||||
// The next are two invalid overlong encodings and 30 two-byte sequences.
|
||||
0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, /* c0-c1 + c2-cf */
|
||||
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, /* d0-df */
|
||||
// 16 three-byte sequences.
|
||||
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, /* e0-ef */
|
||||
// 5 four-byte sequences, followed by sequences that could only encode
|
||||
// code points outside of the Unicode range.
|
||||
4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; /* f0-f4 + f5-ff */
|
||||
// clang-format on
|
||||
return lengths[first];
|
||||
}
|
||||
|
||||
|
||||
static inline bool IsContinuationCharacter(byte chr) {
|
||||
return chr >= 0x80 && chr <= 0xBF;
|
||||
}
|
||||
|
||||
// This method decodes an UTF-8 value according to RFC 3629 and
|
||||
// https://encoding.spec.whatwg.org/#utf-8-decoder .
|
||||
uchar Utf8::CalculateValue(const byte* str, size_t max_length, size_t* cursor) {
|
||||
DCHECK_GT(max_length, 0);
|
||||
DCHECK_GT(str[0], kMaxOneByteChar);
|
||||
|
||||
size_t length = NonASCIISequenceLength(str[0]);
|
||||
State state = State::kAccept;
|
||||
Utf8IncrementalBuffer buffer = 0;
|
||||
uchar t;
|
||||
|
||||
// Check continuation characters.
|
||||
size_t max_count = std::min(length, max_length);
|
||||
size_t count = 1;
|
||||
while (count < max_count && IsContinuationCharacter(str[count])) {
|
||||
count++;
|
||||
}
|
||||
size_t i = 0;
|
||||
do {
|
||||
t = ValueOfIncremental(str[i], &i, &state, &buffer);
|
||||
} while (i < max_length && t == kIncomplete);
|
||||
|
||||
if (length >= 3 && count < 2) {
|
||||
// Not enough continuation bytes to check overlong sequences.
|
||||
*cursor += 1;
|
||||
return kBadChar;
|
||||
}
|
||||
|
||||
// Check overly long sequences & other conditions.
|
||||
if (length == 3) {
|
||||
if (str[0] == 0xE0 && (str[1] < 0xA0 || str[1] > 0xBF)) {
|
||||
// Overlong three-byte sequence? The first byte generates a kBadChar.
|
||||
*cursor += 1;
|
||||
return kBadChar;
|
||||
} else if (str[0] == 0xED && (str[1] < 0x80 || str[1] > 0x9F)) {
|
||||
// High and low surrogate halves? The first byte generates a kBadChar.
|
||||
*cursor += 1;
|
||||
return kBadChar;
|
||||
}
|
||||
} else if (length == 4) {
|
||||
if (str[0] == 0xF0 && (str[1] < 0x90 || str[1] > 0xBF)) {
|
||||
// Overlong four-byte sequence. The first byte generates a kBadChar.
|
||||
*cursor += 1;
|
||||
return kBadChar;
|
||||
} else if (str[0] == 0xF4 && (str[1] < 0x80 || str[1] > 0x8F)) {
|
||||
// Code points outside of the Unicode range. The first byte generates a
|
||||
// kBadChar.
|
||||
*cursor += 1;
|
||||
return kBadChar;
|
||||
}
|
||||
}
|
||||
|
||||
*cursor += count;
|
||||
|
||||
if (count != length) {
|
||||
// Not enough continuation characters.
|
||||
return kBadChar;
|
||||
}
|
||||
|
||||
// All errors have been handled, so we only have to assemble the result.
|
||||
switch (length) {
|
||||
case 2:
|
||||
return ((str[0] << 6) + str[1]) - 0x00003080;
|
||||
case 3:
|
||||
return ((str[0] << 12) + (str[1] << 6) + str[2]) - 0x000E2080;
|
||||
case 4:
|
||||
return ((str[0] << 18) + (str[1] << 12) + (str[2] << 6) + str[3]) -
|
||||
0x03C82080;
|
||||
}
|
||||
|
||||
UNREACHABLE();
|
||||
*cursor += i;
|
||||
return (state == State::kAccept) ? t : kBadChar;
|
||||
}
|
||||
|
||||
/*
|
||||
Overlong sequence detection: Since Blink's TextCodecUTF8 rejects multi-byte
|
||||
characters which could be expressed with less bytes, we must too.
|
||||
|
||||
Each continuation byte (10xxxxxx) carries 6 bits of payload. The lead bytes of
|
||||
1, 2, 3 and 4-byte characters are 0xxxxxxx, 110xxxxx, 1110xxxx and 11110xxx, and
|
||||
carry 7, 5, 4, and 3 bits of payload, respectively.
|
||||
|
||||
Thus, a two-byte character can contain 11 bits of payload, a three-byte
|
||||
character 16, and a four-byte character 21.
|
||||
|
||||
If we encounter a two-byte character which contains 7 bits or less, a three-byte
|
||||
character which contains 11 bits or less, or a four-byte character which
|
||||
contains 16 bits or less, we reject the character and generate a kBadChar for
|
||||
each of the bytes. This is because Blink handles overlong sequences by rejecting
|
||||
the first byte of the character (returning kBadChar); thus the rest are lonely
|
||||
continuation bytes and generate a kBadChar each.
|
||||
*/
|
||||
|
||||
uchar Utf8::ValueOfIncremental(byte next, Utf8IncrementalBuffer* buffer) {
|
||||
// Decodes UTF-8 bytes incrementally, allowing the decoding of bytes as they
|
||||
// stream in. This **must** be followed by a call to ValueOfIncrementalFinish
|
||||
// when the stream is complete, to ensure incomplete sequences are handled.
|
||||
uchar Utf8::ValueOfIncremental(byte next, size_t* cursor, State* state,
|
||||
Utf8IncrementalBuffer* buffer) {
|
||||
DCHECK_NOT_NULL(buffer);
|
||||
State old_state = *state;
|
||||
*cursor += 1;
|
||||
|
||||
// The common case: 1-byte Utf8 (and no incomplete char in the buffer)
|
||||
if (V8_LIKELY(next <= kMaxOneByteChar && *buffer == 0)) {
|
||||
if (V8_LIKELY(next <= kMaxOneByteChar && old_state == State::kAccept)) {
|
||||
DCHECK_EQ(0u, *buffer);
|
||||
return static_cast<uchar>(next);
|
||||
}
|
||||
|
||||
if (*buffer == 0) {
|
||||
// We're at the start of a new character.
|
||||
uint32_t kind = NonASCIISequenceLength(next);
|
||||
CHECK_LE(kind, 4);
|
||||
if (kind >= 2) {
|
||||
// Start of 2..4 byte character, and no buffer.
|
||||
// So we're at the lead byte of a 2/3/4 sequence, or we're at a continuation
|
||||
// char in that sequence.
|
||||
Utf8DfaDecoder::Decode(next, state, buffer);
|
||||
|
||||
// The mask for the lower bits depends on the kind, and is
|
||||
// 0x1F, 0x0F, 0x07 for kinds 2, 3, 4 respectively. We can get that
|
||||
// with one shift.
|
||||
uint8_t mask = 0x7F >> kind;
|
||||
|
||||
// Store the kind in the top nibble, and kind - 1 (i.e., remaining bytes)
|
||||
// in 2nd nibble, and the value in the bottom three. The 2nd nibble is
|
||||
// intended as a counter about how many bytes are still needed.
|
||||
uint32_t character_info = kind << 28 | (kind - 1) << 24;
|
||||
DCHECK_EQ(character_info & mask, 0);
|
||||
*buffer = character_info | (next & mask);
|
||||
return kIncomplete;
|
||||
} else {
|
||||
// No buffer, and not the start of a 1-byte char (handled at the
|
||||
// beginning), and not the start of a 2..4 byte char (or the start of an
|
||||
// overlong / invalid sequence)? Bad char.
|
||||
switch (*state) {
|
||||
case State::kAccept: {
|
||||
uchar t = *buffer;
|
||||
*buffer = 0;
|
||||
return kBadChar;
|
||||
}
|
||||
} else if (*buffer <= 0xFF) {
|
||||
// We have one unprocessed byte left (from the last else case in this if
|
||||
// statement).
|
||||
uchar previous = *buffer;
|
||||
*buffer = 0;
|
||||
uchar t = ValueOfIncremental(previous, buffer);
|
||||
if (t == kIncomplete) {
|
||||
// If we have an incomplete character, process both the previous and the
|
||||
// next byte at once.
|
||||
return ValueOfIncremental(next, buffer);
|
||||
} else {
|
||||
// Otherwise, process the previous byte and save the next byte for next
|
||||
// time.
|
||||
DCHECK_EQ(0u, *buffer);
|
||||
*buffer = next;
|
||||
return t;
|
||||
}
|
||||
} else if (IsContinuationCharacter(next)) {
|
||||
// We're inside of a character, as described by buffer.
|
||||
|
||||
// How many bytes (excluding this one) do we still expect?
|
||||
uint8_t bytes_expected = *buffer >> 28;
|
||||
uint8_t bytes_left = (*buffer >> 24) & 0x0F;
|
||||
|
||||
// Two-byte overlong sequence detection is handled by
|
||||
// NonASCIISequenceLength, so we don't need to check anything here.
|
||||
if (bytes_expected == 3 && bytes_left == 2) {
|
||||
// Check that there are at least 12 bytes of payload.
|
||||
uint8_t lead_payload = *buffer & (0x7F >> bytes_expected);
|
||||
DCHECK_LE(lead_payload, 0xF);
|
||||
if (lead_payload == 0 && next < 0xA0) {
|
||||
// 0xA0 = 0b10100000 (payload: 100000). Overlong sequence: 0 bits from
|
||||
// the first byte, at most 5 from the second byte, and at most 6 from
|
||||
// the third -> in total at most 11.
|
||||
|
||||
*buffer = next;
|
||||
return kBadChar;
|
||||
} else if (lead_payload == 0xD && next > 0x9F) {
|
||||
// The resulting code point would be on a range which is reserved for
|
||||
// UTF-16 surrogate halves.
|
||||
*buffer = next;
|
||||
return kBadChar;
|
||||
}
|
||||
} else if (bytes_expected == 4 && bytes_left == 3) {
|
||||
// Check that there are at least 17 bytes of payload.
|
||||
uint8_t lead_payload = *buffer & (0x7F >> bytes_expected);
|
||||
|
||||
// If the lead byte was bigger than 0xF4 (payload: 4), it's not a start of
|
||||
// any valid character, and this is detected by NonASCIISequenceLength.
|
||||
DCHECK_LE(lead_payload, 0x4);
|
||||
if (lead_payload == 0 && next < 0x90) {
|
||||
// 0x90 = 10010000 (payload 10000). Overlong sequence: 0 bits from the
|
||||
// first byte, at most 4 from the second byte, at most 12 from the third
|
||||
// and fourth bytes -> in total at most 16.
|
||||
*buffer = next;
|
||||
return kBadChar;
|
||||
} else if (lead_payload == 4 && next > 0x8F) {
|
||||
// Invalid code point; value greater than 0b100001111000000000000
|
||||
// (0x10FFFF).
|
||||
*buffer = next;
|
||||
return kBadChar;
|
||||
}
|
||||
}
|
||||
|
||||
bytes_left--;
|
||||
// Update the value.
|
||||
uint32_t value = ((*buffer & 0xFFFFFF) << 6) | (next & 0x3F);
|
||||
if (bytes_left) {
|
||||
*buffer = (bytes_expected << 28 | bytes_left << 24 | value);
|
||||
return kIncomplete;
|
||||
} else {
|
||||
#ifdef DEBUG
|
||||
// Check that overlong sequences were already detected.
|
||||
bool sequence_was_too_long = (bytes_expected == 2 && value < 0x80) ||
|
||||
(bytes_expected == 3 && value < 0x800) ||
|
||||
(bytes_expected == 4 && value < 0x8000);
|
||||
DCHECK(!sequence_was_too_long);
|
||||
#endif
|
||||
case State::kReject:
|
||||
*state = State::kAccept;
|
||||
*buffer = 0;
|
||||
return value;
|
||||
}
|
||||
} else {
|
||||
// Within a character, but not a continuation character? Then the
|
||||
// previous char was a bad char. But we need to save the current
|
||||
// one.
|
||||
*buffer = next;
|
||||
return kBadChar;
|
||||
|
||||
// If we hit a bad byte, we need to determine if we were trying to start
|
||||
// a sequence or continue one. If we were trying to start a sequence,
|
||||
// that means it's just an invalid lead byte and we need to continue to
|
||||
// the next (which we already did above). If we were already in a
|
||||
// sequence, we need to reprocess this same byte after resetting to the
|
||||
// initial state.
|
||||
if (old_state != State::kAccept) {
|
||||
// We were trying to continue a sequence, so let's reprocess this byte
|
||||
// next time.
|
||||
*cursor -= 1;
|
||||
}
|
||||
return kBadChar;
|
||||
|
||||
default:
|
||||
return kIncomplete;
|
||||
}
|
||||
}
|
||||
|
||||
uchar Utf8::ValueOfIncrementalFinish(Utf8IncrementalBuffer* buffer) {
|
||||
DCHECK_NOT_NULL(buffer);
|
||||
if (*buffer == 0) {
|
||||
// Finishes the incremental decoding, ensuring that if an unfinished sequence
|
||||
// is left that it is replaced by a replacement char.
|
||||
uchar Utf8::ValueOfIncrementalFinish(State* state) {
|
||||
if (*state == State::kAccept) {
|
||||
return kBufferEmpty;
|
||||
} else {
|
||||
// Process left-over chars. An incomplete char at the end maps to kBadChar.
|
||||
uchar t = ValueOfIncremental(0, buffer);
|
||||
return (t == kIncomplete) ? kBadChar : t;
|
||||
DCHECK_GT(*state, State::kAccept);
|
||||
*state = State::kAccept;
|
||||
return kBadChar;
|
||||
}
|
||||
}
|
||||
|
||||
bool Utf8::ValidateEncoding(const byte* bytes, size_t length) {
|
||||
const byte* cursor = bytes;
|
||||
const byte* end = bytes + length;
|
||||
|
||||
while (cursor < end) {
|
||||
// Skip over single-byte values.
|
||||
if (*cursor <= kMaxOneByteChar) {
|
||||
++cursor;
|
||||
continue;
|
||||
}
|
||||
|
||||
// Get the length the the character.
|
||||
size_t seq_length = NonASCIISequenceLength(*cursor);
|
||||
// For some invalid characters NonASCIISequenceLength returns 0.
|
||||
if (seq_length == 0) return false;
|
||||
|
||||
const byte* char_end = cursor + seq_length;
|
||||
|
||||
// Return false if we do not have enough bytes for the character.
|
||||
if (char_end > end) return false;
|
||||
|
||||
// Check if the bytes of the character are continuation bytes.
|
||||
for (const byte* i = cursor + 1; i < char_end; ++i) {
|
||||
if (!IsContinuationCharacter(*i)) return false;
|
||||
}
|
||||
|
||||
// Check overly long sequences & other conditions.
|
||||
if (seq_length == 3) {
|
||||
if (cursor[0] == 0xE0 && (cursor[1] < 0xA0 || cursor[1] > 0xBF)) {
|
||||
// Overlong three-byte sequence?
|
||||
return false;
|
||||
} else if (cursor[0] == 0xED && (cursor[1] < 0x80 || cursor[1] > 0x9F)) {
|
||||
// High and low surrogate halves?
|
||||
return false;
|
||||
}
|
||||
} else if (seq_length == 4) {
|
||||
if (cursor[0] == 0xF0 && (cursor[1] < 0x90 || cursor[1] > 0xBF)) {
|
||||
// Overlong four-byte sequence.
|
||||
return false;
|
||||
} else if (cursor[0] == 0xF4 && (cursor[1] < 0x80 || cursor[1] > 0x8F)) {
|
||||
// Code points outside of the Unicode range.
|
||||
return false;
|
||||
}
|
||||
}
|
||||
cursor = char_end;
|
||||
State state = State::kAccept;
|
||||
Utf8IncrementalBuffer throw_away = 0;
|
||||
for (size_t i = 0; i < length && state != State::kReject; i++) {
|
||||
Utf8DfaDecoder::Decode(bytes[i], &state, &throw_away);
|
||||
}
|
||||
return true;
|
||||
return state == State::kAccept;
|
||||
}
|
||||
|
||||
// Uppercase: point.category == 'Lu'
|
||||
|
||||
@ -7,6 +7,7 @@
|
||||
|
||||
#include <sys/types.h>
|
||||
#include "src/globals.h"
|
||||
#include "src/third_party/utf8-decoder/utf8-decoder.h"
|
||||
#include "src/utils.h"
|
||||
/**
|
||||
* \file
|
||||
@ -129,6 +130,8 @@ class Utf16 {
|
||||
|
||||
class V8_EXPORT_PRIVATE Utf8 {
|
||||
public:
|
||||
using State = Utf8DfaDecoder::State;
|
||||
|
||||
static inline uchar Length(uchar chr, int previous);
|
||||
static inline unsigned EncodeOneByte(char* out, uint8_t c);
|
||||
static inline unsigned Encode(char* out,
|
||||
@ -158,9 +161,9 @@ class V8_EXPORT_PRIVATE Utf8 {
|
||||
static inline uchar ValueOf(const byte* str, size_t length, size_t* cursor);
|
||||
|
||||
typedef uint32_t Utf8IncrementalBuffer;
|
||||
static uchar ValueOfIncremental(byte next_byte,
|
||||
static uchar ValueOfIncremental(byte next_byte, size_t* cursor, State* state,
|
||||
Utf8IncrementalBuffer* buffer);
|
||||
static uchar ValueOfIncrementalFinish(Utf8IncrementalBuffer* buffer);
|
||||
static uchar ValueOfIncrementalFinish(State* state);
|
||||
|
||||
// Excludes non-characters from the set of valid code points.
|
||||
static inline bool IsValidCharacter(uchar c);
|
||||
|
||||
@ -1406,6 +1406,7 @@
|
||||
'strtod.h',
|
||||
'ic/stub-cache.cc',
|
||||
'ic/stub-cache.h',
|
||||
'third_party/utf8-decoder/utf8-decoder.h',
|
||||
'tracing/trace-event.cc',
|
||||
'tracing/trace-event.h',
|
||||
'tracing/traced-value.cc',
|
||||
|
||||
@ -19,12 +19,16 @@ static int Ucs2CharLength(unibrow::uchar c) {
|
||||
|
||||
static int Utf8LengthHelper(const char* s) {
|
||||
unibrow::Utf8::Utf8IncrementalBuffer buffer(unibrow::Utf8::kBufferEmpty);
|
||||
unibrow::Utf8::State state = unibrow::Utf8::State::kAccept;
|
||||
|
||||
int length = 0;
|
||||
for (; *s != '\0'; s++) {
|
||||
unibrow::uchar tmp = unibrow::Utf8::ValueOfIncremental(*s, &buffer);
|
||||
size_t i = 0;
|
||||
while (s[i] != '\0') {
|
||||
unibrow::uchar tmp =
|
||||
unibrow::Utf8::ValueOfIncremental(s[i], &i, &state, &buffer);
|
||||
length += Ucs2CharLength(tmp);
|
||||
}
|
||||
unibrow::uchar tmp = unibrow::Utf8::ValueOfIncrementalFinish(&buffer);
|
||||
unibrow::uchar tmp = unibrow::Utf8::ValueOfIncrementalFinish(&state);
|
||||
length += Ucs2CharLength(tmp);
|
||||
return length;
|
||||
}
|
||||
|
||||
@ -37,13 +37,15 @@ void DecodeNormally(const std::vector<byte>& bytes,
|
||||
void DecodeIncrementally(const std::vector<byte>& bytes,
|
||||
std::vector<unibrow::uchar>* output) {
|
||||
unibrow::Utf8::Utf8IncrementalBuffer buffer = 0;
|
||||
for (auto b : bytes) {
|
||||
unibrow::uchar result = unibrow::Utf8::ValueOfIncremental(b, &buffer);
|
||||
unibrow::Utf8::State state = unibrow::Utf8::State::kAccept;
|
||||
for (size_t i = 0; i < bytes.size();) {
|
||||
unibrow::uchar result =
|
||||
unibrow::Utf8::ValueOfIncremental(bytes[i], &i, &state, &buffer);
|
||||
if (result != unibrow::Utf8::kIncomplete) {
|
||||
output->push_back(result);
|
||||
}
|
||||
}
|
||||
unibrow::uchar result = unibrow::Utf8::ValueOfIncrementalFinish(&buffer);
|
||||
unibrow::uchar result = unibrow::Utf8::ValueOfIncrementalFinish(&state);
|
||||
if (result != unibrow::Utf8::kBufferEmpty) {
|
||||
output->push_back(result);
|
||||
}
|
||||
|
||||
Loading…
Reference in New Issue
Block a user