size_t qfind_first_byte_of_sse42(const StringPiece& haystack,
const StringPiece& needles) {
if (UNLIKELY(needles.empty() || haystack.empty())) {
return StringPiece::npos;
} else if (needles.size() <= 16) {
// we can save some unnecessary load instructions by optimizing for
// the common case of needles.size() <= 16
return qfind_first_byte_of_needles16(haystack, needles);
}
size_t index = haystack.size();
for (size_t i = 0; i < haystack.size(); i += 16) {
size_t b = 16;
auto arr1 = __builtin_ia32_loaddqu(haystack.data() + i);
for (size_t j = 0; j < needles.size(); j += 16) {
auto arr2 = __builtin_ia32_loaddqu(needles.data() + j);
auto index = __builtin_ia32_pcmpestri128(arr2, needles.size() - j,
arr1, haystack.size() - i, 0);
b = std::min<size_t>(index, b);
}
if (b < 16) {
return i + b;
}
};
return StringPiece::npos;
}
CAMLprim value mlre2__find_first(value v_regex, value v_sub, value v_str) {
CAMLparam2(v_regex, v_str);
CAMLlocal1(v_retval);
CAMLlocalN(error_args, 2);
const RE2 * re = Regex_val(v_regex);
const char* input = String_val(v_str);
int len = caml_string_length(v_str);
StringPiece str = StringPiece(input, len);
int n = Int_val(v_sub) + 1;
StringPiece * submatches = new StringPiece[n];
assert_valid_sub(re, v_sub);
if (! re->Match(str, 0, str.length(), RE2::UNANCHORED, submatches, n)) {
delete[] submatches;
caml_raise_with_string(*caml_named_value("mlre2__Regex_match_failed"),
re->pattern().c_str());
}
StringPiece * sub = submatches + Int_val(v_sub);
if (!sub->data()) {
delete[] submatches;
error_args[0] = caml_copy_string(re->pattern().c_str());
error_args[1] = v_sub;
caml_raise_with_args(*caml_named_value("mlre2__Regex_submatch_did_not_capture"),
2, error_args);
}
v_retval = caml_alloc_string(sub->length());
memcpy(String_val(v_retval), String_val(v_str) + (sub->data() - input), sub->length());
delete[] submatches;
CAMLreturn(v_retval);
}
void SplitStringKeepEmpty(
const StringPiece& full,
const StringPiece& delim,
std::vector<std::string>* result)
{
// 单个字符的分隔符转调字符版本的分割函数，要快一些
if (delim.length() == 1)
{
SplitStringKeepEmpty(full, delim[0], result);
return;
}
result->clear();
if (full.empty() || delim.empty())
return;
size_t prev_pos = 0;
size_t pos;
std::string token;
while ((pos = full.find(delim, prev_pos)) != std::string::npos)
{
token.assign(full.data() + prev_pos, pos - prev_pos);
result->push_back(token);
prev_pos = pos + delim.length();
}
token.assign(full.data() + prev_pos, full.length() - prev_pos);
result->push_back(token);
}
void DoSplitLines(
const StringPiece& full,
std::vector<StringType>* result,
bool keep_line_endling
)
{
result->clear();
size_t prev_pos = 0;
size_t pos;
StringType token;
while ((pos = full.find('\n', prev_pos)) != std::string::npos)
{
token.assign(full.data() + prev_pos, pos - prev_pos + 1);
if (!keep_line_endling)
RemoveLineEnding(&token);
result->push_back(token);
prev_pos = pos + 1;
}
if (prev_pos < full.size())
{
token.assign(full.data() + prev_pos, full.length() - prev_pos);
if (!keep_line_endling)
RemoveLineEnding(&token);
result->push_back(token);
}
}
bool LocalSequenceFileWriter::WriteRecord(const StringPiece& key, const StringPiece& value) {
int record_size = key.size() + value.size();
int key_len = key.size();
int int_len = sizeof(int);
int size = 2 * int_len + record_size;
int used_bytes = 0;
scoped_array<char> data(new char[size]);
if (m_records_written_after_sync >= kSyncInterval) {
size += int_len + SYNC_HASH_SIZE;
data.reset(new char[size]);
used_bytes += WriteSyncToBuf(data.get() + used_bytes);
m_records_written_after_sync = 0;
}
// format:
// record_size|key_len|key|value
used_bytes += WriteInt(data.get() + used_bytes, record_size);
used_bytes += WriteInt(data.get() + used_bytes, key_len);
memcpy(data.get() + used_bytes, key.data(), key_len);
used_bytes += key_len;
memcpy(data.get() + used_bytes, value.data(), value.size());
used_bytes += value.size();
CHECK(used_bytes == size);
if (!Write(data.get(), used_bytes)) {
return false;
}
++m_records_written_after_sync;
return true;
}
size_t qfind_first_byte_of_sse42(const StringPiece& haystack,
const StringPiece& needles) {
if (UNLIKELY(needles.empty() || haystack.empty())) {
return StringPiece::npos;
} else if (needles.size() <= 16) {
// we can save some unnecessary load instructions by optimizing for
// the common case of needles.size() <= 16
return qfind_first_byte_of_needles16(haystack, needles);
}
if (haystack.size() < 16 &&
PAGE_FOR(haystack.end() - 1) != PAGE_FOR(haystack.data() + 16)) {
// We can't safely SSE-load haystack. Use a different approach.
if (haystack.size() <= 2) {
return qfind_first_of(haystack, needles, asciiCaseSensitive);
}
return qfind_first_byte_of_byteset(haystack, needles);
}
auto ret = scanHaystackBlock<false>(haystack, needles, 0);
if (ret != StringPiece::npos) {
return ret;
}
size_t i = nextAlignedIndex(haystack.data());
for (; i < haystack.size(); i += 16) {
auto ret = scanHaystackBlock<true>(haystack, needles, i);
if (ret != StringPiece::npos) {
return ret;
}
}
return StringPiece::npos;
}
bool RML_RE::Rewrite(string *out, const StringPiece &rewrite,
const StringPiece &text, int *vec, int veclen) const {
for (const char *s = rewrite.data(), *end = s + rewrite.size();
s < end; s++) {
int c = *s;
if (c == '\\') {
c = *++s;
if (isdigit(c)) {
int n = (c - '0');
if (n >= veclen) {
//fprintf(stderr, requested group %d in regexp %.*s\n",
// n, rewrite.size(), rewrite.data());
return false;
}
int start = vec[2 * n];
if (start >= 0)
out->append(text.data() + start, vec[2 * n + 1] - start);
} else if (c == '\\') {
out->push_back('\\');
} else {
//fprintf(stderr, "invalid rewrite pattern: %.*s\n",
// rewrite.size(), rewrite.data());
return false;
}
} else {
out->push_back(c);
}
}
return true;
}
U_CAPI int32_t U_EXPORT2
unum_parseDecimal(const UNumberFormat* fmt,
const UChar* text,
int32_t textLength,
int32_t *parsePos /* 0 = start */,
char *outBuf,
int32_t outBufLength,
UErrorCode *status)
{
if (U_FAILURE(*status)) {
return -1;
}
if ((outBuf == NULL && outBufLength != 0) || outBufLength < 0) {
*status = U_ILLEGAL_ARGUMENT_ERROR;
return -1;
}
Formattable res;
parseRes(res, fmt, text, textLength, parsePos, status);
StringPiece sp = res.getDecimalNumber(*status);
if (U_FAILURE(*status)) {
return -1;
} else if (sp.size() > outBufLength) {
*status = U_BUFFER_OVERFLOW_ERROR;
} else if (sp.size() == outBufLength) {
uprv_strncpy(outBuf, sp.data(), sp.size());
*status = U_STRING_NOT_TERMINATED_WARNING;
} else {
U_ASSERT(outBufLength > 0);
uprv_strcpy(outBuf, sp.data());
}
return sp.size();
}
float ParseFloat(const StringPiece& str) {
char* errIndex;
float value = static_cast<float>(strtod(str.data(), &errIndex));
if (errIndex == str.data()) {
throw util::ParseNumberException(str);
}
return value;
}
void ReplaceAll(std::string* s, const StringPiece& from, const StringPiece& to)
{
size_t pos = 0;
while ((pos = s->find(from.data(), pos, from.size())) != std::string::npos)
{
s->replace(pos, from.size(), to.data(), to.size());
pos += to.size();
}
}
int ParseInt(const StringPiece& str ) {
char* errIndex;
//could wrap?
int value = static_cast<int>(strtol(str.data(), &errIndex,10));
if (errIndex == str.data()) {
throw util::ParseNumberException(str);
}
return value;
}
void Crypto::DigestUpdate(Digest x, const StringPiece &in) {
auto d = FromVoid<CCDigest*>(x);
switch(d->algo) {
case CCDigestAlgo::MD5: CC_MD5_Update (static_cast<CC_MD5_CTX*> (d->v), in.data(), in.size()); break;
case CCDigestAlgo::SHA1: CC_SHA1_Update (static_cast<CC_SHA1_CTX*> (d->v), in.data(), in.size()); break;
case CCDigestAlgo::SHA256: CC_SHA256_Update(static_cast<CC_SHA256_CTX*>(d->v), in.data(), in.size()); break;
case CCDigestAlgo::SHA384: CC_SHA384_Update(static_cast<CC_SHA512_CTX*>(d->v), in.data(), in.size()); break;
case CCDigestAlgo::SHA512: CC_SHA512_Update(static_cast<CC_SHA512_CTX*>(d->v), in.data(), in.size()); break;
default: break;
}
}
static int new_pos(const char *input, StringPiece &remaining, int startpos, StringPiece &match) {
if (remaining.length() < 0) {
return -1;
} else {
/* casting these size_t's to int is safe because StringPiece's track
* their lengths using ints */
size_t first_unexamined = remaining.data() + startpos - input;
size_t first_unmatched = match.data() - input + match.length();
return (int) (first_unexamined > first_unmatched ? first_unexamined : first_unmatched);
}
}
UCollationResult Collator::compareUTF8(const StringPiece &source,
const StringPiece &target,
UErrorCode &status) const {
if(U_FAILURE(status)) {
return UCOL_EQUAL;
}
UCharIterator sIter, tIter;
uiter_setUTF8(&sIter, source.data(), source.length());
uiter_setUTF8(&tIter, target.data(), target.length());
return compare(sIter, tIter, status);
}
/**
* Set the DigitList from a decimal number string.
*
* The incoming string _must_ be nul terminated, even though it is arriving
* as a StringPiece because that is what the decNumber library wants.
* We can get away with this for an internal function; it would not
* be acceptable for a public API.
*/
void
DigitList::set(StringPiece source, UErrorCode &status, uint32_t /*fastpathBits*/) {
if (U_FAILURE(status)) {
return;
}
#if 0
if(fastpathBits==(kFastpathOk|kNoDecimal)) {
int32_t size = source.size();
const char *data = source.data();
int64_t r = 0;
int64_t m = 1;
// fast parse
while(size>0) {
char ch = data[--size];
if(ch=='+') {
break;
} else if(ch=='-') {
r = -r;
break;
} else {
int64_t d = ch-'0';
//printf("CH[%d]=%c, %d, *=%d\n", size,ch, (int)d, (int)m);
r+=(d)*m;
m *= 10;
}
}
//printf("R=%d\n", r);
set(r);
} else
#endif
{
// Figure out a max number of digits to use during the conversion, and
// resize the number up if necessary.
int32_t numDigits = source.length();
if (numDigits > fContext.digits) {
// fContext.digits == fStorage.getCapacity()
decNumber *t = fStorage.resize(numDigits, fStorage.getCapacity());
if (t == NULL) {
status = U_MEMORY_ALLOCATION_ERROR;
return;
}
fDecNumber = t;
fContext.digits = numDigits;
}
fContext.status = 0;
uprv_decNumberFromString(fDecNumber, source.data(), &fContext);
if ((fContext.status & DEC_Conversion_syntax) != 0) {
status = U_DECIMAL_NUMBER_SYNTAX_ERROR;
}
}
internalClear();
}
std::u16string utf8ToUtf16(const StringPiece& utf8) {
ssize_t utf16Length = utf8_to_utf16_length(reinterpret_cast<const uint8_t*>(utf8.data()),
utf8.length());
if (utf16Length <= 0) {
return {};
}
std::u16string utf16;
utf16.resize(utf16Length);
utf8_to_utf16(reinterpret_cast<const uint8_t*>(utf8.data()), utf8.length(), &*utf16.begin());
return utf16;
}
void FDSymbolizePrinter::doPrint(StringPiece sp) {
if (buffer_) {
if (sp.size() > buffer_->tailroom()) {
flush();
writeFull(fd_, sp.data(), sp.size());
} else {
memcpy(buffer_->writableTail(), sp.data(), sp.size());
buffer_->append(sp.size());
}
} else {
writeFull(fd_, sp.data(), sp.size());
}
}
FFState* TargetBigramFeature::Evaluate(const Hypothesis& cur_hypo,
const FFState* prev_state,
ScoreComponentCollection* accumulator) const
{
const TargetBigramState* tbState = dynamic_cast<const TargetBigramState*>(prev_state);
assert(tbState);
// current hypothesis target phrase
const Phrase& targetPhrase = cur_hypo.GetCurrTargetPhrase();
if (targetPhrase.GetSize() == 0) {
return new TargetBigramState(*tbState);
}
// extract all bigrams w1 w2 from current hypothesis
for (size_t i = 0; i < targetPhrase.GetSize(); ++i) {
const Factor* f1 = NULL;
if (i == 0) {
f1 = tbState->GetWord().GetFactor(m_factorType);
} else {
f1 = targetPhrase.GetWord(i-1).GetFactor(m_factorType);
}
const Factor* f2 = targetPhrase.GetWord(i).GetFactor(m_factorType);
const StringPiece w1 = f1->GetString();
const StringPiece w2 = f2->GetString();
// skip bigrams if they don't belong to a given restricted vocabulary
if (m_vocab.size() &&
(FindStringPiece(m_vocab, w1) == m_vocab.end() || FindStringPiece(m_vocab, w2) == m_vocab.end())) {
continue;
}
string name(w1.data(), w1.size());
name += ":";
name.append(w2.data(), w2.size());
accumulator->PlusEquals(this,name,1);
}
if (cur_hypo.GetWordsBitmap().IsComplete()) {
const StringPiece w1 = targetPhrase.GetWord(targetPhrase.GetSize()-1).GetFactor(m_factorType)->GetString();
const string& w2 = EOS_;
if (m_vocab.empty() || (FindStringPiece(m_vocab, w1) != m_vocab.end())) {
string name(w1.data(), w1.size());
name += ":";
name += w2;
accumulator->PlusEquals(this,name,1);
}
return NULL;
}
return new TargetBigramState(targetPhrase.GetWord(targetPhrase.GetSize()-1));
}
size_t qfind_first_byte_of_memchr(const StringPiece& haystack,
const StringPiece& needles) {
size_t best = haystack.size();
for (char needle: needles) {
const void* ptr = memchr(haystack.data(), needle, best);
if (ptr) {
auto found = static_cast<const char*>(ptr) - haystack.data();
best = std::min<size_t>(best, found);
}
}
if (best == haystack.size()) {
return StringPiece::npos;
}
return best;
}
// -------------------------------------
void
DigitList::set(const StringPiece &source, UErrorCode &status) {
if (U_FAILURE(status)) {
return;
}
// Figure out a max number of digits to use during the conversion, and
// resize the number up if necessary.
int32_t numDigits = source.length();
if (numDigits > fContext.digits) {
fContext.digits = numDigits;
char *t = fStorage.resize(sizeof(decNumber) + numDigits, fStorage.getCapacity());
if (t == NULL) {
status = U_MEMORY_ALLOCATION_ERROR;
return;
}
fDecNumber = (decNumber *)fStorage.getAlias();
}
fContext.status = 0;
uprv_decNumberFromString(fDecNumber, source.data(), &fContext);
if ((fContext.status & DEC_Conversion_syntax) != 0) {
status = U_DECIMAL_NUMBER_SYNTAX_ERROR;
}
fHaveDouble = FALSE;
}
void Session::reply(StringPiece msg)
{
if (!noreply_)
{
conn_->send(msg.data(), msg.size());
}
}
bool XmlDocument::XPath(StringPiece const &v, std::wstring *r) {
WStringPiece str;
xmlXPathObjectPtr xpath = xmlXPathEvalExpression((unsigned char const*)v.data(), libxml_stuff->xpath_context);
if (xpath) {
xmlChar *s_ = xmlXPathCastToString(xpath);
if (s_) {
StringPiece s((char*)s_);
wchar_t *output = conv_buf.Alloc(s.length());
int const conv_result = cpcl::TryConvertUTF8_UTF16(s, output, conv_buf.Size());
if (conv_result == -1) {
output = 0;
Trace(CPCL_TRACE_LEVEL_ERROR, "XmlDocument::XPath(%s): utf_to_uc fails", v.as_string().c_str());
}
else {
output = conv_buf.Data() + conv_result;
if (conv_result > ((int)(conv_buf.Size()&INT_MAX)))
Trace(CPCL_TRACE_LEVEL_WARNING, "XmlDocument::XPath(%s): TryConvertUTF8_UTF16 okashi desu ne...", v.as_string().c_str());
}
xmlFree(s_);
if (output) {
//*output = 0;
str = WStringPiece(conv_buf.Data(), output - conv_buf.Data());
}
}
xmlXPathFreeObject(xpath);
}
if (TrimResults)
str = str.trim(TrimChars);
if ((!str.empty()) && (r))
r->assign(str.data(), str.size());
return (!str.empty());
}
bool RML_RE::DoMatchImpl(const StringPiece& text,
Anchor anchor,
int* consumed,
const Arg* const* args,
int n,
int* vec,
int vecsize) const {
assert((1 + n) * 3 <= vecsize); // results + PCRE workspace
int matches = TryMatch(text, 0, anchor, vec, vecsize);
assert(matches >= 0); // TryMatch never returns negatives
if (matches == 0)
return false;
*consumed = vec[1];
if (args == NULL) {
// We are not interested in results
return true;
}
// If we got here, we must have matched the whole pattern.
// We do not need (can not do) any more checks on the value of 'matches' here
// -- see the comment for TryMatch.
for (int i = 0; i < n; i++) {
const int start = vec[2*(i+1)];
const int limit = vec[2*(i+1)+1];
if (!args[i]->Parse(text.data() + start, limit-start)) {
// TODO: Should we indicate what the error was?
return false;
}
}
return true;
}
// For each character in characters_wanted, sets the index corresponding
// to the ASCII code of that character to 1 in table. This is used by
// the m_find.*_of methods below to tell whether or not a character is in
// the lookup table in constant time.
// The argument `table' must be an array that is large enough to hold all
// the possible values of an unsigned char. Thus it should be be declared
// as follows:
// bool table[UCHAR_MAX + 1]
static inline void BuildLookupTable(const StringPiece& characters_wanted,
bool* table) {
const size_type length = characters_wanted.length();
const char* const data = characters_wanted.data();
for (size_type i = 0; i < length; ++i) {
table[static_cast<unsigned char>(data[i])] = true;
}
}
// Do not assert in this function since it is used by the asssertion code!
std::wstring SysMultiByteToWide(const StringPiece& mb, uint32 code_page) {
if (mb.empty())
return std::wstring();
int mb_length = static_cast<int>(mb.length());
// Compute the length of the buffer.
int charcount = MultiByteToWideChar(code_page, 0,
mb.data(), mb_length, NULL, 0);
if (charcount == 0)
return std::wstring();
std::wstring wide;
wide.resize(charcount);
MultiByteToWideChar(code_page, 0, mb.data(), mb_length, &wide[0], charcount);
return wide;
}
bool XmlWriter::StartDocument(StringPiece encoding/*, IOStream *output*/) {
/* unsigned char UTF8_BOM[] = { 0xEF, 0xBB, 0xBF };
output->Write(UTF8_BOM, arraysize(UTF8_BOM)); */
if (encoding.empty())
encoding = StringPieceFromLiteral("UTF-8");
return (xmlTextWriterStartDocument(libxml_stuff->writer, NULL, encoding.data(), NULL) != -1);
}
void Item::resetKey(StringPiece k)
{
assert(k.size() <= 250);
keylen_ = k.size();
receivedBytes_ = 0;
append(k.data(), k.size());
hash_ = boost::hash_range(k.begin(), k.end());
}
U_EXPORT UBool U_EXPORT2
operator==(const StringPiece& x, const StringPiece& y) {
int32_t len = x.size();
if (len != y.size()) {
return false;
}
if (len == 0) {
return true;
}
const char* p = x.data();
const char* p2 = y.data();
// Test last byte in case strings share large common prefix
--len;
if (p[len] != p2[len]) return false;
// At this point we can, but don't have to, ignore the last byte.
return uprv_memcmp(p, p2, len) == 0;
}
void CaseMap::utf8ToUpper(
const char *locale, uint32_t options,
StringPiece src, ByteSink &sink, Edits *edits,
UErrorCode &errorCode) {
ucasemap_mapUTF8(
ustrcase_getCaseLocale(locale), options, UCASEMAP_BREAK_ITERATOR_NULL
src.data(), src.length(),
ucasemap_internalUTF8ToUpper, sink, edits, errorCode);
}
void CaseMap::utf8Fold(
uint32_t options,
StringPiece src, ByteSink &sink, Edits *edits,
UErrorCode &errorCode) {
ucasemap_mapUTF8(
UCASE_LOC_ROOT, options, UCASEMAP_BREAK_ITERATOR_NULL
src.data(), src.length(),
ucasemap_internalUTF8Fold, sink, edits, errorCode);
}
