whytofly
/
qt6-bb10
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Move the UTF conversion methods to qstringconverter 

Separate them from the qutfcodec, so that the codec
can later on be moved out of Qt Core.

Fix the QUtf methods to take qsizetype instead of int
for length arguments.

This also makes it possible to not build QTextCodec into
the bootstrap lib anymore.

Change-Id: I0b4f83139d61b19c651520a2f3a5012aa7e85cb8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
bb10
Lars Knoll 2020-04-17 12:10:21 +02:00
parent f64a6bd638
commit ea0a08c898
23 changed files with 1294 additions and 1239 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
qmake/CMakeLists.txt
View File

@ -40,7 +40,6 @@ qt_add_tool(qmake # special case
../src/3rdparty/pcre2/src/pcre2_ucp.h
../src/3rdparty/pcre2/src/pcre2_valid_utf.c
../src/3rdparty/pcre2/src/pcre2_xclass.c
../src/corelib/codecs/qutfcodec.cpp ../src/corelib/codecs/qutfcodec_p.h
../src/corelib/global/qendian.cpp # special case
../src/corelib/global/qglobal.cpp ../src/corelib/global/qglobal.h
../src/corelib/global/qlibraryinfo.cpp
@ -105,6 +104,7 @@ qt_add_tool(qmake # special case
../src/corelib/tools/qringbuffer.cpp # special case
../src/corelib/text/qstring.cpp ../src/corelib/text/qstring.h
../src/corelib/text/qstringbuilder.cpp ../src/corelib/text/qstringbuilder.h
../src/corelib/text/qstringconverter.cpp ../src/corelib/text/qstringconverter.h ../src/corelib/text/qstringconverter_p.h
../src/corelib/text/qstringlist.cpp ../src/corelib/text/qstringlist.h
../src/corelib/text/qstringmatcher.h
../src/corelib/tools/qvector.h

11
qmake/Makefile.unix
View File

@ -17,7 +17,6 @@ OBJS = \
#qt code (please keep in order matching DEPEND_SRC)
QOBJS = \
qutfcodec.o \
qendian.o qglobal.o qlogging.o qmalloc.o qnumeric.o qoperatingsystemversion.o qrandom.o \
qabstractfileengine.o qbuffer.o qdatastream.o qdebug.o \
qdir.o qdiriterator.o \
@ -32,7 +31,7 @@ QOBJS = \
qcalendar.o qgregoriancalendar.o qromancalendar.o \
qcryptographichash.o qdatetime.o qhash.o \
qlocale.o qlocale_tools.o qmap.o qregularexpression.o qregexp.o qringbuffer.o \
qstringbuilder.o qstring.o qstringlist.o qversionnumber.o \
qstringbuilder.o qstring.o qstringconverter.o qstringlist.o qversionnumber.o \
qvsnprintf.o qxmlstream.o qxmlutils.o \
pcre2_auto_possess.o pcre2_chartables.o pcre2_compile.o pcre2_config.o \
pcre2_context.o pcre2_dfa_match.o pcre2_error.o pcre2_extuni.o \
@ -74,7 +73,6 @@ DEPEND_SRC = \
$(QMKGENSRC)/win32/msvc_vcxproj.cpp \
$(QMKGENSRC)/win32/winmakefile.cpp \
$(QMKGENSRC)/xmloutput.cpp \
$(SOURCE_PATH)/src/corelib/codecs/qutfcodec.cpp \
$(SOURCE_PATH)/src/corelib/global/qendian.cpp \
$(SOURCE_PATH)/src/corelib/global/qglobal.cpp \
$(SOURCE_PATH)/src/corelib/global/qlibraryinfo.cpp \
@ -122,6 +120,7 @@ DEPEND_SRC = \
$(SOURCE_PATH)/src/corelib/text/qregularexpression.cpp \
$(SOURCE_PATH)/src/corelib/text/qregexp.cpp \
$(SOURCE_PATH)/src/corelib/text/qstringbuilder.cpp \
$(SOURCE_PATH)/src/corelib/text/qstringconverter.cpp \
$(SOURCE_PATH)/src/corelib/text/qstring.cpp \
$(SOURCE_PATH)/src/corelib/text/qstringlist.cpp \
$(SOURCE_PATH)/src/corelib/text/qvsnprintf.cpp \
@ -380,15 +379,15 @@ qoperatingsystemversion_darwin.o: $(SOURCE_PATH)/src/corelib/global/qoperatingsy
qcore_foundation.o: $(SOURCE_PATH)/src/corelib/kernel/qcore_foundation.mm
$(CXX) -c -o $@ $(CXXFLAGS) $<
qutfcodec.o: $(SOURCE_PATH)/src/corelib/codecs/qutfcodec.cpp
$(CXX) -c -o $@ $(CXXFLAGS) $<
qstring.o: $(SOURCE_PATH)/src/corelib/text/qstring.cpp
$(CXX) -c -o $@ $(CXXFLAGS) $<
qstringbuilder.o: $(SOURCE_PATH)/src/corelib/text/qstringbuilder.cpp
$(CXX) -c -o $@ $(CXXFLAGS) $<
qstringconverter.o: $(SOURCE_PATH)/src/corelib/text/qstringconverter.cpp
$(CXX) -c -o $@ $(CXXFLAGS) $<
qlocale.o: $(SOURCE_PATH)/src/corelib/text/qlocale.cpp
$(CXX) -c -o $@ $(CXXFLAGS) $<

2
qmake/Makefile.win32
View File

@ -104,8 +104,8 @@ QTOBJS= \
qoperatingsystemversion_win.obj \
qregexp.obj \
qromancalendar.obj \
qutfcodec.obj \
qstring.obj \
qstringconverter.obj \
qstringlist.obj \
qstringbuilder.obj \
qsystemerror.obj \

5
qmake/qmake.pro
View File

@ -159,11 +159,11 @@ SOURCES += \
qsettings.cpp \
qstring.cpp \
qstringbuilder.cpp \
qstringconverter.cpp \
qstringlist.cpp \
qsystemerror.cpp \
qtemporaryfile.cpp \
qtextstream.cpp \
qutfcodec.cpp \
quuid.cpp \
qvariant.cpp \
qversionnumber.cpp \
@ -217,12 +217,13 @@ HEADERS += \
qromancalendar_p.h \
qstring.h \
qstringbuilder.h \
qstringconverter_p.h \
qstringconverter.h \
qstringlist.h \
qstringmatcher.h \
qsystemerror_p.h \
qtemporaryfile.h \
qtextstream.h \
qutfcodec_p.h \
quuid.h \
qvector.h \
qversionnumber.h \

1
src/corelib/CMakeLists.txt
View File

@ -169,6 +169,7 @@ qt_add_module(Core
text/qstring.cpp text/qstring.h
text/qstring_compat.cpp
text/qstringalgorithms.h text/qstringalgorithms_p.h
text/qstringconverter.cpp text/qstringconverter.h text/qstringconverter_p.h
text/qstringbuilder.cpp text/qstringbuilder.h
text/qstringiterator_p.h
text/qstringlist.cpp text/qstringlist.h

940
src/corelib/codecs/qutfcodec.cpp
View File

@ -48,946 +48,6 @@
QT_BEGIN_NAMESPACE
enum { Endian = 0, Data = 1 };
static const uchar utf8bom[] = { 0xef, 0xbb, 0xbf };
#if (defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2)) \
|| (defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64))
static Q_ALWAYS_INLINE uint qBitScanReverse(unsigned v) noexcept
{
uint result = qCountLeadingZeroBits(v);
// Now Invert the result: clz will count *down* from the msb to the lsb, so the msb index is 31
// and the lsb index is 0. The result for _bit_scan_reverse is expected to be the index when
// counting up: msb index is 0 (because it starts there), and the lsb index is 31.
result ^= sizeof(unsigned) * 8 - 1;
return result;
}
#endif
#if defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2)
static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end)
{
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
# ifdef __AVX2__
__m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src));
__m128i data1 = _mm256_castsi256_si128(data);
__m128i data2 = _mm256_extracti128_si256(data, 1);
# else
__m128i data1 = _mm_loadu_si128((const __m128i*)src);
__m128i data2 = _mm_loadu_si128(1+(const __m128i*)src);
# endif
// check if everything is ASCII
// the highest ASCII value is U+007F
// Do the packing directly:
// The PACKUSWB instruction has packs a signed 16-bit integer to an unsigned 8-bit
// with saturation. That is, anything from 0x0100 to 0x7fff is saturated to 0xff,
// while all negatives (0x8000 to 0xffff) get saturated to 0x00. To detect non-ASCII,
// we simply do a signed greater-than comparison to 0x00. That means we detect NULs as
// "non-ASCII", but it's an acceptable compromise.
__m128i packed = _mm_packus_epi16(data1, data2);
__m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128());
// store, even if there are non-ASCII characters here
_mm_storeu_si128((__m128i*)dst, packed);
// n will contain 1 bit set per character in [data1, data2] that is non-ASCII (or NUL)
ushort n = ~_mm_movemask_epi8(nonAscii);
if (n) {
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
n = qCountTrailingZeroBits(n);
dst += n;
src += n;
return false;
}
}
if (end - src >= 8) {
// do eight characters at a time
__m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src));
__m128i packed = _mm_packus_epi16(data, data);
__m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128());
// store even non-ASCII
_mm_storel_epi64(reinterpret_cast<__m128i *>(dst), packed);
uchar n = ~_mm_movemask_epi8(nonAscii);
if (n) {
nextAscii = src + qBitScanReverse(n) + 1;
n = qCountTrailingZeroBits(n);
dst += n;
src += n;
return false;
}
}
return src == end;
}
static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end)
{
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
__m128i data = _mm_loadu_si128((const __m128i*)src);
#ifdef __AVX2__
const int BitSpacing = 2;
// load and zero extend to an YMM register
const __m256i extended = _mm256_cvtepu8_epi16(data);
uint n = _mm256_movemask_epi8(extended);
if (!n) {
// store
_mm256_storeu_si256((__m256i*)dst, extended);
continue;
}
#else
const int BitSpacing = 1;
// check if everything is ASCII
// movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII
uint n = _mm_movemask_epi8(data);
if (!n) {
// unpack
_mm_storeu_si128((__m128i*)dst, _mm_unpacklo_epi8(data, _mm_setzero_si128()));
_mm_storeu_si128(1+(__m128i*)dst, _mm_unpackhi_epi8(data, _mm_setzero_si128()));
continue;
}
#endif
// copy the front part that is still ASCII
while (!(n & 1)) {
*dst++ = *src++;
n >>= BitSpacing;
}
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
n = qBitScanReverse(n);
nextAscii = src + (n / BitSpacing) + 1;
return false;
}
if (end - src >= 8) {
__m128i data = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src));
uint n = _mm_movemask_epi8(data) & 0xff;
if (!n) {
// unpack and store
_mm_storeu_si128(reinterpret_cast<__m128i *>(dst), _mm_unpacklo_epi8(data, _mm_setzero_si128()));
} else {
while (!(n & 1)) {
*dst++ = *src++;
n >>= 1;
}
n = qBitScanReverse(n);
nextAscii = src + n + 1;
return false;
}
}
return src == end;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
#ifdef __AVX2__
// do 32 characters at a time
// (this is similar to simdTestMask in qstring.cpp)
const __m256i mask = _mm256_set1_epi8(0x80);
for ( ; end - src >= 32; src += 32) {
__m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src));
if (_mm256_testz_si256(mask, data))
continue;
uint n = _mm256_movemask_epi8(data);
Q_ASSUME(n);
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
#endif
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16) {
__m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i*>(src));
// check if everything is ASCII
// movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII
uint n = _mm_movemask_epi8(data);
if (!n)
continue;
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
// do four characters at a time
for ( ; end - src >= 4; src += 4) {
quint32 data = qFromUnaligned<quint32>(src);
data &= 0x80808080U;
if (!data)
continue;
// We don't try to guess which of the three bytes is ASCII and which
// one isn't. The chance that at least two of them are non-ASCII is
// better than 75%.
nextAscii = src;
return src;
}
nextAscii = end;
return src;
}
#elif defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64) // vaddv is only available on Aarch64
static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end)
{
uint16x8_t maxAscii = vdupq_n_u16(0x7f);
uint16x8_t mask1 = { 1, 1 << 2, 1 << 4, 1 << 6, 1 << 8, 1 << 10, 1 << 12, 1 << 14 };
uint16x8_t mask2 = vshlq_n_u16(mask1, 1);
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
// load 2 lanes (or: "load interleaved")
uint16x8x2_t in = vld2q_u16(src);
// check if any of the elements > 0x7f, select 1 bit per element (element 0 -> bit 0, element 1 -> bit 1, etc),
// add those together into a scalar, and merge the scalars.
uint16_t nonAscii = vaddvq_u16(vandq_u16(vcgtq_u16(in.val[0], maxAscii), mask1))
| vaddvq_u16(vandq_u16(vcgtq_u16(in.val[1], maxAscii), mask2));
// merge the two lanes by shifting the values of the second by 8 and inserting them
uint16x8_t out = vsliq_n_u16(in.val[0], in.val[1], 8);
// store, even if there are non-ASCII characters here
vst1q_u8(dst, vreinterpretq_u8_u16(out));
if (nonAscii) {
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(nonAscii) + 1;
nonAscii = qCountTrailingZeroBits(nonAscii);
dst += nonAscii;
src += nonAscii;
return false;
}
}
return src == end;
}
static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end)
{
// do eight characters at a time
uint8x8_t msb_mask = vdup_n_u8(0x80);
uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 };
for ( ; end - src >= 8; src += 8, dst += 8) {
uint8x8_t c = vld1_u8(src);
uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask));
if (!n) {
// store
vst1q_u16(dst, vmovl_u8(c));
continue;
}
// copy the front part that is still ASCII
while (!(n & 1)) {
*dst++ = *src++;
n >>= 1;
}
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
n = qBitScanReverse(n);
nextAscii = src + n + 1;
return false;
}
return src == end;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
// The SIMD code below is untested, so just force an early return until
// we've had the time to verify it works.
nextAscii = end;
return src;
// do eight characters at a time
uint8x8_t msb_mask = vdup_n_u8(0x80);
uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 };
for ( ; end - src >= 8; src += 8) {
uint8x8_t c = vld1_u8(src);
uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask));
if (!n)
continue;
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
nextAscii = end;
return src;
}
#else
static inline bool simdEncodeAscii(uchar *, const ushort *, const ushort *, const ushort *)
{
return false;
}
static inline bool simdDecodeAscii(ushort *, const uchar *, const uchar *, const uchar *)
{
return false;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
nextAscii = end;
return src;
}
#endif
QByteArray QUtf8::convertFromUnicode(const QChar *uc, int len)
{
// create a QByteArray with the worst case scenario size
QByteArray result(len * 3, Qt::Uninitialized);
uchar *dst = reinterpret_cast<uchar *>(const_cast<char *>(result.constData()));
const ushort *src = reinterpret_cast<const ushort *>(uc);
const ushort *const end = src + len;
while (src != end) {
const ushort *nextAscii = end;
if (simdEncodeAscii(dst, nextAscii, src, end))
break;
do {
ushort uc = *src++;
int res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, dst, src, end);
if (res < 0) {
// encoding error - append '?'
*dst++ = '?';
}
} while (src < nextAscii);
}
result.truncate(dst - reinterpret_cast<uchar *>(const_cast<char *>(result.constData())));
return result;
}
QByteArray QUtf8::convertFromUnicode(const QChar *uc, int len, QTextCodec::ConverterState *state)
{
uchar replacement = '?';
int rlen = 3*len;
int surrogate_high = -1;
if (state) {
if (state->flags & QTextCodec::ConvertInvalidToNull)
replacement = 0;
if (!(state->flags & QTextCodec::IgnoreHeader))
rlen += 3;
if (state->remainingChars)
surrogate_high = state->state_data[0];
}
QByteArray rstr(rlen, Qt::Uninitialized);
uchar *cursor = reinterpret_cast<uchar *>(const_cast<char *>(rstr.constData()));
const ushort *src = reinterpret_cast<const ushort *>(uc);
const ushort *const end = src + len;
int invalid = 0;
if (state && !(state->flags & QTextCodec::IgnoreHeader)) {
// append UTF-8 BOM
*cursor++ = utf8bom[0];
*cursor++ = utf8bom[1];
*cursor++ = utf8bom[2];
}
const ushort *nextAscii = src;
while (src != end) {
int res;
ushort uc;
if (surrogate_high != -1) {
uc = surrogate_high;
surrogate_high = -1;
res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, cursor, src, end);
} else {
if (src >= nextAscii && simdEncodeAscii(cursor, nextAscii, src, end))
break;
uc = *src++;
res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, cursor, src, end);
}
if (Q_LIKELY(res >= 0))
continue;
if (res == QUtf8BaseTraits::Error) {
// encoding error
++invalid;
*cursor++ = replacement;
} else if (res == QUtf8BaseTraits::EndOfString) {
surrogate_high = uc;
break;
}
}
rstr.resize(cursor - (const uchar*)rstr.constData());
if (state) {
state->invalidChars += invalid;
state->flags |= QTextCodec::IgnoreHeader;
state->remainingChars = 0;
if (surrogate_high >= 0) {
state->remainingChars = 1;
state->state_data[0] = surrogate_high;
}
}
return rstr;
}
QString QUtf8::convertToUnicode(const char *chars, int len)
{
// UTF-8 to UTF-16 always needs the exact same number of words or less:
// UTF-8 UTF-16
// 1 byte 1 word
// 2 bytes 1 word
// 3 bytes 1 word
// 4 bytes 2 words (one surrogate pair)
// That is, we'll use the full buffer if the input is US-ASCII (1-byte UTF-8),
// half the buffer for U+0080-U+07FF text (e.g., Greek, Cyrillic, Arabic) or
// non-BMP text, and one third of the buffer for U+0800-U+FFFF text (e.g, CJK).
//
// The table holds for invalid sequences too: we'll insert one replacement char
// per invalid byte.
QString result(len, Qt::Uninitialized);
QChar *data = const_cast<QChar*>(result.constData()); // we know we're not shared
const QChar *end = convertToUnicode(data, chars, len);
result.truncate(end - data);
return result;
}
/*!
\since 5.7
\overload
Converts the UTF-8 sequence of \a len octets beginning at \a chars to
a sequence of QChar starting at \a buffer. The buffer is expected to be
large enough to hold the result. An upper bound for the size of the
buffer is \a len QChars.
If, during decoding, an error occurs, a QChar::ReplacementCharacter is
written.
Returns a pointer to one past the last QChar written.
This function never throws.
*/
QChar *QUtf8::convertToUnicode(QChar *buffer, const char *chars, int len) noexcept
{
ushort *dst = reinterpret_cast<ushort *>(buffer);
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
// attempt to do a full decoding in SIMD
const uchar *nextAscii = end;
if (!simdDecodeAscii(dst, nextAscii, src, end)) {
// at least one non-ASCII entry
// check if we failed to decode the UTF-8 BOM; if so, skip it
if (Q_UNLIKELY(src == reinterpret_cast<const uchar *>(chars))
&& end - src >= 3
&& Q_UNLIKELY(src[0] == utf8bom[0] && src[1] == utf8bom[1] && src[2] == utf8bom[2])) {
src += 3;
}
while (src < end) {
nextAscii = end;
if (simdDecodeAscii(dst, nextAscii, src, end))
break;
do {
uchar b = *src++;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, dst, src, end);
if (res < 0) {
// decoding error
*dst++ = QChar::ReplacementCharacter;
}
} while (src < nextAscii);
}
}
return reinterpret_cast<QChar *>(dst);
}
QString QUtf8::convertToUnicode(const char *chars, int len, QTextCodec::ConverterState *state)
{
bool headerdone = false;
ushort replacement = QChar::ReplacementCharacter;
int invalid = 0;
int res;
uchar ch = 0;
// See above for buffer requirements for stateless decoding. However, that
// fails if the state is not empty. The following situations can add to the
// requirements:
// state contains chars starts with requirement
// 1 of 2 bytes valid continuation 0
// 2 of 3 bytes same 0
// 3 bytes of 4 same +1 (need to insert surrogate pair)
// 1 of 2 bytes invalid continuation +1 (need to insert replacement and restart)
// 2 of 3 bytes same +1 (same)
// 3 of 4 bytes same +1 (same)
QString result(len + 1, Qt::Uninitialized);
ushort *dst = reinterpret_cast<ushort *>(const_cast<QChar *>(result.constData()));
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
if (state) {
if (state->flags & QTextCodec::IgnoreHeader)
headerdone = true;
if (state->flags & QTextCodec::ConvertInvalidToNull)
replacement = QChar::Null;
if (state->remainingChars) {
// handle incoming state first
uchar remainingCharsData[4]; // longest UTF-8 sequence possible
int remainingCharsCount = state->remainingChars;
int newCharsToCopy = qMin<int>(sizeof(remainingCharsData) - remainingCharsCount, end - src);
memset(remainingCharsData, 0, sizeof(remainingCharsData));
memcpy(remainingCharsData, &state->state_data[0], remainingCharsCount);
memcpy(remainingCharsData + remainingCharsCount, src, newCharsToCopy);
const uchar *begin = &remainingCharsData[1];
res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(remainingCharsData[0], dst, begin,
static_cast<const uchar *>(remainingCharsData) + remainingCharsCount + newCharsToCopy);
if (res == QUtf8BaseTraits::Error || (res == QUtf8BaseTraits::EndOfString && len == 0)) {
// special case for len == 0:
// if we were supplied an empty string, terminate the previous, unfinished sequence with error
++invalid;
*dst++ = replacement;
} else if (res == QUtf8BaseTraits::EndOfString) {
// if we got EndOfString again, then there were too few bytes in src;
// copy to our state and return
state->remainingChars = remainingCharsCount + newCharsToCopy;
memcpy(&state->state_data[0], remainingCharsData, state->remainingChars);
return QString();
} else if (!headerdone && res >= 0) {
// eat the UTF-8 BOM
headerdone = true;
if (dst[-1] == 0xfeff)
--dst;
}
// adjust src now that we have maybe consumed a few chars
if (res >= 0) {
Q_ASSERT(res > remainingCharsCount);
src += res - remainingCharsCount;
}
}
}
// main body, stateless decoding
res = 0;
const uchar *nextAscii = src;
const uchar *start = src;
while (res >= 0 && src < end) {
if (src >= nextAscii && simdDecodeAscii(dst, nextAscii, src, end))
break;
ch = *src++;
res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(ch, dst, src, end);
if (!headerdone && res >= 0) {
headerdone = true;
if (src == start + 3) { // 3 == sizeof(utf8-bom)
// eat the UTF-8 BOM (it can only appear at the beginning of the string).
if (dst[-1] == 0xfeff)
--dst;
}
}
if (res == QUtf8BaseTraits::Error) {
res = 0;
++invalid;
*dst++ = replacement;
}
}
if (!state && res == QUtf8BaseTraits::EndOfString) {
// unterminated UTF sequence
*dst++ = QChar::ReplacementCharacter;
while (src++ < end)
*dst++ = QChar::ReplacementCharacter;
}
result.truncate(dst - (const ushort *)result.unicode());
if (state) {
state->invalidChars += invalid;
if (headerdone)
state->flags |= QTextCodec::IgnoreHeader;
if (res == QUtf8BaseTraits::EndOfString) {
--src; // unread the byte in ch
state->remainingChars = end - src;
memcpy(&state->state_data[0], src, end - src);
} else {
state->remainingChars = 0;
}
}
return result;
}
struct QUtf8NoOutputTraits : public QUtf8BaseTraitsNoAscii
{
struct NoOutput {};
static void appendUtf16(const NoOutput &, ushort) {}
static void appendUcs4(const NoOutput &, uint) {}
};
QUtf8::ValidUtf8Result QUtf8::isValidUtf8(const char *chars, qsizetype len)
{
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
const uchar *nextAscii = src;
bool isValidAscii = true;
while (src < end) {
if (src >= nextAscii)
src = simdFindNonAscii(src, end, nextAscii);
if (src == end)
break;
do {
uchar b = *src++;
if ((b & 0x80) == 0)
continue;
isValidAscii = false;
QUtf8NoOutputTraits::NoOutput output;
int res = QUtf8Functions::fromUtf8<QUtf8NoOutputTraits>(b, output, src, end);
if (res < 0) {
// decoding error
return { false, false };
}
} while (src < nextAscii);
}
return { true, isValidAscii };
}
int QUtf8::compareUtf8(const char *utf8, qsizetype u8len, const QChar *utf16, int u16len)
{
uint uc1, uc2;
auto src1 = reinterpret_cast<const uchar *>(utf8);
auto end1 = src1 + u8len;
QStringIterator src2(utf16, utf16 + u16len);
while (src1 < end1 && src2.hasNext()) {
uchar b = *src1++;
uint *output = &uc1;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, output, src1, end1);
if (res < 0) {
// decoding error
uc1 = QChar::ReplacementCharacter;
}
uc2 = src2.next();
if (uc1 != uc2)
return int(uc1) - int(uc2);
}
// the shorter string sorts first
return (end1 > src1) - int(src2.hasNext());
}
int QUtf8::compareUtf8(const char *utf8, qsizetype u8len, QLatin1String s)
{
uint uc1;
auto src1 = reinterpret_cast<const uchar *>(utf8);
auto end1 = src1 + u8len;
auto src2 = reinterpret_cast<const uchar *>(s.latin1());
auto end2 = src2 + s.size();
while (src1 < end1 && src2 < end2) {
uchar b = *src1++;
uint *output = &uc1;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, output, src1, end1);
if (res < 0) {
// decoding error
uc1 = QChar::ReplacementCharacter;
}
uint uc2 = *src2++;
if (uc1 != uc2)
return int(uc1) - int(uc2);
}
// the shorter string sorts first
return (end1 > src1) - (end2 > src2);
}
QByteArray QUtf16::convertFromUnicode(const QChar *uc, int len, QTextCodec::ConverterState *state, DataEndianness e)
{
DataEndianness endian = e;
int length = 2*len;
if (!state || (!(state->flags & QTextCodec::IgnoreHeader))) {
length += 2;
}
if (e == DetectEndianness) {
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
}
QByteArray d;
d.resize(length);
char *data = d.data();
if (!state || !(state->flags & QTextCodec::IgnoreHeader)) {
QChar bom(QChar::ByteOrderMark);
if (endian == BigEndianness)
qToBigEndian(bom.unicode(), data);
else
qToLittleEndian(bom.unicode(), data);
data += 2;
}
if (endian == BigEndianness)
qToBigEndian<ushort>(uc, len, data);
else
qToLittleEndian<ushort>(uc, len, data);
if (state) {
state->remainingChars = 0;
state->flags |= QTextCodec::IgnoreHeader;
}
return d;
}
QString QUtf16::convertToUnicode(const char *chars, int len, QTextCodec::ConverterState *state, DataEndianness e)
{
DataEndianness endian = e;
bool half = false;
uchar buf = 0;
bool headerdone = false;
if (state) {
headerdone = state->flags & QTextCodec::IgnoreHeader;
if (endian == DetectEndianness)
endian = (DataEndianness)state->state_data[Endian];
if (state->remainingChars) {
half = true;
buf = state->state_data[Data];
}
}
if (headerdone && endian == DetectEndianness)
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
QString result(len, Qt::Uninitialized); // worst case
QChar *qch = (QChar *)result.data();
while (len--) {
if (half) {
QChar ch;
if (endian == LittleEndianness) {
ch.setRow(*chars++);
ch.setCell(buf);
} else {
ch.setRow(buf);
ch.setCell(*chars++);
}
if (!headerdone) {
headerdone = true;
if (endian == DetectEndianness) {
if (ch == QChar::ByteOrderSwapped) {
endian = LittleEndianness;
} else if (ch == QChar::ByteOrderMark) {
endian = BigEndianness;
} else {
if (QSysInfo::ByteOrder == QSysInfo::BigEndian) {
endian = BigEndianness;
} else {
endian = LittleEndianness;
ch = QChar::fromUcs2((ch.unicode() >> 8) | ((ch.unicode() & 0xff) << 8));
}
*qch++ = ch;
}
} else if (ch != QChar::ByteOrderMark) {
*qch++ = ch;
}
} else {
*qch++ = ch;
}
half = false;
} else {
buf = *chars++;
half = true;
}
}
result.truncate(qch - result.unicode());
if (state) {
if (headerdone)
state->flags |= QTextCodec::IgnoreHeader;
state->state_data[Endian] = endian;
if (half) {
state->remainingChars = 1;
state->state_data[Data] = buf;
} else {
state->remainingChars = 0;
state->state_data[Data] = 0;
}
}
return result;
}
QByteArray QUtf32::convertFromUnicode(const QChar *uc, int len, QTextCodec::ConverterState *state, DataEndianness e)
{
DataEndianness endian = e;
int length = 4*len;
if (!state || (!(state->flags & QTextCodec::IgnoreHeader))) {
length += 4;
}
if (e == DetectEndianness) {
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
}
QByteArray d(length, Qt::Uninitialized);
char *data = d.data();
if (!state || !(state->flags & QTextCodec::IgnoreHeader)) {
if (endian == BigEndianness) {
data[0] = 0;
data[1] = 0;
data[2] = (char)0xfe;
data[3] = (char)0xff;
} else {
data[0] = (char)0xff;
data[1] = (char)0xfe;
data[2] = 0;
data[3] = 0;
}
data += 4;
}
QStringIterator i(uc, uc + len);
if (endian == BigEndianness) {
while (i.hasNext()) {
uint cp = i.next();
qToBigEndian(cp, data);
data += 4;
}
} else {
while (i.hasNext()) {
uint cp = i.next();
qToLittleEndian(cp, data);
data += 4;
}
}
if (state) {
state->remainingChars = 0;
state->flags |= QTextCodec::IgnoreHeader;
}
return d;
}
QString QUtf32::convertToUnicode(const char *chars, int len, QTextCodec::ConverterState *state, DataEndianness e)
{
DataEndianness endian = e;
uchar tuple[4];
int num = 0;
bool headerdone = false;
if (state) {
headerdone = state->flags & QTextCodec::IgnoreHeader;
if (endian == DetectEndianness) {
endian = (DataEndianness)state->state_data[Endian];
}
num = state->remainingChars;
memcpy(tuple, &state->state_data[Data], 4);
}
if (headerdone && endian == DetectEndianness)
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
QString result;
result.resize((num + len) >> 2 << 1); // worst case
QChar *qch = (QChar *)result.data();
const char *end = chars + len;
while (chars < end) {
tuple[num++] = *chars++;
if (num == 4) {
if (!headerdone) {
headerdone = true;
if (endian == DetectEndianness) {
if (tuple[0] == 0xff && tuple[1] == 0xfe && tuple[2] == 0 && tuple[3] == 0 && endian != BigEndianness) {
endian = LittleEndianness;
num = 0;
continue;
} else if (tuple[0] == 0 && tuple[1] == 0 && tuple[2] == 0xfe && tuple[3] == 0xff && endian != LittleEndianness) {
endian = BigEndianness;
num = 0;
continue;
} else if (QSysInfo::ByteOrder == QSysInfo::BigEndian) {
endian = BigEndianness;
} else {
endian = LittleEndianness;
}
} else if (((endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple)) == QChar::ByteOrderMark) {
num = 0;
continue;
}
}
uint code = (endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple);
for (char16_t c : QChar::fromUcs4(code))
*qch++ = c;
num = 0;
}
}
result.truncate(qch - result.unicode());
if (state) {
if (headerdone)
state->flags |= QTextCodec::IgnoreHeader;
state->state_data[Endian] = endian;
state->remainingChars = num;
memcpy(&state->state_data[Data], tuple, 4);
}
return result;
}
QString qFromUtfEncoded(const QByteArray &ba)
{
const int arraySize = ba.size();
const uchar *buf = reinterpret_cast<const uchar *>(ba.constData());
const uint bom = 0xfeff;
if (arraySize > 3) {
uint uc = qFromUnaligned<uint>(buf);
if (uc == qToBigEndian(bom) || uc == qToLittleEndian(bom))
return QUtf32::convertToUnicode(ba.constData(), ba.length(), nullptr); // utf-32
}
if (arraySize > 1) {
ushort uc = qFromUnaligned<ushort>(buf);
if (uc == qToBigEndian(ushort(bom)) || qToLittleEndian(ushort(bom)))
return QUtf16::convertToUnicode(ba.constData(), ba.length(), nullptr); // utf-16
}
return QUtf8::convertToUnicode(ba.constData(), ba.length());
}
#if QT_CONFIG(textcodec)
QUtf8Codec::~QUtf8Codec()

262
src/corelib/codecs/qutfcodec_p.h
View File

@ -60,271 +60,11 @@
#include "QtCore/qtextcodec.h"
#endif
#include "private/qstringconverter_p.h"
#include "private/qtextcodec_p.h"
QT_BEGIN_NAMESPACE
struct QUtf8BaseTraits
{
static const bool isTrusted = false;
static const bool allowNonCharacters = true;
static const bool skipAsciiHandling = false;
static const int Error = -1;
static const int EndOfString = -2;
static bool isValidCharacter(uint u)
{ return int(u) >= 0; }
static void appendByte(uchar *&ptr, uchar b)
{ *ptr++ = b; }
static uchar peekByte(const uchar *ptr, int n = 0)
{ return ptr[n]; }
static qptrdiff availableBytes(const uchar *ptr, const uchar *end)
{ return end - ptr; }
static void advanceByte(const uchar *&ptr, int n = 1)
{ ptr += n; }
static void appendUtf16(ushort *&ptr, ushort uc)
{ *ptr++ = uc; }
static void appendUcs4(ushort *&ptr, uint uc)
{
appendUtf16(ptr, QChar::highSurrogate(uc));
appendUtf16(ptr, QChar::lowSurrogate(uc));
}
static ushort peekUtf16(const ushort *ptr, int n = 0)
{ return ptr[n]; }
static qptrdiff availableUtf16(const ushort *ptr, const ushort *end)
{ return end - ptr; }
static void advanceUtf16(const ushort *&ptr, int n = 1)
{ ptr += n; }
// it's possible to output to UCS-4 too
static void appendUtf16(uint *&ptr, ushort uc)
{ *ptr++ = uc; }
static void appendUcs4(uint *&ptr, uint uc)
{ *ptr++ = uc; }
};
struct QUtf8BaseTraitsNoAscii : public QUtf8BaseTraits
{
static const bool skipAsciiHandling = true;
};
namespace QUtf8Functions
{
/// returns 0 on success; errors can only happen if \a u is a surrogate:
/// Error if \a u is a low surrogate;
/// if \a u is a high surrogate, Error if the next isn't a low one,
/// EndOfString if we run into the end of the string.
template <typename Traits, typename OutputPtr, typename InputPtr> inline
int toUtf8(ushort u, OutputPtr &dst, InputPtr &src, InputPtr end)
{
if (!Traits::skipAsciiHandling && u < 0x80) {
// U+0000 to U+007F (US-ASCII) - one byte
Traits::appendByte(dst, uchar(u));
return 0;
} else if (u < 0x0800) {
// U+0080 to U+07FF - two bytes
// first of two bytes
Traits::appendByte(dst, 0xc0 | uchar(u >> 6));
} else {
if (!QChar::isSurrogate(u)) {
// U+0800 to U+FFFF (except U+D800-U+DFFF) - three bytes
if (!Traits::allowNonCharacters && QChar::isNonCharacter(u))
return Traits::Error;
// first of three bytes
Traits::appendByte(dst, 0xe0 | uchar(u >> 12));
} else {
// U+10000 to U+10FFFF - four bytes
// need to get one extra codepoint
if (Traits::availableUtf16(src, end) == 0)
return Traits::EndOfString;
ushort low = Traits::peekUtf16(src);
if (!QChar::isHighSurrogate(u))
return Traits::Error;
if (!QChar::isLowSurrogate(low))
return Traits::Error;
Traits::advanceUtf16(src);
uint ucs4 = QChar::surrogateToUcs4(u, low);
if (!Traits::allowNonCharacters && QChar::isNonCharacter(ucs4))
return Traits::Error;
// first byte
Traits::appendByte(dst, 0xf0 | (uchar(ucs4 >> 18) & 0xf));
// second of four bytes
Traits::appendByte(dst, 0x80 | (uchar(ucs4 >> 12) & 0x3f));
// for the rest of the bytes
u = ushort(ucs4);
}
// second to last byte
Traits::appendByte(dst, 0x80 | (uchar(u >> 6) & 0x3f));
}
// last byte
Traits::appendByte(dst, 0x80 | (u & 0x3f));
return 0;
}
inline bool isContinuationByte(uchar b)
{
return (b & 0xc0) == 0x80;
}
/// returns the number of characters consumed (including \a b) in case of success;
/// returns negative in case of error: Traits::Error or Traits::EndOfString
template <typename Traits, typename OutputPtr, typename InputPtr> inline
int fromUtf8(uchar b, OutputPtr &dst, InputPtr &src, InputPtr end)
{
int charsNeeded;
uint min_uc;
uint uc;
if (!Traits::skipAsciiHandling && b < 0x80) {
// US-ASCII
Traits::appendUtf16(dst, b);
return 1;
}
if (!Traits::isTrusted && Q_UNLIKELY(b <= 0xC1)) {
// an UTF-8 first character must be at least 0xC0
// however, all 0xC0 and 0xC1 first bytes can only produce overlong sequences
return Traits::Error;
} else if (b < 0xe0) {
charsNeeded = 2;
min_uc = 0x80;
uc = b & 0x1f;
} else if (b < 0xf0) {
charsNeeded = 3;
min_uc = 0x800;
uc = b & 0x0f;
} else if (b < 0xf5) {
charsNeeded = 4;
min_uc = 0x10000;
uc = b & 0x07;
} else {
// the last Unicode character is U+10FFFF
// it's encoded in UTF-8 as "\xF4\x8F\xBF\xBF"
// therefore, a byte higher than 0xF4 is not the UTF-8 first byte
return Traits::Error;
}
int bytesAvailable = Traits::availableBytes(src, end);
if (Q_UNLIKELY(bytesAvailable < charsNeeded - 1)) {
// it's possible that we have an error instead of just unfinished bytes
if (bytesAvailable > 0 && !isContinuationByte(Traits::peekByte(src, 0)))
return Traits::Error;
if (bytesAvailable > 1 && !isContinuationByte(Traits::peekByte(src, 1)))
return Traits::Error;
return Traits::EndOfString;
}
// first continuation character
b = Traits::peekByte(src, 0);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
if (charsNeeded > 2) {
// second continuation character
b = Traits::peekByte(src, 1);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
if (charsNeeded > 3) {
// third continuation character
b = Traits::peekByte(src, 2);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
}
}
// we've decoded something; safety-check it
if (!Traits::isTrusted) {
if (uc < min_uc)
return Traits::Error;
if (QChar::isSurrogate(uc) || uc > QChar::LastValidCodePoint)
return Traits::Error;
if (!Traits::allowNonCharacters && QChar::isNonCharacter(uc))
return Traits::Error;
}
// write the UTF-16 sequence
if (!QChar::requiresSurrogates(uc)) {
// UTF-8 decoded and no surrogates are required
// detach if necessary
Traits::appendUtf16(dst, ushort(uc));
} else {
// UTF-8 decoded to something that requires a surrogate pair
Traits::appendUcs4(dst, uc);
}
Traits::advanceByte(src, charsNeeded - 1);
return charsNeeded;
}
}
enum DataEndianness
{
DetectEndianness,
BigEndianness,
LittleEndianness
};
struct QUtf8
{
static QChar *convertToUnicode(QChar *, const char *, int) noexcept;
static QString convertToUnicode(const char *, int);
static QString convertToUnicode(const char *, int, QTextCodec::ConverterState *);
static QByteArray convertFromUnicode(const QChar *, int);
static QByteArray convertFromUnicode(const QChar *, int, QTextCodec::ConverterState *);
struct ValidUtf8Result {
bool isValidUtf8;
bool isValidAscii;
};
static ValidUtf8Result isValidUtf8(const char *, qsizetype);
static int compareUtf8(const char *, qsizetype, const QChar *, int);
static int compareUtf8(const char *, qsizetype, QLatin1String s);
};
struct QUtf16
{
static QString convertToUnicode(const char *, int, QTextCodec::ConverterState *, DataEndianness = DetectEndianness);
static QByteArray convertFromUnicode(const QChar *, int, QTextCodec::ConverterState *, DataEndianness = DetectEndianness);
};
struct QUtf32
{
static QString convertToUnicode(const char *, int, QTextCodec::ConverterState *, DataEndianness = DetectEndianness);
static QByteArray convertFromUnicode(const QChar *, int, QTextCodec::ConverterState *, DataEndianness = DetectEndianness);
};
/*
Converts from different utf encodings looking at a possible byte order mark at the
beginning of the string. If no BOM exists, utf-8 is assumed.
*/
QString Q_CORE_EXPORT qFromUtfEncoded(const QByteArray &ba);
#if QT_CONFIG(textcodec)
class QUtf8Codec : public QTextCodec {

5
src/corelib/global/qconfig-bootstrapped.h
View File

@ -141,18 +141,17 @@
#define QT_FEATURE_zstd -1
#endif
#define QT_FEATURE_textcodec -1
#ifdef QT_BUILD_QMAKE
#define QT_FEATURE_commandlineparser -1
#define QT_NO_COMPRESS
#define QT_JSON_READONLY
#define QT_FEATURE_settings 1
#define QT_NO_STANDARDPATHS
#define QT_FEATURE_textcodec -1
#else
#define QT_FEATURE_codecs -1
#define QT_FEATURE_commandlineparser 1
#define QT_FEATURE_settings -1
#define QT_FEATURE_textcodec 1
#endif
#endif // QT_BOOTSTRAPPED

2
src/corelib/io/qfilesystemiterator_unix.cpp
View File

@ -42,7 +42,7 @@
#if QT_CONFIG(textcodec)
# include <qtextcodec.h>
# include <private/qutfcodec_p.h>
# include <private/qstringconverter_p.h>
#endif
#ifndef QT_NO_FILESYSTEMITERATOR

2
src/corelib/io/qurlrecode.cpp
View File

@ -38,7 +38,7 @@
****************************************************************************/
#include "qurl.h"
#include "private/qutfcodec_p.h"
#include "private/qstringconverter_p.h"
#include "private/qtools_p.h"
#include "private/qsimd_p.h"

2
src/corelib/serialization/qcborstreamreader.cpp
View File

@ -44,7 +44,7 @@
#include <private/qbytearray_p.h>
#include <private/qnumeric_p.h>
#include <private/qutfcodec_p.h>
#include <private/qstringconverter_p.h>
#include <qdebug.h>
#include <qstack.h>

2
src/corelib/serialization/qcborvalue_p.h
View File

@ -54,7 +54,7 @@
#include "qcborvalue.h"
#include <private/qglobal_p.h>
#include <private/qutfcodec_p.h>
#include <private/qstringconverter_p.h>
#include <math.h>

2
src/corelib/serialization/qjsonparser.cpp
View File

@ -44,7 +44,7 @@
#include <qdebug.h>
#include "qjsonparser_p.h"
#include "qjson_p.h"
#include "private/qutfcodec_p.h"
#include "private/qstringconverter_p.h"
#include "private/qcborvalue_p.h"
#include "private/qnumeric_p.h"

2
src/corelib/serialization/qjsonwriter.cpp
View File

@ -42,7 +42,7 @@
#include <qlocale.h>
#include "qjsonwriter_p.h"
#include "qjson_p.h"
#include "private/qutfcodec_p.h"
#include "private/qstringconverter_p.h"
#include <private/qnumeric_p.h>
#include <private/qcborvalue_p.h>

2
src/corelib/text/qstring.cpp
View File

@ -48,7 +48,7 @@
#if QT_CONFIG(textcodec)
#include <qtextcodec.h>
#endif
#include <private/qutfcodec_p.h>
#include <private/qstringconverter_p.h>
#include "qlocale_tools_p.h"
#include "private/qsimd_p.h"
#include <qnumeric.h>

2
src/corelib/text/qstringbuilder.cpp
View File

@ -38,7 +38,7 @@
****************************************************************************/
#include "qstringbuilder.h"
#include <private/qutfcodec_p.h>
#include <private/qstringconverter_p.h>
QT_BEGIN_NAMESPACE

949
src/corelib/text/qstringconverter.cpp
View File

@ -39,10 +39,954 @@
****************************************************************************/
#include <qstringconverter.h>
#include <private/qutfcodec_p.h>
#include <private/qstringconverter_p.h>
#include "qendian.h"
#include "private/qsimd_p.h"
#include "private/qstringiterator_p.h"
QT_BEGIN_NAMESPACE
enum { Endian = 0, Data = 1 };
static const uchar utf8bom[] = { 0xef, 0xbb, 0xbf };
#if (defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2)) \
|| (defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64))
static Q_ALWAYS_INLINE uint qBitScanReverse(unsigned v) noexcept
{
uint result = qCountLeadingZeroBits(v);
// Now Invert the result: clz will count *down* from the msb to the lsb, so the msb index is 31
// and the lsb index is 0. The result for _bit_scan_reverse is expected to be the index when
// counting up: msb index is 0 (because it starts there), and the lsb index is 31.
result ^= sizeof(unsigned) * 8 - 1;
return result;
}
#endif
#if defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2)
static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end)
{
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
# ifdef __AVX2__
__m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src));
__m128i data1 = _mm256_castsi256_si128(data);
__m128i data2 = _mm256_extracti128_si256(data, 1);
# else
__m128i data1 = _mm_loadu_si128((const __m128i*)src);
__m128i data2 = _mm_loadu_si128(1+(const __m128i*)src);
# endif
// check if everything is ASCII
// the highest ASCII value is U+007F
// Do the packing directly:
// The PACKUSWB instruction has packs a signed 16-bit integer to an unsigned 8-bit
// with saturation. That is, anything from 0x0100 to 0x7fff is saturated to 0xff,
// while all negatives (0x8000 to 0xffff) get saturated to 0x00. To detect non-ASCII,
// we simply do a signed greater-than comparison to 0x00. That means we detect NULs as
// "non-ASCII", but it's an acceptable compromise.
__m128i packed = _mm_packus_epi16(data1, data2);
__m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128());
// store, even if there are non-ASCII characters here
_mm_storeu_si128((__m128i*)dst, packed);
// n will contain 1 bit set per character in [data1, data2] that is non-ASCII (or NUL)
ushort n = ~_mm_movemask_epi8(nonAscii);
if (n) {
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
n = qCountTrailingZeroBits(n);
dst += n;
src += n;
return false;
}
}
if (end - src >= 8) {
// do eight characters at a time
__m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src));
__m128i packed = _mm_packus_epi16(data, data);
__m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128());
// store even non-ASCII
_mm_storel_epi64(reinterpret_cast<__m128i *>(dst), packed);
uchar n = ~_mm_movemask_epi8(nonAscii);
if (n) {
nextAscii = src + qBitScanReverse(n) + 1;
n = qCountTrailingZeroBits(n);
dst += n;
src += n;
return false;
}
}
return src == end;
}
static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end)
{
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
__m128i data = _mm_loadu_si128((const __m128i*)src);
#ifdef __AVX2__
const int BitSpacing = 2;
// load and zero extend to an YMM register
const __m256i extended = _mm256_cvtepu8_epi16(data);
uint n = _mm256_movemask_epi8(extended);
if (!n) {
// store
_mm256_storeu_si256((__m256i*)dst, extended);
continue;
}
#else
const int BitSpacing = 1;
// check if everything is ASCII
// movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII
uint n = _mm_movemask_epi8(data);
if (!n) {
// unpack
_mm_storeu_si128((__m128i*)dst, _mm_unpacklo_epi8(data, _mm_setzero_si128()));
_mm_storeu_si128(1+(__m128i*)dst, _mm_unpackhi_epi8(data, _mm_setzero_si128()));
continue;
}
#endif
// copy the front part that is still ASCII
while (!(n & 1)) {
*dst++ = *src++;
n >>= BitSpacing;
}
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
n = qBitScanReverse(n);
nextAscii = src + (n / BitSpacing) + 1;
return false;
}
if (end - src >= 8) {
__m128i data = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src));
uint n = _mm_movemask_epi8(data) & 0xff;
if (!n) {
// unpack and store
_mm_storeu_si128(reinterpret_cast<__m128i *>(dst), _mm_unpacklo_epi8(data, _mm_setzero_si128()));
} else {
while (!(n & 1)) {
*dst++ = *src++;
n >>= 1;
}
n = qBitScanReverse(n);
nextAscii = src + n + 1;
return false;
}
}
return src == end;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
#ifdef __AVX2__
// do 32 characters at a time
// (this is similar to simdTestMask in qstring.cpp)
const __m256i mask = _mm256_set1_epi8(0x80);
for ( ; end - src >= 32; src += 32) {
__m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src));
if (_mm256_testz_si256(mask, data))
continue;
uint n = _mm256_movemask_epi8(data);
Q_ASSUME(n);
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
#endif
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16) {
__m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i*>(src));
// check if everything is ASCII
// movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII
uint n = _mm_movemask_epi8(data);
if (!n)
continue;
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
// do four characters at a time
for ( ; end - src >= 4; src += 4) {
quint32 data = qFromUnaligned<quint32>(src);
data &= 0x80808080U;
if (!data)
continue;
// We don't try to guess which of the three bytes is ASCII and which
// one isn't. The chance that at least two of them are non-ASCII is
// better than 75%.
nextAscii = src;
return src;
}
nextAscii = end;
return src;
}
#elif defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64) // vaddv is only available on Aarch64
static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end)
{
uint16x8_t maxAscii = vdupq_n_u16(0x7f);
uint16x8_t mask1 = { 1, 1 << 2, 1 << 4, 1 << 6, 1 << 8, 1 << 10, 1 << 12, 1 << 14 };
uint16x8_t mask2 = vshlq_n_u16(mask1, 1);
// do sixteen characters at a time
for ( ; end - src >= 16; src += 16, dst += 16) {
// load 2 lanes (or: "load interleaved")
uint16x8x2_t in = vld2q_u16(src);
// check if any of the elements > 0x7f, select 1 bit per element (element 0 -> bit 0, element 1 -> bit 1, etc),
// add those together into a scalar, and merge the scalars.
uint16_t nonAscii = vaddvq_u16(vandq_u16(vcgtq_u16(in.val[0], maxAscii), mask1))
| vaddvq_u16(vandq_u16(vcgtq_u16(in.val[1], maxAscii), mask2));
// merge the two lanes by shifting the values of the second by 8 and inserting them
uint16x8_t out = vsliq_n_u16(in.val[0], in.val[1], 8);
// store, even if there are non-ASCII characters here
vst1q_u8(dst, vreinterpretq_u8_u16(out));
if (nonAscii) {
// find the next probable ASCII character
// we don't want to load 32 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(nonAscii) + 1;
nonAscii = qCountTrailingZeroBits(nonAscii);
dst += nonAscii;
src += nonAscii;
return false;
}
}
return src == end;
}
static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end)
{
// do eight characters at a time
uint8x8_t msb_mask = vdup_n_u8(0x80);
uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 };
for ( ; end - src >= 8; src += 8, dst += 8) {
uint8x8_t c = vld1_u8(src);
uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask));
if (!n) {
// store
vst1q_u16(dst, vmovl_u8(c));
continue;
}
// copy the front part that is still ASCII
while (!(n & 1)) {
*dst++ = *src++;
n >>= 1;
}
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
n = qBitScanReverse(n);
nextAscii = src + n + 1;
return false;
}
return src == end;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
// The SIMD code below is untested, so just force an early return until
// we've had the time to verify it works.
nextAscii = end;
return src;
// do eight characters at a time
uint8x8_t msb_mask = vdup_n_u8(0x80);
uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 };
for ( ; end - src >= 8; src += 8) {
uint8x8_t c = vld1_u8(src);
uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask));
if (!n)
continue;
// find the next probable ASCII character
// we don't want to load 16 bytes again in this loop if we know there are non-ASCII
// characters still coming
nextAscii = src + qBitScanReverse(n) + 1;
// return the non-ASCII character
return src + qCountTrailingZeroBits(n);
}
nextAscii = end;
return src;
}
#else
static inline bool simdEncodeAscii(uchar *, const ushort *, const ushort *, const ushort *)
{
return false;
}
static inline bool simdDecodeAscii(ushort *, const uchar *, const uchar *, const uchar *)
{
return false;
}
static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii)
{
nextAscii = end;
return src;
}
#endif
QByteArray QUtf8::convertFromUnicode(const QChar *uc, qsizetype len)
{
// create a QByteArray with the worst case scenario size
QByteArray result(len * 3, Qt::Uninitialized);
uchar *dst = reinterpret_cast<uchar *>(const_cast<char *>(result.constData()));
const ushort *src = reinterpret_cast<const ushort *>(uc);
const ushort *const end = src + len;
while (src != end) {
const ushort *nextAscii = end;
if (simdEncodeAscii(dst, nextAscii, src, end))
break;
do {
ushort uc = *src++;
int res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, dst, src, end);
if (res < 0) {
// encoding error - append '?'
*dst++ = '?';
}
} while (src < nextAscii);
}
result.truncate(dst - reinterpret_cast<uchar *>(const_cast<char *>(result.constData())));
return result;
}
QByteArray QUtf8::convertFromUnicode(const QChar *uc, qsizetype len, QStringConverter::State *state)
{
uchar replacement = '?';
qsizetype rlen = 3*len;
int surrogate_high = -1;
if (state) {
if (state->flags & QStringConverter::ConvertInvalidToNull)
replacement = 0;
if (!(state->flags & QStringConverter::IgnoreHeader))
rlen += 3;
if (state->remainingChars)
surrogate_high = state->state_data[0];
}
QByteArray rstr(rlen, Qt::Uninitialized);
uchar *cursor = reinterpret_cast<uchar *>(const_cast<char *>(rstr.constData()));
const ushort *src = reinterpret_cast<const ushort *>(uc);
const ushort *const end = src + len;
int invalid = 0;
if (state && !(state->flags & QStringConverter::IgnoreHeader)) {
// append UTF-8 BOM
*cursor++ = utf8bom[0];
*cursor++ = utf8bom[1];
*cursor++ = utf8bom[2];
}
const ushort *nextAscii = src;
while (src != end) {
int res;
ushort uc;
if (surrogate_high != -1) {
uc = surrogate_high;
surrogate_high = -1;
res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, cursor, src, end);
} else {
if (src >= nextAscii && simdEncodeAscii(cursor, nextAscii, src, end))
break;
uc = *src++;
res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, cursor, src, end);
}
if (Q_LIKELY(res >= 0))
continue;
if (res == QUtf8BaseTraits::Error) {
// encoding error
++invalid;
*cursor++ = replacement;
} else if (res == QUtf8BaseTraits::EndOfString) {
surrogate_high = uc;
break;
}
}
rstr.resize(cursor - (const uchar*)rstr.constData());
if (state) {
state->invalidChars += invalid;
state->flags |= QStringConverter::IgnoreHeader;
state->remainingChars = 0;
if (surrogate_high >= 0) {
state->remainingChars = 1;
state->state_data[0] = surrogate_high;
}
}
return rstr;
}
QString QUtf8::convertToUnicode(const char *chars, qsizetype len)
{
// UTF-8 to UTF-16 always needs the exact same number of words or less:
// UTF-8 UTF-16
// 1 byte 1 word
// 2 bytes 1 word
// 3 bytes 1 word
// 4 bytes 2 words (one surrogate pair)
// That is, we'll use the full buffer if the input is US-ASCII (1-byte UTF-8),
// half the buffer for U+0080-U+07FF text (e.g., Greek, Cyrillic, Arabic) or
// non-BMP text, and one third of the buffer for U+0800-U+FFFF text (e.g, CJK).
//
// The table holds for invalid sequences too: we'll insert one replacement char
// per invalid byte.
QString result(len, Qt::Uninitialized);
QChar *data = const_cast<QChar*>(result.constData()); // we know we're not shared
const QChar *end = convertToUnicode(data, chars, len);
result.truncate(end - data);
return result;
}
/*!
\since 5.7
\overload
Converts the UTF-8 sequence of \a len octets beginning at \a chars to
a sequence of QChar starting at \a buffer. The buffer is expected to be
large enough to hold the result. An upper bound for the size of the
buffer is \a len QChars.
If, during decoding, an error occurs, a QChar::ReplacementCharacter is
written.
Returns a pointer to one past the last QChar written.
This function never throws.
*/
QChar *QUtf8::convertToUnicode(QChar *buffer, const char *chars, qsizetype len) noexcept
{
ushort *dst = reinterpret_cast<ushort *>(buffer);
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
// attempt to do a full decoding in SIMD
const uchar *nextAscii = end;
if (!simdDecodeAscii(dst, nextAscii, src, end)) {
// at least one non-ASCII entry
// check if we failed to decode the UTF-8 BOM; if so, skip it
if (Q_UNLIKELY(src == reinterpret_cast<const uchar *>(chars))
&& end - src >= 3
&& Q_UNLIKELY(src[0] == utf8bom[0] && src[1] == utf8bom[1] && src[2] == utf8bom[2])) {
src += 3;
}
while (src < end) {
nextAscii = end;
if (simdDecodeAscii(dst, nextAscii, src, end))
break;
do {
uchar b = *src++;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, dst, src, end);
if (res < 0) {
// decoding error
*dst++ = QChar::ReplacementCharacter;
}
} while (src < nextAscii);
}
}
return reinterpret_cast<QChar *>(dst);
}
QString QUtf8::convertToUnicode(const char *chars, qsizetype len, QStringConverter::State *state)
{
bool headerdone = false;
ushort replacement = QChar::ReplacementCharacter;
int invalid = 0;
int res;
uchar ch = 0;
// See above for buffer requirements for stateless decoding. However, that
// fails if the state is not empty. The following situations can add to the
// requirements:
// state contains chars starts with requirement
// 1 of 2 bytes valid continuation 0
// 2 of 3 bytes same 0
// 3 bytes of 4 same +1 (need to insert surrogate pair)
// 1 of 2 bytes invalid continuation +1 (need to insert replacement and restart)
// 2 of 3 bytes same +1 (same)
// 3 of 4 bytes same +1 (same)
QString result(len + 1, Qt::Uninitialized);
ushort *dst = reinterpret_cast<ushort *>(const_cast<QChar *>(result.constData()));
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
if (state) {
if (state->flags & QStringConverter::IgnoreHeader)
headerdone = true;
if (state->flags & QStringConverter::ConvertInvalidToNull)
replacement = QChar::Null;
if (state->remainingChars) {
// handle incoming state first
uchar remainingCharsData[4]; // longest UTF-8 sequence possible
qsizetype remainingCharsCount = state->remainingChars;
qsizetype newCharsToCopy = qMin<int>(sizeof(remainingCharsData) - remainingCharsCount, end - src);
memset(remainingCharsData, 0, sizeof(remainingCharsData));
memcpy(remainingCharsData, &state->state_data[0], remainingCharsCount);
memcpy(remainingCharsData + remainingCharsCount, src, newCharsToCopy);
const uchar *begin = &remainingCharsData[1];
res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(remainingCharsData[0], dst, begin,
static_cast<const uchar *>(remainingCharsData) + remainingCharsCount + newCharsToCopy);
if (res == QUtf8BaseTraits::Error || (res == QUtf8BaseTraits::EndOfString && len == 0)) {
// special case for len == 0:
// if we were supplied an empty string, terminate the previous, unfinished sequence with error
++invalid;
*dst++ = replacement;
} else if (res == QUtf8BaseTraits::EndOfString) {
// if we got EndOfString again, then there were too few bytes in src;
// copy to our state and return
state->remainingChars = remainingCharsCount + newCharsToCopy;
memcpy(&state->state_data[0], remainingCharsData, state->remainingChars);
return QString();
} else if (!headerdone && res >= 0) {
// eat the UTF-8 BOM
headerdone = true;
if (dst[-1] == 0xfeff)
--dst;
}
// adjust src now that we have maybe consumed a few chars
if (res >= 0) {
Q_ASSERT(res > remainingCharsCount);
src += res - remainingCharsCount;
}
}
}
// main body, stateless decoding
res = 0;
const uchar *nextAscii = src;
const uchar *start = src;
while (res >= 0 && src < end) {
if (src >= nextAscii && simdDecodeAscii(dst, nextAscii, src, end))
break;
ch = *src++;
res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(ch, dst, src, end);
if (!headerdone && res >= 0) {
headerdone = true;
if (src == start + 3) { // 3 == sizeof(utf8-bom)
// eat the UTF-8 BOM (it can only appear at the beginning of the string).
if (dst[-1] == 0xfeff)
--dst;
}
}
if (res == QUtf8BaseTraits::Error) {
res = 0;
++invalid;
*dst++ = replacement;
}
}
if (!state && res == QUtf8BaseTraits::EndOfString) {
// unterminated UTF sequence
*dst++ = QChar::ReplacementCharacter;
while (src++ < end)
*dst++ = QChar::ReplacementCharacter;
}
result.truncate(dst - (const ushort *)result.unicode());
if (state) {
state->invalidChars += invalid;
if (headerdone)
state->flags |= QStringConverter::IgnoreHeader;
if (res == QUtf8BaseTraits::EndOfString) {
--src; // unread the byte in ch
state->remainingChars = end - src;
memcpy(&state->state_data[0], src, end - src);
} else {
state->remainingChars = 0;
}
}
return result;
}
struct QUtf8NoOutputTraits : public QUtf8BaseTraitsNoAscii
{
struct NoOutput {};
static void appendUtf16(const NoOutput &, ushort) {}
static void appendUcs4(const NoOutput &, uint) {}
};
QUtf8::ValidUtf8Result QUtf8::isValidUtf8(const char *chars, qsizetype len)
{
const uchar *src = reinterpret_cast<const uchar *>(chars);
const uchar *end = src + len;
const uchar *nextAscii = src;
bool isValidAscii = true;
while (src < end) {
if (src >= nextAscii)
src = simdFindNonAscii(src, end, nextAscii);
if (src == end)
break;
do {
uchar b = *src++;
if ((b & 0x80) == 0)
continue;
isValidAscii = false;
QUtf8NoOutputTraits::NoOutput output;
int res = QUtf8Functions::fromUtf8<QUtf8NoOutputTraits>(b, output, src, end);
if (res < 0) {
// decoding error
return { false, false };
}
} while (src < nextAscii);
}
return { true, isValidAscii };
}
int QUtf8::compareUtf8(const char *utf8, qsizetype u8len, const QChar *utf16, qsizetype u16len)
{
uint uc1, uc2;
auto src1 = reinterpret_cast<const uchar *>(utf8);
auto end1 = src1 + u8len;
QStringIterator src2(utf16, utf16 + u16len);
while (src1 < end1 && src2.hasNext()) {
uchar b = *src1++;
uint *output = &uc1;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, output, src1, end1);
if (res < 0) {
// decoding error
uc1 = QChar::ReplacementCharacter;
}
uc2 = src2.next();
if (uc1 != uc2)
return int(uc1) - int(uc2);
}
// the shorter string sorts first
return (end1 > src1) - int(src2.hasNext());
}
int QUtf8::compareUtf8(const char *utf8, qsizetype u8len, QLatin1String s)
{
uint uc1;
auto src1 = reinterpret_cast<const uchar *>(utf8);
auto end1 = src1 + u8len;
auto src2 = reinterpret_cast<const uchar *>(s.latin1());
auto end2 = src2 + s.size();
while (src1 < end1 && src2 < end2) {
uchar b = *src1++;
uint *output = &uc1;
int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, output, src1, end1);
if (res < 0) {
// decoding error
uc1 = QChar::ReplacementCharacter;
}
uint uc2 = *src2++;
if (uc1 != uc2)
return int(uc1) - int(uc2);
}
// the shorter string sorts first
return (end1 > src1) - (end2 > src2);
}
QByteArray QUtf16::convertFromUnicode(const QChar *uc, qsizetype len, QStringConverter::State *state, DataEndianness e)
{
DataEndianness endian = e;
qsizetype length = 2*len;
if (!state || (!(state->flags & QStringConverter::IgnoreHeader))) {
length += 2;
}
if (e == DetectEndianness) {
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
}
QByteArray d;
d.resize(length);
char *data = d.data();
if (!state || !(state->flags & QStringConverter::IgnoreHeader)) {
QChar bom(QChar::ByteOrderMark);
if (endian == BigEndianness)
qToBigEndian(bom.unicode(), data);
else
qToLittleEndian(bom.unicode(), data);
data += 2;
}
if (endian == BigEndianness)
qToBigEndian<ushort>(uc, len, data);
else
qToLittleEndian<ushort>(uc, len, data);
if (state) {
state->remainingChars = 0;
state->flags |= QStringConverter::IgnoreHeader;
}
return d;
}
QString QUtf16::convertToUnicode(const char *chars, qsizetype len, QStringConverter::State *state, DataEndianness e)
{
DataEndianness endian = e;
bool half = false;
uchar buf = 0;
bool headerdone = false;
if (state) {
headerdone = state->flags & QStringConverter::IgnoreHeader;
if (endian == DetectEndianness)
endian = (DataEndianness)state->state_data[Endian];
if (state->remainingChars) {
half = true;
buf = state->state_data[Data];
}
}
if (headerdone && endian == DetectEndianness)
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
QString result(len, Qt::Uninitialized); // worst case
QChar *qch = (QChar *)result.data();
while (len--) {
if (half) {
QChar ch;
if (endian == LittleEndianness) {
ch.setRow(*chars++);
ch.setCell(buf);
} else {
ch.setRow(buf);
ch.setCell(*chars++);
}
if (!headerdone) {
headerdone = true;
if (endian == DetectEndianness) {
if (ch == QChar::ByteOrderSwapped) {
endian = LittleEndianness;
} else if (ch == QChar::ByteOrderMark) {
endian = BigEndianness;
} else {
if (QSysInfo::ByteOrder == QSysInfo::BigEndian) {
endian = BigEndianness;
} else {
endian = LittleEndianness;
ch = QChar::fromUcs2((ch.unicode() >> 8) | ((ch.unicode() & 0xff) << 8));
}
*qch++ = ch;
}
} else if (ch != QChar::ByteOrderMark) {
*qch++ = ch;
}
} else {
*qch++ = ch;
}
half = false;
} else {
buf = *chars++;
half = true;
}
}
result.truncate(qch - result.unicode());
if (state) {
if (headerdone)
state->flags |= QStringConverter::IgnoreHeader;
state->state_data[Endian] = endian;
if (half) {
state->remainingChars = 1;
state->state_data[Data] = buf;
} else {
state->remainingChars = 0;
state->state_data[Data] = 0;
}
}
return result;
}
QByteArray QUtf32::convertFromUnicode(const QChar *uc, qsizetype len, QStringConverter::State *state, DataEndianness e)
{
DataEndianness endian = e;
qsizetype length = 4*len;
if (!state || (!(state->flags & QStringConverter::IgnoreHeader))) {
length += 4;
}
if (e == DetectEndianness) {
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
}
QByteArray d(length, Qt::Uninitialized);
char *data = d.data();
if (!state || !(state->flags & QStringConverter::IgnoreHeader)) {
if (endian == BigEndianness) {
data[0] = 0;
data[1] = 0;
data[2] = (char)0xfe;
data[3] = (char)0xff;
} else {
data[0] = (char)0xff;
data[1] = (char)0xfe;
data[2] = 0;
data[3] = 0;
}
data += 4;
}
QStringIterator i(uc, uc + len);
if (endian == BigEndianness) {
while (i.hasNext()) {
uint cp = i.next();
qToBigEndian(cp, data);
data += 4;
}
} else {
while (i.hasNext()) {
uint cp = i.next();
qToLittleEndian(cp, data);
data += 4;
}
}
if (state) {
state->remainingChars = 0;
state->flags |= QStringConverter::IgnoreHeader;
}
return d;
}
QString QUtf32::convertToUnicode(const char *chars, qsizetype len, QStringConverter::State *state, DataEndianness e)
{
DataEndianness endian = e;
uchar tuple[4];
int num = 0;
bool headerdone = false;
if (state) {
headerdone = state->flags & QStringConverter::IgnoreHeader;
if (endian == DetectEndianness) {
endian = (DataEndianness)state->state_data[Endian];
}
num = state->remainingChars;
memcpy(tuple, &state->state_data[Data], 4);
}
if (headerdone && endian == DetectEndianness)
endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness;
QString result;
result.resize((num + len) >> 2 << 1); // worst case
QChar *qch = (QChar *)result.data();
const char *end = chars + len;
while (chars < end) {
tuple[num++] = *chars++;
if (num == 4) {
if (!headerdone) {
headerdone = true;
if (endian == DetectEndianness) {
if (tuple[0] == 0xff && tuple[1] == 0xfe && tuple[2] == 0 && tuple[3] == 0 && endian != BigEndianness) {
endian = LittleEndianness;
num = 0;
continue;
} else if (tuple[0] == 0 && tuple[1] == 0 && tuple[2] == 0xfe && tuple[3] == 0xff && endian != LittleEndianness) {
endian = BigEndianness;
num = 0;
continue;
} else if (QSysInfo::ByteOrder == QSysInfo::BigEndian) {
endian = BigEndianness;
} else {
endian = LittleEndianness;
}
} else if (((endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple)) == QChar::ByteOrderMark) {
num = 0;
continue;
}
}
uint code = (endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple);
for (char16_t c : QChar::fromUcs4(code))
*qch++ = c;
num = 0;
}
}
result.truncate(qch - result.unicode());
if (state) {
if (headerdone)
state->flags |= QStringConverter::IgnoreHeader;
state->state_data[Endian] = endian;
state->remainingChars = num;
memcpy(&state->state_data[Data], tuple, 4);
}
return result;
}
QString qFromUtfEncoded(const QByteArray &ba)
{
const qsizetype arraySize = ba.size();
const uchar *buf = reinterpret_cast<const uchar *>(ba.constData());
const uint bom = 0xfeff;
if (arraySize > 3) {
uint uc = qFromUnaligned<uint>(buf);
if (uc == qToBigEndian(bom) || uc == qToLittleEndian(bom))
return QUtf32::convertToUnicode(ba.constData(), ba.length(), nullptr); // utf-32
}
if (arraySize > 1) {
ushort uc = qFromUnaligned<ushort>(buf);
if (uc == qToBigEndian(ushort(bom)) || qToLittleEndian(ushort(bom)))
return QUtf16::convertToUnicode(ba.constData(), ba.length(), nullptr); // utf-16
}
return QUtf8::convertToUnicode(ba.constData(), ba.length());
}
/*!
\enum QStringConverter::Flag
@ -60,7 +1004,8 @@ void QStringConverter::State::clear()
{
if (clearFn)
clearFn(this);
state_data[0] = state_data[1] = state_data[2] = state_data[3] = 0;
else
state_data[0] = state_data[1] = state_data[2] = state_data[3] = 0;
remainingChars = 0;
invalidChars = 0;
}

323
src/corelib/text/qstringconverter_p.h Normal file
View File

@ -0,0 +1,323 @@
/****************************************************************************
**
** Copyright (C) 2018 The Qt Company Ltd.
** Copyright (C) 2018 Intel Corporation.
** Contact: https://www.qt.io/licensing/
**
** This file is part of the QtCore module of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** Commercial License Usage
** Licensees holding valid commercial Qt licenses may use this file in
** accordance with the commercial license agreement provided with the
** Software or, alternatively, in accordance with the terms contained in
** a written agreement between you and The Qt Company. For licensing terms
** and conditions see https://www.qt.io/terms-conditions. For further
** information use the contact form at https://www.qt.io/contact-us.
**
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 3 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL3 included in the
** packaging of this file. Please review the following information to
** ensure the GNU Lesser General Public License version 3 requirements
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 2.0 or (at your option) the GNU General
** Public license version 3 or any later version approved by the KDE Free
** Qt Foundation. The licenses are as published by the Free Software
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3
** included in the packaging of this file. Please review the following
** information to ensure the GNU General Public License requirements will
** be met: https://www.gnu.org/licenses/gpl-2.0.html and
** https://www.gnu.org/licenses/gpl-3.0.html.
**
** $QT_END_LICENSE$
**
****************************************************************************/
#ifndef QSTRINGCONVERTER_P_H
#define QSTRINGCONVERTER_P_H
//
// W A R N I N G
// -------------
//
// This file is not part of the Qt API. It exists purely as an
// implementation detail. This header file may change from version to
// version without notice, or even be removed.
//
// We mean it.
//
#include <QtCore/qstring.h>
#include <QtCore/qendian.h>
#include <QtCore/qstringconverter.h>
QT_BEGIN_NAMESPACE
struct QUtf8BaseTraits
{
static const bool isTrusted = false;
static const bool allowNonCharacters = true;
static const bool skipAsciiHandling = false;
static const int Error = -1;
static const int EndOfString = -2;
static bool isValidCharacter(uint u)
{ return int(u) >= 0; }
static void appendByte(uchar *&ptr, uchar b)
{ *ptr++ = b; }
static uchar peekByte(const uchar *ptr, int n = 0)
{ return ptr[n]; }
static qptrdiff availableBytes(const uchar *ptr, const uchar *end)
{ return end - ptr; }
static void advanceByte(const uchar *&ptr, int n = 1)
{ ptr += n; }
static void appendUtf16(ushort *&ptr, ushort uc)
{ *ptr++ = uc; }
static void appendUcs4(ushort *&ptr, uint uc)
{
appendUtf16(ptr, QChar::highSurrogate(uc));
appendUtf16(ptr, QChar::lowSurrogate(uc));
}
static ushort peekUtf16(const ushort *ptr, int n = 0)
{ return ptr[n]; }
static qptrdiff availableUtf16(const ushort *ptr, const ushort *end)
{ return end - ptr; }
static void advanceUtf16(const ushort *&ptr, int n = 1)
{ ptr += n; }
// it's possible to output to UCS-4 too
static void appendUtf16(uint *&ptr, ushort uc)
{ *ptr++ = uc; }
static void appendUcs4(uint *&ptr, uint uc)
{ *ptr++ = uc; }
};
struct QUtf8BaseTraitsNoAscii : public QUtf8BaseTraits
{
static const bool skipAsciiHandling = true;
};
namespace QUtf8Functions
{
/// returns 0 on success; errors can only happen if \a u is a surrogate:
/// Error if \a u is a low surrogate;
/// if \a u is a high surrogate, Error if the next isn't a low one,
/// EndOfString if we run into the end of the string.
template <typename Traits, typename OutputPtr, typename InputPtr> inline
int toUtf8(ushort u, OutputPtr &dst, InputPtr &src, InputPtr end)
{
if (!Traits::skipAsciiHandling && u < 0x80) {
// U+0000 to U+007F (US-ASCII) - one byte
Traits::appendByte(dst, uchar(u));
return 0;
} else if (u < 0x0800) {
// U+0080 to U+07FF - two bytes
// first of two bytes
Traits::appendByte(dst, 0xc0 | uchar(u >> 6));
} else {
if (!QChar::isSurrogate(u)) {
// U+0800 to U+FFFF (except U+D800-U+DFFF) - three bytes
if (!Traits::allowNonCharacters && QChar::isNonCharacter(u))
return Traits::Error;
// first of three bytes
Traits::appendByte(dst, 0xe0 | uchar(u >> 12));
} else {
// U+10000 to U+10FFFF - four bytes
// need to get one extra codepoint
if (Traits::availableUtf16(src, end) == 0)
return Traits::EndOfString;
ushort low = Traits::peekUtf16(src);
if (!QChar::isHighSurrogate(u))
return Traits::Error;
if (!QChar::isLowSurrogate(low))
return Traits::Error;
Traits::advanceUtf16(src);
uint ucs4 = QChar::surrogateToUcs4(u, low);
if (!Traits::allowNonCharacters && QChar::isNonCharacter(ucs4))
return Traits::Error;
// first byte
Traits::appendByte(dst, 0xf0 | (uchar(ucs4 >> 18) & 0xf));
// second of four bytes
Traits::appendByte(dst, 0x80 | (uchar(ucs4 >> 12) & 0x3f));
// for the rest of the bytes
u = ushort(ucs4);
}
// second to last byte
Traits::appendByte(dst, 0x80 | (uchar(u >> 6) & 0x3f));
}
// last byte
Traits::appendByte(dst, 0x80 | (u & 0x3f));
return 0;
}
inline bool isContinuationByte(uchar b)
{
return (b & 0xc0) == 0x80;
}
/// returns the number of characters consumed (including \a b) in case of success;
/// returns negative in case of error: Traits::Error or Traits::EndOfString
template <typename Traits, typename OutputPtr, typename InputPtr> inline
int fromUtf8(uchar b, OutputPtr &dst, InputPtr &src, InputPtr end)
{
int charsNeeded;
uint min_uc;
uint uc;
if (!Traits::skipAsciiHandling && b < 0x80) {
// US-ASCII
Traits::appendUtf16(dst, b);
return 1;
}
if (!Traits::isTrusted && Q_UNLIKELY(b <= 0xC1)) {
// an UTF-8 first character must be at least 0xC0
// however, all 0xC0 and 0xC1 first bytes can only produce overlong sequences
return Traits::Error;
} else if (b < 0xe0) {
charsNeeded = 2;
min_uc = 0x80;
uc = b & 0x1f;
} else if (b < 0xf0) {
charsNeeded = 3;
min_uc = 0x800;
uc = b & 0x0f;
} else if (b < 0xf5) {
charsNeeded = 4;
min_uc = 0x10000;
uc = b & 0x07;
} else {
// the last Unicode character is U+10FFFF
// it's encoded in UTF-8 as "\xF4\x8F\xBF\xBF"
// therefore, a byte higher than 0xF4 is not the UTF-8 first byte
return Traits::Error;
}
int bytesAvailable = Traits::availableBytes(src, end);
if (Q_UNLIKELY(bytesAvailable < charsNeeded - 1)) {
// it's possible that we have an error instead of just unfinished bytes
if (bytesAvailable > 0 && !isContinuationByte(Traits::peekByte(src, 0)))
return Traits::Error;
if (bytesAvailable > 1 && !isContinuationByte(Traits::peekByte(src, 1)))
return Traits::Error;
return Traits::EndOfString;
}
// first continuation character
b = Traits::peekByte(src, 0);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
if (charsNeeded > 2) {
// second continuation character
b = Traits::peekByte(src, 1);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
if (charsNeeded > 3) {
// third continuation character
b = Traits::peekByte(src, 2);
if (!isContinuationByte(b))
return Traits::Error;
uc <<= 6;
uc |= b & 0x3f;
}
}
// we've decoded something; safety-check it
if (!Traits::isTrusted) {
if (uc < min_uc)
return Traits::Error;
if (QChar::isSurrogate(uc) || uc > QChar::LastValidCodePoint)
return Traits::Error;
if (!Traits::allowNonCharacters && QChar::isNonCharacter(uc))
return Traits::Error;
}
// write the UTF-16 sequence
if (!QChar::requiresSurrogates(uc)) {
// UTF-8 decoded and no surrogates are required
// detach if necessary
Traits::appendUtf16(dst, ushort(uc));
} else {
// UTF-8 decoded to something that requires a surrogate pair
Traits::appendUcs4(dst, uc);
}
Traits::advanceByte(src, charsNeeded - 1);
return charsNeeded;
}
}
enum DataEndianness
{
DetectEndianness,
BigEndianness,
LittleEndianness
};
struct QUtf8
{
static QChar *convertToUnicode(QChar *, const char *, qsizetype) noexcept;
static QString convertToUnicode(const char *, qsizetype);
static QString convertToUnicode(const char *, qsizetype, QStringConverter::State *);
static QByteArray convertFromUnicode(const QChar *, qsizetype);
static QByteArray convertFromUnicode(const QChar *, qsizetype, QStringConverter::State *);
struct ValidUtf8Result {
bool isValidUtf8;
bool isValidAscii;
};
static ValidUtf8Result isValidUtf8(const char *, qsizetype);
static int compareUtf8(const char *, qsizetype, const QChar *, qsizetype);
static int compareUtf8(const char *, qsizetype, QLatin1String s);
};
struct QUtf16
{
static QString convertToUnicode(const char *, qsizetype, QStringConverter::State *, DataEndianness = DetectEndianness);
static QByteArray convertFromUnicode(const QChar *, qsizetype, QStringConverter::State *, DataEndianness = DetectEndianness);
};
struct QUtf32
{
static QString convertToUnicode(const char *, qsizetype, QStringConverter::State *, DataEndianness = DetectEndianness);
static QByteArray convertFromUnicode(const QChar *, qsizetype, QStringConverter::State *, DataEndianness = DetectEndianness);
};
/*
Converts from different utf encodings looking at a possible byte order mark at the
beginning of the string. If no BOM exists, utf-8 is assumed.
*/
Q_CORE_EXPORT QString qFromUtfEncoded(const QByteArray &ba);
QT_END_NAMESPACE
#endif // QSTRINGCONVERTER_P_H

1
src/corelib/text/text.pri
View File

@ -20,6 +20,7 @@ HEADERS += \
text/qstringalgorithms_p.h \
text/qstringbuilder.h \
text/qstringconverter.h \
text/qstringconverter_p.h \
text/qstringiterator_p.h \
text/qstringlist.h \
text/qstringliteral.h \

4
src/gui/kernel/qclipboard.cpp
View File

@ -46,9 +46,7 @@
#include "qvariant.h"
#include "qbuffer.h"
#include "qimage.h"
#if QT_CONFIG(textcodec)
#include "private/qutfcodec_p.h"
#endif
#include "private/qstringconverter_p.h"
#include "private/qguiapplication_p.h"
#include <qpa/qplatformintegration.h>

4
src/tools/bootstrap/.prev_CMakeLists.txt
View File

@ -41,9 +41,6 @@ qt_add_module(Bootstrap
../../3rdparty/pcre2/src/pcre2_ucp.h
../../3rdparty/pcre2/src/pcre2_valid_utf.c
../../3rdparty/pcre2/src/pcre2_xclass.c
../../corelib/codecs/qlatincodec.cpp
../../corelib/codecs/qtextcodec.cpp
../../corelib/codecs/qutfcodec.cpp
../../corelib/global/qendian.cpp
../../corelib/global/qglobal.cpp
../../corelib/global/qlogging.cpp
@ -109,7 +106,6 @@ qt_add_module(Bootstrap
../../corelib/text/qstringbuilder.cpp
../../corelib/text/qstringconverter.cpp
../../corelib/text/qstringlist.cpp
../../corelib/text/qstringview.cpp
../../corelib/text/qvsnprintf.cpp
../../corelib/time/qcalendar.cpp
../../corelib/time/qdatetime.cpp

4
src/tools/bootstrap/CMakeLists.txt
View File

@ -42,9 +42,6 @@ qt_extend_target(Bootstrap
../../3rdparty/pcre2/src/pcre2_ucp.h
../../3rdparty/pcre2/src/pcre2_valid_utf.c
../../3rdparty/pcre2/src/pcre2_xclass.c
../../corelib/codecs/qlatincodec.cpp
../../corelib/codecs/qtextcodec.cpp
../../corelib/codecs/qutfcodec.cpp
../../corelib/global/qendian.cpp
../../corelib/global/qglobal.cpp
../../corelib/global/qlogging.cpp
@ -110,7 +107,6 @@ qt_extend_target(Bootstrap
../../corelib/text/qstringbuilder.cpp
../../corelib/text/qstringconverter.cpp
../../corelib/text/qstringlist.cpp
../../corelib/text/qstringview.cpp
../../corelib/text/qvsnprintf.cpp
../../corelib/time/qcalendar.cpp
../../corelib/time/qdatetime.cpp

4
src/tools/bootstrap/bootstrap.pro
View File

@ -28,9 +28,6 @@ INCLUDEPATH += \
$$PWD/../../3rdparty/pcre2/src
SOURCES += \
../../corelib/codecs/qlatincodec.cpp \
../../corelib/codecs/qtextcodec.cpp \
../../corelib/codecs/qutfcodec.cpp \
../../corelib/global/qendian.cpp \
../../corelib/global/qglobal.cpp \
../../corelib/global/qlogging.cpp \
@ -96,7 +93,6 @@ SOURCES += \
../../corelib/text/qstringconverter.cpp \
../../corelib/text/qstring_compat.cpp \
../../corelib/text/qstringlist.cpp \
../../corelib/text/qstringview.cpp \
../../corelib/text/qvsnprintf.cpp \
../../corelib/time/qcalendar.cpp \
../../corelib/time/qdatetime.cpp \