1 | /**************************************************************************** |
2 | ** |
3 | ** Copyright (C) 2020 The Qt Company Ltd. |
4 | ** Copyright (C) 2020 Intel Corporation. |
5 | ** Contact: https://www.qt.io/licensing/ |
6 | ** |
7 | ** This file is part of the QtCore module of the Qt Toolkit. |
8 | ** |
9 | ** $QT_BEGIN_LICENSE:LGPL$ |
10 | ** Commercial License Usage |
11 | ** Licensees holding valid commercial Qt licenses may use this file in |
12 | ** accordance with the commercial license agreement provided with the |
13 | ** Software or, alternatively, in accordance with the terms contained in |
14 | ** a written agreement between you and The Qt Company. For licensing terms |
15 | ** and conditions see https://www.qt.io/terms-conditions. For further |
16 | ** information use the contact form at https://www.qt.io/contact-us. |
17 | ** |
18 | ** GNU Lesser General Public License Usage |
19 | ** Alternatively, this file may be used under the terms of the GNU Lesser |
20 | ** General Public License version 3 as published by the Free Software |
21 | ** Foundation and appearing in the file LICENSE.LGPL3 included in the |
22 | ** packaging of this file. Please review the following information to |
23 | ** ensure the GNU Lesser General Public License version 3 requirements |
24 | ** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. |
25 | ** |
26 | ** GNU General Public License Usage |
27 | ** Alternatively, this file may be used under the terms of the GNU |
28 | ** General Public License version 2.0 or (at your option) the GNU General |
29 | ** Public license version 3 or any later version approved by the KDE Free |
30 | ** Qt Foundation. The licenses are as published by the Free Software |
31 | ** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 |
32 | ** included in the packaging of this file. Please review the following |
33 | ** information to ensure the GNU General Public License requirements will |
34 | ** be met: https://www.gnu.org/licenses/gpl-2.0.html and |
35 | ** https://www.gnu.org/licenses/gpl-3.0.html. |
36 | ** |
37 | ** $QT_END_LICENSE$ |
38 | ** |
39 | ****************************************************************************/ |
40 | |
41 | #include <qstringconverter.h> |
42 | #include <private/qstringconverter_p.h> |
43 | #include "qendian.h" |
44 | |
45 | #include "private/qsimd_p.h" |
46 | #include "private/qstringiterator_p.h" |
47 | #include "qbytearraymatcher.h" |
48 | |
49 | #ifdef Q_OS_WIN |
50 | #include <qt_windows.h> |
51 | #endif |
52 | |
53 | #if __has_include(<bit>) && __cplusplus > 201703L |
54 | #include <bit> |
55 | #endif |
56 | |
57 | QT_BEGIN_NAMESPACE |
58 | |
59 | enum { Endian = 0, Data = 1 }; |
60 | |
61 | static const uchar utf8bom[] = { 0xef, 0xbb, 0xbf }; |
62 | |
63 | #if (defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2)) \ |
64 | || (defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64)) |
65 | static Q_ALWAYS_INLINE uint qBitScanReverse(unsigned v) noexcept |
66 | { |
67 | #if defined(__cpp_lib_int_pow2) && __cpp_lib_int_pow2 >= 202002L |
68 | return std::bit_width(v) - 1; |
69 | #else |
70 | uint result = qCountLeadingZeroBits(v); |
71 | // Now Invert the result: clz will count *down* from the msb to the lsb, so the msb index is 31 |
72 | // and the lsb index is 0. The result for _bit_scan_reverse is expected to be the index when |
73 | // counting up: msb index is 0 (because it starts there), and the lsb index is 31. |
74 | result ^= sizeof(unsigned) * 8 - 1; |
75 | return result; |
76 | #endif |
77 | } |
78 | #endif |
79 | |
80 | #if defined(__SSE2__) && defined(QT_COMPILER_SUPPORTS_SSE2) |
81 | static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end) |
82 | { |
83 | // do sixteen characters at a time |
84 | for ( ; end - src >= 16; src += 16, dst += 16) { |
85 | # ifdef __AVX2__ |
86 | __m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src)); |
87 | __m128i data1 = _mm256_castsi256_si128(data); |
88 | __m128i data2 = _mm256_extracti128_si256(data, 1); |
89 | # else |
90 | __m128i data1 = _mm_loadu_si128((const __m128i*)src); |
91 | __m128i data2 = _mm_loadu_si128(1+(const __m128i*)src); |
92 | # endif |
93 | |
94 | // check if everything is ASCII |
95 | // the highest ASCII value is U+007F |
96 | // Do the packing directly: |
97 | // The PACKUSWB instruction has packs a signed 16-bit integer to an unsigned 8-bit |
98 | // with saturation. That is, anything from 0x0100 to 0x7fff is saturated to 0xff, |
99 | // while all negatives (0x8000 to 0xffff) get saturated to 0x00. To detect non-ASCII, |
100 | // we simply do a signed greater-than comparison to 0x00. That means we detect NULs as |
101 | // "non-ASCII", but it's an acceptable compromise. |
102 | __m128i packed = _mm_packus_epi16(data1, data2); |
103 | __m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128()); |
104 | |
105 | // store, even if there are non-ASCII characters here |
106 | _mm_storeu_si128((__m128i*)dst, packed); |
107 | |
108 | // n will contain 1 bit set per character in [data1, data2] that is non-ASCII (or NUL) |
109 | ushort n = ~_mm_movemask_epi8(nonAscii); |
110 | if (n) { |
111 | // find the next probable ASCII character |
112 | // we don't want to load 32 bytes again in this loop if we know there are non-ASCII |
113 | // characters still coming |
114 | nextAscii = src + qBitScanReverse(n) + 1; |
115 | |
116 | n = qCountTrailingZeroBits(n); |
117 | dst += n; |
118 | src += n; |
119 | return false; |
120 | } |
121 | } |
122 | |
123 | if (end - src >= 8) { |
124 | // do eight characters at a time |
125 | __m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src)); |
126 | __m128i packed = _mm_packus_epi16(data, data); |
127 | __m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128()); |
128 | |
129 | // store even non-ASCII |
130 | _mm_storel_epi64(reinterpret_cast<__m128i *>(dst), packed); |
131 | |
132 | uchar n = ~_mm_movemask_epi8(nonAscii); |
133 | if (n) { |
134 | nextAscii = src + qBitScanReverse(n) + 1; |
135 | n = qCountTrailingZeroBits(n); |
136 | dst += n; |
137 | src += n; |
138 | return false; |
139 | } |
140 | } |
141 | |
142 | return src == end; |
143 | } |
144 | |
145 | static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end) |
146 | { |
147 | // do sixteen characters at a time |
148 | for ( ; end - src >= 16; src += 16, dst += 16) { |
149 | __m128i data = _mm_loadu_si128((const __m128i*)src); |
150 | |
151 | #ifdef __AVX2__ |
152 | const int BitSpacing = 2; |
153 | // load and zero extend to an YMM register |
154 | const __m256i extended = _mm256_cvtepu8_epi16(data); |
155 | |
156 | uint n = _mm256_movemask_epi8(extended); |
157 | if (!n) { |
158 | // store |
159 | _mm256_storeu_si256((__m256i*)dst, extended); |
160 | continue; |
161 | } |
162 | #else |
163 | const int BitSpacing = 1; |
164 | |
165 | // check if everything is ASCII |
166 | // movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII |
167 | uint n = _mm_movemask_epi8(data); |
168 | if (!n) { |
169 | // unpack |
170 | _mm_storeu_si128((__m128i*)dst, _mm_unpacklo_epi8(data, _mm_setzero_si128())); |
171 | _mm_storeu_si128(1+(__m128i*)dst, _mm_unpackhi_epi8(data, _mm_setzero_si128())); |
172 | continue; |
173 | } |
174 | #endif |
175 | |
176 | // copy the front part that is still ASCII |
177 | while (!(n & 1)) { |
178 | *dst++ = *src++; |
179 | n >>= BitSpacing; |
180 | } |
181 | |
182 | // find the next probable ASCII character |
183 | // we don't want to load 16 bytes again in this loop if we know there are non-ASCII |
184 | // characters still coming |
185 | n = qBitScanReverse(n); |
186 | nextAscii = src + (n / BitSpacing) + 1; |
187 | return false; |
188 | |
189 | } |
190 | |
191 | if (end - src >= 8) { |
192 | __m128i data = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src)); |
193 | uint n = _mm_movemask_epi8(data) & 0xff; |
194 | if (!n) { |
195 | // unpack and store |
196 | _mm_storeu_si128(reinterpret_cast<__m128i *>(dst), _mm_unpacklo_epi8(data, _mm_setzero_si128())); |
197 | } else { |
198 | while (!(n & 1)) { |
199 | *dst++ = *src++; |
200 | n >>= 1; |
201 | } |
202 | |
203 | n = qBitScanReverse(n); |
204 | nextAscii = src + n + 1; |
205 | return false; |
206 | } |
207 | } |
208 | |
209 | return src == end; |
210 | } |
211 | |
212 | static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii) |
213 | { |
214 | #ifdef __AVX2__ |
215 | // do 32 characters at a time |
216 | // (this is similar to simdTestMask in qstring.cpp) |
217 | const __m256i mask = _mm256_set1_epi8(0x80); |
218 | for ( ; end - src >= 32; src += 32) { |
219 | __m256i data = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src)); |
220 | if (_mm256_testz_si256(mask, data)) |
221 | continue; |
222 | |
223 | uint n = _mm256_movemask_epi8(data); |
224 | Q_ASSUME(n); |
225 | |
226 | // find the next probable ASCII character |
227 | // we don't want to load 32 bytes again in this loop if we know there are non-ASCII |
228 | // characters still coming |
229 | nextAscii = src + qBitScanReverse(n) + 1; |
230 | |
231 | // return the non-ASCII character |
232 | return src + qCountTrailingZeroBits(n); |
233 | } |
234 | #endif |
235 | |
236 | // do sixteen characters at a time |
237 | for ( ; end - src >= 16; src += 16) { |
238 | __m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i*>(src)); |
239 | |
240 | // check if everything is ASCII |
241 | // movemask extracts the high bit of every byte, so n is non-zero if something isn't ASCII |
242 | uint n = _mm_movemask_epi8(data); |
243 | if (!n) |
244 | continue; |
245 | |
246 | // find the next probable ASCII character |
247 | // we don't want to load 16 bytes again in this loop if we know there are non-ASCII |
248 | // characters still coming |
249 | nextAscii = src + qBitScanReverse(n) + 1; |
250 | |
251 | // return the non-ASCII character |
252 | return src + qCountTrailingZeroBits(n); |
253 | } |
254 | |
255 | // do four characters at a time |
256 | for ( ; end - src >= 4; src += 4) { |
257 | quint32 data = qFromUnaligned<quint32>(src); |
258 | data &= 0x80808080U; |
259 | if (!data) |
260 | continue; |
261 | |
262 | // We don't try to guess which of the three bytes is ASCII and which |
263 | // one isn't. The chance that at least two of them are non-ASCII is |
264 | // better than 75%. |
265 | nextAscii = src; |
266 | return src; |
267 | } |
268 | nextAscii = end; |
269 | return src; |
270 | } |
271 | |
272 | // Compare only the US-ASCII beginning of [src8, end8) and [src16, end16) |
273 | // and advance src8 and src16 to the first character that could not be compared |
274 | static void simdCompareAscii(const char8_t *&src8, const char8_t *end8, const char16_t *&src16, const char16_t *end16) |
275 | { |
276 | int bitSpacing = 1; |
277 | qptrdiff len = qMin(end8 - src8, end16 - src16); |
278 | qptrdiff offset = 0; |
279 | uint mask = 0; |
280 | |
281 | // do sixteen characters at a time |
282 | for ( ; offset + 16 < len; offset += 16) { |
283 | __m128i data8 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src8 + offset)); |
284 | #ifdef __AVX2__ |
285 | // AVX2 version, use 256-bit registers and VPMOVXZBW |
286 | __m256i data16 = _mm256_loadu_si256(reinterpret_cast<const __m256i *>(src16 + offset)); |
287 | |
288 | // expand US-ASCII as if it were Latin1 and confirm it's US-ASCII |
289 | __m256i datax8 = _mm256_cvtepu8_epi16(data8); |
290 | mask = _mm256_movemask_epi8(datax8); |
291 | if (mask) |
292 | break; |
293 | |
294 | // compare Latin1 to UTF-16 |
295 | __m256i latin1cmp = _mm256_cmpeq_epi16(datax8, data16); |
296 | mask = ~_mm256_movemask_epi8(latin1cmp); |
297 | if (mask) |
298 | break; |
299 | #else |
300 | // non-AVX2 code |
301 | __m128i datalo16 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src16 + offset)); |
302 | __m128i datahi16 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src16 + offset) + 1); |
303 | |
304 | // expand US-ASCII as if it were Latin1, we'll confirm later |
305 | __m128i datalo8 = _mm_unpacklo_epi8(data8, _mm_setzero_si128()); |
306 | __m128i datahi8 = _mm_unpackhi_epi8(data8, _mm_setzero_si128()); |
307 | |
308 | // compare Latin1 to UTF-16 |
309 | __m128i latin1cmplo = _mm_cmpeq_epi16(datalo8, datalo16); |
310 | __m128i latin1cmphi = _mm_cmpeq_epi16(datahi8, datahi16); |
311 | mask = _mm_movemask_epi8(latin1cmphi) << 16; |
312 | mask |= ushort(_mm_movemask_epi8(latin1cmplo)); |
313 | mask = ~mask; |
314 | if (mask) |
315 | break; |
316 | |
317 | // confirm it was US-ASCII |
318 | mask = _mm_movemask_epi8(data8); |
319 | if (mask) { |
320 | bitSpacing = 0; |
321 | break; |
322 | } |
323 | #endif |
324 | } |
325 | |
326 | // helper for comparing 4 or 8 characters |
327 | auto cmp_lt_16 = [&mask, &offset](int n, __m128i data8, __m128i data16) { |
328 | // n = 4 -> sizemask = 0xff |
329 | // n = 8 -> sizemask = 0xffff |
330 | unsigned sizemask = (1U << (2 * n)) - 1; |
331 | |
332 | // expand as if Latin1 |
333 | data8 = _mm_unpacklo_epi8(data8, _mm_setzero_si128()); |
334 | |
335 | // compare and confirm it's US-ASCII |
336 | __m128i latin1cmp = _mm_cmpeq_epi16(data8, data16); |
337 | mask = ~_mm_movemask_epi8(latin1cmp) & sizemask; |
338 | mask |= _mm_movemask_epi8(data8); |
339 | if (mask == 0) |
340 | offset += n; |
341 | }; |
342 | |
343 | // do eight characters at a time |
344 | if (mask == 0 && offset + 8 < len) { |
345 | __m128i data8 = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src8 + offset)); |
346 | __m128i data16 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src16 + offset)); |
347 | cmp_lt_16(8, data8, data16); |
348 | } |
349 | |
350 | // do four characters |
351 | if (mask == 0 && offset + 4 < len) { |
352 | __m128i data8 = _mm_cvtsi32_si128(qFromUnaligned<quint32>(src8 + offset)); |
353 | __m128i data16 = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src16 + offset)); |
354 | cmp_lt_16(4, data8, data16); |
355 | } |
356 | |
357 | // correct the source pointers to point to the first character we couldn't deal with |
358 | if (mask) |
359 | offset += qCountTrailingZeroBits(mask) >> bitSpacing; |
360 | src8 += offset; |
361 | src16 += offset; |
362 | } |
363 | #elif defined(__ARM_NEON__) && defined(Q_PROCESSOR_ARM_64) // vaddv is only available on Aarch64 |
364 | static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const ushort *&src, const ushort *end) |
365 | { |
366 | uint16x8_t maxAscii = vdupq_n_u16(0x7f); |
367 | uint16x8_t mask1 = { 1, 1 << 2, 1 << 4, 1 << 6, 1 << 8, 1 << 10, 1 << 12, 1 << 14 }; |
368 | uint16x8_t mask2 = vshlq_n_u16(mask1, 1); |
369 | |
370 | // do sixteen characters at a time |
371 | for ( ; end - src >= 16; src += 16, dst += 16) { |
372 | // load 2 lanes (or: "load interleaved") |
373 | uint16x8x2_t in = vld2q_u16(src); |
374 | |
375 | // check if any of the elements > 0x7f, select 1 bit per element (element 0 -> bit 0, element 1 -> bit 1, etc), |
376 | // add those together into a scalar, and merge the scalars. |
377 | uint16_t nonAscii = vaddvq_u16(vandq_u16(vcgtq_u16(in.val[0], maxAscii), mask1)) |
378 | | vaddvq_u16(vandq_u16(vcgtq_u16(in.val[1], maxAscii), mask2)); |
379 | |
380 | // merge the two lanes by shifting the values of the second by 8 and inserting them |
381 | uint16x8_t out = vsliq_n_u16(in.val[0], in.val[1], 8); |
382 | |
383 | // store, even if there are non-ASCII characters here |
384 | vst1q_u8(dst, vreinterpretq_u8_u16(out)); |
385 | |
386 | if (nonAscii) { |
387 | // find the next probable ASCII character |
388 | // we don't want to load 32 bytes again in this loop if we know there are non-ASCII |
389 | // characters still coming |
390 | nextAscii = src + qBitScanReverse(nonAscii) + 1; |
391 | |
392 | nonAscii = qCountTrailingZeroBits(nonAscii); |
393 | dst += nonAscii; |
394 | src += nonAscii; |
395 | return false; |
396 | } |
397 | } |
398 | return src == end; |
399 | } |
400 | |
401 | static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const uchar *&src, const uchar *end) |
402 | { |
403 | // do eight characters at a time |
404 | uint8x8_t msb_mask = vdup_n_u8(0x80); |
405 | uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 }; |
406 | for ( ; end - src >= 8; src += 8, dst += 8) { |
407 | uint8x8_t c = vld1_u8(src); |
408 | uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask)); |
409 | if (!n) { |
410 | // store |
411 | vst1q_u16(dst, vmovl_u8(c)); |
412 | continue; |
413 | } |
414 | |
415 | // copy the front part that is still ASCII |
416 | while (!(n & 1)) { |
417 | *dst++ = *src++; |
418 | n >>= 1; |
419 | } |
420 | |
421 | // find the next probable ASCII character |
422 | // we don't want to load 16 bytes again in this loop if we know there are non-ASCII |
423 | // characters still coming |
424 | n = qBitScanReverse(n); |
425 | nextAscii = src + n + 1; |
426 | return false; |
427 | |
428 | } |
429 | return src == end; |
430 | } |
431 | |
432 | static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii) |
433 | { |
434 | // The SIMD code below is untested, so just force an early return until |
435 | // we've had the time to verify it works. |
436 | nextAscii = end; |
437 | return src; |
438 | |
439 | // do eight characters at a time |
440 | uint8x8_t msb_mask = vdup_n_u8(0x80); |
441 | uint8x8_t add_mask = { 1, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7 }; |
442 | for ( ; end - src >= 8; src += 8) { |
443 | uint8x8_t c = vld1_u8(src); |
444 | uint8_t n = vaddv_u8(vand_u8(vcge_u8(c, msb_mask), add_mask)); |
445 | if (!n) |
446 | continue; |
447 | |
448 | // find the next probable ASCII character |
449 | // we don't want to load 16 bytes again in this loop if we know there are non-ASCII |
450 | // characters still coming |
451 | nextAscii = src + qBitScanReverse(n) + 1; |
452 | |
453 | // return the non-ASCII character |
454 | return src + qCountTrailingZeroBits(n); |
455 | } |
456 | nextAscii = end; |
457 | return src; |
458 | } |
459 | |
460 | static void simdCompareAscii(const char8_t *&, const char8_t *, const char16_t *&, const char16_t *) |
461 | { |
462 | } |
463 | #else |
464 | static inline bool simdEncodeAscii(uchar *, const ushort *, const ushort *, const ushort *) |
465 | { |
466 | return false; |
467 | } |
468 | |
469 | static inline bool simdDecodeAscii(ushort *, const uchar *, const uchar *, const uchar *) |
470 | { |
471 | return false; |
472 | } |
473 | |
474 | static inline const uchar *simdFindNonAscii(const uchar *src, const uchar *end, const uchar *&nextAscii) |
475 | { |
476 | nextAscii = end; |
477 | return src; |
478 | } |
479 | |
480 | static void simdCompareAscii(const char8_t *&, const char8_t *, const char16_t *&, const char16_t *) |
481 | { |
482 | } |
483 | #endif |
484 | |
485 | enum { = 1 }; |
486 | |
487 | QByteArray QUtf8::convertFromUnicode(QStringView in) |
488 | { |
489 | qsizetype len = in.size(); |
490 | |
491 | // create a QByteArray with the worst case scenario size |
492 | QByteArray result(len * 3, Qt::Uninitialized); |
493 | uchar *dst = reinterpret_cast<uchar *>(const_cast<char *>(result.constData())); |
494 | const ushort *src = reinterpret_cast<const ushort *>(in.data()); |
495 | const ushort *const end = src + len; |
496 | |
497 | while (src != end) { |
498 | const ushort *nextAscii = end; |
499 | if (simdEncodeAscii(dst, nextAscii, src, end)) |
500 | break; |
501 | |
502 | do { |
503 | ushort u = *src++; |
504 | int res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(u, dst, src, end); |
505 | if (res < 0) { |
506 | // encoding error - append '?' |
507 | *dst++ = '?'; |
508 | } |
509 | } while (src < nextAscii); |
510 | } |
511 | |
512 | result.truncate(dst - reinterpret_cast<uchar *>(const_cast<char *>(result.constData()))); |
513 | return result; |
514 | } |
515 | |
516 | QByteArray QUtf8::convertFromUnicode(QStringView in, QStringConverterBase::State *state) |
517 | { |
518 | QByteArray ba(3*in.size() +3, Qt::Uninitialized); |
519 | char *end = convertFromUnicode(ba.data(), in, state); |
520 | ba.truncate(end - ba.data()); |
521 | return ba; |
522 | } |
523 | |
524 | char *QUtf8::convertFromUnicode(char *out, QStringView in, QStringConverter::State *state) |
525 | { |
526 | Q_ASSERT(state); |
527 | const QChar *uc = in.data(); |
528 | qsizetype len = in.length(); |
529 | if (!len) |
530 | return out; |
531 | |
532 | auto appendReplacementChar = [state](uchar *cursor) -> uchar * { |
533 | if (state->flags & QStringConverter::Flag::ConvertInvalidToNull) { |
534 | *cursor++ = 0; |
535 | } else { |
536 | // QChar::replacement encoded in utf8 |
537 | *cursor++ = 0xef; |
538 | *cursor++ = 0xbf; |
539 | *cursor++ = 0xbd; |
540 | } |
541 | return cursor; |
542 | }; |
543 | |
544 | uchar *cursor = reinterpret_cast<uchar *>(out); |
545 | const ushort *src = reinterpret_cast<const ushort *>(uc); |
546 | const ushort *const end = src + len; |
547 | |
548 | if (!(state->flags & QStringDecoder::Flag::Stateless)) { |
549 | if (state->remainingChars) { |
550 | int res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(state->state_data[0], cursor, src, end); |
551 | if (res < 0) |
552 | cursor = appendReplacementChar(cursor); |
553 | state->state_data[0] = 0; |
554 | state->remainingChars = 0; |
555 | } else if (!(state->internalState & HeaderDone) && state->flags & QStringConverter::Flag::WriteBom) { |
556 | // append UTF-8 BOM |
557 | *cursor++ = utf8bom[0]; |
558 | *cursor++ = utf8bom[1]; |
559 | *cursor++ = utf8bom[2]; |
560 | state->internalState |= HeaderDone; |
561 | } |
562 | } |
563 | |
564 | while (src != end) { |
565 | const ushort *nextAscii = end; |
566 | if (simdEncodeAscii(cursor, nextAscii, src, end)) |
567 | break; |
568 | |
569 | do { |
570 | ushort uc = *src++; |
571 | int res = QUtf8Functions::toUtf8<QUtf8BaseTraits>(uc, cursor, src, end); |
572 | if (Q_LIKELY(res >= 0)) |
573 | continue; |
574 | |
575 | if (res == QUtf8BaseTraits::Error) { |
576 | // encoding error |
577 | ++state->invalidChars; |
578 | cursor = appendReplacementChar(cursor); |
579 | } else if (res == QUtf8BaseTraits::EndOfString) { |
580 | if (state->flags & QStringConverter::Flag::Stateless) { |
581 | ++state->invalidChars; |
582 | cursor = appendReplacementChar(cursor); |
583 | } else { |
584 | state->remainingChars = 1; |
585 | state->state_data[0] = uc; |
586 | } |
587 | return reinterpret_cast<char *>(cursor); |
588 | } |
589 | } while (src < nextAscii); |
590 | } |
591 | |
592 | return reinterpret_cast<char *>(cursor); |
593 | } |
594 | |
595 | QString QUtf8::convertToUnicode(QByteArrayView in) |
596 | { |
597 | // UTF-8 to UTF-16 always needs the exact same number of words or less: |
598 | // UTF-8 UTF-16 |
599 | // 1 byte 1 word |
600 | // 2 bytes 1 word |
601 | // 3 bytes 1 word |
602 | // 4 bytes 2 words (one surrogate pair) |
603 | // That is, we'll use the full buffer if the input is US-ASCII (1-byte UTF-8), |
604 | // half the buffer for U+0080-U+07FF text (e.g., Greek, Cyrillic, Arabic) or |
605 | // non-BMP text, and one third of the buffer for U+0800-U+FFFF text (e.g, CJK). |
606 | // |
607 | // The table holds for invalid sequences too: we'll insert one replacement char |
608 | // per invalid byte. |
609 | QString result(in.size(), Qt::Uninitialized); |
610 | QChar *data = const_cast<QChar*>(result.constData()); // we know we're not shared |
611 | const QChar *end = convertToUnicode(data, in); |
612 | result.truncate(end - data); |
613 | return result; |
614 | } |
615 | |
616 | /*! |
617 | \since 5.7 |
618 | \overload |
619 | |
620 | Converts the UTF-8 sequence of bytes viewed by \a in to a sequence of |
621 | QChar starting at \a buffer. The buffer is expected to be large enough |
622 | to hold the result. An upper bound for the size of the buffer is |
623 | \c in.size() QChars. |
624 | |
625 | If, during decoding, an error occurs, a QChar::ReplacementCharacter is |
626 | written. |
627 | |
628 | Returns a pointer to one past the last QChar written. |
629 | |
630 | This function never throws. |
631 | */ |
632 | |
633 | QChar *QUtf8::convertToUnicode(QChar *buffer, QByteArrayView in) noexcept |
634 | { |
635 | ushort *dst = reinterpret_cast<ushort *>(buffer); |
636 | const uchar *const start = reinterpret_cast<const uchar *>(in.data()); |
637 | const uchar *src = start; |
638 | const uchar *end = src + in.size(); |
639 | |
640 | // attempt to do a full decoding in SIMD |
641 | const uchar *nextAscii = end; |
642 | if (!simdDecodeAscii(dst, nextAscii, src, end)) { |
643 | // at least one non-ASCII entry |
644 | // check if we failed to decode the UTF-8 BOM; if so, skip it |
645 | if (Q_UNLIKELY(src == start) |
646 | && end - src >= 3 |
647 | && Q_UNLIKELY(src[0] == utf8bom[0] && src[1] == utf8bom[1] && src[2] == utf8bom[2])) { |
648 | src += 3; |
649 | } |
650 | |
651 | while (src < end) { |
652 | nextAscii = end; |
653 | if (simdDecodeAscii(dst, nextAscii, src, end)) |
654 | break; |
655 | |
656 | do { |
657 | uchar b = *src++; |
658 | int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, dst, src, end); |
659 | if (res < 0) { |
660 | // decoding error |
661 | *dst++ = QChar::ReplacementCharacter; |
662 | } |
663 | } while (src < nextAscii); |
664 | } |
665 | } |
666 | |
667 | return reinterpret_cast<QChar *>(dst); |
668 | } |
669 | |
670 | QString QUtf8::convertToUnicode(QByteArrayView in, QStringConverter::State *state) |
671 | { |
672 | // See above for buffer requirements for stateless decoding. However, that |
673 | // fails if the state is not empty. The following situations can add to the |
674 | // requirements: |
675 | // state contains chars starts with requirement |
676 | // 1 of 2 bytes valid continuation 0 |
677 | // 2 of 3 bytes same 0 |
678 | // 3 bytes of 4 same +1 (need to insert surrogate pair) |
679 | // 1 of 2 bytes invalid continuation +1 (need to insert replacement and restart) |
680 | // 2 of 3 bytes same +1 (same) |
681 | // 3 of 4 bytes same +1 (same) |
682 | QString result(in.size() + 1, Qt::Uninitialized); |
683 | QChar *end = convertToUnicode(result.data(), in, state); |
684 | result.truncate(end - result.constData()); |
685 | return result; |
686 | } |
687 | |
688 | QChar *QUtf8::convertToUnicode(QChar *out, QByteArrayView in, QStringConverter::State *state) |
689 | { |
690 | qsizetype len = in.size(); |
691 | |
692 | Q_ASSERT(state); |
693 | if (!len) |
694 | return out; |
695 | |
696 | |
697 | ushort replacement = QChar::ReplacementCharacter; |
698 | if (state->flags & QStringConverter::Flag::ConvertInvalidToNull) |
699 | replacement = QChar::Null; |
700 | |
701 | int res; |
702 | uchar ch = 0; |
703 | |
704 | ushort *dst = reinterpret_cast<ushort *>(out); |
705 | const uchar *src = reinterpret_cast<const uchar *>(in.data()); |
706 | const uchar *end = src + len; |
707 | |
708 | if (!(state->flags & QStringConverter::Flag::Stateless)) { |
709 | bool = state->internalState & HeaderDone || state->flags & QStringConverter::Flag::ConvertInitialBom; |
710 | if (state->remainingChars || !headerdone) { |
711 | // handle incoming state first |
712 | uchar remainingCharsData[4]; // longest UTF-8 sequence possible |
713 | qsizetype remainingCharsCount = state->remainingChars; |
714 | qsizetype newCharsToCopy = qMin<qsizetype>(sizeof(remainingCharsData) - remainingCharsCount, end - src); |
715 | |
716 | memset(remainingCharsData, 0, sizeof(remainingCharsData)); |
717 | memcpy(remainingCharsData, &state->state_data[0], remainingCharsCount); |
718 | memcpy(remainingCharsData + remainingCharsCount, src, newCharsToCopy); |
719 | |
720 | const uchar *begin = &remainingCharsData[1]; |
721 | res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(remainingCharsData[0], dst, begin, |
722 | static_cast<const uchar *>(remainingCharsData) + remainingCharsCount + newCharsToCopy); |
723 | if (res == QUtf8BaseTraits::Error) { |
724 | ++state->invalidChars; |
725 | *dst++ = replacement; |
726 | ++src; |
727 | } else if (res == QUtf8BaseTraits::EndOfString) { |
728 | // if we got EndOfString again, then there were too few bytes in src; |
729 | // copy to our state and return |
730 | state->remainingChars = remainingCharsCount + newCharsToCopy; |
731 | memcpy(&state->state_data[0], remainingCharsData, state->remainingChars); |
732 | return out; |
733 | } else if (!headerdone) { |
734 | // eat the UTF-8 BOM |
735 | if (dst[-1] == 0xfeff) |
736 | --dst; |
737 | } |
738 | state->internalState |= HeaderDone; |
739 | |
740 | // adjust src now that we have maybe consumed a few chars |
741 | if (res >= 0) { |
742 | Q_ASSERT(res > remainingCharsCount); |
743 | src += res - remainingCharsCount; |
744 | } |
745 | } |
746 | } else if (!(state->flags & QStringConverter::Flag::ConvertInitialBom)) { |
747 | // stateless, remove initial BOM |
748 | if (len > 2 && src[0] == utf8bom[0] && src[1] == utf8bom[1] && src[2] == utf8bom[2]) |
749 | // skip BOM |
750 | src += 3; |
751 | } |
752 | |
753 | // main body, stateless decoding |
754 | res = 0; |
755 | const uchar *nextAscii = src; |
756 | while (res >= 0 && src < end) { |
757 | if (src >= nextAscii && simdDecodeAscii(dst, nextAscii, src, end)) |
758 | break; |
759 | |
760 | ch = *src++; |
761 | res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(ch, dst, src, end); |
762 | if (res == QUtf8BaseTraits::Error) { |
763 | res = 0; |
764 | ++state->invalidChars; |
765 | *dst++ = replacement; |
766 | } |
767 | } |
768 | |
769 | if (res == QUtf8BaseTraits::EndOfString) { |
770 | // unterminated UTF sequence |
771 | if (state->flags & QStringConverter::Flag::Stateless) { |
772 | *dst++ = QChar::ReplacementCharacter; |
773 | ++state->invalidChars; |
774 | while (src++ < end) { |
775 | *dst++ = QChar::ReplacementCharacter; |
776 | ++state->invalidChars; |
777 | } |
778 | state->remainingChars = 0; |
779 | } else { |
780 | --src; // unread the byte in ch |
781 | state->remainingChars = end - src; |
782 | memcpy(&state->state_data[0], src, end - src); |
783 | } |
784 | } else { |
785 | state->remainingChars = 0; |
786 | } |
787 | |
788 | return reinterpret_cast<QChar *>(dst); |
789 | } |
790 | |
791 | struct QUtf8NoOutputTraits : public QUtf8BaseTraitsNoAscii |
792 | { |
793 | struct NoOutput {}; |
794 | static void appendUtf16(const NoOutput &, ushort) {} |
795 | static void appendUcs4(const NoOutput &, uint) {} |
796 | }; |
797 | |
798 | QUtf8::ValidUtf8Result QUtf8::isValidUtf8(QByteArrayView in) |
799 | { |
800 | const uchar *src = reinterpret_cast<const uchar *>(in.data()); |
801 | const uchar *end = src + in.size(); |
802 | const uchar *nextAscii = src; |
803 | bool isValidAscii = true; |
804 | |
805 | while (src < end) { |
806 | if (src >= nextAscii) |
807 | src = simdFindNonAscii(src, end, nextAscii); |
808 | if (src == end) |
809 | break; |
810 | |
811 | do { |
812 | uchar b = *src++; |
813 | if ((b & 0x80) == 0) |
814 | continue; |
815 | |
816 | isValidAscii = false; |
817 | QUtf8NoOutputTraits::NoOutput output; |
818 | int res = QUtf8Functions::fromUtf8<QUtf8NoOutputTraits>(b, output, src, end); |
819 | if (res < 0) { |
820 | // decoding error |
821 | return { false, false }; |
822 | } |
823 | } while (src < nextAscii); |
824 | } |
825 | |
826 | return { true, isValidAscii }; |
827 | } |
828 | |
829 | int QUtf8::compareUtf8(QByteArrayView utf8, QStringView utf16) noexcept |
830 | { |
831 | auto src1 = reinterpret_cast<const char8_t *>(utf8.data()); |
832 | auto end1 = src1 + utf8.size(); |
833 | auto src2 = reinterpret_cast<const char16_t *>(utf16.data()); |
834 | auto end2 = src2 + utf16.size(); |
835 | |
836 | do { |
837 | simdCompareAscii(src1, end1, src2, end2); |
838 | |
839 | if (src1 < end1 && src2 < end2) { |
840 | char32_t uc1 = *src1++; |
841 | char32_t uc2 = *src2++; |
842 | |
843 | if (uc1 >= 0x80) { |
844 | char32_t *output = &uc1; |
845 | int res = QUtf8Functions::fromUtf8<QUtf8BaseTraitsNoAscii>(uc1, output, src1, end1); |
846 | if (res < 0) { |
847 | // decoding error |
848 | uc1 = QChar::ReplacementCharacter; |
849 | } |
850 | |
851 | // Only decode the UTF-16 surrogate pair if the UTF-8 code point |
852 | // wasn't US-ASCII (a surrogate cannot match US-ASCII). |
853 | if (QChar::isHighSurrogate(uc2) && src2 < end2 && QChar::isLowSurrogate(*src2)) |
854 | uc2 = QChar::surrogateToUcs4(uc2, *src2++); |
855 | } |
856 | |
857 | if (uc1 != uc2) |
858 | return int(uc1) - int(uc2); |
859 | } |
860 | } while (src1 < end1 && src2 < end2); |
861 | |
862 | // the shorter string sorts first |
863 | return (end1 > src1) - int(end2 > src2); |
864 | } |
865 | |
866 | int QUtf8::compareUtf8(QByteArrayView utf8, QLatin1String s) |
867 | { |
868 | uint uc1 = QChar::Null; |
869 | auto src1 = reinterpret_cast<const uchar *>(utf8.data()); |
870 | auto end1 = src1 + utf8.size(); |
871 | auto src2 = reinterpret_cast<const uchar *>(s.latin1()); |
872 | auto end2 = src2 + s.size(); |
873 | |
874 | while (src1 < end1 && src2 < end2) { |
875 | uchar b = *src1++; |
876 | uint *output = &uc1; |
877 | int res = QUtf8Functions::fromUtf8<QUtf8BaseTraits>(b, output, src1, end1); |
878 | if (res < 0) { |
879 | // decoding error |
880 | uc1 = QChar::ReplacementCharacter; |
881 | } |
882 | |
883 | uint uc2 = *src2++; |
884 | if (uc1 != uc2) |
885 | return int(uc1) - int(uc2); |
886 | } |
887 | |
888 | // the shorter string sorts first |
889 | return (end1 > src1) - (end2 > src2); |
890 | } |
891 | |
892 | QByteArray QUtf16::convertFromUnicode(QStringView in, QStringConverter::State *state, DataEndianness endian) |
893 | { |
894 | bool writeBom = !(state->internalState & HeaderDone) && state->flags & QStringConverter::Flag::WriteBom; |
895 | qsizetype length = 2 * in.size(); |
896 | if (writeBom) |
897 | length += 2; |
898 | |
899 | QByteArray d(length, Qt::Uninitialized); |
900 | char *end = convertFromUnicode(d.data(), in, state, endian); |
901 | Q_ASSERT(end - d.constData() == d.length()); |
902 | Q_UNUSED(end); |
903 | return d; |
904 | } |
905 | |
906 | char *QUtf16::convertFromUnicode(char *out, QStringView in, QStringConverter::State *state, DataEndianness endian) |
907 | { |
908 | Q_ASSERT(state); |
909 | bool writeBom = !(state->internalState & HeaderDone) && state->flags & QStringConverter::Flag::WriteBom; |
910 | |
911 | if (endian == DetectEndianness) |
912 | endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness; |
913 | |
914 | if (writeBom) { |
915 | QChar bom(QChar::ByteOrderMark); |
916 | if (endian == BigEndianness) |
917 | qToBigEndian(bom.unicode(), out); |
918 | else |
919 | qToLittleEndian(bom.unicode(), out); |
920 | out += 2; |
921 | } |
922 | if (endian == BigEndianness) |
923 | qToBigEndian<ushort>(in.data(), in.length(), out); |
924 | else |
925 | qToLittleEndian<ushort>(in.data(), in.length(), out); |
926 | |
927 | state->remainingChars = 0; |
928 | state->internalState |= HeaderDone; |
929 | return out + 2*in.length(); |
930 | } |
931 | |
932 | QString QUtf16::convertToUnicode(QByteArrayView in, QStringConverter::State *state, DataEndianness endian) |
933 | { |
934 | QString result((in.size() + 1) >> 1, Qt::Uninitialized); // worst case |
935 | QChar *qch = convertToUnicode(result.data(), in, state, endian); |
936 | result.truncate(qch - result.constData()); |
937 | return result; |
938 | } |
939 | |
940 | QChar *QUtf16::convertToUnicode(QChar *out, QByteArrayView in, QStringConverter::State *state, DataEndianness endian) |
941 | { |
942 | qsizetype len = in.size(); |
943 | const char *chars = in.data(); |
944 | |
945 | Q_ASSERT(state); |
946 | |
947 | if (endian == DetectEndianness) |
948 | endian = (DataEndianness)state->state_data[Endian]; |
949 | |
950 | const char *end = chars + len; |
951 | |
952 | // make sure we can decode at least one char |
953 | if (state->remainingChars + len < 2) { |
954 | if (len) { |
955 | Q_ASSERT(state->remainingChars == 0 && len == 1); |
956 | state->remainingChars = 1; |
957 | state->state_data[Data] = *chars; |
958 | } |
959 | return out; |
960 | } |
961 | |
962 | bool = state && state->internalState & HeaderDone; |
963 | if (state->flags & QStringConverter::Flag::ConvertInitialBom) |
964 | headerdone = true; |
965 | |
966 | if (!headerdone || state->remainingChars) { |
967 | uchar buf; |
968 | if (state->remainingChars) |
969 | buf = state->state_data[Data]; |
970 | else |
971 | buf = *chars++; |
972 | |
973 | // detect BOM, set endianness |
974 | state->internalState |= HeaderDone; |
975 | QChar ch(buf, *chars++); |
976 | if (endian == DetectEndianness) { |
977 | if (ch == QChar::ByteOrderSwapped) { |
978 | endian = BigEndianness; |
979 | } else if (ch == QChar::ByteOrderMark) { |
980 | endian = LittleEndianness; |
981 | } else { |
982 | if (QSysInfo::ByteOrder == QSysInfo::BigEndian) { |
983 | endian = BigEndianness; |
984 | } else { |
985 | endian = LittleEndianness; |
986 | } |
987 | } |
988 | } |
989 | if (endian == BigEndianness) |
990 | ch = QChar::fromUcs2((ch.unicode() >> 8) | ((ch.unicode() & 0xff) << 8)); |
991 | if (headerdone || ch != QChar::ByteOrderMark) |
992 | *out++ = ch; |
993 | } else if (endian == DetectEndianness) { |
994 | endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness; |
995 | } |
996 | |
997 | int nPairs = (end - chars) >> 1; |
998 | if (endian == BigEndianness) |
999 | qFromBigEndian<ushort>(chars, nPairs, out); |
1000 | else |
1001 | qFromLittleEndian<ushort>(chars, nPairs, out); |
1002 | out += nPairs; |
1003 | |
1004 | state->state_data[Endian] = endian; |
1005 | state->remainingChars = 0; |
1006 | if ((end - chars) & 1) { |
1007 | if (state->flags & QStringConverter::Flag::Stateless) { |
1008 | *out++ = state->flags & QStringConverter::Flag::ConvertInvalidToNull ? QChar::Null : QChar::ReplacementCharacter; |
1009 | } else { |
1010 | state->remainingChars = 1; |
1011 | state->state_data[Data] = *(end - 1); |
1012 | } |
1013 | } else { |
1014 | state->state_data[Data] = 0; |
1015 | } |
1016 | |
1017 | return out; |
1018 | } |
1019 | |
1020 | QByteArray QUtf32::convertFromUnicode(QStringView in, QStringConverter::State *state, DataEndianness endian) |
1021 | { |
1022 | bool writeBom = !(state->internalState & HeaderDone) && state->flags & QStringConverter::Flag::WriteBom; |
1023 | int length = 4*in.size(); |
1024 | if (writeBom) |
1025 | length += 4; |
1026 | QByteArray ba(length, Qt::Uninitialized); |
1027 | char *end = convertFromUnicode(ba.data(), in, state, endian); |
1028 | Q_ASSERT(end - ba.constData() == length); |
1029 | Q_UNUSED(end); |
1030 | return ba; |
1031 | } |
1032 | |
1033 | char *QUtf32::convertFromUnicode(char *out, QStringView in, QStringConverter::State *state, DataEndianness endian) |
1034 | { |
1035 | Q_ASSERT(state); |
1036 | |
1037 | bool writeBom = !(state->internalState & HeaderDone) && state->flags & QStringConverter::Flag::WriteBom; |
1038 | qsizetype length = 4*in.length(); |
1039 | if (writeBom) |
1040 | length += 4; |
1041 | |
1042 | if (endian == DetectEndianness) |
1043 | endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness; |
1044 | |
1045 | if (writeBom) { |
1046 | if (endian == BigEndianness) { |
1047 | out[0] = 0; |
1048 | out[1] = 0; |
1049 | out[2] = (char)0xfe; |
1050 | out[3] = (char)0xff; |
1051 | } else { |
1052 | out[0] = (char)0xff; |
1053 | out[1] = (char)0xfe; |
1054 | out[2] = 0; |
1055 | out[3] = 0; |
1056 | } |
1057 | out += 4; |
1058 | state->internalState |= HeaderDone; |
1059 | } |
1060 | |
1061 | const QChar *uc = in.data(); |
1062 | const QChar *end = in.data() + in.length(); |
1063 | QChar ch; |
1064 | uint ucs4; |
1065 | if (state->remainingChars == 1) { |
1066 | ch = state->state_data[Data]; |
1067 | // this is ugly, but shortcuts a whole lot of logic that would otherwise be required |
1068 | state->remainingChars = 0; |
1069 | goto decode_surrogate; |
1070 | } |
1071 | |
1072 | while (uc < end) { |
1073 | ch = *uc++; |
1074 | if (Q_LIKELY(!ch.isSurrogate())) { |
1075 | ucs4 = ch.unicode(); |
1076 | } else if (Q_LIKELY(ch.isHighSurrogate())) { |
1077 | decode_surrogate: |
1078 | if (uc == end) { |
1079 | if (state->flags & QStringConverter::Flag::Stateless) { |
1080 | ucs4 = state->flags & QStringConverter::Flag::ConvertInvalidToNull ? 0 : QChar::ReplacementCharacter; |
1081 | } else { |
1082 | state->remainingChars = 1; |
1083 | state->state_data[Data] = ch.unicode(); |
1084 | return out; |
1085 | } |
1086 | } else if (uc->isLowSurrogate()) { |
1087 | ucs4 = QChar::surrogateToUcs4(ch, *uc++); |
1088 | } else { |
1089 | ucs4 = state->flags & QStringConverter::Flag::ConvertInvalidToNull ? 0 : QChar::ReplacementCharacter; |
1090 | } |
1091 | } else { |
1092 | ucs4 = state->flags & QStringConverter::Flag::ConvertInvalidToNull ? 0 : QChar::ReplacementCharacter; |
1093 | } |
1094 | if (endian == BigEndianness) |
1095 | qToBigEndian(ucs4, out); |
1096 | else |
1097 | qToLittleEndian(ucs4, out); |
1098 | out += 4; |
1099 | } |
1100 | |
1101 | return out; |
1102 | } |
1103 | |
1104 | QString QUtf32::convertToUnicode(QByteArrayView in, QStringConverter::State *state, DataEndianness endian) |
1105 | { |
1106 | QString result; |
1107 | result.resize((in.size() + 7) >> 1); // worst case |
1108 | QChar *end = convertToUnicode(result.data(), in, state, endian); |
1109 | result.truncate(end - result.constData()); |
1110 | return result; |
1111 | } |
1112 | |
1113 | QChar *QUtf32::convertToUnicode(QChar *out, QByteArrayView in, QStringConverter::State *state, DataEndianness endian) |
1114 | { |
1115 | qsizetype len = in.size(); |
1116 | const char *chars = in.data(); |
1117 | |
1118 | Q_ASSERT(state); |
1119 | if (endian == DetectEndianness) |
1120 | endian = (DataEndianness)state->state_data[Endian]; |
1121 | |
1122 | const char *end = chars + len; |
1123 | |
1124 | uchar tuple[4]; |
1125 | memcpy(tuple, &state->state_data[Data], 4); |
1126 | |
1127 | // make sure we can decode at least one char |
1128 | if (state->remainingChars + len < 4) { |
1129 | if (len) { |
1130 | while (chars < end) { |
1131 | tuple[state->remainingChars] = *chars; |
1132 | ++state->remainingChars; |
1133 | ++chars; |
1134 | } |
1135 | Q_ASSERT(state->remainingChars < 4); |
1136 | memcpy(&state->state_data[Data], tuple, 4); |
1137 | } |
1138 | return out; |
1139 | } |
1140 | |
1141 | bool = state->internalState & HeaderDone; |
1142 | if (state->flags & QStringConverter::Flag::ConvertInitialBom) |
1143 | headerdone = true; |
1144 | |
1145 | int num = state->remainingChars; |
1146 | state->remainingChars = 0; |
1147 | |
1148 | if (!headerdone || endian == DetectEndianness || num) { |
1149 | while (num < 4) |
1150 | tuple[num++] = *chars++; |
1151 | if (endian == DetectEndianness) { |
1152 | if (tuple[0] == 0xff && tuple[1] == 0xfe && tuple[2] == 0 && tuple[3] == 0) { |
1153 | endian = LittleEndianness; |
1154 | } else if (tuple[0] == 0 && tuple[1] == 0 && tuple[2] == 0xfe && tuple[3] == 0xff) { |
1155 | endian = BigEndianness; |
1156 | } else if (QSysInfo::ByteOrder == QSysInfo::BigEndian) { |
1157 | endian = BigEndianness; |
1158 | } else { |
1159 | endian = LittleEndianness; |
1160 | } |
1161 | } |
1162 | uint code = (endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple); |
1163 | if (headerdone || code != QChar::ByteOrderMark) { |
1164 | if (QChar::requiresSurrogates(code)) { |
1165 | *out++ = QChar(QChar::highSurrogate(code)); |
1166 | *out++ = QChar(QChar::lowSurrogate(code)); |
1167 | } else { |
1168 | *out++ = QChar(code); |
1169 | } |
1170 | } |
1171 | num = 0; |
1172 | } else if (endian == DetectEndianness) { |
1173 | endian = (QSysInfo::ByteOrder == QSysInfo::BigEndian) ? BigEndianness : LittleEndianness; |
1174 | } |
1175 | state->state_data[Endian] = endian; |
1176 | state->internalState |= HeaderDone; |
1177 | |
1178 | while (chars < end) { |
1179 | tuple[num++] = *chars++; |
1180 | if (num == 4) { |
1181 | uint code = (endian == BigEndianness) ? qFromBigEndian<quint32>(tuple) : qFromLittleEndian<quint32>(tuple); |
1182 | for (char16_t c : QChar::fromUcs4(code)) |
1183 | *out++ = c; |
1184 | num = 0; |
1185 | } |
1186 | } |
1187 | |
1188 | if (num) { |
1189 | if (state->flags & QStringDecoder::Flag::Stateless) { |
1190 | *out++ = QChar::ReplacementCharacter; |
1191 | } else { |
1192 | state->state_data[Endian] = endian; |
1193 | state->remainingChars = num; |
1194 | memcpy(&state->state_data[Data], tuple, 4); |
1195 | } |
1196 | } |
1197 | |
1198 | return out; |
1199 | } |
1200 | |
1201 | #if defined(Q_OS_WIN) && !defined(QT_BOOTSTRAPPED) |
1202 | static QString convertToUnicodeCharByChar(QByteArrayView in, QStringConverter::State *state) |
1203 | { |
1204 | qsizetype length = in.size(); |
1205 | const char *chars = in.data(); |
1206 | |
1207 | Q_ASSERT(state); |
1208 | if (state->flags & QStringConverter::Flag::Stateless) // temporary |
1209 | state = nullptr; |
1210 | |
1211 | if (!chars || !length) |
1212 | return QString(); |
1213 | |
1214 | int copyLocation = 0; |
1215 | int extra = 2; |
1216 | if (state && state->remainingChars) { |
1217 | copyLocation = state->remainingChars; |
1218 | extra += copyLocation; |
1219 | } |
1220 | int newLength = length + extra; |
1221 | char *mbcs = new char[newLength]; |
1222 | //ensure that we have a NULL terminated string |
1223 | mbcs[newLength-1] = 0; |
1224 | mbcs[newLength-2] = 0; |
1225 | memcpy(&(mbcs[copyLocation]), chars, length); |
1226 | if (copyLocation) { |
1227 | //copy the last character from the state |
1228 | mbcs[0] = (char)state->state_data[0]; |
1229 | state->remainingChars = 0; |
1230 | } |
1231 | const char *mb = mbcs; |
1232 | const char *next = 0; |
1233 | QString s; |
1234 | while ((next = CharNextExA(CP_ACP, mb, 0)) != mb) { |
1235 | wchar_t wc[2] ={0}; |
1236 | int charlength = next - mb; |
1237 | int len = MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED|MB_ERR_INVALID_CHARS, mb, charlength, wc, 2); |
1238 | if (len>0) { |
1239 | s.append(QChar(wc[0])); |
1240 | } else { |
1241 | int r = GetLastError(); |
1242 | //check if the character being dropped is the last character |
1243 | if (r == ERROR_NO_UNICODE_TRANSLATION && mb == (mbcs+newLength -3) && state) { |
1244 | state->remainingChars = 1; |
1245 | state->state_data[0] = (char)*mb; |
1246 | } |
1247 | } |
1248 | mb = next; |
1249 | } |
1250 | delete [] mbcs; |
1251 | return s; |
1252 | } |
1253 | |
1254 | |
1255 | QString QLocal8Bit::convertToUnicode(QByteArrayView in, QStringConverter::State *state) |
1256 | { |
1257 | qsizetype length = in.size(); |
1258 | |
1259 | Q_ASSERT(length < INT_MAX); // ### FIXME |
1260 | const char *mb = in.data(); |
1261 | int mblen = length; |
1262 | |
1263 | if (!mb || !mblen) |
1264 | return QString(); |
1265 | |
1266 | QVarLengthArray<wchar_t, 4096> wc(4096); |
1267 | int len; |
1268 | QString sp; |
1269 | bool prepend = false; |
1270 | char state_data = 0; |
1271 | int remainingChars = 0; |
1272 | |
1273 | //save the current state information |
1274 | if (state) { |
1275 | state_data = (char)state->state_data[0]; |
1276 | remainingChars = state->remainingChars; |
1277 | } |
1278 | |
1279 | //convert the pending character (if available) |
1280 | if (state && remainingChars) { |
1281 | char prev[3] = {0}; |
1282 | prev[0] = state_data; |
1283 | prev[1] = mb[0]; |
1284 | remainingChars = 0; |
1285 | len = MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED, |
1286 | prev, 2, wc.data(), wc.length()); |
1287 | if (len) { |
1288 | sp.append(QChar(wc[0])); |
1289 | if (mblen == 1) { |
1290 | state->remainingChars = 0; |
1291 | return sp; |
1292 | } |
1293 | prepend = true; |
1294 | mb++; |
1295 | mblen--; |
1296 | wc[0] = 0; |
1297 | } |
1298 | } |
1299 | |
1300 | while (!(len=MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED|MB_ERR_INVALID_CHARS, |
1301 | mb, mblen, wc.data(), wc.length()))) { |
1302 | int r = GetLastError(); |
1303 | if (r == ERROR_INSUFFICIENT_BUFFER) { |
1304 | const int wclen = MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED, |
1305 | mb, mblen, 0, 0); |
1306 | wc.resize(wclen); |
1307 | } else if (r == ERROR_NO_UNICODE_TRANSLATION) { |
1308 | //find the last non NULL character |
1309 | while (mblen > 1 && !(mb[mblen-1])) |
1310 | mblen--; |
1311 | //check whether, we hit an invalid character in the middle |
1312 | if ((mblen <= 1) || (remainingChars && state_data)) |
1313 | return convertToUnicodeCharByChar(in, state); |
1314 | //Remove the last character and try again... |
1315 | state_data = mb[mblen-1]; |
1316 | remainingChars = 1; |
1317 | mblen--; |
1318 | } else { |
1319 | // Fail. |
1320 | qWarning("MultiByteToWideChar: Cannot convert multibyte text" ); |
1321 | break; |
1322 | } |
1323 | } |
1324 | |
1325 | if (len <= 0) |
1326 | return QString(); |
1327 | |
1328 | if (wc[len-1] == 0) // len - 1: we don't want terminator |
1329 | --len; |
1330 | |
1331 | //save the new state information |
1332 | if (state) { |
1333 | state->state_data[0] = (char)state_data; |
1334 | state->remainingChars = remainingChars; |
1335 | } |
1336 | QString s((QChar*)wc.data(), len); |
1337 | if (prepend) { |
1338 | return sp+s; |
1339 | } |
1340 | return s; |
1341 | } |
1342 | |
1343 | QByteArray QLocal8Bit::convertFromUnicode(QStringView in, QStringConverter::State *state) |
1344 | { |
1345 | const QChar *ch = in.data(); |
1346 | qsizetype uclen = in.size(); |
1347 | |
1348 | Q_ASSERT(uclen < INT_MAX); // ### FIXME |
1349 | Q_ASSERT(state); |
1350 | Q_UNUSED(state); // ### Fixme |
1351 | if (state->flags & QStringConverter::Flag::Stateless) // temporary |
1352 | state = nullptr; |
1353 | |
1354 | if (!ch) |
1355 | return QByteArray(); |
1356 | if (uclen == 0) |
1357 | return QByteArray("" ); |
1358 | BOOL used_def; |
1359 | QByteArray mb(4096, 0); |
1360 | int len; |
1361 | while (!(len=WideCharToMultiByte(CP_ACP, 0, (const wchar_t*)ch, uclen, |
1362 | mb.data(), mb.size()-1, 0, &used_def))) |
1363 | { |
1364 | int r = GetLastError(); |
1365 | if (r == ERROR_INSUFFICIENT_BUFFER) { |
1366 | mb.resize(1+WideCharToMultiByte(CP_ACP, 0, |
1367 | (const wchar_t*)ch, uclen, |
1368 | 0, 0, 0, &used_def)); |
1369 | // and try again... |
1370 | } else { |
1371 | // Fail. Probably can't happen in fact (dwFlags is 0). |
1372 | #ifndef QT_NO_DEBUG |
1373 | // Can't use qWarning(), as it'll recurse to handle %ls |
1374 | fprintf(stderr, |
1375 | "WideCharToMultiByte: Cannot convert multibyte text (error %d): %ls\n" , |
1376 | r, reinterpret_cast<const wchar_t*>(QString(ch, uclen).utf16())); |
1377 | #endif |
1378 | break; |
1379 | } |
1380 | } |
1381 | mb.resize(len); |
1382 | return mb; |
1383 | } |
1384 | #endif |
1385 | |
1386 | void QStringConverter::State::clear() |
1387 | { |
1388 | if (clearFn) |
1389 | clearFn(this); |
1390 | else |
1391 | state_data[0] = state_data[1] = state_data[2] = state_data[3] = 0; |
1392 | remainingChars = 0; |
1393 | invalidChars = 0; |
1394 | internalState = 0; |
1395 | } |
1396 | |
1397 | static QChar *fromUtf16(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1398 | { |
1399 | return QUtf16::convertToUnicode(out, in, state, DetectEndianness); |
1400 | } |
1401 | |
1402 | static char *toUtf16(char *out, QStringView in, QStringConverter::State *state) |
1403 | { |
1404 | return QUtf16::convertFromUnicode(out, in, state, DetectEndianness); |
1405 | } |
1406 | |
1407 | static QChar *fromUtf16BE(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1408 | { |
1409 | return QUtf16::convertToUnicode(out, in, state, BigEndianness); |
1410 | } |
1411 | |
1412 | static char *toUtf16BE(char *out, QStringView in, QStringConverter::State *state) |
1413 | { |
1414 | return QUtf16::convertFromUnicode(out, in, state, BigEndianness); |
1415 | } |
1416 | |
1417 | static QChar *fromUtf16LE(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1418 | { |
1419 | return QUtf16::convertToUnicode(out, in, state, LittleEndianness); |
1420 | } |
1421 | |
1422 | static char *toUtf16LE(char *out, QStringView in, QStringConverter::State *state) |
1423 | { |
1424 | return QUtf16::convertFromUnicode(out, in, state, LittleEndianness); |
1425 | } |
1426 | |
1427 | static QChar *fromUtf32(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1428 | { |
1429 | return QUtf32::convertToUnicode(out, in, state, DetectEndianness); |
1430 | } |
1431 | |
1432 | static char *toUtf32(char *out, QStringView in, QStringConverter::State *state) |
1433 | { |
1434 | return QUtf32::convertFromUnicode(out, in, state, DetectEndianness); |
1435 | } |
1436 | |
1437 | static QChar *fromUtf32BE(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1438 | { |
1439 | return QUtf32::convertToUnicode(out, in, state, BigEndianness); |
1440 | } |
1441 | |
1442 | static char *toUtf32BE(char *out, QStringView in, QStringConverter::State *state) |
1443 | { |
1444 | return QUtf32::convertFromUnicode(out, in, state, BigEndianness); |
1445 | } |
1446 | |
1447 | static QChar *fromUtf32LE(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1448 | { |
1449 | return QUtf32::convertToUnicode(out, in, state, LittleEndianness); |
1450 | } |
1451 | |
1452 | static char *toUtf32LE(char *out, QStringView in, QStringConverter::State *state) |
1453 | { |
1454 | return QUtf32::convertFromUnicode(out, in, state, LittleEndianness); |
1455 | } |
1456 | |
1457 | void qt_from_latin1(char16_t *dst, const char *str, size_t size) noexcept; |
1458 | |
1459 | static QChar *fromLatin1(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1460 | { |
1461 | Q_ASSERT(state); |
1462 | Q_UNUSED(state); |
1463 | |
1464 | qt_from_latin1(reinterpret_cast<char16_t *>(out), in.data(), size_t(in.size())); |
1465 | return out + in.size(); |
1466 | } |
1467 | |
1468 | |
1469 | static char *toLatin1(char *out, QStringView in, QStringConverter::State *state) |
1470 | { |
1471 | Q_ASSERT(state); |
1472 | if (state->flags & QStringConverter::Flag::Stateless) // temporary |
1473 | state = nullptr; |
1474 | |
1475 | const char replacement = (state && state->flags & QStringConverter::Flag::ConvertInvalidToNull) ? 0 : '?'; |
1476 | int invalid = 0; |
1477 | for (qsizetype i = 0; i < in.length(); ++i) { |
1478 | if (in[i] > QChar(0xff)) { |
1479 | *out = replacement; |
1480 | ++invalid; |
1481 | } else { |
1482 | *out = (char)in[i].cell(); |
1483 | } |
1484 | ++out; |
1485 | } |
1486 | if (state) |
1487 | state->invalidChars += invalid; |
1488 | return out; |
1489 | } |
1490 | |
1491 | static QChar *fromLocal8Bit(QChar *out, QByteArrayView in, QStringConverter::State *state) |
1492 | { |
1493 | QString s = QLocal8Bit::convertToUnicode(in, state); |
1494 | memcpy(out, s.constData(), s.length()*sizeof(QChar)); |
1495 | return out + s.length(); |
1496 | } |
1497 | |
1498 | static char *toLocal8Bit(char *out, QStringView in, QStringConverter::State *state) |
1499 | { |
1500 | QByteArray s = QLocal8Bit::convertFromUnicode(in, state); |
1501 | memcpy(out, s.constData(), s.length()); |
1502 | return out + s.length(); |
1503 | } |
1504 | |
1505 | |
1506 | static qsizetype fromUtf8Len(qsizetype l) { return l + 1; } |
1507 | static qsizetype toUtf8Len(qsizetype l) { return 3*(l + 1); } |
1508 | |
1509 | static qsizetype fromUtf16Len(qsizetype l) { return l/2 + 2; } |
1510 | static qsizetype toUtf16Len(qsizetype l) { return 2*(l + 1); } |
1511 | |
1512 | static qsizetype fromUtf32Len(qsizetype l) { return l/2 + 2; } |
1513 | static qsizetype toUtf32Len(qsizetype l) { return 4*(l + 1); } |
1514 | |
1515 | static qsizetype fromLatin1Len(qsizetype l) { return l + 1; } |
1516 | static qsizetype toLatin1Len(qsizetype l) { return l + 1; } |
1517 | |
1518 | |
1519 | |
1520 | /*! |
1521 | \class QStringConverterBase |
1522 | \internal |
1523 | |
1524 | Just a common base class for QStringConverter and QTextCodec |
1525 | */ |
1526 | |
1527 | /*! |
1528 | \class QStringConverter |
1529 | \inmodule QtCore |
1530 | \brief The QStringConverter class provides a base class for encoding and decoding text. |
1531 | \reentrant |
1532 | \ingroup i18n |
1533 | |
1534 | Qt uses UTF-16 to store, draw and manipulate strings. In many |
1535 | situations you may wish to deal with data that uses a different |
1536 | encoding. Most text data transferred over files and network connections is encoded |
1537 | in UTF-8. |
1538 | |
1539 | The QStringConverter class is a base class for the \l {QStringEncoder} and |
1540 | \l {QStringDecoder} classes that help with converting between different |
1541 | text encodings. QStringDecoder can decode a string from an encoded representation |
1542 | into UTF-16, the format Qt uses internally. QStringEncoder does the opposite |
1543 | operation, encoding UTF-16 encoded data (usually in the form of a QString) to |
1544 | the requested encoding. |
1545 | |
1546 | The supported encodings are: |
1547 | |
1548 | \list |
1549 | \li UTF-8 |
1550 | \li UTF-16 |
1551 | \li UTF-16BE |
1552 | \li UTF-16LE |
1553 | \li UTF-32 |
1554 | \li UTF-32BE |
1555 | \li UTF-32LE |
1556 | \li ISO-8859-1 (Latin-1) |
1557 | \li The system encoding |
1558 | \endlist |
1559 | |
1560 | \l {QStringConverter}s can be used as follows to convert some encoded |
1561 | string to and from UTF-16. |
1562 | |
1563 | Suppose you have some string encoded in UTF-8, and |
1564 | want to convert it to a QString. The simple way |
1565 | to do it is to use a \l {QStringDecoder} like this: |
1566 | |
1567 | \snippet code/src_corelib_text_qstringconverter.cpp 0 |
1568 | |
1569 | After this, \c string holds the text in decoded form. |
1570 | Converting a string from Unicode to the local encoding is just as |
1571 | easy using the \l {QStringEncoder} class: |
1572 | |
1573 | \snippet code/src_corelib_text_qstringconverter.cpp 1 |
1574 | |
1575 | To read or write text files in various encodings, use QTextStream and |
1576 | its \l{QTextStream::setEncoding()}{setEncoding()} function. |
1577 | |
1578 | Some care must be taken when trying to convert the data in chunks, |
1579 | for example, when receiving it over a network. In such cases it is |
1580 | possible that a multi-byte character will be split over two |
1581 | chunks. At best this might result in the loss of a character and |
1582 | at worst cause the entire conversion to fail. |
1583 | |
1584 | Both QStringEncoder and QStringDecoder make this easy, by tracking |
1585 | this in an internal state. So simply calling the encoder or decoder |
1586 | again with the next chunk of data will automatically continue encoding |
1587 | or decoding the data correctly: |
1588 | |
1589 | \snippet code/src_corelib_text_qstringconverter.cpp 2 |
1590 | |
1591 | The QStringDecoder object maintains state between chunks and therefore |
1592 | works correctly even if a multi-byte character is split between |
1593 | chunks. |
1594 | |
1595 | QStringConverter objects can't be copied because of their internal state, but |
1596 | can be moved. |
1597 | |
1598 | \sa QTextStream, QStringDecoder, QStringEncoder |
1599 | */ |
1600 | |
1601 | /*! |
1602 | \enum QStringConverter::Flag |
1603 | |
1604 | \value Default Default conversion rules apply. |
1605 | \value ConvertInvalidToNull If this flag is set, each invalid input |
1606 | character is output as a null character. If it is not set, |
1607 | invalid input characters are represented as QChar::ReplacementCharacter |
1608 | if the output encoding can represent that character, otherwise as a question mark. |
1609 | \value WriteBom When converting from a QString to an output encoding, write a QChar::ByteOrderMark as the first |
1610 | character if the output encoding supports this. This is the case for UTF-8, UTF-16 and UTF-32 |
1611 | encodings. |
1612 | \value ConvertInitialBom When converting from an input encoding to a QString the QStringDecoder usually skips an |
1613 | leading QChar::ByteOrderMark. When this flag is set, the byte order mark will not be |
1614 | skipped, but converted to utf-16 and inserted at the start of the created QString. |
1615 | \value Stateless Ignore possible converter states between different function calls |
1616 | to encode or decode strings. This will also cause the QStringConverter to raise an error if an incomplete |
1617 | sequence of data is encountered. |
1618 | */ |
1619 | |
1620 | /*! |
1621 | \enum QStringConverter::Encoding |
1622 | \value Utf8 Create a converter to or from UTF-8 |
1623 | \value Utf16 Create a converter to or from UTF-16. When decoding, the byte order will get automatically |
1624 | detected by a leading byte order mark. If none exists or when encoding, the system byte order will |
1625 | be assumed. |
1626 | \value Utf16BE Create a converter to or from big endian UTF-16. |
1627 | \value Utf16LE Create a converter to or from litte endian UTF-16. |
1628 | \value Utf32 Create a converter to or from UTF-32. When decoding, the byte order will get automatically |
1629 | detected by a leading byte order mark. If none exists or when encoding, the system byte order will |
1630 | be assumed. |
1631 | \value Utf32BE Create a converter to or from big endian UTF-32. |
1632 | \value Utf32LE Create a converter to or from litte endian UTF-32. |
1633 | \value Latin1 Create a converter to or from ISO-8859-1 (Latin1). |
1634 | \value System Create a converter to or from the underlying encoding of the |
1635 | operating systems locale. This is always assumed to be UTF-8 for Unix based |
1636 | systems. On Windows, this converts to and from the locale code page. |
1637 | */ |
1638 | |
1639 | /*! |
1640 | \struct QStringConverter::Interface |
1641 | \internal |
1642 | */ |
1643 | |
1644 | const QStringConverter::Interface QStringConverter::encodingInterfaces[QStringConverter::LastEncoding + 1] = |
1645 | { |
1646 | { "UTF-8" , QUtf8::convertToUnicode, fromUtf8Len, QUtf8::convertFromUnicode, toUtf8Len }, |
1647 | { "UTF-16" , fromUtf16, fromUtf16Len, toUtf16, toUtf16Len }, |
1648 | { "UTF-16LE" , fromUtf16LE, fromUtf16Len, toUtf16LE, toUtf16Len }, |
1649 | { "UTF-16BE" , fromUtf16BE, fromUtf16Len, toUtf16BE, toUtf16Len }, |
1650 | { "UTF-32" , fromUtf32, fromUtf32Len, toUtf32, toUtf32Len }, |
1651 | { "UTF-32LE" , fromUtf32LE, fromUtf32Len, toUtf32LE, toUtf32Len }, |
1652 | { "UTF-32BE" , fromUtf32BE, fromUtf32Len, toUtf32BE, toUtf32Len }, |
1653 | { "ISO-8859-1" , fromLatin1, fromLatin1Len, toLatin1, toLatin1Len }, |
1654 | { "Locale" , fromLocal8Bit, fromUtf8Len, toLocal8Bit, toUtf8Len } |
1655 | }; |
1656 | |
1657 | // match names case insensitive and skipping '-' and '_' |
1658 | static bool nameMatch(const char *a, const char *b) |
1659 | { |
1660 | while (*a && *b) { |
1661 | if (*a == '-' || *a == '_') { |
1662 | ++a; |
1663 | continue; |
1664 | } |
1665 | if (*b == '-' || *b == '_') { |
1666 | ++b; |
1667 | continue; |
1668 | } |
1669 | if (toupper(*a) != toupper(*b)) |
1670 | return false; |
1671 | ++a; |
1672 | ++b; |
1673 | } |
1674 | return !*a && !*b; |
1675 | } |
1676 | |
1677 | |
1678 | /*! |
1679 | \fn constexpr QStringConverter::QStringConverter() |
1680 | \internal |
1681 | */ |
1682 | |
1683 | /*! |
1684 | \fn constexpr QStringConverter::QStringConverter(Encoding, Flags) |
1685 | \internal |
1686 | */ |
1687 | |
1688 | /*! |
1689 | \internal |
1690 | */ |
1691 | QStringConverter::QStringConverter(const char *name, Flags f) |
1692 | : iface(nullptr), state(f) |
1693 | { |
1694 | auto e = encodingForName(name); |
1695 | if (e) |
1696 | iface = encodingInterfaces + int(e.value()); |
1697 | } |
1698 | |
1699 | /*! |
1700 | \fn bool QStringConverter::isValid() const |
1701 | |
1702 | Returns true if this is a valid string converter that can be used for encoding or |
1703 | decoding text. |
1704 | |
1705 | Default constructed string converters or converters constructed with an unsupported |
1706 | name are not valid. |
1707 | */ |
1708 | |
1709 | /*! |
1710 | \fn void QStringConverter::resetState() |
1711 | |
1712 | Resets the internal state of the converter, clearing potential errors or partial |
1713 | conversions. |
1714 | */ |
1715 | |
1716 | /*! |
1717 | \fn bool QStringConverter::hasError() const |
1718 | |
1719 | Returns true if a conversion could not correctly convert a character. This could for example |
1720 | get triggered by an invalid UTF-8 sequence or when a character can't get converted due to |
1721 | limitations in the target encoding. |
1722 | */ |
1723 | |
1724 | /*! |
1725 | \fn const char *QStringConverter::name() const |
1726 | |
1727 | Returns the canonical name of the encoding this QStringConverter can encode or decode. |
1728 | Returns a nullptr if the converter is not valid. |
1729 | |
1730 | \sa isValid() |
1731 | */ |
1732 | |
1733 | /*! |
1734 | Returns an optional encoding for \a name. The optional is empty if the name could |
1735 | not get converted to a valid encoding. |
1736 | */ |
1737 | std::optional<QStringConverter::Encoding> QStringConverter::encodingForName(const char *name) |
1738 | { |
1739 | for (int i = 0; i < LastEncoding + 1; ++i) { |
1740 | if (nameMatch(encodingInterfaces[i].name, name)) |
1741 | return QStringConverter::Encoding(i); |
1742 | } |
1743 | if (nameMatch(name, "latin1" )) |
1744 | return QStringConverter::Latin1; |
1745 | return std::nullopt; |
1746 | } |
1747 | |
1748 | /*! |
1749 | Returns the encoding for the content of \a data if it can be determined. |
1750 | \a expectedFirstCharacter can be passed as an additional hint to help determine |
1751 | the encoding. |
1752 | |
1753 | The returned optional is empty, if the encoding is unclear. |
1754 | */ |
1755 | std::optional<QStringConverter::Encoding> QStringConverter::encodingForData(QByteArrayView data, char16_t expectedFirstCharacter) |
1756 | { |
1757 | qsizetype arraySize = data.size(); |
1758 | if (arraySize > 3) { |
1759 | uint uc = qFromUnaligned<uint>(data.data()); |
1760 | if (uc == qToBigEndian(uint(QChar::ByteOrderMark))) |
1761 | return QStringConverter::Utf32BE; |
1762 | if (uc == qToLittleEndian(uint(QChar::ByteOrderMark))) |
1763 | return QStringConverter::Utf32LE; |
1764 | if (expectedFirstCharacter) { |
1765 | // catch also anything starting with the expected character |
1766 | if (qToLittleEndian(uc) == expectedFirstCharacter) |
1767 | return QStringConverter::Utf32LE; |
1768 | else if (qToBigEndian(uc) == expectedFirstCharacter) |
1769 | return QStringConverter::Utf32BE; |
1770 | } |
1771 | } |
1772 | |
1773 | if (arraySize > 2) { |
1774 | if (memcmp(data.data(), utf8bom, sizeof(utf8bom)) == 0) |
1775 | return QStringConverter::Utf8; |
1776 | } |
1777 | |
1778 | if (arraySize > 1) { |
1779 | ushort uc = qFromUnaligned<ushort>(data.data()); |
1780 | if (uc == qToBigEndian(ushort(QChar::ByteOrderMark))) |
1781 | return QStringConverter::Utf16BE; |
1782 | if (uc == qToLittleEndian(ushort(QChar::ByteOrderMark))) |
1783 | return QStringConverter::Utf16LE; |
1784 | if (expectedFirstCharacter) { |
1785 | // catch also anything starting with the expected character |
1786 | if (qToLittleEndian(uc) == expectedFirstCharacter) |
1787 | return QStringConverter::Utf16LE; |
1788 | else if (qToBigEndian(uc) == expectedFirstCharacter) |
1789 | return QStringConverter::Utf16BE; |
1790 | } |
1791 | } |
1792 | return std::nullopt; |
1793 | } |
1794 | |
1795 | /*! |
1796 | Tries to determine the encoding of the HTML in \a data by looking at leading byte |
1797 | order marks or a charset specifier in the HTML meta tag. If the optional is empty, |
1798 | the encoding specified is not supported by QStringConverter. If no encoding is |
1799 | detected, the method returns Utf8. |
1800 | */ |
1801 | std::optional<QStringConverter::Encoding> QStringConverter::encodingForHtml(QByteArrayView data) |
1802 | { |
1803 | // determine charset |
1804 | auto encoding = encodingForData(data); |
1805 | if (encoding) |
1806 | // trust the initial BOM |
1807 | return encoding; |
1808 | |
1809 | QByteArray = data.first(qMin(data.size(), qsizetype(1024))).toByteArray().toLower(); |
1810 | int pos = header.indexOf("meta " ); |
1811 | if (pos != -1) { |
1812 | pos = header.indexOf("charset=" , pos); |
1813 | if (pos != -1) { |
1814 | pos += int(qstrlen("charset=" )); |
1815 | if (pos < header.size() && (header.at(pos) == '\"' || header.at(pos) == '\'')) |
1816 | ++pos; |
1817 | |
1818 | int pos2 = pos; |
1819 | // The attribute can be closed with either """, "'", ">" or "/", |
1820 | // none of which are valid charset characters. |
1821 | while (++pos2 < header.size()) { |
1822 | char ch = header.at(pos2); |
1823 | if (ch == '\"' || ch == '\'' || ch == '>' || ch == '/') { |
1824 | QByteArray name = header.mid(pos, pos2 - pos); |
1825 | int colon = name.indexOf(':'); |
1826 | if (colon > 0) |
1827 | name = name.left(colon); |
1828 | name = name.simplified(); |
1829 | if (name == "unicode" ) // QTBUG-41998, ICU will return UTF-16. |
1830 | name = QByteArrayLiteral("UTF-8" ); |
1831 | if (!name.isEmpty()) |
1832 | return encodingForName(name); |
1833 | } |
1834 | } |
1835 | } |
1836 | } |
1837 | return Utf8; |
1838 | } |
1839 | |
1840 | /*! |
1841 | Returns the canonical name for \a encoding. |
1842 | */ |
1843 | const char *QStringConverter::nameForEncoding(QStringConverter::Encoding e) |
1844 | { |
1845 | return encodingInterfaces[int(e)].name; |
1846 | } |
1847 | |
1848 | /*! |
1849 | \class QStringEncoder |
1850 | \inmodule QtCore |
1851 | \brief The QStringEncoder class provides a state-based encoder for text. |
1852 | \reentrant |
1853 | \ingroup i18n |
1854 | |
1855 | A text encoder converts text from Qt's internal representation into an encoded |
1856 | text format using a specific encoding. |
1857 | |
1858 | Converting a string from Unicode to the local encoding can be achieved |
1859 | using the following code: |
1860 | |
1861 | \snippet code/src_corelib_text_qstringconverter.cpp 1 |
1862 | |
1863 | The encoder remembers any state that is required between calls, so converting |
1864 | data received in chunks, for example, when receiving it over a network, is just as |
1865 | easy, by calling the encoder whenever new data is available: |
1866 | |
1867 | \snippet code/src_corelib_text_qstringconverter.cpp 3 |
1868 | |
1869 | The QStringEncoder object maintains state between chunks and therefore |
1870 | works correctly even if a UTF-16 surrogate character is split between |
1871 | chunks. |
1872 | |
1873 | QStringEncoder objects can't be copied because of their internal state, but |
1874 | can be moved. |
1875 | |
1876 | \sa QStringConverter, QStringDecoder |
1877 | */ |
1878 | |
1879 | /*! |
1880 | \fn constexpr QStringEncoder::QStringEncoder(const Interface *i) |
1881 | \internal |
1882 | */ |
1883 | |
1884 | /*! |
1885 | \fn constexpr QStringEncoder::QStringEncoder() |
1886 | |
1887 | Default constructs an encoder. The default encoder is not valid, |
1888 | and can't be used for converting text. |
1889 | */ |
1890 | |
1891 | /*! |
1892 | \fn constexpr QStringEncoder::QStringEncoder(Encoding encoding, Flags flags = Flag::Default) |
1893 | |
1894 | Creates an encoder object using \a encoding and \a flags. |
1895 | */ |
1896 | |
1897 | /*! |
1898 | \fn constexpr QStringEncoder::QStringEncoder(const char *name, Flags flags = Flag::Default) |
1899 | |
1900 | Creates an encoder object using \a name and \a flags. |
1901 | If \a name is not the name of a known encoding an invalid converter will get created. |
1902 | |
1903 | \sa isValid() |
1904 | */ |
1905 | |
1906 | /*! |
1907 | \fn QByteArray QStringEncoder::operator()(const QString &in) |
1908 | \fn QByteArray QStringEncoder::encode(const QString &in) |
1909 | |
1910 | Converts \a in and returns the data as a byte array. |
1911 | */ |
1912 | |
1913 | /*! |
1914 | \fn QByteArray QStringEncoder::operator()(QStringView in) |
1915 | \fn QByteArray QStringEncoder::encode(QStringView in) |
1916 | \overload |
1917 | |
1918 | Converts \a in and returns the data as a byte array. |
1919 | */ |
1920 | |
1921 | /*! |
1922 | \fn QByteArray QStringEncoder::operator()(const QChar *in, qsizetype length) |
1923 | \fn QByteArray QStringEncoder::encode(const QChar *in, qsizetype length) |
1924 | \overload |
1925 | |
1926 | Converts \a length QChars from \a in and returns the data as a byte array. |
1927 | */ |
1928 | |
1929 | /*! |
1930 | \fn qsizetype QStringEncoder::requiredSpace(qsizetype inputLength) const |
1931 | |
1932 | Returns the maximum amount of characters required to be able to process |
1933 | \a inputLength decoded data. |
1934 | |
1935 | \sa appendToBuffer |
1936 | */ |
1937 | |
1938 | /*! |
1939 | \fn char *QStringEncoder::appendToBuffer(char *out, const QChar *in, qsizetype length) |
1940 | |
1941 | Encodes \a length QChars from \a in and writes the encoded result into the buffer |
1942 | starting at \a out. Returns a pointer to the end of data written. |
1943 | |
1944 | \a out needs to be large enough to be able to hold all the decoded data. Use |
1945 | \l{requiredSpace} to determine the maximum size requirements to be able to encode |
1946 | a QChar buffer of \a length. |
1947 | |
1948 | \sa requiredSpace |
1949 | */ |
1950 | |
1951 | /*! |
1952 | \class QStringDecoder |
1953 | \inmodule QtCore |
1954 | \brief The QStringDecoder class provides a state-based decoder for text. |
1955 | \reentrant |
1956 | \ingroup i18n |
1957 | |
1958 | A text decoder converts text an encoded text format that uses a specific encoding |
1959 | into Qt's internal representation. |
1960 | |
1961 | Converting encoded data into a QString can be achieved |
1962 | using the following code: |
1963 | |
1964 | \snippet code/src_corelib_text_qstringconverter.cpp 0 |
1965 | |
1966 | The decoder remembers any state that is required between calls, so converting |
1967 | data received in chunks, for example, when receiving it over a network, is just as |
1968 | easy, by calling the decoder whenever new data is available: |
1969 | |
1970 | \snippet code/src_corelib_text_qstringconverter.cpp 2 |
1971 | |
1972 | The QStringDecoder object maintains state between chunks and therefore |
1973 | works correctly even if chunks are split in the middle of a multi-byte character |
1974 | sequence. |
1975 | |
1976 | QStringDecoder objects can't be copied because of their internal state, but |
1977 | can be moved. |
1978 | |
1979 | \sa QStringConverter, QStringEncoder |
1980 | */ |
1981 | |
1982 | /*! |
1983 | \fn constexpr QStringDecoder::QStringDecoder(const Interface *i) |
1984 | \internal |
1985 | */ |
1986 | |
1987 | /*! |
1988 | \fn constexpr QStringDecoder::QStringDecoder() |
1989 | |
1990 | Default constructs an decoder. The default decoder is not valid, |
1991 | and can't be used for converting text. |
1992 | */ |
1993 | |
1994 | /*! |
1995 | \fn constexpr QStringDecoder::QStringDecoder(Encoding encoding, Flags flags = Flag::Default) |
1996 | |
1997 | Creates an decoder object using \a encoding and \a flags. |
1998 | */ |
1999 | |
2000 | /*! |
2001 | \fn constexpr QStringDecoder::QStringDecoder(const char *name, Flags flags = Flag::Default) |
2002 | |
2003 | Creates an decoder object using \a name and \a flags. |
2004 | If \a name is not the name of a known encoding an invalid converter will get created. |
2005 | |
2006 | \sa isValid() |
2007 | */ |
2008 | |
2009 | /*! |
2010 | \fn QString QStringDecoder::operator()(const QByteArray &ba) |
2011 | \fn QString QStringDecoder::decode(const QByteArray &ba) |
2012 | |
2013 | Converts \a ba and returns the data as a QString. |
2014 | */ |
2015 | |
2016 | /*! |
2017 | \fn QString QStringDecoder::operator()(const char *in, qsizetype size) |
2018 | \fn QString QStringDecoder::decode(const char *in, qsizetype size) |
2019 | \overload |
2020 | |
2021 | Converts a byte array containing the first \a size bytes of the array \a in |
2022 | and returns the data as a QString. |
2023 | */ |
2024 | |
2025 | /*! |
2026 | \fn QString QStringDecoder::operator()(const char *chars) |
2027 | \fn QString QStringDecoder::decode(const char *chars) |
2028 | \overload |
2029 | |
2030 | Converts \a chars and returns the data as a QString. \a chars is assumed to |
2031 | point to a \c{\0}-terminated string and its length is determined dynamically. |
2032 | */ |
2033 | |
2034 | /*! |
2035 | \fn qsizetype QStringDecoder::requiredSpace(qsizetype inputLength) const |
2036 | |
2037 | Returns the maximum amount of UTF-16 code units required to be able to process |
2038 | \a inputLength encoded data. |
2039 | |
2040 | \sa appendToBuffer |
2041 | */ |
2042 | |
2043 | /*! |
2044 | \fn QChar *QStringDecoder::appendToBuffer(QChar *out, QByteArrayView in) |
2045 | |
2046 | Decodes the sequence of bytes viewed by \a in and writes the decoded result into |
2047 | the buffer starting at \a out. Returns a pointer to the end of data written. |
2048 | |
2049 | \a out needs to be large enough to be able to hold all the decoded data. Use |
2050 | \l{requiredSpace} to determine the maximum size requirements to decode an encoded |
2051 | data buffer of \c in.size() bytes. |
2052 | |
2053 | \sa requiredSpace |
2054 | */ |
2055 | |
2056 | QT_END_NAMESPACE |
2057 | |