Line data Source code
1 : /* stringlib: codec implementations */
2 :
3 : #if !STRINGLIB_IS_UNICODE
4 : # error "codecs.h is specific to Unicode"
5 : #endif
6 :
7 : #include "pycore_bitutils.h" // _Py_bswap32()
8 :
9 : /* Mask to quickly check whether a C 'size_t' contains a
10 : non-ASCII, UTF8-encoded char. */
11 : #if (SIZEOF_SIZE_T == 8)
12 : # define ASCII_CHAR_MASK 0x8080808080808080ULL
13 : #elif (SIZEOF_SIZE_T == 4)
14 : # define ASCII_CHAR_MASK 0x80808080U
15 : #else
16 : # error C 'size_t' size should be either 4 or 8!
17 : #endif
18 :
19 : /* 10xxxxxx */
20 : #define IS_CONTINUATION_BYTE(ch) ((ch) >= 0x80 && (ch) < 0xC0)
21 :
22 : Py_LOCAL_INLINE(Py_UCS4)
23 808360 : STRINGLIB(utf8_decode)(const char **inptr, const char *end,
24 : STRINGLIB_CHAR *dest,
25 : Py_ssize_t *outpos)
26 : {
27 : Py_UCS4 ch;
28 808360 : const char *s = *inptr;
29 808360 : STRINGLIB_CHAR *p = dest + *outpos;
30 :
31 4643027 : while (s < end) {
32 4437657 : ch = (unsigned char)*s;
33 :
34 4437657 : if (ch < 0x80) {
35 : /* Fast path for runs of ASCII characters. Given that common UTF-8
36 : input will consist of an overwhelming majority of ASCII
37 : characters, we try to optimize for this case by checking
38 : as many characters as a C 'size_t' can contain.
39 : First, check if we can do an aligned read, as most CPUs have
40 : a penalty for unaligned reads.
41 : */
42 1645830 : if (_Py_IS_ALIGNED(s, ALIGNOF_SIZE_T)) {
43 : /* Help register allocation */
44 201628 : const char *_s = s;
45 201628 : STRINGLIB_CHAR *_p = p;
46 3552107 : while (_s + SIZEOF_SIZE_T <= end) {
47 : /* Read a whole size_t at a time (either 4 or 8 bytes),
48 : and do a fast unrolled copy if it only contains ASCII
49 : characters. */
50 3435945 : size_t value = *(const size_t *) _s;
51 3435945 : if (value & ASCII_CHAR_MASK)
52 85472 : break;
53 : #if PY_LITTLE_ENDIAN
54 3350480 : _p[0] = (STRINGLIB_CHAR)(value & 0xFFu);
55 3350480 : _p[1] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
56 3350480 : _p[2] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
57 3350480 : _p[3] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
58 : # if SIZEOF_SIZE_T == 8
59 3350480 : _p[4] = (STRINGLIB_CHAR)((value >> 32) & 0xFFu);
60 3350480 : _p[5] = (STRINGLIB_CHAR)((value >> 40) & 0xFFu);
61 3350480 : _p[6] = (STRINGLIB_CHAR)((value >> 48) & 0xFFu);
62 3350480 : _p[7] = (STRINGLIB_CHAR)((value >> 56) & 0xFFu);
63 : # endif
64 : #else
65 : # if SIZEOF_SIZE_T == 8
66 : _p[0] = (STRINGLIB_CHAR)((value >> 56) & 0xFFu);
67 : _p[1] = (STRINGLIB_CHAR)((value >> 48) & 0xFFu);
68 : _p[2] = (STRINGLIB_CHAR)((value >> 40) & 0xFFu);
69 : _p[3] = (STRINGLIB_CHAR)((value >> 32) & 0xFFu);
70 : _p[4] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
71 : _p[5] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
72 : _p[6] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
73 : _p[7] = (STRINGLIB_CHAR)(value & 0xFFu);
74 : # else
75 : _p[0] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
76 : _p[1] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
77 : _p[2] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
78 : _p[3] = (STRINGLIB_CHAR)(value & 0xFFu);
79 : # endif
80 : #endif
81 3350480 : _s += SIZEOF_SIZE_T;
82 3350480 : _p += SIZEOF_SIZE_T;
83 : }
84 201628 : s = _s;
85 201628 : p = _p;
86 201628 : if (s == end)
87 7978 : break;
88 193650 : ch = (unsigned char)*s;
89 : }
90 1637852 : if (ch < 0x80) {
91 1632922 : s++;
92 1632922 : *p++ = ch;
93 1632922 : continue;
94 : }
95 : }
96 :
97 2796764 : if (ch < 0xE0) {
98 : /* \xC2\x80-\xDF\xBF -- 0080-07FF */
99 : Py_UCS4 ch2;
100 387399 : if (ch < 0xC2) {
101 : /* invalid sequence
102 : \x80-\xBF -- continuation byte
103 : \xC0-\xC1 -- fake 0000-007F */
104 971 : goto InvalidStart;
105 : }
106 386428 : if (end - s < 2) {
107 : /* unexpected end of data: the caller will decide whether
108 : it's an error or not */
109 2217 : break;
110 : }
111 384211 : ch2 = (unsigned char)s[1];
112 384211 : if (!IS_CONTINUATION_BYTE(ch2))
113 : /* invalid continuation byte */
114 1106 : goto InvalidContinuation1;
115 383105 : ch = (ch << 6) + ch2 -
116 : ((0xC0 << 6) + 0x80);
117 383105 : assert ((ch > 0x007F) && (ch <= 0x07FF));
118 383105 : s += 2;
119 147520 : if (STRINGLIB_MAX_CHAR <= 0x007F ||
120 : (STRINGLIB_MAX_CHAR < 0x07FF && ch > STRINGLIB_MAX_CHAR))
121 : /* Out-of-range */
122 139583 : goto Return;
123 243522 : *p++ = ch;
124 243522 : continue;
125 : }
126 :
127 2409362 : if (ch < 0xF0) {
128 : /* \xE0\xA0\x80-\xEF\xBF\xBF -- 0800-FFFF */
129 : Py_UCS4 ch2, ch3;
130 1250548 : if (end - s < 3) {
131 : /* unexpected end of data: the caller will decide whether
132 : it's an error or not */
133 119993 : if (end - s < 2)
134 58166 : break;
135 61827 : ch2 = (unsigned char)s[1];
136 61827 : if (!IS_CONTINUATION_BYTE(ch2) ||
137 : (ch2 < 0xA0 ? ch == 0xE0 : ch == 0xED))
138 : /* for clarification see comments below */
139 119 : goto InvalidContinuation1;
140 61708 : break;
141 : }
142 1130555 : ch2 = (unsigned char)s[1];
143 1130555 : ch3 = (unsigned char)s[2];
144 1130555 : if (!IS_CONTINUATION_BYTE(ch2)) {
145 : /* invalid continuation byte */
146 8059 : goto InvalidContinuation1;
147 : }
148 1122496 : if (ch == 0xE0) {
149 20950 : if (ch2 < 0xA0)
150 : /* invalid sequence
151 : \xE0\x80\x80-\xE0\x9F\xBF -- fake 0000-0800 */
152 68 : goto InvalidContinuation1;
153 1101546 : } else if (ch == 0xED && ch2 >= 0xA0) {
154 : /* Decoding UTF-8 sequences in range \xED\xA0\x80-\xED\xBF\xBF
155 : will result in surrogates in range D800-DFFF. Surrogates are
156 : not valid UTF-8 so they are rejected.
157 : See https://www.unicode.org/versions/Unicode5.2.0/ch03.pdf
158 : (table 3-7) and http://www.rfc-editor.org/rfc/rfc3629.txt */
159 1487 : goto InvalidContinuation1;
160 : }
161 1120941 : if (!IS_CONTINUATION_BYTE(ch3)) {
162 : /* invalid continuation byte */
163 288 : goto InvalidContinuation2;
164 : }
165 1120653 : ch = (ch << 12) + (ch2 << 6) + ch3 -
166 : ((0xE0 << 12) + (0x80 << 6) + 0x80);
167 1120653 : assert ((ch > 0x07FF) && (ch <= 0xFFFF));
168 1120653 : s += 3;
169 : if (STRINGLIB_MAX_CHAR <= 0x07FF ||
170 : (STRINGLIB_MAX_CHAR < 0xFFFF && ch > STRINGLIB_MAX_CHAR))
171 : /* Out-of-range */
172 246326 : goto Return;
173 874327 : *p++ = ch;
174 874327 : continue;
175 : }
176 :
177 1158810 : if (ch < 0xF5) {
178 : /* \xF0\x90\x80\x80-\xF4\x8F\xBF\xBF -- 10000-10FFFF */
179 : Py_UCS4 ch2, ch3, ch4;
180 1154816 : if (end - s < 4) {
181 : /* unexpected end of data: the caller will decide whether
182 : it's an error or not */
183 6656 : if (end - s < 2)
184 6140 : break;
185 516 : ch2 = (unsigned char)s[1];
186 516 : if (!IS_CONTINUATION_BYTE(ch2) ||
187 : (ch2 < 0x90 ? ch == 0xF0 : ch == 0xF4))
188 : /* for clarification see comments below */
189 127 : goto InvalidContinuation1;
190 389 : if (end - s < 3)
191 125 : break;
192 264 : ch3 = (unsigned char)s[2];
193 264 : if (!IS_CONTINUATION_BYTE(ch3))
194 111 : goto InvalidContinuation2;
195 153 : break;
196 : }
197 1148162 : ch2 = (unsigned char)s[1];
198 1148162 : ch3 = (unsigned char)s[2];
199 1148162 : ch4 = (unsigned char)s[3];
200 1148162 : if (!IS_CONTINUATION_BYTE(ch2)) {
201 : /* invalid continuation byte */
202 94 : goto InvalidContinuation1;
203 : }
204 1148068 : if (ch == 0xF0) {
205 295313 : if (ch2 < 0x90)
206 : /* invalid sequence
207 : \xF0\x80\x80\x80-\xF0\x8F\xBF\xBF -- fake 0000-FFFF */
208 36 : goto InvalidContinuation1;
209 852751 : } else if (ch == 0xF4 && ch2 >= 0x90) {
210 : /* invalid sequence
211 : \xF4\x90\x80\x80- -- 110000- overflow */
212 100 : goto InvalidContinuation1;
213 : }
214 1147932 : if (!IS_CONTINUATION_BYTE(ch3)) {
215 : /* invalid continuation byte */
216 98 : goto InvalidContinuation2;
217 : }
218 1147834 : if (!IS_CONTINUATION_BYTE(ch4)) {
219 : /* invalid continuation byte */
220 369 : goto InvalidContinuation3;
221 : }
222 1147465 : ch = (ch << 18) + (ch2 << 12) + (ch3 << 6) + ch4 -
223 : ((0xF0 << 18) + (0x80 << 12) + (0x80 << 6) + 0x80);
224 1147465 : assert ((ch > 0xFFFF) && (ch <= 0x10FFFF));
225 1147465 : s += 4;
226 : if (STRINGLIB_MAX_CHAR <= 0xFFFF ||
227 : (STRINGLIB_MAX_CHAR < 0x10FFFF && ch > STRINGLIB_MAX_CHAR))
228 : /* Out-of-range */
229 63565 : goto Return;
230 1083900 : *p++ = ch;
231 1083900 : continue;
232 : }
233 3998 : goto InvalidStart;
234 : }
235 341855 : ch = 0;
236 808360 : Return:
237 808360 : *inptr = s;
238 808360 : *outpos = p - dest;
239 808360 : return ch;
240 4969 : InvalidStart:
241 4969 : ch = 1;
242 4969 : goto Return;
243 11196 : InvalidContinuation1:
244 11196 : ch = 2;
245 11196 : goto Return;
246 497 : InvalidContinuation2:
247 497 : ch = 3;
248 497 : goto Return;
249 369 : InvalidContinuation3:
250 369 : ch = 4;
251 369 : goto Return;
252 : }
253 :
254 : #undef ASCII_CHAR_MASK
255 :
256 :
257 : /* UTF-8 encoder specialized for a Unicode kind to avoid the slow
258 : PyUnicode_READ() macro. Delete some parts of the code depending on the kind:
259 : UCS-1 strings don't need to handle surrogates for example. */
260 : Py_LOCAL_INLINE(char *)
261 1498671 : STRINGLIB(utf8_encoder)(_PyBytesWriter *writer,
262 : PyObject *unicode,
263 : const STRINGLIB_CHAR *data,
264 : Py_ssize_t size,
265 : _Py_error_handler error_handler,
266 : const char *errors)
267 : {
268 : Py_ssize_t i; /* index into data of next input character */
269 : char *p; /* next free byte in output buffer */
270 : #if STRINGLIB_SIZEOF_CHAR > 1
271 1311449 : PyObject *error_handler_obj = NULL;
272 1311449 : PyObject *exc = NULL;
273 1311449 : PyObject *rep = NULL;
274 : #endif
275 : #if STRINGLIB_SIZEOF_CHAR == 1
276 187222 : const Py_ssize_t max_char_size = 2;
277 : #elif STRINGLIB_SIZEOF_CHAR == 2
278 247869 : const Py_ssize_t max_char_size = 3;
279 : #else /* STRINGLIB_SIZEOF_CHAR == 4 */
280 1063580 : const Py_ssize_t max_char_size = 4;
281 : #endif
282 :
283 1498671 : assert(size >= 0);
284 1498671 : if (size > PY_SSIZE_T_MAX / max_char_size) {
285 : /* integer overflow */
286 0 : PyErr_NoMemory();
287 0 : return NULL;
288 : }
289 :
290 1498671 : _PyBytesWriter_Init(writer);
291 1498671 : p = _PyBytesWriter_Alloc(writer, size * max_char_size);
292 1498671 : if (p == NULL)
293 0 : return NULL;
294 :
295 104179600 : for (i = 0; i < size;) {
296 102681300 : Py_UCS4 ch = data[i++];
297 :
298 102681300 : if (ch < 0x80) {
299 : /* Encode ASCII */
300 92471000 : *p++ = (char) ch;
301 :
302 : }
303 : else
304 : #if STRINGLIB_SIZEOF_CHAR > 1
305 9837690 : if (ch < 0x0800)
306 : #endif
307 : {
308 : /* Encode Latin-1 */
309 449777 : *p++ = (char)(0xc0 | (ch >> 6));
310 449777 : *p++ = (char)(0x80 | (ch & 0x3f));
311 : }
312 : #if STRINGLIB_SIZEOF_CHAR > 1
313 9760630 : else if (Py_UNICODE_IS_SURROGATE(ch)) {
314 : Py_ssize_t startpos, endpos, newpos;
315 : Py_ssize_t k;
316 13114 : if (error_handler == _Py_ERROR_UNKNOWN) {
317 4461 : error_handler = _Py_GetErrorHandler(errors);
318 : }
319 :
320 13114 : startpos = i-1;
321 13114 : endpos = startpos+1;
322 :
323 19901 : while ((endpos < size) && Py_UNICODE_IS_SURROGATE(data[endpos]))
324 6787 : endpos++;
325 :
326 : /* Only overallocate the buffer if it's not the last write */
327 13114 : writer->overallocate = (endpos < size);
328 :
329 13114 : switch (error_handler)
330 : {
331 4 : case _Py_ERROR_REPLACE:
332 4 : memset(p, '?', endpos - startpos);
333 4 : p += (endpos - startpos);
334 : /* fall through */
335 8 : case _Py_ERROR_IGNORE:
336 8 : i += (endpos - startpos - 1);
337 8 : break;
338 :
339 8448 : case _Py_ERROR_SURROGATEPASS:
340 23218 : for (k=startpos; k<endpos; k++) {
341 14770 : ch = data[k];
342 14770 : *p++ = (char)(0xe0 | (ch >> 12));
343 14770 : *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
344 14770 : *p++ = (char)(0x80 | (ch & 0x3f));
345 : }
346 8448 : i += (endpos - startpos - 1);
347 8448 : break;
348 :
349 5 : case _Py_ERROR_BACKSLASHREPLACE:
350 : /* subtract preallocated bytes */
351 5 : writer->min_size -= max_char_size * (endpos - startpos);
352 5 : p = backslashreplace(writer, p,
353 : unicode, startpos, endpos);
354 5 : if (p == NULL)
355 382 : goto error;
356 5 : i += (endpos - startpos - 1);
357 5 : break;
358 :
359 2 : case _Py_ERROR_XMLCHARREFREPLACE:
360 : /* subtract preallocated bytes */
361 2 : writer->min_size -= max_char_size * (endpos - startpos);
362 2 : p = xmlcharrefreplace(writer, p,
363 : unicode, startpos, endpos);
364 2 : if (p == NULL)
365 0 : goto error;
366 2 : i += (endpos - startpos - 1);
367 2 : break;
368 :
369 4269 : case _Py_ERROR_SURROGATEESCAPE:
370 8574 : for (k=startpos; k<endpos; k++) {
371 4359 : ch = data[k];
372 4359 : if (!(0xDC80 <= ch && ch <= 0xDCFF))
373 : break;
374 4305 : *p++ = (char)(ch & 0xff);
375 : }
376 4269 : if (k >= endpos) {
377 4215 : i += (endpos - startpos - 1);
378 4215 : break;
379 : }
380 54 : startpos = k;
381 54 : assert(startpos < endpos);
382 : /* fall through */
383 : default:
384 436 : rep = unicode_encode_call_errorhandler(
385 : errors, &error_handler_obj, "utf-8", "surrogates not allowed",
386 : unicode, &exc, startpos, endpos, &newpos);
387 436 : if (!rep)
388 379 : goto error;
389 :
390 57 : if (newpos < startpos) {
391 51 : writer->overallocate = 1;
392 51 : p = _PyBytesWriter_Prepare(writer, p,
393 51 : max_char_size * (startpos - newpos));
394 51 : if (p == NULL)
395 0 : goto error;
396 : }
397 : else {
398 : /* subtract preallocated bytes */
399 6 : writer->min_size -= max_char_size * (newpos - startpos);
400 : /* Only overallocate the buffer if it's not the last write */
401 6 : writer->overallocate = (newpos < size);
402 : }
403 :
404 57 : if (PyBytes_Check(rep)) {
405 2 : p = _PyBytesWriter_WriteBytes(writer, p,
406 1 : PyBytes_AS_STRING(rep),
407 : PyBytes_GET_SIZE(rep));
408 : }
409 : else {
410 : /* rep is unicode */
411 56 : if (PyUnicode_READY(rep) < 0)
412 0 : goto error;
413 :
414 56 : if (!PyUnicode_IS_ASCII(rep)) {
415 3 : raise_encode_exception(&exc, "utf-8", unicode,
416 : startpos, endpos,
417 : "surrogates not allowed");
418 3 : goto error;
419 : }
420 :
421 106 : p = _PyBytesWriter_WriteBytes(writer, p,
422 53 : PyUnicode_DATA(rep),
423 : PyUnicode_GET_LENGTH(rep));
424 : }
425 :
426 54 : if (p == NULL)
427 0 : goto error;
428 54 : Py_CLEAR(rep);
429 :
430 54 : i = newpos;
431 : }
432 :
433 : /* If overallocation was disabled, ensure that it was the last
434 : write. Otherwise, we missed an optimization */
435 12732 : assert(writer->overallocate || i == size);
436 : }
437 : else
438 : #if STRINGLIB_SIZEOF_CHAR > 2
439 8527880 : if (ch < 0x10000)
440 : #endif
441 : {
442 1324229 : *p++ = (char)(0xe0 | (ch >> 12));
443 1324229 : *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
444 1324229 : *p++ = (char)(0x80 | (ch & 0x3f));
445 : }
446 : #if STRINGLIB_SIZEOF_CHAR > 2
447 : else /* ch >= 0x10000 */
448 : {
449 8423290 : assert(ch <= MAX_UNICODE);
450 : /* Encode UCS4 Unicode ordinals */
451 8423290 : *p++ = (char)(0xf0 | (ch >> 18));
452 8423290 : *p++ = (char)(0x80 | ((ch >> 12) & 0x3f));
453 8423290 : *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
454 8423290 : *p++ = (char)(0x80 | (ch & 0x3f));
455 : }
456 : #endif /* STRINGLIB_SIZEOF_CHAR > 2 */
457 : #endif /* STRINGLIB_SIZEOF_CHAR > 1 */
458 : }
459 :
460 : #if STRINGLIB_SIZEOF_CHAR > 1
461 1311067 : Py_XDECREF(error_handler_obj);
462 1311067 : Py_XDECREF(exc);
463 : #endif
464 1498289 : return p;
465 :
466 : #if STRINGLIB_SIZEOF_CHAR > 1
467 382 : error:
468 382 : Py_XDECREF(rep);
469 382 : Py_XDECREF(error_handler_obj);
470 382 : Py_XDECREF(exc);
471 382 : return NULL;
472 : #endif
473 : }
474 :
475 : /* The pattern for constructing UCS2-repeated masks. */
476 : #if SIZEOF_LONG == 8
477 : # define UCS2_REPEAT_MASK 0x0001000100010001ul
478 : #elif SIZEOF_LONG == 4
479 : # define UCS2_REPEAT_MASK 0x00010001ul
480 : #else
481 : # error C 'long' size should be either 4 or 8!
482 : #endif
483 :
484 : /* The mask for fast checking. */
485 : #if STRINGLIB_SIZEOF_CHAR == 1
486 : /* The mask for fast checking of whether a C 'long' contains a
487 : non-ASCII or non-Latin1 UTF16-encoded characters. */
488 : # define FAST_CHAR_MASK (UCS2_REPEAT_MASK * (0xFFFFu & ~STRINGLIB_MAX_CHAR))
489 : #else
490 : /* The mask for fast checking of whether a C 'long' may contain
491 : UTF16-encoded surrogate characters. This is an efficient heuristic,
492 : assuming that non-surrogate characters with a code point >= 0x8000 are
493 : rare in most input.
494 : */
495 : # define FAST_CHAR_MASK (UCS2_REPEAT_MASK * 0x8000u)
496 : #endif
497 : /* The mask for fast byte-swapping. */
498 : #define STRIPPED_MASK (UCS2_REPEAT_MASK * 0x00FFu)
499 : /* Swap bytes. */
500 : #define SWAB(value) ((((value) >> 8) & STRIPPED_MASK) | \
501 : (((value) & STRIPPED_MASK) << 8))
502 :
503 : Py_LOCAL_INLINE(Py_UCS4)
504 35617 : STRINGLIB(utf16_decode)(const unsigned char **inptr, const unsigned char *e,
505 : STRINGLIB_CHAR *dest, Py_ssize_t *outpos,
506 : int native_ordering)
507 : {
508 : Py_UCS4 ch;
509 35617 : const unsigned char *q = *inptr;
510 35617 : STRINGLIB_CHAR *p = dest + *outpos;
511 : /* Offsets from q for retrieving byte pairs in the right order. */
512 : #if PY_LITTLE_ENDIAN
513 35617 : int ihi = !!native_ordering, ilo = !native_ordering;
514 : #else
515 : int ihi = !native_ordering, ilo = !!native_ordering;
516 : #endif
517 35617 : --e;
518 :
519 149276 : while (q < e) {
520 : Py_UCS4 ch2;
521 : /* First check for possible aligned read of a C 'long'. Unaligned
522 : reads are more expensive, better to defer to another iteration. */
523 120891 : if (_Py_IS_ALIGNED(q, ALIGNOF_LONG)) {
524 : /* Fast path for runs of in-range non-surrogate chars. */
525 34190 : const unsigned char *_q = q;
526 238329 : while (_q + SIZEOF_LONG <= e) {
527 208001 : unsigned long block = * (const unsigned long *) _q;
528 208001 : if (native_ordering) {
529 : /* Can use buffer directly */
530 139938 : if (block & FAST_CHAR_MASK)
531 2545 : break;
532 : }
533 : else {
534 : /* Need to byte-swap */
535 68063 : if (block & SWAB(FAST_CHAR_MASK))
536 1317 : break;
537 : #if STRINGLIB_SIZEOF_CHAR == 1
538 61077 : block >>= 8;
539 : #else
540 5669 : block = SWAB(block);
541 : #endif
542 : }
543 : #if PY_LITTLE_ENDIAN
544 : # if SIZEOF_LONG == 4
545 : p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu);
546 : p[1] = (STRINGLIB_CHAR)(block >> 16);
547 : # elif SIZEOF_LONG == 8
548 204139 : p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu);
549 204139 : p[1] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu);
550 204139 : p[2] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu);
551 204139 : p[3] = (STRINGLIB_CHAR)(block >> 48);
552 : # endif
553 : #else
554 : # if SIZEOF_LONG == 4
555 : p[0] = (STRINGLIB_CHAR)(block >> 16);
556 : p[1] = (STRINGLIB_CHAR)(block & 0xFFFFu);
557 : # elif SIZEOF_LONG == 8
558 : p[0] = (STRINGLIB_CHAR)(block >> 48);
559 : p[1] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu);
560 : p[2] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu);
561 : p[3] = (STRINGLIB_CHAR)(block & 0xFFFFu);
562 : # endif
563 : #endif
564 204139 : _q += SIZEOF_LONG;
565 204139 : p += SIZEOF_LONG / 2;
566 : }
567 34190 : q = _q;
568 34190 : if (q >= e)
569 88 : break;
570 : }
571 :
572 120803 : ch = (q[ihi] << 8) | q[ilo];
573 120803 : q += 2;
574 120803 : if (!Py_UNICODE_IS_SURROGATE(ch)) {
575 : #if STRINGLIB_SIZEOF_CHAR < 2
576 99890 : if (ch > STRINGLIB_MAX_CHAR)
577 : /* Out-of-range */
578 6930 : goto Return;
579 : #endif
580 113639 : *p++ = (STRINGLIB_CHAR)ch;
581 113639 : continue;
582 : }
583 :
584 : /* UTF-16 code pair: */
585 234 : if (!Py_UNICODE_IS_HIGH_SURROGATE(ch))
586 105 : goto IllegalEncoding;
587 129 : if (q >= e)
588 44 : goto UnexpectedEnd;
589 85 : ch2 = (q[ihi] << 8) | q[ilo];
590 85 : q += 2;
591 85 : if (!Py_UNICODE_IS_LOW_SURROGATE(ch2))
592 12 : goto IllegalSurrogate;
593 73 : ch = Py_UNICODE_JOIN_SURROGATES(ch, ch2);
594 : #if STRINGLIB_SIZEOF_CHAR < 4
595 : /* Out-of-range */
596 53 : goto Return;
597 : #else
598 20 : *p++ = (STRINGLIB_CHAR)ch;
599 : #endif
600 : }
601 28473 : ch = 0;
602 35617 : Return:
603 35617 : *inptr = q;
604 35617 : *outpos = p - dest;
605 35617 : return ch;
606 44 : UnexpectedEnd:
607 44 : ch = 1;
608 44 : goto Return;
609 105 : IllegalEncoding:
610 105 : ch = 2;
611 105 : goto Return;
612 12 : IllegalSurrogate:
613 12 : ch = 3;
614 12 : goto Return;
615 : }
616 : #undef UCS2_REPEAT_MASK
617 : #undef FAST_CHAR_MASK
618 : #undef STRIPPED_MASK
619 : #undef SWAB
620 :
621 :
622 : #if STRINGLIB_MAX_CHAR >= 0x80
623 : Py_LOCAL_INLINE(Py_ssize_t)
624 5979 : STRINGLIB(utf16_encode)(const STRINGLIB_CHAR *in,
625 : Py_ssize_t len,
626 : unsigned short **outptr,
627 : int native_ordering)
628 : {
629 5979 : unsigned short *out = *outptr;
630 5979 : const STRINGLIB_CHAR *end = in + len;
631 : #if STRINGLIB_SIZEOF_CHAR == 1
632 2912 : if (native_ordering) {
633 2010 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
634 125307 : while (in < unrolled_end) {
635 123297 : out[0] = in[0];
636 123297 : out[1] = in[1];
637 123297 : out[2] = in[2];
638 123297 : out[3] = in[3];
639 123297 : in += 4; out += 4;
640 : }
641 4149 : while (in < end) {
642 2139 : *out++ = *in++;
643 : }
644 : } else {
645 : # define SWAB2(CH) ((CH) << 8) /* high byte is zero */
646 902 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
647 56481 : while (in < unrolled_end) {
648 55579 : out[0] = SWAB2(in[0]);
649 55579 : out[1] = SWAB2(in[1]);
650 55579 : out[2] = SWAB2(in[2]);
651 55579 : out[3] = SWAB2(in[3]);
652 55579 : in += 4; out += 4;
653 : }
654 1856 : while (in < end) {
655 954 : Py_UCS4 ch = *in++;
656 954 : *out++ = SWAB2((Py_UCS2)ch);
657 : }
658 : #undef SWAB2
659 : }
660 2912 : *outptr = out;
661 2912 : return len;
662 : #else
663 3067 : if (native_ordering) {
664 : #if STRINGLIB_MAX_CHAR < 0x10000
665 2026 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
666 42796 : while (in < unrolled_end) {
667 : /* check if any character is a surrogate character */
668 40774 : if (((in[0] ^ 0xd800) &
669 40774 : (in[1] ^ 0xd800) &
670 40774 : (in[2] ^ 0xd800) &
671 40774 : (in[3] ^ 0xd800) & 0xf800) == 0)
672 4 : break;
673 40770 : out[0] = in[0];
674 40770 : out[1] = in[1];
675 40770 : out[2] = in[2];
676 40770 : out[3] = in[3];
677 40770 : in += 4; out += 4;
678 : }
679 : #endif
680 4977 : while (in < end) {
681 : Py_UCS4 ch;
682 2998 : ch = *in++;
683 2998 : if (ch < 0xd800)
684 2842 : *out++ = ch;
685 156 : else if (ch < 0xe000)
686 : /* reject surrogate characters (U+D800-U+DFFF) */
687 87 : goto fail;
688 : #if STRINGLIB_MAX_CHAR >= 0x10000
689 69 : else if (ch >= 0x10000) {
690 61 : out[0] = Py_UNICODE_HIGH_SURROGATE(ch);
691 61 : out[1] = Py_UNICODE_LOW_SURROGATE(ch);
692 61 : out += 2;
693 : }
694 : #endif
695 : else
696 8 : *out++ = ch;
697 : }
698 : } else {
699 : #define SWAB2(CH) (((CH) << 8) | ((CH) >> 8))
700 : #if STRINGLIB_MAX_CHAR < 0x10000
701 979 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
702 19826 : while (in < unrolled_end) {
703 : /* check if any character is a surrogate character */
704 18849 : if (((in[0] ^ 0xd800) &
705 18849 : (in[1] ^ 0xd800) &
706 18849 : (in[2] ^ 0xd800) &
707 18849 : (in[3] ^ 0xd800) & 0xf800) == 0)
708 2 : break;
709 18847 : out[0] = SWAB2(in[0]);
710 18847 : out[1] = SWAB2(in[1]);
711 18847 : out[2] = SWAB2(in[2]);
712 18847 : out[3] = SWAB2(in[3]);
713 18847 : in += 4; out += 4;
714 : }
715 : #endif
716 2452 : while (in < end) {
717 1471 : Py_UCS4 ch = *in++;
718 1471 : if (ch < 0xd800)
719 1414 : *out++ = SWAB2((Py_UCS2)ch);
720 57 : else if (ch < 0xe000)
721 : /* reject surrogate characters (U+D800-U+DFFF) */
722 20 : goto fail;
723 : #if STRINGLIB_MAX_CHAR >= 0x10000
724 37 : else if (ch >= 0x10000) {
725 33 : Py_UCS2 ch1 = Py_UNICODE_HIGH_SURROGATE(ch);
726 33 : Py_UCS2 ch2 = Py_UNICODE_LOW_SURROGATE(ch);
727 33 : out[0] = SWAB2(ch1);
728 33 : out[1] = SWAB2(ch2);
729 33 : out += 2;
730 : }
731 : #endif
732 : else
733 4 : *out++ = SWAB2((Py_UCS2)ch);
734 : }
735 : #undef SWAB2
736 : }
737 2960 : *outptr = out;
738 2960 : return len;
739 107 : fail:
740 107 : *outptr = out;
741 107 : return len - (end - in + 1);
742 : #endif
743 : }
744 :
745 : static inline uint32_t
746 294777 : STRINGLIB(SWAB4)(STRINGLIB_CHAR ch)
747 : {
748 294777 : uint32_t word = ch;
749 : #if STRINGLIB_SIZEOF_CHAR == 1
750 : /* high bytes are zero */
751 219139 : return (word << 24);
752 : #elif STRINGLIB_SIZEOF_CHAR == 2
753 : /* high bytes are zero */
754 75595 : return ((word & 0x00FFu) << 24) | ((word & 0xFF00u) << 8);
755 : #else
756 43 : return _Py_bswap32(word);
757 : #endif
758 : }
759 :
760 : Py_LOCAL_INLINE(Py_ssize_t)
761 2618 : STRINGLIB(utf32_encode)(const STRINGLIB_CHAR *in,
762 : Py_ssize_t len,
763 : uint32_t **outptr,
764 : int native_ordering)
765 : {
766 2618 : uint32_t *out = *outptr;
767 2618 : const STRINGLIB_CHAR *end = in + len;
768 2618 : if (native_ordering) {
769 1817 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
770 162449 : while (in < unrolled_end) {
771 : #if STRINGLIB_SIZEOF_CHAR > 1
772 : /* check if any character is a surrogate character */
773 40756 : if (((in[0] ^ 0xd800) &
774 40756 : (in[1] ^ 0xd800) &
775 40756 : (in[2] ^ 0xd800) &
776 40756 : (in[3] ^ 0xd800) & 0xf800) == 0)
777 12 : break;
778 : #endif
779 160632 : out[0] = in[0];
780 160632 : out[1] = in[1];
781 160632 : out[2] = in[2];
782 160632 : out[3] = in[3];
783 160632 : in += 4; out += 4;
784 : }
785 3697 : while (in < end) {
786 : Py_UCS4 ch;
787 1969 : ch = *in++;
788 : #if STRINGLIB_SIZEOF_CHAR > 1
789 568 : if (Py_UNICODE_IS_SURROGATE(ch)) {
790 : /* reject surrogate characters (U+D800-U+DFFF) */
791 89 : goto fail;
792 : }
793 : #endif
794 1880 : *out++ = ch;
795 : }
796 : } else {
797 801 : const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
798 74285 : while (in < unrolled_end) {
799 : #if STRINGLIB_SIZEOF_CHAR > 1
800 : /* check if any character is a surrogate character */
801 18856 : if (((in[0] ^ 0xd800) &
802 18856 : (in[1] ^ 0xd800) &
803 18856 : (in[2] ^ 0xd800) &
804 18856 : (in[3] ^ 0xd800) & 0xf800) == 0)
805 6 : break;
806 : #endif
807 73484 : out[0] = STRINGLIB(SWAB4)(in[0]);
808 73484 : out[1] = STRINGLIB(SWAB4)(in[1]);
809 73484 : out[2] = STRINGLIB(SWAB4)(in[2]);
810 73484 : out[3] = STRINGLIB(SWAB4)(in[3]);
811 73484 : in += 4; out += 4;
812 : }
813 1642 : while (in < end) {
814 861 : Py_UCS4 ch = *in++;
815 : #if STRINGLIB_SIZEOF_CHAR > 1
816 258 : if (Py_UNICODE_IS_SURROGATE(ch)) {
817 : /* reject surrogate characters (U+D800-U+DFFF) */
818 20 : goto fail;
819 : }
820 : #endif
821 841 : *out++ = STRINGLIB(SWAB4)(ch);
822 : }
823 : }
824 2509 : *outptr = out;
825 2509 : return len;
826 : #if STRINGLIB_SIZEOF_CHAR > 1
827 109 : fail:
828 109 : *outptr = out;
829 109 : return len - (end - in + 1);
830 : #endif
831 : }
832 :
833 : #endif
|