| 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 |
STRINGLIB(utf8_decode)(const char **inptr, const char *end, |
| 24 |
STRINGLIB_CHAR *dest, |
| 25 |
Py_ssize_t *outpos) |
| 26 |
{ |
| 27 |
Py_UCS4 ch; |
| 28 |
const char *s = *inptr; |
| 29 |
STRINGLIB_CHAR *p = dest + *outpos; |
| 30 |
|
| 31 |
while (s < end) { |
| 32 |
ch = (unsigned char)*s; |
| 33 |
|
| 34 |
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 |
if (_Py_IS_ALIGNED(s, ALIGNOF_SIZE_T)) { |
| 43 |
/* Help register allocation */ |
| 44 |
const char *_s = s; |
| 45 |
STRINGLIB_CHAR *_p = p; |
| 46 |
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 |
size_t value = *(const size_t *) _s; |
| 51 |
if (value & ASCII_CHAR_MASK) |
| 52 |
break; |
| 53 |
#if PY_LITTLE_ENDIAN |
| 54 |
_p[0] = (STRINGLIB_CHAR)(value & 0xFFu); |
| 55 |
_p[1] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu); |
| 56 |
_p[2] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu); |
| 57 |
_p[3] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu); |
| 58 |
# if SIZEOF_SIZE_T == 8 |
| 59 |
_p[4] = (STRINGLIB_CHAR)((value >> 32) & 0xFFu); |
| 60 |
_p[5] = (STRINGLIB_CHAR)((value >> 40) & 0xFFu); |
| 61 |
_p[6] = (STRINGLIB_CHAR)((value >> 48) & 0xFFu); |
| 62 |
_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 |
_s += SIZEOF_SIZE_T; |
| 82 |
_p += SIZEOF_SIZE_T; |
| 83 |
} |
| 84 |
s = _s; |
| 85 |
p = _p; |
| 86 |
if (s == end) |
| 87 |
break; |
| 88 |
ch = (unsigned char)*s; |
| 89 |
} |
| 90 |
if (ch < 0x80) { |
| 91 |
s++; |
| 92 |
*p++ = ch; |
| 93 |
continue; |
| 94 |
} |
| 95 |
} |
| 96 |
|
| 97 |
if (ch < 0xE0) { |
| 98 |
/* \xC2\x80-\xDF\xBF -- 0080-07FF */ |
| 99 |
Py_UCS4 ch2; |
| 100 |
if (ch < 0xC2) { |
| 101 |
/* invalid sequence |
| 102 |
\x80-\xBF -- continuation byte |
| 103 |
\xC0-\xC1 -- fake 0000-007F */ |
| 104 |
goto InvalidStart; |
| 105 |
} |
| 106 |
if (end - s < 2) { |
| 107 |
/* unexpected end of data: the caller will decide whether |
| 108 |
it's an error or not */ |
| 109 |
break; |
| 110 |
} |
| 111 |
ch2 = (unsigned char)s[1]; |
| 112 |
if (!IS_CONTINUATION_BYTE(ch2)) |
| 113 |
/* invalid continuation byte */ |
| 114 |
goto InvalidContinuation1; |
| 115 |
ch = (ch << 6) + ch2 - |
| 116 |
((0xC0 << 6) + 0x80); |
| 117 |
assert ((ch > 0x007F) && (ch <= 0x07FF)); |
| 118 |
s += 2; |
| 119 |
if (STRINGLIB_MAX_CHAR <= 0x007F || |
| 120 |
(STRINGLIB_MAX_CHAR < 0x07FF && ch > STRINGLIB_MAX_CHAR)) |
| 121 |
/* Out-of-range */ |
| 122 |
goto Return; |
| 123 |
*p++ = ch; |
| 124 |
continue; |
| 125 |
} |
| 126 |
|
| 127 |
if (ch < 0xF0) { |
| 128 |
/* \xE0\xA0\x80-\xEF\xBF\xBF -- 0800-FFFF */ |
| 129 |
Py_UCS4 ch2, ch3; |
| 130 |
if (end - s < 3) { |
| 131 |
/* unexpected end of data: the caller will decide whether |
| 132 |
it's an error or not */ |
| 133 |
if (end - s < 2) |
| 134 |
break; |
| 135 |
ch2 = (unsigned char)s[1]; |
| 136 |
if (!IS_CONTINUATION_BYTE(ch2) || |
| 137 |
(ch2 < 0xA0 ? ch == 0xE0 : ch == 0xED)) |
| 138 |
/* for clarification see comments below */ |
| 139 |
goto InvalidContinuation1; |
| 140 |
break; |
| 141 |
} |
| 142 |
ch2 = (unsigned char)s[1]; |
| 143 |
ch3 = (unsigned char)s[2]; |
| 144 |
if (!IS_CONTINUATION_BYTE(ch2)) { |
| 145 |
/* invalid continuation byte */ |
| 146 |
goto InvalidContinuation1; |
| 147 |
} |
| 148 |
if (ch == 0xE0) { |
| 149 |
if (ch2 < 0xA0) |
| 150 |
/* invalid sequence |
| 151 |
\xE0\x80\x80-\xE0\x9F\xBF -- fake 0000-0800 */ |
| 152 |
goto InvalidContinuation1; |
| 153 |
} 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 |
goto InvalidContinuation1; |
| 160 |
} |
| 161 |
if (!IS_CONTINUATION_BYTE(ch3)) { |
| 162 |
/* invalid continuation byte */ |
| 163 |
goto InvalidContinuation2; |
| 164 |
} |
| 165 |
ch = (ch << 12) + (ch2 << 6) + ch3 - |
| 166 |
((0xE0 << 12) + (0x80 << 6) + 0x80); |
| 167 |
assert ((ch > 0x07FF) && (ch <= 0xFFFF)); |
| 168 |
s += 3; |
| 169 |
if (STRINGLIB_MAX_CHAR <= 0x07FF || |
| 170 |
(STRINGLIB_MAX_CHAR < 0xFFFF && ch > STRINGLIB_MAX_CHAR)) |
| 171 |
/* Out-of-range */ |
| 172 |
goto Return; |
| 173 |
*p++ = ch; |
| 174 |
continue; |
| 175 |
} |
| 176 |
|
| 177 |
if (ch < 0xF5) { |
| 178 |
/* \xF0\x90\x80\x80-\xF4\x8F\xBF\xBF -- 10000-10FFFF */ |
| 179 |
Py_UCS4 ch2, ch3, ch4; |
| 180 |
if (end - s < 4) { |
| 181 |
/* unexpected end of data: the caller will decide whether |
| 182 |
it's an error or not */ |
| 183 |
if (end - s < 2) |
| 184 |
break; |
| 185 |
ch2 = (unsigned char)s[1]; |
| 186 |
if (!IS_CONTINUATION_BYTE(ch2) || |
| 187 |
(ch2 < 0x90 ? ch == 0xF0 : ch == 0xF4)) |
| 188 |
/* for clarification see comments below */ |
| 189 |
goto InvalidContinuation1; |
| 190 |
if (end - s < 3) |
| 191 |
break; |
| 192 |
ch3 = (unsigned char)s[2]; |
| 193 |
if (!IS_CONTINUATION_BYTE(ch3)) |
| 194 |
goto InvalidContinuation2; |
| 195 |
break; |
| 196 |
} |
| 197 |
ch2 = (unsigned char)s[1]; |
| 198 |
ch3 = (unsigned char)s[2]; |
| 199 |
ch4 = (unsigned char)s[3]; |
| 200 |
if (!IS_CONTINUATION_BYTE(ch2)) { |
| 201 |
/* invalid continuation byte */ |
| 202 |
goto InvalidContinuation1; |
| 203 |
} |
| 204 |
if (ch == 0xF0) { |
| 205 |
if (ch2 < 0x90) |
| 206 |
/* invalid sequence |
| 207 |
\xF0\x80\x80\x80-\xF0\x8F\xBF\xBF -- fake 0000-FFFF */ |
| 208 |
goto InvalidContinuation1; |
| 209 |
} else if (ch == 0xF4 && ch2 >= 0x90) { |
| 210 |
/* invalid sequence |
| 211 |
\xF4\x90\x80\x80- -- 110000- overflow */ |
| 212 |
goto InvalidContinuation1; |
| 213 |
} |
| 214 |
if (!IS_CONTINUATION_BYTE(ch3)) { |
| 215 |
/* invalid continuation byte */ |
| 216 |
goto InvalidContinuation2; |
| 217 |
} |
| 218 |
if (!IS_CONTINUATION_BYTE(ch4)) { |
| 219 |
/* invalid continuation byte */ |
| 220 |
goto InvalidContinuation3; |
| 221 |
} |
| 222 |
ch = (ch << 18) + (ch2 << 12) + (ch3 << 6) + ch4 - |
| 223 |
((0xF0 << 18) + (0x80 << 12) + (0x80 << 6) + 0x80); |
| 224 |
assert ((ch > 0xFFFF) && (ch <= 0x10FFFF)); |
| 225 |
s += 4; |
| 226 |
if (STRINGLIB_MAX_CHAR <= 0xFFFF || |
| 227 |
(STRINGLIB_MAX_CHAR < 0x10FFFF && ch > STRINGLIB_MAX_CHAR)) |
| 228 |
/* Out-of-range */ |
| 229 |
goto Return; |
| 230 |
*p++ = ch; |
| 231 |
continue; |
| 232 |
} |
| 233 |
goto InvalidStart; |
| 234 |
} |
| 235 |
ch = 0; |
| 236 |
Return: |
| 237 |
*inptr = s; |
| 238 |
*outpos = p - dest; |
| 239 |
return ch; |
| 240 |
InvalidStart: |
| 241 |
ch = 1; |
| 242 |
goto Return; |
| 243 |
InvalidContinuation1: |
| 244 |
ch = 2; |
| 245 |
goto Return; |
| 246 |
InvalidContinuation2: |
| 247 |
ch = 3; |
| 248 |
goto Return; |
| 249 |
InvalidContinuation3: |
| 250 |
ch = 4; |
| 251 |
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 |
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 |
PyObject *error_handler_obj = NULL; |
| 272 |
PyObject *exc = NULL; |
| 273 |
PyObject *rep = NULL; |
| 274 |
#endif |
| 275 |
#if STRINGLIB_SIZEOF_CHAR == 1 |
| 276 |
const Py_ssize_t max_char_size = 2; |
| 277 |
#elif STRINGLIB_SIZEOF_CHAR == 2 |
| 278 |
const Py_ssize_t max_char_size = 3; |
| 279 |
#else /* STRINGLIB_SIZEOF_CHAR == 4 */ |
| 280 |
const Py_ssize_t max_char_size = 4; |
| 281 |
#endif |
| 282 |
|
| 283 |
assert(size >= 0); |
| 284 |
if (size > PY_SSIZE_T_MAX / max_char_size) { |
| 285 |
/* integer overflow */ |
| 286 |
PyErr_NoMemory(); |
| 287 |
return NULL; |
| 288 |
} |
| 289 |
|
| 290 |
_PyBytesWriter_Init(writer); |
| 291 |
p = _PyBytesWriter_Alloc(writer, size * max_char_size); |
| 292 |
if (p == NULL) |
| 293 |
return NULL; |
| 294 |
|
| 295 |
for (i = 0; i < size;) { |
| 296 |
Py_UCS4 ch = data[i++]; |
| 297 |
|
| 298 |
if (ch < 0x80) { |
| 299 |
/* Encode ASCII */ |
| 300 |
*p++ = (char) ch; |
| 301 |
|
| 302 |
} |
| 303 |
else |
| 304 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 305 |
if (ch < 0x0800) |
| 306 |
#endif |
| 307 |
{ |
| 308 |
/* Encode Latin-1 */ |
| 309 |
*p++ = (char)(0xc0 | (ch >> 6)); |
| 310 |
*p++ = (char)(0x80 | (ch & 0x3f)); |
| 311 |
} |
| 312 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 313 |
else if (Py_UNICODE_IS_SURROGATE(ch)) { |
| 314 |
Py_ssize_t startpos, endpos, newpos; |
| 315 |
Py_ssize_t k; |
| 316 |
if (error_handler == _Py_ERROR_UNKNOWN) { |
| 317 |
error_handler = _Py_GetErrorHandler(errors); |
| 318 |
} |
| 319 |
|
| 320 |
startpos = i-1; |
| 321 |
endpos = startpos+1; |
| 322 |
|
| 323 |
while ((endpos < size) && Py_UNICODE_IS_SURROGATE(data[endpos])) |
| 324 |
endpos++; |
| 325 |
|
| 326 |
/* Only overallocate the buffer if it's not the last write */ |
| 327 |
writer->overallocate = (endpos < size); |
| 328 |
|
| 329 |
switch (error_handler) |
| 330 |
{ |
| 331 |
case _Py_ERROR_REPLACE: |
| 332 |
memset(p, '?', endpos - startpos); |
| 333 |
p += (endpos - startpos); |
| 334 |
_Py_FALLTHROUGH; |
| 335 |
case _Py_ERROR_IGNORE: |
| 336 |
i += (endpos - startpos - 1); |
| 337 |
break; |
| 338 |
|
| 339 |
case _Py_ERROR_SURROGATEPASS: |
| 340 |
for (k=startpos; k<endpos; k++) { |
| 341 |
ch = data[k]; |
| 342 |
*p++ = (char)(0xe0 | (ch >> 12)); |
| 343 |
*p++ = (char)(0x80 | ((ch >> 6) & 0x3f)); |
| 344 |
*p++ = (char)(0x80 | (ch & 0x3f)); |
| 345 |
} |
| 346 |
i += (endpos - startpos - 1); |
| 347 |
break; |
| 348 |
|
| 349 |
case _Py_ERROR_BACKSLASHREPLACE: |
| 350 |
/* subtract preallocated bytes */ |
| 351 |
writer->min_size -= max_char_size * (endpos - startpos); |
| 352 |
p = backslashreplace(writer, p, |
| 353 |
unicode, startpos, endpos); |
| 354 |
if (p == NULL) |
| 355 |
goto error; |
| 356 |
i += (endpos - startpos - 1); |
| 357 |
break; |
| 358 |
|
| 359 |
case _Py_ERROR_XMLCHARREFREPLACE: |
| 360 |
/* subtract preallocated bytes */ |
| 361 |
writer->min_size -= max_char_size * (endpos - startpos); |
| 362 |
p = xmlcharrefreplace(writer, p, |
| 363 |
unicode, startpos, endpos); |
| 364 |
if (p == NULL) |
| 365 |
goto error; |
| 366 |
i += (endpos - startpos - 1); |
| 367 |
break; |
| 368 |
|
| 369 |
case _Py_ERROR_SURROGATEESCAPE: |
| 370 |
for (k=startpos; k<endpos; k++) { |
| 371 |
ch = data[k]; |
| 372 |
if (!(0xDC80 <= ch && ch <= 0xDCFF)) |
| 373 |
break; |
| 374 |
*p++ = (char)(ch & 0xff); |
| 375 |
} |
| 376 |
if (k >= endpos) { |
| 377 |
i += (endpos - startpos - 1); |
| 378 |
break; |
| 379 |
} |
| 380 |
startpos = k; |
| 381 |
assert(startpos < endpos); |
| 382 |
_Py_FALLTHROUGH; |
| 383 |
default: |
| 384 |
rep = unicode_encode_call_errorhandler( |
| 385 |
errors, &error_handler_obj, "utf-8", "surrogates not allowed", |
| 386 |
unicode, &exc, startpos, endpos, &newpos); |
| 387 |
if (!rep) |
| 388 |
goto error; |
| 389 |
|
| 390 |
if (newpos < startpos) { |
| 391 |
writer->overallocate = 1; |
| 392 |
p = _PyBytesWriter_Prepare(writer, p, |
| 393 |
max_char_size * (startpos - newpos)); |
| 394 |
if (p == NULL) |
| 395 |
goto error; |
| 396 |
} |
| 397 |
else { |
| 398 |
/* subtract preallocated bytes */ |
| 399 |
writer->min_size -= max_char_size * (newpos - startpos); |
| 400 |
/* Only overallocate the buffer if it's not the last write */ |
| 401 |
writer->overallocate = (newpos < size); |
| 402 |
} |
| 403 |
|
| 404 |
if (PyBytes_Check(rep)) { |
| 405 |
p = _PyBytesWriter_WriteBytes(writer, p, |
| 406 |
PyBytes_AS_STRING(rep), |
| 407 |
PyBytes_GET_SIZE(rep)); |
| 408 |
} |
| 409 |
else { |
| 410 |
/* rep is unicode */ |
| 411 |
if (!PyUnicode_IS_ASCII(rep)) { |
| 412 |
raise_encode_exception(&exc, "utf-8", unicode, |
| 413 |
startpos, endpos, |
| 414 |
"surrogates not allowed"); |
| 415 |
goto error; |
| 416 |
} |
| 417 |
|
| 418 |
p = _PyBytesWriter_WriteBytes(writer, p, |
| 419 |
PyUnicode_DATA(rep), |
| 420 |
PyUnicode_GET_LENGTH(rep)); |
| 421 |
} |
| 422 |
|
| 423 |
if (p == NULL) |
| 424 |
goto error; |
| 425 |
Py_CLEAR(rep); |
| 426 |
|
| 427 |
i = newpos; |
| 428 |
} |
| 429 |
|
| 430 |
/* If overallocation was disabled, ensure that it was the last |
| 431 |
write. Otherwise, we missed an optimization */ |
| 432 |
assert(writer->overallocate || i == size); |
| 433 |
} |
| 434 |
else |
| 435 |
#if STRINGLIB_SIZEOF_CHAR > 2 |
| 436 |
if (ch < 0x10000) |
| 437 |
#endif |
| 438 |
{ |
| 439 |
*p++ = (char)(0xe0 | (ch >> 12)); |
| 440 |
*p++ = (char)(0x80 | ((ch >> 6) & 0x3f)); |
| 441 |
*p++ = (char)(0x80 | (ch & 0x3f)); |
| 442 |
} |
| 443 |
#if STRINGLIB_SIZEOF_CHAR > 2 |
| 444 |
else /* ch >= 0x10000 */ |
| 445 |
{ |
| 446 |
assert(ch <= MAX_UNICODE); |
| 447 |
/* Encode UCS4 Unicode ordinals */ |
| 448 |
*p++ = (char)(0xf0 | (ch >> 18)); |
| 449 |
*p++ = (char)(0x80 | ((ch >> 12) & 0x3f)); |
| 450 |
*p++ = (char)(0x80 | ((ch >> 6) & 0x3f)); |
| 451 |
*p++ = (char)(0x80 | (ch & 0x3f)); |
| 452 |
} |
| 453 |
#endif /* STRINGLIB_SIZEOF_CHAR > 2 */ |
| 454 |
#endif /* STRINGLIB_SIZEOF_CHAR > 1 */ |
| 455 |
} |
| 456 |
|
| 457 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 458 |
Py_XDECREF(error_handler_obj); |
| 459 |
Py_XDECREF(exc); |
| 460 |
#endif |
| 461 |
return p; |
| 462 |
|
| 463 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 464 |
error: |
| 465 |
Py_XDECREF(rep); |
| 466 |
Py_XDECREF(error_handler_obj); |
| 467 |
Py_XDECREF(exc); |
| 468 |
return NULL; |
| 469 |
#endif |
| 470 |
} |
| 471 |
|
| 472 |
/* The pattern for constructing UCS2-repeated masks. */ |
| 473 |
#if SIZEOF_LONG == 8 |
| 474 |
# define UCS2_REPEAT_MASK 0x0001000100010001ul |
| 475 |
#elif SIZEOF_LONG == 4 |
| 476 |
# define UCS2_REPEAT_MASK 0x00010001ul |
| 477 |
#else |
| 478 |
# error C 'long' size should be either 4 or 8! |
| 479 |
#endif |
| 480 |
|
| 481 |
/* The mask for fast checking. */ |
| 482 |
#if STRINGLIB_SIZEOF_CHAR == 1 |
| 483 |
/* The mask for fast checking of whether a C 'long' contains a |
| 484 |
non-ASCII or non-Latin1 UTF16-encoded characters. */ |
| 485 |
# define FAST_CHAR_MASK (UCS2_REPEAT_MASK * (0xFFFFu & ~STRINGLIB_MAX_CHAR)) |
| 486 |
#else |
| 487 |
/* The mask for fast checking of whether a C 'long' may contain |
| 488 |
UTF16-encoded surrogate characters. This is an efficient heuristic, |
| 489 |
assuming that non-surrogate characters with a code point >= 0x8000 are |
| 490 |
rare in most input. |
| 491 |
*/ |
| 492 |
# define FAST_CHAR_MASK (UCS2_REPEAT_MASK * 0x8000u) |
| 493 |
#endif |
| 494 |
/* The mask for fast byte-swapping. */ |
| 495 |
#define STRIPPED_MASK (UCS2_REPEAT_MASK * 0x00FFu) |
| 496 |
/* Swap bytes. */ |
| 497 |
#define SWAB(value) ((((value) >> 8) & STRIPPED_MASK) | \ |
| 498 |
(((value) & STRIPPED_MASK) << 8)) |
| 499 |
|
| 500 |
Py_LOCAL_INLINE(Py_UCS4) |
| 501 |
STRINGLIB(utf16_decode)(const unsigned char **inptr, const unsigned char *e, |
| 502 |
STRINGLIB_CHAR *dest, Py_ssize_t *outpos, |
| 503 |
int native_ordering) |
| 504 |
{ |
| 505 |
Py_UCS4 ch; |
| 506 |
const unsigned char *q = *inptr; |
| 507 |
STRINGLIB_CHAR *p = dest + *outpos; |
| 508 |
/* Offsets from q for retrieving byte pairs in the right order. */ |
| 509 |
#if PY_LITTLE_ENDIAN |
| 510 |
int ihi = !!native_ordering, ilo = !native_ordering; |
| 511 |
#else |
| 512 |
int ihi = !native_ordering, ilo = !!native_ordering; |
| 513 |
#endif |
| 514 |
--e; |
| 515 |
|
| 516 |
while (q < e) { |
| 517 |
Py_UCS4 ch2; |
| 518 |
/* First check for possible aligned read of a C 'long'. Unaligned |
| 519 |
reads are more expensive, better to defer to another iteration. */ |
| 520 |
if (_Py_IS_ALIGNED(q, ALIGNOF_LONG)) { |
| 521 |
/* Fast path for runs of in-range non-surrogate chars. */ |
| 522 |
const unsigned char *_q = q; |
| 523 |
while (_q + SIZEOF_LONG <= e) { |
| 524 |
unsigned long block = * (const unsigned long *) _q; |
| 525 |
if (native_ordering) { |
| 526 |
/* Can use buffer directly */ |
| 527 |
if (block & FAST_CHAR_MASK) |
| 528 |
break; |
| 529 |
} |
| 530 |
else { |
| 531 |
/* Need to byte-swap */ |
| 532 |
if (block & SWAB(FAST_CHAR_MASK)) |
| 533 |
break; |
| 534 |
#if STRINGLIB_SIZEOF_CHAR == 1 |
| 535 |
block >>= 8; |
| 536 |
#else |
| 537 |
block = SWAB(block); |
| 538 |
#endif |
| 539 |
} |
| 540 |
#if PY_LITTLE_ENDIAN |
| 541 |
# if SIZEOF_LONG == 4 |
| 542 |
p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu); |
| 543 |
p[1] = (STRINGLIB_CHAR)(block >> 16); |
| 544 |
# elif SIZEOF_LONG == 8 |
| 545 |
p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu); |
| 546 |
p[1] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu); |
| 547 |
p[2] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu); |
| 548 |
p[3] = (STRINGLIB_CHAR)(block >> 48); |
| 549 |
# endif |
| 550 |
#else |
| 551 |
# if SIZEOF_LONG == 4 |
| 552 |
p[0] = (STRINGLIB_CHAR)(block >> 16); |
| 553 |
p[1] = (STRINGLIB_CHAR)(block & 0xFFFFu); |
| 554 |
# elif SIZEOF_LONG == 8 |
| 555 |
p[0] = (STRINGLIB_CHAR)(block >> 48); |
| 556 |
p[1] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu); |
| 557 |
p[2] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu); |
| 558 |
p[3] = (STRINGLIB_CHAR)(block & 0xFFFFu); |
| 559 |
# endif |
| 560 |
#endif |
| 561 |
_q += SIZEOF_LONG; |
| 562 |
p += SIZEOF_LONG / 2; |
| 563 |
} |
| 564 |
q = _q; |
| 565 |
if (q >= e) |
| 566 |
break; |
| 567 |
} |
| 568 |
|
| 569 |
ch = (q[ihi] << 8) | q[ilo]; |
| 570 |
q += 2; |
| 571 |
if (!Py_UNICODE_IS_SURROGATE(ch)) { |
| 572 |
#if STRINGLIB_SIZEOF_CHAR < 2 |
| 573 |
if (ch > STRINGLIB_MAX_CHAR) |
| 574 |
/* Out-of-range */ |
| 575 |
goto Return; |
| 576 |
#endif |
| 577 |
*p++ = (STRINGLIB_CHAR)ch; |
| 578 |
continue; |
| 579 |
} |
| 580 |
|
| 581 |
/* UTF-16 code pair: */ |
| 582 |
if (!Py_UNICODE_IS_HIGH_SURROGATE(ch)) |
| 583 |
goto IllegalEncoding; |
| 584 |
if (q >= e) |
| 585 |
goto UnexpectedEnd; |
| 586 |
ch2 = (q[ihi] << 8) | q[ilo]; |
| 587 |
q += 2; |
| 588 |
if (!Py_UNICODE_IS_LOW_SURROGATE(ch2)) |
| 589 |
goto IllegalSurrogate; |
| 590 |
ch = Py_UNICODE_JOIN_SURROGATES(ch, ch2); |
| 591 |
#if STRINGLIB_SIZEOF_CHAR < 4 |
| 592 |
/* Out-of-range */ |
| 593 |
goto Return; |
| 594 |
#else |
| 595 |
*p++ = (STRINGLIB_CHAR)ch; |
| 596 |
#endif |
| 597 |
} |
| 598 |
ch = 0; |
| 599 |
Return: |
| 600 |
*inptr = q; |
| 601 |
*outpos = p - dest; |
| 602 |
return ch; |
| 603 |
UnexpectedEnd: |
| 604 |
ch = 1; |
| 605 |
goto Return; |
| 606 |
IllegalEncoding: |
| 607 |
ch = 2; |
| 608 |
goto Return; |
| 609 |
IllegalSurrogate: |
| 610 |
ch = 3; |
| 611 |
goto Return; |
| 612 |
} |
| 613 |
#undef UCS2_REPEAT_MASK |
| 614 |
#undef FAST_CHAR_MASK |
| 615 |
#undef STRIPPED_MASK |
| 616 |
#undef SWAB |
| 617 |
|
| 618 |
|
| 619 |
#if STRINGLIB_MAX_CHAR >= 0x80 |
| 620 |
Py_LOCAL_INLINE(Py_ssize_t) |
| 621 |
STRINGLIB(utf16_encode)(const STRINGLIB_CHAR *in, |
| 622 |
Py_ssize_t len, |
| 623 |
unsigned short **outptr, |
| 624 |
int native_ordering) |
| 625 |
{ |
| 626 |
unsigned short *out = *outptr; |
| 627 |
const STRINGLIB_CHAR *end = in + len; |
| 628 |
#if STRINGLIB_SIZEOF_CHAR == 1 |
| 629 |
if (native_ordering) { |
| 630 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 631 |
while (in < unrolled_end) { |
| 632 |
out[0] = in[0]; |
| 633 |
out[1] = in[1]; |
| 634 |
out[2] = in[2]; |
| 635 |
out[3] = in[3]; |
| 636 |
in += 4; out += 4; |
| 637 |
} |
| 638 |
while (in < end) { |
| 639 |
*out++ = *in++; |
| 640 |
} |
| 641 |
} else { |
| 642 |
# define SWAB2(CH) ((CH) << 8) /* high byte is zero */ |
| 643 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 644 |
while (in < unrolled_end) { |
| 645 |
out[0] = SWAB2(in[0]); |
| 646 |
out[1] = SWAB2(in[1]); |
| 647 |
out[2] = SWAB2(in[2]); |
| 648 |
out[3] = SWAB2(in[3]); |
| 649 |
in += 4; out += 4; |
| 650 |
} |
| 651 |
while (in < end) { |
| 652 |
Py_UCS4 ch = *in++; |
| 653 |
*out++ = SWAB2((Py_UCS2)ch); |
| 654 |
} |
| 655 |
#undef SWAB2 |
| 656 |
} |
| 657 |
*outptr = out; |
| 658 |
return len; |
| 659 |
#else |
| 660 |
if (native_ordering) { |
| 661 |
#if STRINGLIB_MAX_CHAR < 0x10000 |
| 662 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 663 |
while (in < unrolled_end) { |
| 664 |
/* check if any character is a surrogate character */ |
| 665 |
if (((in[0] ^ 0xd800) & |
| 666 |
(in[1] ^ 0xd800) & |
| 667 |
(in[2] ^ 0xd800) & |
| 668 |
(in[3] ^ 0xd800) & 0xf800) == 0) |
| 669 |
break; |
| 670 |
out[0] = in[0]; |
| 671 |
out[1] = in[1]; |
| 672 |
out[2] = in[2]; |
| 673 |
out[3] = in[3]; |
| 674 |
in += 4; out += 4; |
| 675 |
} |
| 676 |
#endif |
| 677 |
while (in < end) { |
| 678 |
Py_UCS4 ch; |
| 679 |
ch = *in++; |
| 680 |
if (ch < 0xd800) |
| 681 |
*out++ = ch; |
| 682 |
else if (ch < 0xe000) |
| 683 |
/* reject surrogate characters (U+D800-U+DFFF) */ |
| 684 |
goto fail; |
| 685 |
#if STRINGLIB_MAX_CHAR >= 0x10000 |
| 686 |
else if (ch >= 0x10000) { |
| 687 |
out[0] = Py_UNICODE_HIGH_SURROGATE(ch); |
| 688 |
out[1] = Py_UNICODE_LOW_SURROGATE(ch); |
| 689 |
out += 2; |
| 690 |
} |
| 691 |
#endif |
| 692 |
else |
| 693 |
*out++ = ch; |
| 694 |
} |
| 695 |
} else { |
| 696 |
#define SWAB2(CH) (((CH) << 8) | ((CH) >> 8)) |
| 697 |
#if STRINGLIB_MAX_CHAR < 0x10000 |
| 698 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 699 |
while (in < unrolled_end) { |
| 700 |
/* check if any character is a surrogate character */ |
| 701 |
if (((in[0] ^ 0xd800) & |
| 702 |
(in[1] ^ 0xd800) & |
| 703 |
(in[2] ^ 0xd800) & |
| 704 |
(in[3] ^ 0xd800) & 0xf800) == 0) |
| 705 |
break; |
| 706 |
out[0] = SWAB2(in[0]); |
| 707 |
out[1] = SWAB2(in[1]); |
| 708 |
out[2] = SWAB2(in[2]); |
| 709 |
out[3] = SWAB2(in[3]); |
| 710 |
in += 4; out += 4; |
| 711 |
} |
| 712 |
#endif |
| 713 |
while (in < end) { |
| 714 |
Py_UCS4 ch = *in++; |
| 715 |
if (ch < 0xd800) |
| 716 |
*out++ = SWAB2((Py_UCS2)ch); |
| 717 |
else if (ch < 0xe000) |
| 718 |
/* reject surrogate characters (U+D800-U+DFFF) */ |
| 719 |
goto fail; |
| 720 |
#if STRINGLIB_MAX_CHAR >= 0x10000 |
| 721 |
else if (ch >= 0x10000) { |
| 722 |
Py_UCS2 ch1 = Py_UNICODE_HIGH_SURROGATE(ch); |
| 723 |
Py_UCS2 ch2 = Py_UNICODE_LOW_SURROGATE(ch); |
| 724 |
out[0] = SWAB2(ch1); |
| 725 |
out[1] = SWAB2(ch2); |
| 726 |
out += 2; |
| 727 |
} |
| 728 |
#endif |
| 729 |
else |
| 730 |
*out++ = SWAB2((Py_UCS2)ch); |
| 731 |
} |
| 732 |
#undef SWAB2 |
| 733 |
} |
| 734 |
*outptr = out; |
| 735 |
return len; |
| 736 |
fail: |
| 737 |
*outptr = out; |
| 738 |
return len - (end - in + 1); |
| 739 |
#endif |
| 740 |
} |
| 741 |
|
| 742 |
static inline uint32_t |
| 743 |
STRINGLIB(SWAB4)(STRINGLIB_CHAR ch) |
| 744 |
{ |
| 745 |
uint32_t word = ch; |
| 746 |
#if STRINGLIB_SIZEOF_CHAR == 1 |
| 747 |
/* high bytes are zero */ |
| 748 |
return (word << 24); |
| 749 |
#elif STRINGLIB_SIZEOF_CHAR == 2 |
| 750 |
/* high bytes are zero */ |
| 751 |
return ((word & 0x00FFu) << 24) | ((word & 0xFF00u) << 8); |
| 752 |
#else |
| 753 |
return _Py_bswap32(word); |
| 754 |
#endif |
| 755 |
} |
| 756 |
|
| 757 |
Py_LOCAL_INLINE(Py_ssize_t) |
| 758 |
STRINGLIB(utf32_encode)(const STRINGLIB_CHAR *in, |
| 759 |
Py_ssize_t len, |
| 760 |
uint32_t **outptr, |
| 761 |
int native_ordering) |
| 762 |
{ |
| 763 |
uint32_t *out = *outptr; |
| 764 |
const STRINGLIB_CHAR *end = in + len; |
| 765 |
if (native_ordering) { |
| 766 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 767 |
while (in < unrolled_end) { |
| 768 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 769 |
/* check if any character is a surrogate character */ |
| 770 |
if (((in[0] ^ 0xd800) & |
| 771 |
(in[1] ^ 0xd800) & |
| 772 |
(in[2] ^ 0xd800) & |
| 773 |
(in[3] ^ 0xd800) & 0xf800) == 0) |
| 774 |
break; |
| 775 |
#endif |
| 776 |
out[0] = in[0]; |
| 777 |
out[1] = in[1]; |
| 778 |
out[2] = in[2]; |
| 779 |
out[3] = in[3]; |
| 780 |
in += 4; out += 4; |
| 781 |
} |
| 782 |
while (in < end) { |
| 783 |
Py_UCS4 ch; |
| 784 |
ch = *in++; |
| 785 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 786 |
if (Py_UNICODE_IS_SURROGATE(ch)) { |
| 787 |
/* reject surrogate characters (U+D800-U+DFFF) */ |
| 788 |
goto fail; |
| 789 |
} |
| 790 |
#endif |
| 791 |
*out++ = ch; |
| 792 |
} |
| 793 |
} else { |
| 794 |
const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4); |
| 795 |
while (in < unrolled_end) { |
| 796 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 797 |
/* check if any character is a surrogate character */ |
| 798 |
if (((in[0] ^ 0xd800) & |
| 799 |
(in[1] ^ 0xd800) & |
| 800 |
(in[2] ^ 0xd800) & |
| 801 |
(in[3] ^ 0xd800) & 0xf800) == 0) |
| 802 |
break; |
| 803 |
#endif |
| 804 |
out[0] = STRINGLIB(SWAB4)(in[0]); |
| 805 |
out[1] = STRINGLIB(SWAB4)(in[1]); |
| 806 |
out[2] = STRINGLIB(SWAB4)(in[2]); |
| 807 |
out[3] = STRINGLIB(SWAB4)(in[3]); |
| 808 |
in += 4; out += 4; |
| 809 |
} |
| 810 |
while (in < end) { |
| 811 |
Py_UCS4 ch = *in++; |
| 812 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 813 |
if (Py_UNICODE_IS_SURROGATE(ch)) { |
| 814 |
/* reject surrogate characters (U+D800-U+DFFF) */ |
| 815 |
goto fail; |
| 816 |
} |
| 817 |
#endif |
| 818 |
*out++ = STRINGLIB(SWAB4)(ch); |
| 819 |
} |
| 820 |
} |
| 821 |
*outptr = out; |
| 822 |
return len; |
| 823 |
#if STRINGLIB_SIZEOF_CHAR > 1 |
| 824 |
fail: |
| 825 |
*outptr = out; |
| 826 |
return len - (end - in + 1); |
| 827 |
#endif |
| 828 |
} |
| 829 |
|
| 830 |
#endif |