1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ******************************************************************************* |
5 | * |
6 | * Copyright (C) 2008-2011, International Business Machines |
7 | * Corporation, Google and others. All Rights Reserved. |
8 | * |
9 | ******************************************************************************* |
10 | */ |
11 | // Author : eldawy@google.com (Mohamed Eldawy) |
12 | // ucnvsel.cpp |
13 | // |
14 | // Purpose: To generate a list of encodings capable of handling |
15 | // a given Unicode text |
16 | // |
17 | // Started 09-April-2008 |
18 | |
19 | /** |
20 | * \file |
21 | * |
22 | * This is an implementation of an encoding selector. |
23 | * The goal is, given a unicode string, find the encodings |
24 | * this string can be mapped to. To make processing faster |
25 | * a trie is built when you call ucnvsel_open() that |
26 | * stores all encodings a codepoint can map to |
27 | */ |
28 | |
29 | #include "unicode/ucnvsel.h" |
30 | |
31 | #if !UCONFIG_NO_CONVERSION |
32 | |
33 | #include <string.h> |
34 | |
35 | #include "unicode/uchar.h" |
36 | #include "unicode/uniset.h" |
37 | #include "unicode/ucnv.h" |
38 | #include "unicode/ustring.h" |
39 | #include "unicode/uchriter.h" |
40 | #include "utrie2.h" |
41 | #include "propsvec.h" |
42 | #include "uassert.h" |
43 | #include "ucmndata.h" |
44 | #include "udataswp.h" |
45 | #include "uenumimp.h" |
46 | #include "cmemory.h" |
47 | #include "cstring.h" |
48 | |
49 | U_NAMESPACE_USE |
50 | |
51 | struct UConverterSelector { |
52 | UTrie2 *trie; // 16 bit trie containing offsets into pv |
53 | uint32_t* pv; // table of bits! |
54 | int32_t pvCount; |
55 | char** encodings; // which encodings did user ask to use? |
56 | int32_t encodingsCount; |
57 | int32_t encodingStrLength; |
58 | uint8_t* swapped; |
59 | UBool ownPv, ownEncodingStrings; |
60 | }; |
61 | |
62 | static void generateSelectorData(UConverterSelector* result, |
63 | UPropsVectors *upvec, |
64 | const USet* excludedCodePoints, |
65 | const UConverterUnicodeSet whichSet, |
66 | UErrorCode* status) { |
67 | if (U_FAILURE(*status)) { |
68 | return; |
69 | } |
70 | |
71 | int32_t columns = (result->encodingsCount+31)/32; |
72 | |
73 | // set errorValue to all-ones |
74 | for (int32_t col = 0; col < columns; col++) { |
75 | upvec_setValue(upvec, UPVEC_ERROR_VALUE_CP, UPVEC_ERROR_VALUE_CP, |
76 | col, static_cast<uint32_t>(~0), static_cast<uint32_t>(~0), status); |
77 | } |
78 | |
79 | for (int32_t i = 0; i < result->encodingsCount; ++i) { |
80 | uint32_t mask; |
81 | uint32_t column; |
82 | int32_t item_count; |
83 | int32_t j; |
84 | UConverter* test_converter = ucnv_open(result->encodings[i], status); |
85 | if (U_FAILURE(*status)) { |
86 | return; |
87 | } |
88 | USet* unicode_point_set; |
89 | unicode_point_set = uset_open(1, 0); // empty set |
90 | |
91 | ucnv_getUnicodeSet(test_converter, unicode_point_set, |
92 | whichSet, status); |
93 | if (U_FAILURE(*status)) { |
94 | ucnv_close(test_converter); |
95 | return; |
96 | } |
97 | |
98 | column = i / 32; |
99 | mask = 1 << (i%32); |
100 | // now iterate over intervals on set i! |
101 | item_count = uset_getItemCount(unicode_point_set); |
102 | |
103 | for (j = 0; j < item_count; ++j) { |
104 | UChar32 start_char; |
105 | UChar32 end_char; |
106 | UErrorCode smallStatus = U_ZERO_ERROR; |
107 | uset_getItem(unicode_point_set, j, &start_char, &end_char, NULL, 0, |
108 | &smallStatus); |
109 | if (U_FAILURE(smallStatus)) { |
110 | // this will be reached for the converters that fill the set with |
111 | // strings. Those should be ignored by our system |
112 | } else { |
113 | upvec_setValue(upvec, start_char, end_char, column, static_cast<uint32_t>(~0), mask, |
114 | status); |
115 | } |
116 | } |
117 | ucnv_close(test_converter); |
118 | uset_close(unicode_point_set); |
119 | if (U_FAILURE(*status)) { |
120 | return; |
121 | } |
122 | } |
123 | |
124 | // handle excluded encodings! Simply set their values to all 1's in the upvec |
125 | if (excludedCodePoints) { |
126 | int32_t item_count = uset_getItemCount(excludedCodePoints); |
127 | for (int32_t j = 0; j < item_count; ++j) { |
128 | UChar32 start_char; |
129 | UChar32 end_char; |
130 | |
131 | uset_getItem(excludedCodePoints, j, &start_char, &end_char, NULL, 0, |
132 | status); |
133 | for (int32_t col = 0; col < columns; col++) { |
134 | upvec_setValue(upvec, start_char, end_char, col, static_cast<uint32_t>(~0), static_cast<uint32_t>(~0), |
135 | status); |
136 | } |
137 | } |
138 | } |
139 | |
140 | // alright. Now, let's put things in the same exact form you'd get when you |
141 | // unserialize things. |
142 | result->trie = upvec_compactToUTrie2WithRowIndexes(upvec, status); |
143 | result->pv = upvec_cloneArray(upvec, &result->pvCount, NULL, status); |
144 | result->pvCount *= columns; // number of uint32_t = rows * columns |
145 | result->ownPv = TRUE; |
146 | } |
147 | |
148 | /* open a selector. If converterListSize is 0, build for all converters. |
149 | If excludedCodePoints is NULL, don't exclude any codepoints */ |
150 | U_CAPI UConverterSelector* U_EXPORT2 |
151 | ucnvsel_open(const char* const* converterList, int32_t converterListSize, |
152 | const USet* excludedCodePoints, |
153 | const UConverterUnicodeSet whichSet, UErrorCode* status) { |
154 | // check if already failed |
155 | if (U_FAILURE(*status)) { |
156 | return NULL; |
157 | } |
158 | // ensure args make sense! |
159 | if (converterListSize < 0 || (converterList == NULL && converterListSize != 0)) { |
160 | *status = U_ILLEGAL_ARGUMENT_ERROR; |
161 | return NULL; |
162 | } |
163 | |
164 | // allocate a new converter |
165 | LocalUConverterSelectorPointer newSelector( |
166 | (UConverterSelector*)uprv_malloc(sizeof(UConverterSelector))); |
167 | if (newSelector.isNull()) { |
168 | *status = U_MEMORY_ALLOCATION_ERROR; |
169 | return NULL; |
170 | } |
171 | uprv_memset(newSelector.getAlias(), 0, sizeof(UConverterSelector)); |
172 | |
173 | if (converterListSize == 0) { |
174 | converterList = NULL; |
175 | converterListSize = ucnv_countAvailable(); |
176 | } |
177 | newSelector->encodings = |
178 | (char**)uprv_malloc(converterListSize * sizeof(char*)); |
179 | if (!newSelector->encodings) { |
180 | *status = U_MEMORY_ALLOCATION_ERROR; |
181 | return NULL; |
182 | } |
183 | newSelector->encodings[0] = NULL; // now we can call ucnvsel_close() |
184 | |
185 | // make a backup copy of the list of converters |
186 | int32_t totalSize = 0; |
187 | int32_t i; |
188 | for (i = 0; i < converterListSize; i++) { |
189 | totalSize += |
190 | (int32_t)uprv_strlen(converterList != NULL ? converterList[i] : ucnv_getAvailableName(i)) + 1; |
191 | } |
192 | // 4-align the totalSize to 4-align the size of the serialized form |
193 | int32_t encodingStrPadding = totalSize & 3; |
194 | if (encodingStrPadding != 0) { |
195 | encodingStrPadding = 4 - encodingStrPadding; |
196 | } |
197 | newSelector->encodingStrLength = totalSize += encodingStrPadding; |
198 | char* allStrings = (char*) uprv_malloc(totalSize); |
199 | if (!allStrings) { |
200 | *status = U_MEMORY_ALLOCATION_ERROR; |
201 | return NULL; |
202 | } |
203 | |
204 | for (i = 0; i < converterListSize; i++) { |
205 | newSelector->encodings[i] = allStrings; |
206 | uprv_strcpy(newSelector->encodings[i], |
207 | converterList != NULL ? converterList[i] : ucnv_getAvailableName(i)); |
208 | allStrings += uprv_strlen(newSelector->encodings[i]) + 1; |
209 | } |
210 | while (encodingStrPadding > 0) { |
211 | *allStrings++ = 0; |
212 | --encodingStrPadding; |
213 | } |
214 | |
215 | newSelector->ownEncodingStrings = TRUE; |
216 | newSelector->encodingsCount = converterListSize; |
217 | UPropsVectors *upvec = upvec_open((converterListSize+31)/32, status); |
218 | generateSelectorData(newSelector.getAlias(), upvec, excludedCodePoints, whichSet, status); |
219 | upvec_close(upvec); |
220 | |
221 | if (U_FAILURE(*status)) { |
222 | return NULL; |
223 | } |
224 | |
225 | return newSelector.orphan(); |
226 | } |
227 | |
228 | /* close opened selector */ |
229 | U_CAPI void U_EXPORT2 |
230 | ucnvsel_close(UConverterSelector *sel) { |
231 | if (!sel) { |
232 | return; |
233 | } |
234 | if (sel->ownEncodingStrings) { |
235 | uprv_free(sel->encodings[0]); |
236 | } |
237 | uprv_free(sel->encodings); |
238 | if (sel->ownPv) { |
239 | uprv_free(sel->pv); |
240 | } |
241 | utrie2_close(sel->trie); |
242 | uprv_free(sel->swapped); |
243 | uprv_free(sel); |
244 | } |
245 | |
246 | static const UDataInfo dataInfo = { |
247 | sizeof(UDataInfo), |
248 | 0, |
249 | |
250 | U_IS_BIG_ENDIAN, |
251 | U_CHARSET_FAMILY, |
252 | U_SIZEOF_UCHAR, |
253 | 0, |
254 | |
255 | { 0x43, 0x53, 0x65, 0x6c }, /* dataFormat="CSel" */ |
256 | { 1, 0, 0, 0 }, /* formatVersion */ |
257 | { 0, 0, 0, 0 } /* dataVersion */ |
258 | }; |
259 | |
260 | enum { |
261 | UCNVSEL_INDEX_TRIE_SIZE, // trie size in bytes |
262 | UCNVSEL_INDEX_PV_COUNT, // number of uint32_t in the bit vectors |
263 | UCNVSEL_INDEX_NAMES_COUNT, // number of encoding names |
264 | UCNVSEL_INDEX_NAMES_LENGTH, // number of encoding name bytes including padding |
265 | UCNVSEL_INDEX_SIZE = 15, // bytes following the DataHeader |
266 | UCNVSEL_INDEX_COUNT = 16 |
267 | }; |
268 | |
269 | /* |
270 | * Serialized form of a UConverterSelector, formatVersion 1: |
271 | * |
272 | * The serialized form begins with a standard ICU DataHeader with a UDataInfo |
273 | * as the template above. |
274 | * This is followed by: |
275 | * int32_t indexes[UCNVSEL_INDEX_COUNT]; // see index entry constants above |
276 | * serialized UTrie2; // indexes[UCNVSEL_INDEX_TRIE_SIZE] bytes |
277 | * uint32_t pv[indexes[UCNVSEL_INDEX_PV_COUNT]]; // bit vectors |
278 | * char* encodingNames[indexes[UCNVSEL_INDEX_NAMES_LENGTH]]; // NUL-terminated strings + padding |
279 | */ |
280 | |
281 | /* serialize a selector */ |
282 | U_CAPI int32_t U_EXPORT2 |
283 | ucnvsel_serialize(const UConverterSelector* sel, |
284 | void* buffer, int32_t bufferCapacity, UErrorCode* status) { |
285 | // check if already failed |
286 | if (U_FAILURE(*status)) { |
287 | return 0; |
288 | } |
289 | // ensure args make sense! |
290 | uint8_t *p = (uint8_t *)buffer; |
291 | if (bufferCapacity < 0 || |
292 | (bufferCapacity > 0 && (p == NULL || (U_POINTER_MASK_LSB(p, 3) != 0))) |
293 | ) { |
294 | *status = U_ILLEGAL_ARGUMENT_ERROR; |
295 | return 0; |
296 | } |
297 | // add up the size of the serialized form |
298 | int32_t serializedTrieSize = utrie2_serialize(sel->trie, NULL, 0, status); |
299 | if (*status != U_BUFFER_OVERFLOW_ERROR && U_FAILURE(*status)) { |
300 | return 0; |
301 | } |
302 | *status = U_ZERO_ERROR; |
303 | |
304 | DataHeader ; |
305 | uprv_memset(&header, 0, sizeof(header)); |
306 | header.dataHeader.headerSize = (uint16_t)((sizeof(header) + 15) & ~15); |
307 | header.dataHeader.magic1 = 0xda; |
308 | header.dataHeader.magic2 = 0x27; |
309 | uprv_memcpy(&header.info, &dataInfo, sizeof(dataInfo)); |
310 | |
311 | int32_t indexes[UCNVSEL_INDEX_COUNT] = { |
312 | serializedTrieSize, |
313 | sel->pvCount, |
314 | sel->encodingsCount, |
315 | sel->encodingStrLength |
316 | }; |
317 | |
318 | int32_t totalSize = |
319 | header.dataHeader.headerSize + |
320 | (int32_t)sizeof(indexes) + |
321 | serializedTrieSize + |
322 | sel->pvCount * 4 + |
323 | sel->encodingStrLength; |
324 | indexes[UCNVSEL_INDEX_SIZE] = totalSize - header.dataHeader.headerSize; |
325 | if (totalSize > bufferCapacity) { |
326 | *status = U_BUFFER_OVERFLOW_ERROR; |
327 | return totalSize; |
328 | } |
329 | // ok, save! |
330 | int32_t length = header.dataHeader.headerSize; |
331 | uprv_memcpy(p, &header, sizeof(header)); |
332 | uprv_memset(p + sizeof(header), 0, length - sizeof(header)); |
333 | p += length; |
334 | |
335 | length = (int32_t)sizeof(indexes); |
336 | uprv_memcpy(p, indexes, length); |
337 | p += length; |
338 | |
339 | utrie2_serialize(sel->trie, p, serializedTrieSize, status); |
340 | p += serializedTrieSize; |
341 | |
342 | length = sel->pvCount * 4; |
343 | uprv_memcpy(p, sel->pv, length); |
344 | p += length; |
345 | |
346 | uprv_memcpy(p, sel->encodings[0], sel->encodingStrLength); |
347 | p += sel->encodingStrLength; |
348 | |
349 | return totalSize; |
350 | } |
351 | |
352 | /** |
353 | * swap a selector into the desired Endianness and Asciiness of |
354 | * the system. Just as FYI, selectors are always saved in the format |
355 | * of the system that created them. They are only converted if used |
356 | * on another system. In other words, selectors created on different |
357 | * system can be different even if the params are identical (endianness |
358 | * and Asciiness differences only) |
359 | * |
360 | * @param ds pointer to data swapper containing swapping info |
361 | * @param inData pointer to incoming data |
362 | * @param length length of inData in bytes |
363 | * @param outData pointer to output data. Capacity should |
364 | * be at least equal to capacity of inData |
365 | * @param status an in/out ICU UErrorCode |
366 | * @return 0 on failure, number of bytes swapped on success |
367 | * number of bytes swapped can be smaller than length |
368 | */ |
369 | static int32_t |
370 | ucnvsel_swap(const UDataSwapper *ds, |
371 | const void *inData, int32_t length, |
372 | void *outData, UErrorCode *status) { |
373 | /* udata_swapDataHeader checks the arguments */ |
374 | int32_t = udata_swapDataHeader(ds, inData, length, outData, status); |
375 | if(U_FAILURE(*status)) { |
376 | return 0; |
377 | } |
378 | |
379 | /* check data format and format version */ |
380 | const UDataInfo *pInfo = (const UDataInfo *)((const char *)inData + 4); |
381 | if(!( |
382 | pInfo->dataFormat[0] == 0x43 && /* dataFormat="CSel" */ |
383 | pInfo->dataFormat[1] == 0x53 && |
384 | pInfo->dataFormat[2] == 0x65 && |
385 | pInfo->dataFormat[3] == 0x6c |
386 | )) { |
387 | udata_printError(ds, "ucnvsel_swap(): data format %02x.%02x.%02x.%02x is not recognized as UConverterSelector data\n" , |
388 | pInfo->dataFormat[0], pInfo->dataFormat[1], |
389 | pInfo->dataFormat[2], pInfo->dataFormat[3]); |
390 | *status = U_INVALID_FORMAT_ERROR; |
391 | return 0; |
392 | } |
393 | if(pInfo->formatVersion[0] != 1) { |
394 | udata_printError(ds, "ucnvsel_swap(): format version %02x is not supported\n" , |
395 | pInfo->formatVersion[0]); |
396 | *status = U_UNSUPPORTED_ERROR; |
397 | return 0; |
398 | } |
399 | |
400 | if(length >= 0) { |
401 | length -= headerSize; |
402 | if(length < 16*4) { |
403 | udata_printError(ds, "ucnvsel_swap(): too few bytes (%d after header) for UConverterSelector data\n" , |
404 | length); |
405 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
406 | return 0; |
407 | } |
408 | } |
409 | |
410 | const uint8_t *inBytes = (const uint8_t *)inData + headerSize; |
411 | uint8_t *outBytes = (uint8_t *)outData + headerSize; |
412 | |
413 | /* read the indexes */ |
414 | const int32_t *inIndexes = (const int32_t *)inBytes; |
415 | int32_t indexes[16]; |
416 | int32_t i; |
417 | for(i = 0; i < 16; ++i) { |
418 | indexes[i] = udata_readInt32(ds, inIndexes[i]); |
419 | } |
420 | |
421 | /* get the total length of the data */ |
422 | int32_t size = indexes[UCNVSEL_INDEX_SIZE]; |
423 | if(length >= 0) { |
424 | if(length < size) { |
425 | udata_printError(ds, "ucnvsel_swap(): too few bytes (%d after header) for all of UConverterSelector data\n" , |
426 | length); |
427 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
428 | return 0; |
429 | } |
430 | |
431 | /* copy the data for inaccessible bytes */ |
432 | if(inBytes != outBytes) { |
433 | uprv_memcpy(outBytes, inBytes, size); |
434 | } |
435 | |
436 | int32_t offset = 0, count; |
437 | |
438 | /* swap the int32_t indexes[] */ |
439 | count = UCNVSEL_INDEX_COUNT*4; |
440 | ds->swapArray32(ds, inBytes, count, outBytes, status); |
441 | offset += count; |
442 | |
443 | /* swap the UTrie2 */ |
444 | count = indexes[UCNVSEL_INDEX_TRIE_SIZE]; |
445 | utrie2_swap(ds, inBytes + offset, count, outBytes + offset, status); |
446 | offset += count; |
447 | |
448 | /* swap the uint32_t pv[] */ |
449 | count = indexes[UCNVSEL_INDEX_PV_COUNT]*4; |
450 | ds->swapArray32(ds, inBytes + offset, count, outBytes + offset, status); |
451 | offset += count; |
452 | |
453 | /* swap the encoding names */ |
454 | count = indexes[UCNVSEL_INDEX_NAMES_LENGTH]; |
455 | ds->swapInvChars(ds, inBytes + offset, count, outBytes + offset, status); |
456 | offset += count; |
457 | |
458 | U_ASSERT(offset == size); |
459 | } |
460 | |
461 | return headerSize + size; |
462 | } |
463 | |
464 | /* unserialize a selector */ |
465 | U_CAPI UConverterSelector* U_EXPORT2 |
466 | ucnvsel_openFromSerialized(const void* buffer, int32_t length, UErrorCode* status) { |
467 | // check if already failed |
468 | if (U_FAILURE(*status)) { |
469 | return NULL; |
470 | } |
471 | // ensure args make sense! |
472 | const uint8_t *p = (const uint8_t *)buffer; |
473 | if (length <= 0 || |
474 | (length > 0 && (p == NULL || (U_POINTER_MASK_LSB(p, 3) != 0))) |
475 | ) { |
476 | *status = U_ILLEGAL_ARGUMENT_ERROR; |
477 | return NULL; |
478 | } |
479 | // header |
480 | if (length < 32) { |
481 | // not even enough space for a minimal header |
482 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
483 | return NULL; |
484 | } |
485 | const DataHeader * = (const DataHeader *)p; |
486 | if (!( |
487 | pHeader->dataHeader.magic1==0xda && |
488 | pHeader->dataHeader.magic2==0x27 && |
489 | pHeader->info.dataFormat[0] == 0x43 && |
490 | pHeader->info.dataFormat[1] == 0x53 && |
491 | pHeader->info.dataFormat[2] == 0x65 && |
492 | pHeader->info.dataFormat[3] == 0x6c |
493 | )) { |
494 | /* header not valid or dataFormat not recognized */ |
495 | *status = U_INVALID_FORMAT_ERROR; |
496 | return NULL; |
497 | } |
498 | if (pHeader->info.formatVersion[0] != 1) { |
499 | *status = U_UNSUPPORTED_ERROR; |
500 | return NULL; |
501 | } |
502 | uint8_t* swapped = NULL; |
503 | if (pHeader->info.isBigEndian != U_IS_BIG_ENDIAN || |
504 | pHeader->info.charsetFamily != U_CHARSET_FAMILY |
505 | ) { |
506 | // swap the data |
507 | UDataSwapper *ds = |
508 | udata_openSwapperForInputData(p, length, U_IS_BIG_ENDIAN, U_CHARSET_FAMILY, status); |
509 | int32_t totalSize = ucnvsel_swap(ds, p, -1, NULL, status); |
510 | if (U_FAILURE(*status)) { |
511 | udata_closeSwapper(ds); |
512 | return NULL; |
513 | } |
514 | if (length < totalSize) { |
515 | udata_closeSwapper(ds); |
516 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
517 | return NULL; |
518 | } |
519 | swapped = (uint8_t*)uprv_malloc(totalSize); |
520 | if (swapped == NULL) { |
521 | udata_closeSwapper(ds); |
522 | *status = U_MEMORY_ALLOCATION_ERROR; |
523 | return NULL; |
524 | } |
525 | ucnvsel_swap(ds, p, length, swapped, status); |
526 | udata_closeSwapper(ds); |
527 | if (U_FAILURE(*status)) { |
528 | uprv_free(swapped); |
529 | return NULL; |
530 | } |
531 | p = swapped; |
532 | pHeader = (const DataHeader *)p; |
533 | } |
534 | if (length < (pHeader->dataHeader.headerSize + 16 * 4)) { |
535 | // not even enough space for the header and the indexes |
536 | uprv_free(swapped); |
537 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
538 | return NULL; |
539 | } |
540 | p += pHeader->dataHeader.headerSize; |
541 | length -= pHeader->dataHeader.headerSize; |
542 | // indexes |
543 | const int32_t *indexes = (const int32_t *)p; |
544 | if (length < indexes[UCNVSEL_INDEX_SIZE]) { |
545 | uprv_free(swapped); |
546 | *status = U_INDEX_OUTOFBOUNDS_ERROR; |
547 | return NULL; |
548 | } |
549 | p += UCNVSEL_INDEX_COUNT * 4; |
550 | // create and populate the selector object |
551 | UConverterSelector* sel = (UConverterSelector*)uprv_malloc(sizeof(UConverterSelector)); |
552 | char **encodings = |
553 | (char **)uprv_malloc( |
554 | indexes[UCNVSEL_INDEX_NAMES_COUNT] * sizeof(char *)); |
555 | if (sel == NULL || encodings == NULL) { |
556 | uprv_free(swapped); |
557 | uprv_free(sel); |
558 | uprv_free(encodings); |
559 | *status = U_MEMORY_ALLOCATION_ERROR; |
560 | return NULL; |
561 | } |
562 | uprv_memset(sel, 0, sizeof(UConverterSelector)); |
563 | sel->pvCount = indexes[UCNVSEL_INDEX_PV_COUNT]; |
564 | sel->encodings = encodings; |
565 | sel->encodingsCount = indexes[UCNVSEL_INDEX_NAMES_COUNT]; |
566 | sel->encodingStrLength = indexes[UCNVSEL_INDEX_NAMES_LENGTH]; |
567 | sel->swapped = swapped; |
568 | // trie |
569 | sel->trie = utrie2_openFromSerialized(UTRIE2_16_VALUE_BITS, |
570 | p, indexes[UCNVSEL_INDEX_TRIE_SIZE], NULL, |
571 | status); |
572 | p += indexes[UCNVSEL_INDEX_TRIE_SIZE]; |
573 | if (U_FAILURE(*status)) { |
574 | ucnvsel_close(sel); |
575 | return NULL; |
576 | } |
577 | // bit vectors |
578 | sel->pv = (uint32_t *)p; |
579 | p += sel->pvCount * 4; |
580 | // encoding names |
581 | char* s = (char*)p; |
582 | for (int32_t i = 0; i < sel->encodingsCount; ++i) { |
583 | sel->encodings[i] = s; |
584 | s += uprv_strlen(s) + 1; |
585 | } |
586 | p += sel->encodingStrLength; |
587 | |
588 | return sel; |
589 | } |
590 | |
591 | // a bunch of functions for the enumeration thingie! Nothing fancy here. Just |
592 | // iterate over the selected encodings |
593 | struct Enumerator { |
594 | int16_t* index; |
595 | int16_t length; |
596 | int16_t cur; |
597 | const UConverterSelector* sel; |
598 | }; |
599 | |
600 | U_CDECL_BEGIN |
601 | |
602 | static void U_CALLCONV |
603 | ucnvsel_close_selector_iterator(UEnumeration *enumerator) { |
604 | uprv_free(((Enumerator*)(enumerator->context))->index); |
605 | uprv_free(enumerator->context); |
606 | uprv_free(enumerator); |
607 | } |
608 | |
609 | |
610 | static int32_t U_CALLCONV |
611 | ucnvsel_count_encodings(UEnumeration *enumerator, UErrorCode *status) { |
612 | // check if already failed |
613 | if (U_FAILURE(*status)) { |
614 | return 0; |
615 | } |
616 | return ((Enumerator*)(enumerator->context))->length; |
617 | } |
618 | |
619 | |
620 | static const char* U_CALLCONV ucnvsel_next_encoding(UEnumeration* enumerator, |
621 | int32_t* resultLength, |
622 | UErrorCode* status) { |
623 | // check if already failed |
624 | if (U_FAILURE(*status)) { |
625 | return NULL; |
626 | } |
627 | |
628 | int16_t cur = ((Enumerator*)(enumerator->context))->cur; |
629 | const UConverterSelector* sel; |
630 | const char* result; |
631 | if (cur >= ((Enumerator*)(enumerator->context))->length) { |
632 | return NULL; |
633 | } |
634 | sel = ((Enumerator*)(enumerator->context))->sel; |
635 | result = sel->encodings[((Enumerator*)(enumerator->context))->index[cur] ]; |
636 | ((Enumerator*)(enumerator->context))->cur++; |
637 | if (resultLength) { |
638 | *resultLength = (int32_t)uprv_strlen(result); |
639 | } |
640 | return result; |
641 | } |
642 | |
643 | static void U_CALLCONV ucnvsel_reset_iterator(UEnumeration* enumerator, |
644 | UErrorCode* status) { |
645 | // check if already failed |
646 | if (U_FAILURE(*status)) { |
647 | return ; |
648 | } |
649 | ((Enumerator*)(enumerator->context))->cur = 0; |
650 | } |
651 | |
652 | U_CDECL_END |
653 | |
654 | |
655 | static const UEnumeration defaultEncodings = { |
656 | NULL, |
657 | NULL, |
658 | ucnvsel_close_selector_iterator, |
659 | ucnvsel_count_encodings, |
660 | uenum_unextDefault, |
661 | ucnvsel_next_encoding, |
662 | ucnvsel_reset_iterator |
663 | }; |
664 | |
665 | |
666 | // internal fn to intersect two sets of masks |
667 | // returns whether the mask has reduced to all zeros |
668 | static UBool intersectMasks(uint32_t* dest, const uint32_t* source1, int32_t len) { |
669 | int32_t i; |
670 | uint32_t oredDest = 0; |
671 | for (i = 0 ; i < len ; ++i) { |
672 | oredDest |= (dest[i] &= source1[i]); |
673 | } |
674 | return oredDest == 0; |
675 | } |
676 | |
677 | // internal fn to count how many 1's are there in a mask |
678 | // algorithm taken from http://graphics.stanford.edu/~seander/bithacks.html |
679 | static int16_t countOnes(uint32_t* mask, int32_t len) { |
680 | int32_t i, totalOnes = 0; |
681 | for (i = 0 ; i < len ; ++i) { |
682 | uint32_t ent = mask[i]; |
683 | for (; ent; totalOnes++) |
684 | { |
685 | ent &= ent - 1; // clear the least significant bit set |
686 | } |
687 | } |
688 | return static_cast<int16_t>(totalOnes); |
689 | } |
690 | |
691 | |
692 | /* internal function! */ |
693 | static UEnumeration *selectForMask(const UConverterSelector* sel, |
694 | uint32_t *theMask, UErrorCode *status) { |
695 | LocalMemory<uint32_t> mask(theMask); |
696 | // this is the context we will use. Store a table of indices to which |
697 | // encodings are legit. |
698 | LocalMemory<Enumerator> result(static_cast<Enumerator *>(uprv_malloc(sizeof(Enumerator)))); |
699 | if (result.isNull()) { |
700 | *status = U_MEMORY_ALLOCATION_ERROR; |
701 | return nullptr; |
702 | } |
703 | result->index = nullptr; // this will be allocated later! |
704 | result->length = result->cur = 0; |
705 | result->sel = sel; |
706 | |
707 | LocalMemory<UEnumeration> en(static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration)))); |
708 | if (en.isNull()) { |
709 | // TODO(markus): Combine Enumerator and UEnumeration into one struct. |
710 | *status = U_MEMORY_ALLOCATION_ERROR; |
711 | return nullptr; |
712 | } |
713 | memcpy(en.getAlias(), &defaultEncodings, sizeof(UEnumeration)); |
714 | |
715 | int32_t columns = (sel->encodingsCount+31)/32; |
716 | int16_t numOnes = countOnes(mask.getAlias(), columns); |
717 | // now, we know the exact space we need for index |
718 | if (numOnes > 0) { |
719 | result->index = static_cast<int16_t*>(uprv_malloc(numOnes * sizeof(int16_t))); |
720 | if (result->index == nullptr) { |
721 | *status = U_MEMORY_ALLOCATION_ERROR; |
722 | return nullptr; |
723 | } |
724 | int32_t i, j; |
725 | int16_t k = 0; |
726 | for (j = 0 ; j < columns; j++) { |
727 | uint32_t v = mask[j]; |
728 | for (i = 0 ; i < 32 && k < sel->encodingsCount; i++, k++) { |
729 | if ((v & 1) != 0) { |
730 | result->index[result->length++] = k; |
731 | } |
732 | v >>= 1; |
733 | } |
734 | } |
735 | } //otherwise, index will remain NULL (and will never be touched by |
736 | //the enumerator code anyway) |
737 | en->context = result.orphan(); |
738 | return en.orphan(); |
739 | } |
740 | |
741 | /* check a string against the selector - UTF16 version */ |
742 | U_CAPI UEnumeration * U_EXPORT2 |
743 | ucnvsel_selectForString(const UConverterSelector* sel, |
744 | const UChar *s, int32_t length, UErrorCode *status) { |
745 | // check if already failed |
746 | if (U_FAILURE(*status)) { |
747 | return NULL; |
748 | } |
749 | // ensure args make sense! |
750 | if (sel == NULL || (s == NULL && length != 0)) { |
751 | *status = U_ILLEGAL_ARGUMENT_ERROR; |
752 | return NULL; |
753 | } |
754 | |
755 | int32_t columns = (sel->encodingsCount+31)/32; |
756 | uint32_t* mask = (uint32_t*) uprv_malloc(columns * 4); |
757 | if (mask == NULL) { |
758 | *status = U_MEMORY_ALLOCATION_ERROR; |
759 | return NULL; |
760 | } |
761 | uprv_memset(mask, ~0, columns *4); |
762 | |
763 | if(s!=NULL) { |
764 | const UChar *limit; |
765 | if (length >= 0) { |
766 | limit = s + length; |
767 | } else { |
768 | limit = NULL; |
769 | } |
770 | |
771 | while (limit == NULL ? *s != 0 : s != limit) { |
772 | UChar32 c; |
773 | uint16_t pvIndex; |
774 | UTRIE2_U16_NEXT16(sel->trie, s, limit, c, pvIndex); |
775 | if (intersectMasks(mask, sel->pv+pvIndex, columns)) { |
776 | break; |
777 | } |
778 | } |
779 | } |
780 | return selectForMask(sel, mask, status); |
781 | } |
782 | |
783 | /* check a string against the selector - UTF8 version */ |
784 | U_CAPI UEnumeration * U_EXPORT2 |
785 | ucnvsel_selectForUTF8(const UConverterSelector* sel, |
786 | const char *s, int32_t length, UErrorCode *status) { |
787 | // check if already failed |
788 | if (U_FAILURE(*status)) { |
789 | return NULL; |
790 | } |
791 | // ensure args make sense! |
792 | if (sel == NULL || (s == NULL && length != 0)) { |
793 | *status = U_ILLEGAL_ARGUMENT_ERROR; |
794 | return NULL; |
795 | } |
796 | |
797 | int32_t columns = (sel->encodingsCount+31)/32; |
798 | uint32_t* mask = (uint32_t*) uprv_malloc(columns * 4); |
799 | if (mask == NULL) { |
800 | *status = U_MEMORY_ALLOCATION_ERROR; |
801 | return NULL; |
802 | } |
803 | uprv_memset(mask, ~0, columns *4); |
804 | |
805 | if (length < 0) { |
806 | length = (int32_t)uprv_strlen(s); |
807 | } |
808 | |
809 | if(s!=NULL) { |
810 | const char *limit = s + length; |
811 | |
812 | while (s != limit) { |
813 | uint16_t pvIndex; |
814 | UTRIE2_U8_NEXT16(sel->trie, s, limit, pvIndex); |
815 | if (intersectMasks(mask, sel->pv+pvIndex, columns)) { |
816 | break; |
817 | } |
818 | } |
819 | } |
820 | return selectForMask(sel, mask, status); |
821 | } |
822 | |
823 | #endif // !UCONFIG_NO_CONVERSION |
824 | |