claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1 | /* |
| 2 | ******************************************************************************* |
| 3 | * |
| 4 | * Copyright (C) 2009-2010, International Business Machines |
| 5 | * Corporation and others. All Rights Reserved. |
| 6 | * |
| 7 | ******************************************************************************* |
| 8 | * file name: normalizer2impl.cpp |
| 9 | * encoding: US-ASCII |
| 10 | * tab size: 8 (not used) |
| 11 | * indentation:4 |
| 12 | * |
| 13 | * created on: 2009nov22 |
| 14 | * created by: Markus W. Scherer |
| 15 | */ |
| 16 | |
| 17 | #include "unicode/utypes.h" |
| 18 | |
| 19 | #if !UCONFIG_NO_NORMALIZATION |
| 20 | |
| 21 | #include "unicode/normalizer2.h" |
| 22 | #include "unicode/udata.h" |
| 23 | #include "unicode/ustring.h" |
| 24 | #include "cmemory.h" |
| 25 | #include "mutex.h" |
| 26 | #include "normalizer2impl.h" |
| 27 | #include "uassert.h" |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 28 | #include "uhash.h" |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 29 | #include "uset_imp.h" |
| 30 | #include "utrie2.h" |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 31 | #include "uvector.h" |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 32 | |
| 33 | U_NAMESPACE_BEGIN |
| 34 | |
| 35 | // ReorderingBuffer -------------------------------------------------------- *** |
| 36 | |
| 37 | UBool ReorderingBuffer::init(int32_t destCapacity, UErrorCode &errorCode) { |
| 38 | int32_t length=str.length(); |
| 39 | start=str.getBuffer(destCapacity); |
| 40 | if(start==NULL) { |
| 41 | // getBuffer() already did str.setToBogus() |
| 42 | errorCode=U_MEMORY_ALLOCATION_ERROR; |
| 43 | return FALSE; |
| 44 | } |
| 45 | limit=start+length; |
| 46 | remainingCapacity=str.getCapacity()-length; |
| 47 | reorderStart=start; |
| 48 | if(start==limit) { |
| 49 | lastCC=0; |
| 50 | } else { |
| 51 | setIterator(); |
| 52 | lastCC=previousCC(); |
| 53 | // Set reorderStart after the last code point with cc<=1 if there is one. |
| 54 | if(lastCC>1) { |
| 55 | while(previousCC()>1) {} |
| 56 | } |
| 57 | reorderStart=codePointLimit; |
| 58 | } |
| 59 | return TRUE; |
| 60 | } |
| 61 | |
| 62 | UBool ReorderingBuffer::equals(const UChar *otherStart, const UChar *otherLimit) const { |
| 63 | int32_t length=(int32_t)(limit-start); |
| 64 | return |
| 65 | length==(int32_t)(otherLimit-otherStart) && |
| 66 | 0==u_memcmp(start, otherStart, length); |
| 67 | } |
| 68 | |
| 69 | UBool ReorderingBuffer::appendSupplementary(UChar32 c, uint8_t cc, UErrorCode &errorCode) { |
| 70 | if(remainingCapacity<2 && !resize(2, errorCode)) { |
| 71 | return FALSE; |
| 72 | } |
| 73 | if(lastCC<=cc || cc==0) { |
| 74 | limit[0]=U16_LEAD(c); |
| 75 | limit[1]=U16_TRAIL(c); |
| 76 | limit+=2; |
| 77 | lastCC=cc; |
| 78 | if(cc<=1) { |
| 79 | reorderStart=limit; |
| 80 | } |
| 81 | } else { |
| 82 | insert(c, cc); |
| 83 | } |
| 84 | remainingCapacity-=2; |
| 85 | return TRUE; |
| 86 | } |
| 87 | |
| 88 | UBool ReorderingBuffer::append(const UChar *s, int32_t length, |
| 89 | uint8_t leadCC, uint8_t trailCC, |
| 90 | UErrorCode &errorCode) { |
| 91 | if(length==0) { |
| 92 | return TRUE; |
| 93 | } |
| 94 | if(remainingCapacity<length && !resize(length, errorCode)) { |
| 95 | return FALSE; |
| 96 | } |
| 97 | remainingCapacity-=length; |
| 98 | if(lastCC<=leadCC || leadCC==0) { |
| 99 | if(trailCC<=1) { |
| 100 | reorderStart=limit+length; |
| 101 | } else if(leadCC<=1) { |
| 102 | reorderStart=limit+1; // Ok if not a code point boundary. |
| 103 | } |
| 104 | const UChar *sLimit=s+length; |
| 105 | do { *limit++=*s++; } while(s!=sLimit); |
| 106 | lastCC=trailCC; |
| 107 | } else { |
| 108 | int32_t i=0; |
| 109 | UChar32 c; |
| 110 | U16_NEXT(s, i, length, c); |
| 111 | insert(c, leadCC); // insert first code point |
| 112 | while(i<length) { |
| 113 | U16_NEXT(s, i, length, c); |
| 114 | if(i<length) { |
| 115 | // s must be in NFD, otherwise we need to use getCC(). |
| 116 | leadCC=Normalizer2Impl::getCCFromYesOrMaybe(impl.getNorm16(c)); |
| 117 | } else { |
| 118 | leadCC=trailCC; |
| 119 | } |
| 120 | append(c, leadCC, errorCode); |
| 121 | } |
| 122 | } |
| 123 | return TRUE; |
| 124 | } |
| 125 | |
| 126 | UBool ReorderingBuffer::appendZeroCC(UChar32 c, UErrorCode &errorCode) { |
| 127 | int32_t cpLength=U16_LENGTH(c); |
| 128 | if(remainingCapacity<cpLength && !resize(cpLength, errorCode)) { |
| 129 | return FALSE; |
| 130 | } |
| 131 | remainingCapacity-=cpLength; |
| 132 | if(cpLength==1) { |
| 133 | *limit++=(UChar)c; |
| 134 | } else { |
| 135 | limit[0]=U16_LEAD(c); |
| 136 | limit[1]=U16_TRAIL(c); |
| 137 | limit+=2; |
| 138 | } |
| 139 | lastCC=0; |
| 140 | reorderStart=limit; |
| 141 | return TRUE; |
| 142 | } |
| 143 | |
| 144 | UBool ReorderingBuffer::appendZeroCC(const UChar *s, const UChar *sLimit, UErrorCode &errorCode) { |
| 145 | if(s==sLimit) { |
| 146 | return TRUE; |
| 147 | } |
| 148 | int32_t length=(int32_t)(sLimit-s); |
| 149 | if(remainingCapacity<length && !resize(length, errorCode)) { |
| 150 | return FALSE; |
| 151 | } |
| 152 | u_memcpy(limit, s, length); |
| 153 | limit+=length; |
| 154 | remainingCapacity-=length; |
| 155 | lastCC=0; |
| 156 | reorderStart=limit; |
| 157 | return TRUE; |
| 158 | } |
| 159 | |
| 160 | void ReorderingBuffer::remove() { |
| 161 | reorderStart=limit=start; |
| 162 | remainingCapacity=str.getCapacity(); |
| 163 | lastCC=0; |
| 164 | } |
| 165 | |
| 166 | void ReorderingBuffer::removeSuffix(int32_t suffixLength) { |
| 167 | if(suffixLength<(limit-start)) { |
| 168 | limit-=suffixLength; |
| 169 | remainingCapacity+=suffixLength; |
| 170 | } else { |
| 171 | limit=start; |
| 172 | remainingCapacity=str.getCapacity(); |
| 173 | } |
| 174 | lastCC=0; |
| 175 | reorderStart=limit; |
| 176 | } |
| 177 | |
| 178 | UBool ReorderingBuffer::resize(int32_t appendLength, UErrorCode &errorCode) { |
| 179 | int32_t reorderStartIndex=(int32_t)(reorderStart-start); |
| 180 | int32_t length=(int32_t)(limit-start); |
| 181 | str.releaseBuffer(length); |
| 182 | int32_t newCapacity=length+appendLength; |
| 183 | int32_t doubleCapacity=2*str.getCapacity(); |
| 184 | if(newCapacity<doubleCapacity) { |
| 185 | newCapacity=doubleCapacity; |
| 186 | } |
| 187 | if(newCapacity<256) { |
| 188 | newCapacity=256; |
| 189 | } |
| 190 | start=str.getBuffer(newCapacity); |
| 191 | if(start==NULL) { |
| 192 | // getBuffer() already did str.setToBogus() |
| 193 | errorCode=U_MEMORY_ALLOCATION_ERROR; |
| 194 | return FALSE; |
| 195 | } |
| 196 | reorderStart=start+reorderStartIndex; |
| 197 | limit=start+length; |
| 198 | remainingCapacity=str.getCapacity()-length; |
| 199 | return TRUE; |
| 200 | } |
| 201 | |
| 202 | void ReorderingBuffer::skipPrevious() { |
| 203 | codePointLimit=codePointStart; |
| 204 | UChar c=*--codePointStart; |
| 205 | if(U16_IS_TRAIL(c) && start<codePointStart && U16_IS_LEAD(*(codePointStart-1))) { |
| 206 | --codePointStart; |
| 207 | } |
| 208 | } |
| 209 | |
| 210 | uint8_t ReorderingBuffer::previousCC() { |
| 211 | codePointLimit=codePointStart; |
| 212 | if(reorderStart>=codePointStart) { |
| 213 | return 0; |
| 214 | } |
| 215 | UChar32 c=*--codePointStart; |
| 216 | if(c<Normalizer2Impl::MIN_CCC_LCCC_CP) { |
| 217 | return 0; |
| 218 | } |
| 219 | |
| 220 | UChar c2; |
| 221 | if(U16_IS_TRAIL(c) && start<codePointStart && U16_IS_LEAD(c2=*(codePointStart-1))) { |
| 222 | --codePointStart; |
| 223 | c=U16_GET_SUPPLEMENTARY(c2, c); |
| 224 | } |
| 225 | return Normalizer2Impl::getCCFromYesOrMaybe(impl.getNorm16(c)); |
| 226 | } |
| 227 | |
| 228 | // Inserts c somewhere before the last character. |
| 229 | // Requires 0<cc<lastCC which implies reorderStart<limit. |
| 230 | void ReorderingBuffer::insert(UChar32 c, uint8_t cc) { |
| 231 | for(setIterator(), skipPrevious(); previousCC()>cc;) {} |
| 232 | // insert c at codePointLimit, after the character with prevCC<=cc |
| 233 | UChar *q=limit; |
| 234 | UChar *r=limit+=U16_LENGTH(c); |
| 235 | do { |
| 236 | *--r=*--q; |
| 237 | } while(codePointLimit!=q); |
| 238 | writeCodePoint(q, c); |
| 239 | if(cc<=1) { |
| 240 | reorderStart=r; |
| 241 | } |
| 242 | } |
| 243 | |
| 244 | // Normalizer2Impl --------------------------------------------------------- *** |
| 245 | |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 246 | struct CanonIterData : public UMemory { |
| 247 | CanonIterData(UErrorCode &errorCode); |
| 248 | ~CanonIterData(); |
| 249 | void addToStartSet(UChar32 origin, UChar32 decompLead, UErrorCode &errorCode); |
| 250 | UTrie2 *trie; |
| 251 | UVector canonStartSets; // contains UnicodeSet * |
| 252 | }; |
| 253 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 254 | Normalizer2Impl::~Normalizer2Impl() { |
| 255 | udata_close(memory); |
| 256 | utrie2_close(normTrie); |
| 257 | UTrie2Singleton(fcdTrieSingleton).deleteInstance(); |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 258 | delete (CanonIterData *)canonIterDataSingleton.fInstance; |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 259 | } |
| 260 | |
| 261 | UBool U_CALLCONV |
| 262 | Normalizer2Impl::isAcceptable(void *context, |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 263 | const char * /* type */, const char * /*name*/, |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 264 | const UDataInfo *pInfo) { |
| 265 | if( |
| 266 | pInfo->size>=20 && |
| 267 | pInfo->isBigEndian==U_IS_BIG_ENDIAN && |
| 268 | pInfo->charsetFamily==U_CHARSET_FAMILY && |
| 269 | pInfo->dataFormat[0]==0x4e && /* dataFormat="Nrm2" */ |
| 270 | pInfo->dataFormat[1]==0x72 && |
| 271 | pInfo->dataFormat[2]==0x6d && |
| 272 | pInfo->dataFormat[3]==0x32 && |
| 273 | pInfo->formatVersion[0]==1 |
| 274 | ) { |
| 275 | Normalizer2Impl *me=(Normalizer2Impl *)context; |
| 276 | uprv_memcpy(me->dataVersion, pInfo->dataVersion, 4); |
| 277 | return TRUE; |
| 278 | } else { |
| 279 | return FALSE; |
| 280 | } |
| 281 | } |
| 282 | |
| 283 | void |
| 284 | Normalizer2Impl::load(const char *packageName, const char *name, UErrorCode &errorCode) { |
| 285 | if(U_FAILURE(errorCode)) { |
| 286 | return; |
| 287 | } |
| 288 | memory=udata_openChoice(packageName, "nrm", name, isAcceptable, this, &errorCode); |
| 289 | if(U_FAILURE(errorCode)) { |
| 290 | return; |
| 291 | } |
| 292 | const uint8_t *inBytes=(const uint8_t *)udata_getMemory(memory); |
| 293 | const int32_t *inIndexes=(const int32_t *)inBytes; |
| 294 | int32_t indexesLength=inIndexes[IX_NORM_TRIE_OFFSET]/4; |
| 295 | if(indexesLength<=IX_MIN_MAYBE_YES) { |
| 296 | errorCode=U_INVALID_FORMAT_ERROR; // Not enough indexes. |
| 297 | return; |
| 298 | } |
| 299 | |
| 300 | minDecompNoCP=inIndexes[IX_MIN_DECOMP_NO_CP]; |
| 301 | minCompNoMaybeCP=inIndexes[IX_MIN_COMP_NO_MAYBE_CP]; |
| 302 | |
| 303 | minYesNo=inIndexes[IX_MIN_YES_NO]; |
| 304 | minNoNo=inIndexes[IX_MIN_NO_NO]; |
| 305 | limitNoNo=inIndexes[IX_LIMIT_NO_NO]; |
| 306 | minMaybeYes=inIndexes[IX_MIN_MAYBE_YES]; |
| 307 | |
| 308 | int32_t offset=inIndexes[IX_NORM_TRIE_OFFSET]; |
| 309 | int32_t nextOffset=inIndexes[IX_EXTRA_DATA_OFFSET]; |
| 310 | normTrie=utrie2_openFromSerialized(UTRIE2_16_VALUE_BITS, |
| 311 | inBytes+offset, nextOffset-offset, NULL, |
| 312 | &errorCode); |
| 313 | if(U_FAILURE(errorCode)) { |
| 314 | return; |
| 315 | } |
| 316 | |
| 317 | offset=nextOffset; |
| 318 | maybeYesCompositions=(const uint16_t *)(inBytes+offset); |
| 319 | extraData=maybeYesCompositions+(MIN_NORMAL_MAYBE_YES-minMaybeYes); |
| 320 | } |
| 321 | |
| 322 | uint8_t Normalizer2Impl::getTrailCCFromCompYesAndZeroCC(const UChar *cpStart, const UChar *cpLimit) const { |
| 323 | UChar32 c; |
| 324 | if(cpStart==(cpLimit-1)) { |
| 325 | c=*cpStart; |
| 326 | } else { |
| 327 | c=U16_GET_SUPPLEMENTARY(cpStart[0], cpStart[1]); |
| 328 | } |
| 329 | uint16_t prevNorm16=getNorm16(c); |
| 330 | if(prevNorm16<=minYesNo) { |
| 331 | return 0; // yesYes and Hangul LV/LVT have ccc=tccc=0 |
| 332 | } else { |
| 333 | return (uint8_t)(*getMapping(prevNorm16)>>8); // tccc from yesNo |
| 334 | } |
| 335 | } |
| 336 | |
| 337 | U_CDECL_BEGIN |
| 338 | |
| 339 | static UBool U_CALLCONV |
| 340 | enumPropertyStartsRange(const void *context, UChar32 start, UChar32 /*end*/, uint32_t /*value*/) { |
| 341 | /* add the start code point to the USet */ |
| 342 | const USetAdder *sa=(const USetAdder *)context; |
| 343 | sa->add(sa->set, start); |
| 344 | return TRUE; |
| 345 | } |
| 346 | |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 347 | static uint32_t U_CALLCONV |
| 348 | segmentStarterMapper(const void * /*context*/, uint32_t value) { |
| 349 | return value&CANON_NOT_SEGMENT_STARTER; |
| 350 | } |
| 351 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 352 | U_CDECL_END |
| 353 | |
| 354 | void |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 355 | Normalizer2Impl::addPropertyStarts(const USetAdder *sa, UErrorCode & /*errorCode*/) const { |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 356 | /* add the start code point of each same-value range of each trie */ |
| 357 | utrie2_enum(normTrie, NULL, enumPropertyStartsRange, sa); |
| 358 | |
| 359 | /* add Hangul LV syllables and LV+1 because of skippables */ |
| 360 | for(UChar c=Hangul::HANGUL_BASE; c<Hangul::HANGUL_LIMIT; c+=Hangul::JAMO_T_COUNT) { |
| 361 | sa->add(sa->set, c); |
| 362 | sa->add(sa->set, c+1); |
| 363 | } |
| 364 | sa->add(sa->set, Hangul::HANGUL_LIMIT); /* add Hangul+1 to continue with other properties */ |
| 365 | } |
| 366 | |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 367 | void |
| 368 | Normalizer2Impl::addCanonIterPropertyStarts(const USetAdder *sa, UErrorCode &errorCode) const { |
| 369 | /* add the start code point of each same-value range of the canonical iterator data trie */ |
| 370 | if(ensureCanonIterData(errorCode)) { |
| 371 | // currently only used for the SEGMENT_STARTER property |
| 372 | utrie2_enum(((CanonIterData *)canonIterDataSingleton.fInstance)->trie, |
| 373 | segmentStarterMapper, enumPropertyStartsRange, sa); |
| 374 | } |
| 375 | } |
| 376 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 377 | const UChar * |
| 378 | Normalizer2Impl::copyLowPrefixFromNulTerminated(const UChar *src, |
| 379 | UChar32 minNeedDataCP, |
| 380 | ReorderingBuffer *buffer, |
| 381 | UErrorCode &errorCode) const { |
| 382 | // Make some effort to support NUL-terminated strings reasonably. |
| 383 | // Take the part of the fast quick check loop that does not look up |
| 384 | // data and check the first part of the string. |
| 385 | // After this prefix, determine the string length to simplify the rest |
| 386 | // of the code. |
| 387 | const UChar *prevSrc=src; |
| 388 | UChar c; |
| 389 | while((c=*src++)<minNeedDataCP && c!=0) {} |
| 390 | // Back out the last character for full processing. |
| 391 | // Copy this prefix. |
| 392 | if(--src!=prevSrc) { |
| 393 | if(buffer!=NULL) { |
| 394 | buffer->appendZeroCC(prevSrc, src, errorCode); |
| 395 | } |
| 396 | } |
| 397 | return src; |
| 398 | } |
| 399 | |
| 400 | // Dual functionality: |
| 401 | // buffer!=NULL: normalize |
| 402 | // buffer==NULL: isNormalized/spanQuickCheckYes |
| 403 | const UChar * |
| 404 | Normalizer2Impl::decompose(const UChar *src, const UChar *limit, |
| 405 | ReorderingBuffer *buffer, |
| 406 | UErrorCode &errorCode) const { |
| 407 | UChar32 minNoCP=minDecompNoCP; |
| 408 | if(limit==NULL) { |
| 409 | src=copyLowPrefixFromNulTerminated(src, minNoCP, buffer, errorCode); |
| 410 | if(U_FAILURE(errorCode)) { |
| 411 | return src; |
| 412 | } |
| 413 | limit=u_strchr(src, 0); |
| 414 | } |
| 415 | |
| 416 | const UChar *prevSrc; |
| 417 | UChar32 c=0; |
| 418 | uint16_t norm16=0; |
| 419 | |
| 420 | // only for quick check |
| 421 | const UChar *prevBoundary=src; |
| 422 | uint8_t prevCC=0; |
| 423 | |
| 424 | for(;;) { |
| 425 | // count code units below the minimum or with irrelevant data for the quick check |
| 426 | for(prevSrc=src; src!=limit;) { |
| 427 | if( (c=*src)<minNoCP || |
| 428 | isMostDecompYesAndZeroCC(norm16=UTRIE2_GET16_FROM_U16_SINGLE_LEAD(normTrie, c)) |
| 429 | ) { |
| 430 | ++src; |
| 431 | } else if(!U16_IS_SURROGATE(c)) { |
| 432 | break; |
| 433 | } else { |
| 434 | UChar c2; |
| 435 | if(U16_IS_SURROGATE_LEAD(c)) { |
| 436 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])) { |
| 437 | c=U16_GET_SUPPLEMENTARY(c, c2); |
| 438 | } |
| 439 | } else /* trail surrogate */ { |
| 440 | if(prevSrc<src && U16_IS_LEAD(c2=*(src-1))) { |
| 441 | --src; |
| 442 | c=U16_GET_SUPPLEMENTARY(c2, c); |
| 443 | } |
| 444 | } |
| 445 | if(isMostDecompYesAndZeroCC(norm16=getNorm16(c))) { |
| 446 | src+=U16_LENGTH(c); |
| 447 | } else { |
| 448 | break; |
| 449 | } |
| 450 | } |
| 451 | } |
| 452 | // copy these code units all at once |
| 453 | if(src!=prevSrc) { |
| 454 | if(buffer!=NULL) { |
| 455 | if(!buffer->appendZeroCC(prevSrc, src, errorCode)) { |
| 456 | break; |
| 457 | } |
| 458 | } else { |
| 459 | prevCC=0; |
| 460 | prevBoundary=src; |
| 461 | } |
| 462 | } |
| 463 | if(src==limit) { |
| 464 | break; |
| 465 | } |
| 466 | |
| 467 | // Check one above-minimum, relevant code point. |
| 468 | src+=U16_LENGTH(c); |
| 469 | if(buffer!=NULL) { |
| 470 | if(!decompose(c, norm16, *buffer, errorCode)) { |
| 471 | break; |
| 472 | } |
| 473 | } else { |
| 474 | if(isDecompYes(norm16)) { |
| 475 | uint8_t cc=getCCFromYesOrMaybe(norm16); |
| 476 | if(prevCC<=cc || cc==0) { |
| 477 | prevCC=cc; |
| 478 | if(cc<=1) { |
| 479 | prevBoundary=src; |
| 480 | } |
| 481 | continue; |
| 482 | } |
| 483 | } |
| 484 | return prevBoundary; // "no" or cc out of order |
| 485 | } |
| 486 | } |
| 487 | return src; |
| 488 | } |
| 489 | |
| 490 | // Decompose a short piece of text which is likely to contain characters that |
| 491 | // fail the quick check loop and/or where the quick check loop's overhead |
| 492 | // is unlikely to be amortized. |
| 493 | // Called by the compose() and makeFCD() implementations. |
| 494 | UBool Normalizer2Impl::decomposeShort(const UChar *src, const UChar *limit, |
| 495 | ReorderingBuffer &buffer, |
| 496 | UErrorCode &errorCode) const { |
| 497 | while(src<limit) { |
| 498 | UChar32 c; |
| 499 | uint16_t norm16; |
| 500 | UTRIE2_U16_NEXT16(normTrie, src, limit, c, norm16); |
| 501 | if(!decompose(c, norm16, buffer, errorCode)) { |
| 502 | return FALSE; |
| 503 | } |
| 504 | } |
| 505 | return TRUE; |
| 506 | } |
| 507 | |
| 508 | UBool Normalizer2Impl::decompose(UChar32 c, uint16_t norm16, |
| 509 | ReorderingBuffer &buffer, |
| 510 | UErrorCode &errorCode) const { |
| 511 | // Only loops for 1:1 algorithmic mappings. |
| 512 | for(;;) { |
| 513 | // get the decomposition and the lead and trail cc's |
| 514 | if(isDecompYes(norm16)) { |
| 515 | // c does not decompose |
| 516 | return buffer.append(c, getCCFromYesOrMaybe(norm16), errorCode); |
| 517 | } else if(isHangul(norm16)) { |
| 518 | // Hangul syllable: decompose algorithmically |
| 519 | UChar jamos[3]; |
| 520 | return buffer.appendZeroCC(jamos, jamos+Hangul::decompose(c, jamos), errorCode); |
| 521 | } else if(isDecompNoAlgorithmic(norm16)) { |
| 522 | c=mapAlgorithmic(c, norm16); |
| 523 | norm16=getNorm16(c); |
| 524 | } else { |
| 525 | // c decomposes, get everything from the variable-length extra data |
| 526 | const uint16_t *mapping=getMapping(norm16); |
| 527 | uint16_t firstUnit=*mapping++; |
| 528 | int32_t length=firstUnit&MAPPING_LENGTH_MASK; |
| 529 | uint8_t leadCC, trailCC; |
| 530 | trailCC=(uint8_t)(firstUnit>>8); |
| 531 | if(firstUnit&MAPPING_HAS_CCC_LCCC_WORD) { |
| 532 | leadCC=(uint8_t)(*mapping++>>8); |
| 533 | } else { |
| 534 | leadCC=0; |
| 535 | } |
| 536 | return buffer.append((const UChar *)mapping, length, leadCC, trailCC, errorCode); |
| 537 | } |
| 538 | } |
| 539 | } |
| 540 | |
| 541 | const UChar * |
| 542 | Normalizer2Impl::getDecomposition(UChar32 c, UChar buffer[4], int32_t &length) const { |
| 543 | const UChar *decomp=NULL; |
| 544 | uint16_t norm16; |
| 545 | for(;;) { |
| 546 | if(c<minDecompNoCP || isDecompYes(norm16=getNorm16(c))) { |
| 547 | // c does not decompose |
| 548 | return decomp; |
| 549 | } else if(isHangul(norm16)) { |
| 550 | // Hangul syllable: decompose algorithmically |
| 551 | length=Hangul::decompose(c, buffer); |
| 552 | return buffer; |
| 553 | } else if(isDecompNoAlgorithmic(norm16)) { |
| 554 | c=mapAlgorithmic(c, norm16); |
| 555 | decomp=buffer; |
| 556 | length=0; |
| 557 | U16_APPEND_UNSAFE(buffer, length, c); |
| 558 | } else { |
| 559 | // c decomposes, get everything from the variable-length extra data |
| 560 | const uint16_t *mapping=getMapping(norm16); |
| 561 | uint16_t firstUnit=*mapping++; |
| 562 | length=firstUnit&MAPPING_LENGTH_MASK; |
| 563 | if(firstUnit&MAPPING_HAS_CCC_LCCC_WORD) { |
| 564 | ++mapping; |
| 565 | } |
| 566 | return (const UChar *)mapping; |
| 567 | } |
| 568 | } |
| 569 | } |
| 570 | |
| 571 | void Normalizer2Impl::decomposeAndAppend(const UChar *src, const UChar *limit, |
| 572 | UBool doDecompose, |
| 573 | ReorderingBuffer &buffer, |
| 574 | UErrorCode &errorCode) const { |
| 575 | if(doDecompose) { |
| 576 | decompose(src, limit, &buffer, errorCode); |
| 577 | return; |
| 578 | } |
| 579 | // Just merge the strings at the boundary. |
| 580 | ForwardUTrie2StringIterator iter(normTrie, src, limit); |
| 581 | uint8_t firstCC, prevCC, cc; |
| 582 | firstCC=prevCC=cc=getCC(iter.next16()); |
| 583 | while(cc!=0) { |
| 584 | prevCC=cc; |
| 585 | cc=getCC(iter.next16()); |
| 586 | }; |
| 587 | buffer.append(src, (int32_t)(iter.codePointStart-src), firstCC, prevCC, errorCode) && |
| 588 | buffer.appendZeroCC(iter.codePointStart, limit, errorCode); |
| 589 | } |
| 590 | |
| 591 | // Note: hasDecompBoundary() could be implemented as aliases to |
| 592 | // hasFCDBoundaryBefore() and hasFCDBoundaryAfter() |
| 593 | // at the cost of building the FCD trie for a decomposition normalizer. |
| 594 | UBool Normalizer2Impl::hasDecompBoundary(UChar32 c, UBool before) const { |
| 595 | for(;;) { |
| 596 | if(c<minDecompNoCP) { |
| 597 | return TRUE; |
| 598 | } |
| 599 | uint16_t norm16=getNorm16(c); |
| 600 | if(isHangul(norm16) || isDecompYesAndZeroCC(norm16)) { |
| 601 | return TRUE; |
| 602 | } else if(norm16>MIN_NORMAL_MAYBE_YES) { |
| 603 | return FALSE; // ccc!=0 |
| 604 | } else if(isDecompNoAlgorithmic(norm16)) { |
| 605 | c=mapAlgorithmic(c, norm16); |
| 606 | } else { |
| 607 | // c decomposes, get everything from the variable-length extra data |
| 608 | const uint16_t *mapping=getMapping(norm16); |
| 609 | uint16_t firstUnit=*mapping++; |
| 610 | if((firstUnit&MAPPING_LENGTH_MASK)==0) { |
| 611 | return FALSE; |
| 612 | } |
| 613 | if(!before) { |
| 614 | // decomp after-boundary: same as hasFCDBoundaryAfter(), |
| 615 | // fcd16<=1 || trailCC==0 |
| 616 | if(firstUnit>0x1ff) { |
| 617 | return FALSE; // trailCC>1 |
| 618 | } |
| 619 | if(firstUnit<=0xff) { |
| 620 | return TRUE; // trailCC==0 |
| 621 | } |
| 622 | // if(trailCC==1) test leadCC==0, same as checking for before-boundary |
| 623 | } |
| 624 | // TRUE if leadCC==0 (hasFCDBoundaryBefore()) |
| 625 | return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (*mapping&0xff00)==0; |
| 626 | } |
| 627 | } |
| 628 | } |
| 629 | |
| 630 | /* |
| 631 | * Finds the recomposition result for |
| 632 | * a forward-combining "lead" character, |
| 633 | * specified with a pointer to its compositions list, |
| 634 | * and a backward-combining "trail" character. |
| 635 | * |
| 636 | * If the lead and trail characters combine, then this function returns |
| 637 | * the following "compositeAndFwd" value: |
| 638 | * Bits 21..1 composite character |
| 639 | * Bit 0 set if the composite is a forward-combining starter |
| 640 | * otherwise it returns -1. |
| 641 | * |
| 642 | * The compositions list has (trail, compositeAndFwd) pair entries, |
| 643 | * encoded as either pairs or triples of 16-bit units. |
| 644 | * The last entry has the high bit of its first unit set. |
| 645 | * |
| 646 | * The list is sorted by ascending trail characters (there are no duplicates). |
| 647 | * A linear search is used. |
| 648 | * |
| 649 | * See normalizer2impl.h for a more detailed description |
| 650 | * of the compositions list format. |
| 651 | */ |
| 652 | int32_t Normalizer2Impl::combine(const uint16_t *list, UChar32 trail) { |
| 653 | uint16_t key1, firstUnit; |
| 654 | if(trail<COMP_1_TRAIL_LIMIT) { |
| 655 | // trail character is 0..33FF |
| 656 | // result entry may have 2 or 3 units |
| 657 | key1=(uint16_t)(trail<<1); |
| 658 | while(key1>(firstUnit=*list)) { |
| 659 | list+=2+(firstUnit&COMP_1_TRIPLE); |
| 660 | } |
| 661 | if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { |
| 662 | if(firstUnit&COMP_1_TRIPLE) { |
| 663 | return ((int32_t)list[1]<<16)|list[2]; |
| 664 | } else { |
| 665 | return list[1]; |
| 666 | } |
| 667 | } |
| 668 | } else { |
| 669 | // trail character is 3400..10FFFF |
| 670 | // result entry has 3 units |
| 671 | key1=(uint16_t)(COMP_1_TRAIL_LIMIT+ |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 672 | (((trail>>COMP_1_TRAIL_SHIFT))& |
| 673 | ~COMP_1_TRIPLE)); |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 674 | uint16_t key2=(uint16_t)(trail<<COMP_2_TRAIL_SHIFT); |
| 675 | uint16_t secondUnit; |
| 676 | for(;;) { |
| 677 | if(key1>(firstUnit=*list)) { |
| 678 | list+=2+(firstUnit&COMP_1_TRIPLE); |
| 679 | } else if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { |
| 680 | if(key2>(secondUnit=list[1])) { |
| 681 | if(firstUnit&COMP_1_LAST_TUPLE) { |
| 682 | break; |
| 683 | } else { |
| 684 | list+=3; |
| 685 | } |
| 686 | } else if(key2==(secondUnit&COMP_2_TRAIL_MASK)) { |
| 687 | return ((int32_t)(secondUnit&~COMP_2_TRAIL_MASK)<<16)|list[2]; |
| 688 | } else { |
| 689 | break; |
| 690 | } |
| 691 | } else { |
| 692 | break; |
| 693 | } |
| 694 | } |
| 695 | } |
| 696 | return -1; |
| 697 | } |
| 698 | |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 699 | /** |
| 700 | * @param list some character's compositions list |
| 701 | * @param set recursively receives the composites from these compositions |
| 702 | */ |
| 703 | void Normalizer2Impl::addComposites(const uint16_t *list, UnicodeSet &set) const { |
| 704 | uint16_t firstUnit; |
| 705 | int32_t compositeAndFwd; |
| 706 | do { |
| 707 | firstUnit=*list; |
| 708 | if((firstUnit&COMP_1_TRIPLE)==0) { |
| 709 | compositeAndFwd=list[1]; |
| 710 | list+=2; |
| 711 | } else { |
| 712 | compositeAndFwd=(((int32_t)list[1]&~COMP_2_TRAIL_MASK)<<16)|list[2]; |
| 713 | list+=3; |
| 714 | } |
| 715 | UChar32 composite=compositeAndFwd>>1; |
| 716 | if((compositeAndFwd&1)!=0) { |
| 717 | addComposites(getCompositionsListForComposite(getNorm16(composite)), set); |
| 718 | } |
| 719 | set.add(composite); |
| 720 | } while((firstUnit&COMP_1_LAST_TUPLE)==0); |
| 721 | } |
| 722 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 723 | /* |
| 724 | * Recomposes the buffer text starting at recomposeStartIndex |
| 725 | * (which is in NFD - decomposed and canonically ordered), |
| 726 | * and truncates the buffer contents. |
| 727 | * |
| 728 | * Note that recomposition never lengthens the text: |
| 729 | * Any character consists of either one or two code units; |
| 730 | * a composition may contain at most one more code unit than the original starter, |
| 731 | * while the combining mark that is removed has at least one code unit. |
| 732 | */ |
| 733 | void Normalizer2Impl::recompose(ReorderingBuffer &buffer, int32_t recomposeStartIndex, |
| 734 | UBool onlyContiguous) const { |
| 735 | UChar *p=buffer.getStart()+recomposeStartIndex; |
| 736 | UChar *limit=buffer.getLimit(); |
| 737 | if(p==limit) { |
| 738 | return; |
| 739 | } |
| 740 | |
| 741 | UChar *starter, *pRemove, *q, *r; |
| 742 | const uint16_t *compositionsList; |
| 743 | UChar32 c, compositeAndFwd; |
| 744 | uint16_t norm16; |
| 745 | uint8_t cc, prevCC; |
| 746 | UBool starterIsSupplementary; |
| 747 | |
| 748 | // Some of the following variables are not used until we have a forward-combining starter |
| 749 | // and are only initialized now to avoid compiler warnings. |
| 750 | compositionsList=NULL; // used as indicator for whether we have a forward-combining starter |
| 751 | starter=NULL; |
| 752 | starterIsSupplementary=FALSE; |
| 753 | prevCC=0; |
| 754 | |
| 755 | for(;;) { |
| 756 | UTRIE2_U16_NEXT16(normTrie, p, limit, c, norm16); |
| 757 | cc=getCCFromYesOrMaybe(norm16); |
| 758 | if( // this character combines backward and |
| 759 | isMaybe(norm16) && |
| 760 | // we have seen a starter that combines forward and |
| 761 | compositionsList!=NULL && |
| 762 | // the backward-combining character is not blocked |
| 763 | (prevCC<cc || prevCC==0) |
| 764 | ) { |
| 765 | if(isJamoVT(norm16)) { |
| 766 | // c is a Jamo V/T, see if we can compose it with the previous character. |
| 767 | if(c<Hangul::JAMO_T_BASE) { |
| 768 | // c is a Jamo Vowel, compose with previous Jamo L and following Jamo T. |
| 769 | UChar prev=(UChar)(*starter-Hangul::JAMO_L_BASE); |
| 770 | if(prev<Hangul::JAMO_L_COUNT) { |
| 771 | pRemove=p-1; |
| 772 | UChar syllable=(UChar) |
| 773 | (Hangul::HANGUL_BASE+ |
| 774 | (prev*Hangul::JAMO_V_COUNT+(c-Hangul::JAMO_V_BASE))* |
| 775 | Hangul::JAMO_T_COUNT); |
| 776 | UChar t; |
| 777 | if(p!=limit && (t=(UChar)(*p-Hangul::JAMO_T_BASE))<Hangul::JAMO_T_COUNT) { |
| 778 | ++p; |
| 779 | syllable+=t; // The next character was a Jamo T. |
| 780 | } |
| 781 | *starter=syllable; |
| 782 | // remove the Jamo V/T |
| 783 | q=pRemove; |
| 784 | r=p; |
| 785 | while(r<limit) { |
| 786 | *q++=*r++; |
| 787 | } |
| 788 | limit=q; |
| 789 | p=pRemove; |
| 790 | } |
| 791 | } |
| 792 | /* |
| 793 | * No "else" for Jamo T: |
| 794 | * Since the input is in NFD, there are no Hangul LV syllables that |
| 795 | * a Jamo T could combine with. |
| 796 | * All Jamo Ts are combined above when handling Jamo Vs. |
| 797 | */ |
| 798 | if(p==limit) { |
| 799 | break; |
| 800 | } |
| 801 | compositionsList=NULL; |
| 802 | continue; |
| 803 | } else if((compositeAndFwd=combine(compositionsList, c))>=0) { |
| 804 | // The starter and the combining mark (c) do combine. |
| 805 | UChar32 composite=compositeAndFwd>>1; |
| 806 | |
| 807 | // Replace the starter with the composite, remove the combining mark. |
| 808 | pRemove=p-U16_LENGTH(c); // pRemove & p: start & limit of the combining mark |
| 809 | if(starterIsSupplementary) { |
| 810 | if(U_IS_SUPPLEMENTARY(composite)) { |
| 811 | // both are supplementary |
| 812 | starter[0]=U16_LEAD(composite); |
| 813 | starter[1]=U16_TRAIL(composite); |
| 814 | } else { |
| 815 | *starter=(UChar)composite; |
| 816 | // The composite is shorter than the starter, |
| 817 | // move the intermediate characters forward one. |
| 818 | starterIsSupplementary=FALSE; |
| 819 | q=starter+1; |
| 820 | r=q+1; |
| 821 | while(r<pRemove) { |
| 822 | *q++=*r++; |
| 823 | } |
| 824 | --pRemove; |
| 825 | } |
| 826 | } else if(U_IS_SUPPLEMENTARY(composite)) { |
| 827 | // The composite is longer than the starter, |
| 828 | // move the intermediate characters back one. |
| 829 | starterIsSupplementary=TRUE; |
| 830 | ++starter; // temporarily increment for the loop boundary |
| 831 | q=pRemove; |
| 832 | r=++pRemove; |
| 833 | while(starter<q) { |
| 834 | *--r=*--q; |
| 835 | } |
| 836 | *starter=U16_TRAIL(composite); |
| 837 | *--starter=U16_LEAD(composite); // undo the temporary increment |
| 838 | } else { |
| 839 | // both are on the BMP |
| 840 | *starter=(UChar)composite; |
| 841 | } |
| 842 | |
| 843 | /* remove the combining mark by moving the following text over it */ |
| 844 | if(pRemove<p) { |
| 845 | q=pRemove; |
| 846 | r=p; |
| 847 | while(r<limit) { |
| 848 | *q++=*r++; |
| 849 | } |
| 850 | limit=q; |
| 851 | p=pRemove; |
| 852 | } |
| 853 | // Keep prevCC because we removed the combining mark. |
| 854 | |
| 855 | if(p==limit) { |
| 856 | break; |
| 857 | } |
| 858 | // Is the composite a starter that combines forward? |
| 859 | if(compositeAndFwd&1) { |
| 860 | compositionsList= |
| 861 | getCompositionsListForComposite(getNorm16(composite)); |
| 862 | } else { |
| 863 | compositionsList=NULL; |
| 864 | } |
| 865 | |
| 866 | // We combined; continue with looking for compositions. |
| 867 | continue; |
| 868 | } |
| 869 | } |
| 870 | |
| 871 | // no combination this time |
| 872 | prevCC=cc; |
| 873 | if(p==limit) { |
| 874 | break; |
| 875 | } |
| 876 | |
| 877 | // If c did not combine, then check if it is a starter. |
| 878 | if(cc==0) { |
| 879 | // Found a new starter. |
| 880 | if((compositionsList=getCompositionsListForDecompYes(norm16))!=NULL) { |
| 881 | // It may combine with something, prepare for it. |
| 882 | if(U_IS_BMP(c)) { |
| 883 | starterIsSupplementary=FALSE; |
| 884 | starter=p-1; |
| 885 | } else { |
| 886 | starterIsSupplementary=TRUE; |
| 887 | starter=p-2; |
| 888 | } |
| 889 | } |
| 890 | } else if(onlyContiguous) { |
| 891 | // FCC: no discontiguous compositions; any intervening character blocks. |
| 892 | compositionsList=NULL; |
| 893 | } |
| 894 | } |
| 895 | buffer.setReorderingLimit(limit); |
| 896 | } |
| 897 | |
| 898 | // Very similar to composeQuickCheck(): Make the same changes in both places if relevant. |
| 899 | // doCompose: normalize |
| 900 | // !doCompose: isNormalized (buffer must be empty and initialized) |
| 901 | UBool |
| 902 | Normalizer2Impl::compose(const UChar *src, const UChar *limit, |
| 903 | UBool onlyContiguous, |
| 904 | UBool doCompose, |
| 905 | ReorderingBuffer &buffer, |
| 906 | UErrorCode &errorCode) const { |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 907 | /* |
| 908 | * prevBoundary points to the last character before the current one |
| 909 | * that has a composition boundary before it with ccc==0 and quick check "yes". |
| 910 | * Keeping track of prevBoundary saves us looking for a composition boundary |
| 911 | * when we find a "no" or "maybe". |
| 912 | * |
| 913 | * When we back out from prevSrc back to prevBoundary, |
| 914 | * then we also remove those same characters (which had been simply copied |
| 915 | * or canonically-order-inserted) from the ReorderingBuffer. |
| 916 | * Therefore, at all times, the [prevBoundary..prevSrc[ source units |
| 917 | * must correspond 1:1 to destination units at the end of the destination buffer. |
| 918 | */ |
| 919 | const UChar *prevBoundary=src; |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 920 | UChar32 minNoMaybeCP=minCompNoMaybeCP; |
| 921 | if(limit==NULL) { |
| 922 | src=copyLowPrefixFromNulTerminated(src, minNoMaybeCP, |
| 923 | doCompose ? &buffer : NULL, |
| 924 | errorCode); |
| 925 | if(U_FAILURE(errorCode)) { |
| 926 | return FALSE; |
| 927 | } |
| 928 | if(prevBoundary<src) { |
| 929 | // Set prevBoundary to the last character in the prefix. |
| 930 | prevBoundary=src-1; |
| 931 | } |
| 932 | limit=u_strchr(src, 0); |
| 933 | } |
| 934 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 935 | const UChar *prevSrc; |
| 936 | UChar32 c=0; |
| 937 | uint16_t norm16=0; |
| 938 | |
| 939 | // only for isNormalized |
| 940 | uint8_t prevCC=0; |
| 941 | |
| 942 | for(;;) { |
| 943 | // count code units below the minimum or with irrelevant data for the quick check |
| 944 | for(prevSrc=src; src!=limit;) { |
| 945 | if( (c=*src)<minNoMaybeCP || |
| 946 | isCompYesAndZeroCC(norm16=UTRIE2_GET16_FROM_U16_SINGLE_LEAD(normTrie, c)) |
| 947 | ) { |
| 948 | ++src; |
| 949 | } else if(!U16_IS_SURROGATE(c)) { |
| 950 | break; |
| 951 | } else { |
| 952 | UChar c2; |
| 953 | if(U16_IS_SURROGATE_LEAD(c)) { |
| 954 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])) { |
| 955 | c=U16_GET_SUPPLEMENTARY(c, c2); |
| 956 | } |
| 957 | } else /* trail surrogate */ { |
| 958 | if(prevSrc<src && U16_IS_LEAD(c2=*(src-1))) { |
| 959 | --src; |
| 960 | c=U16_GET_SUPPLEMENTARY(c2, c); |
| 961 | } |
| 962 | } |
| 963 | if(isCompYesAndZeroCC(norm16=getNorm16(c))) { |
| 964 | src+=U16_LENGTH(c); |
| 965 | } else { |
| 966 | break; |
| 967 | } |
| 968 | } |
| 969 | } |
| 970 | // copy these code units all at once |
| 971 | if(src!=prevSrc) { |
| 972 | if(doCompose) { |
| 973 | if(!buffer.appendZeroCC(prevSrc, src, errorCode)) { |
| 974 | break; |
| 975 | } |
| 976 | } else { |
| 977 | prevCC=0; |
| 978 | } |
| 979 | if(src==limit) { |
| 980 | break; |
| 981 | } |
| 982 | // Set prevBoundary to the last character in the quick check loop. |
| 983 | prevBoundary=src-1; |
| 984 | if( U16_IS_TRAIL(*prevBoundary) && prevSrc<prevBoundary && |
| 985 | U16_IS_LEAD(*(prevBoundary-1)) |
| 986 | ) { |
| 987 | --prevBoundary; |
| 988 | } |
| 989 | // The start of the current character (c). |
| 990 | prevSrc=src; |
| 991 | } else if(src==limit) { |
| 992 | break; |
| 993 | } |
| 994 | |
| 995 | src+=U16_LENGTH(c); |
| 996 | /* |
| 997 | * isCompYesAndZeroCC(norm16) is false, that is, norm16>=minNoNo. |
| 998 | * c is either a "noNo" (has a mapping) or a "maybeYes" (combines backward) |
| 999 | * or has ccc!=0. |
| 1000 | * Check for Jamo V/T, then for regular characters. |
| 1001 | * c is not a Hangul syllable or Jamo L because those have "yes" properties. |
| 1002 | */ |
| 1003 | if(isJamoVT(norm16) && prevBoundary!=prevSrc) { |
| 1004 | UChar prev=*(prevSrc-1); |
| 1005 | UBool needToDecompose=FALSE; |
| 1006 | if(c<Hangul::JAMO_T_BASE) { |
| 1007 | // c is a Jamo Vowel, compose with previous Jamo L and following Jamo T. |
| 1008 | prev=(UChar)(prev-Hangul::JAMO_L_BASE); |
| 1009 | if(prev<Hangul::JAMO_L_COUNT) { |
| 1010 | if(!doCompose) { |
| 1011 | return FALSE; |
| 1012 | } |
| 1013 | UChar syllable=(UChar) |
| 1014 | (Hangul::HANGUL_BASE+ |
| 1015 | (prev*Hangul::JAMO_V_COUNT+(c-Hangul::JAMO_V_BASE))* |
| 1016 | Hangul::JAMO_T_COUNT); |
| 1017 | UChar t; |
| 1018 | if(src!=limit && (t=(UChar)(*src-Hangul::JAMO_T_BASE))<Hangul::JAMO_T_COUNT) { |
| 1019 | ++src; |
| 1020 | syllable+=t; // The next character was a Jamo T. |
| 1021 | prevBoundary=src; |
| 1022 | buffer.setLastChar(syllable); |
| 1023 | continue; |
| 1024 | } |
| 1025 | // If we see L+V+x where x!=T then we drop to the slow path, |
| 1026 | // decompose and recompose. |
| 1027 | // This is to deal with NFKC finding normal L and V but a |
| 1028 | // compatibility variant of a T. We need to either fully compose that |
| 1029 | // combination here (which would complicate the code and may not work |
| 1030 | // with strange custom data) or use the slow path -- or else our replacing |
| 1031 | // two input characters (L+V) with one output character (LV syllable) |
| 1032 | // would violate the invariant that [prevBoundary..prevSrc[ has the same |
| 1033 | // length as what we appended to the buffer since prevBoundary. |
| 1034 | needToDecompose=TRUE; |
| 1035 | } |
| 1036 | } else if(Hangul::isHangulWithoutJamoT(prev)) { |
| 1037 | // c is a Jamo Trailing consonant, |
| 1038 | // compose with previous Hangul LV that does not contain a Jamo T. |
| 1039 | if(!doCompose) { |
| 1040 | return FALSE; |
| 1041 | } |
| 1042 | buffer.setLastChar((UChar)(prev+c-Hangul::JAMO_T_BASE)); |
| 1043 | prevBoundary=src; |
| 1044 | continue; |
| 1045 | } |
| 1046 | if(!needToDecompose) { |
| 1047 | // The Jamo V/T did not compose into a Hangul syllable. |
| 1048 | if(doCompose) { |
| 1049 | if(!buffer.appendBMP((UChar)c, 0, errorCode)) { |
| 1050 | break; |
| 1051 | } |
| 1052 | } else { |
| 1053 | prevCC=0; |
| 1054 | } |
| 1055 | continue; |
| 1056 | } |
| 1057 | } |
| 1058 | /* |
| 1059 | * Source buffer pointers: |
| 1060 | * |
| 1061 | * all done quick check current char not yet |
| 1062 | * "yes" but (c) processed |
| 1063 | * may combine |
| 1064 | * forward |
| 1065 | * [-------------[-------------[-------------[-------------[ |
| 1066 | * | | | | | |
| 1067 | * orig. src prevBoundary prevSrc src limit |
| 1068 | * |
| 1069 | * |
| 1070 | * Destination buffer pointers inside the ReorderingBuffer: |
| 1071 | * |
| 1072 | * all done might take not filled yet |
| 1073 | * characters for |
| 1074 | * reordering |
| 1075 | * [-------------[-------------[-------------[ |
| 1076 | * | | | | |
| 1077 | * start reorderStart limit | |
| 1078 | * +remainingCap.+ |
| 1079 | */ |
| 1080 | if(norm16>=MIN_YES_YES_WITH_CC) { |
| 1081 | uint8_t cc=(uint8_t)norm16; // cc!=0 |
| 1082 | if( onlyContiguous && // FCC |
| 1083 | (doCompose ? buffer.getLastCC() : prevCC)==0 && |
| 1084 | prevBoundary<prevSrc && |
| 1085 | // buffer.getLastCC()==0 && prevBoundary<prevSrc tell us that |
| 1086 | // [prevBoundary..prevSrc[ (which is exactly one character under these conditions) |
| 1087 | // passed the quick check "yes && ccc==0" test. |
| 1088 | // Check whether the last character was a "yesYes" or a "yesNo". |
| 1089 | // If a "yesNo", then we get its trailing ccc from its |
| 1090 | // mapping and check for canonical order. |
| 1091 | // All other cases are ok. |
| 1092 | getTrailCCFromCompYesAndZeroCC(prevBoundary, prevSrc)>cc |
| 1093 | ) { |
| 1094 | // Fails FCD test, need to decompose and contiguously recompose. |
| 1095 | if(!doCompose) { |
| 1096 | return FALSE; |
| 1097 | } |
| 1098 | } else if(doCompose) { |
| 1099 | if(!buffer.append(c, cc, errorCode)) { |
| 1100 | break; |
| 1101 | } |
| 1102 | continue; |
| 1103 | } else if(prevCC<=cc) { |
| 1104 | prevCC=cc; |
| 1105 | continue; |
| 1106 | } else { |
| 1107 | return FALSE; |
| 1108 | } |
| 1109 | } else if(!doCompose && !isMaybeOrNonZeroCC(norm16)) { |
| 1110 | return FALSE; |
| 1111 | } |
| 1112 | |
| 1113 | /* |
| 1114 | * Find appropriate boundaries around this character, |
| 1115 | * decompose the source text from between the boundaries, |
| 1116 | * and recompose it. |
| 1117 | * |
| 1118 | * We may need to remove the last few characters from the ReorderingBuffer |
| 1119 | * to account for source text that was copied or appended |
| 1120 | * but needs to take part in the recomposition. |
| 1121 | */ |
| 1122 | |
| 1123 | /* |
| 1124 | * Find the last composition boundary in [prevBoundary..src[. |
| 1125 | * It is either the decomposition of the current character (at prevSrc), |
| 1126 | * or prevBoundary. |
| 1127 | */ |
| 1128 | if(hasCompBoundaryBefore(c, norm16)) { |
| 1129 | prevBoundary=prevSrc; |
| 1130 | } else if(doCompose) { |
| 1131 | buffer.removeSuffix((int32_t)(prevSrc-prevBoundary)); |
| 1132 | } |
| 1133 | |
| 1134 | // Find the next composition boundary in [src..limit[ - |
| 1135 | // modifies src to point to the next starter. |
| 1136 | src=(UChar *)findNextCompBoundary(src, limit); |
| 1137 | |
| 1138 | // Decompose [prevBoundary..src[ into the buffer and then recompose that part of it. |
| 1139 | int32_t recomposeStartIndex=buffer.length(); |
| 1140 | if(!decomposeShort(prevBoundary, src, buffer, errorCode)) { |
| 1141 | break; |
| 1142 | } |
| 1143 | recompose(buffer, recomposeStartIndex, onlyContiguous); |
| 1144 | if(!doCompose) { |
| 1145 | if(!buffer.equals(prevBoundary, src)) { |
| 1146 | return FALSE; |
| 1147 | } |
| 1148 | buffer.remove(); |
| 1149 | prevCC=0; |
| 1150 | } |
| 1151 | |
| 1152 | // Move to the next starter. We never need to look back before this point again. |
| 1153 | prevBoundary=src; |
| 1154 | } |
| 1155 | return TRUE; |
| 1156 | } |
| 1157 | |
| 1158 | // Very similar to compose(): Make the same changes in both places if relevant. |
| 1159 | // pQCResult==NULL: spanQuickCheckYes |
| 1160 | // pQCResult!=NULL: quickCheck (*pQCResult must be UNORM_YES) |
| 1161 | const UChar * |
| 1162 | Normalizer2Impl::composeQuickCheck(const UChar *src, const UChar *limit, |
| 1163 | UBool onlyContiguous, |
| 1164 | UNormalizationCheckResult *pQCResult) const { |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1165 | /* |
| 1166 | * prevBoundary points to the last character before the current one |
| 1167 | * that has a composition boundary before it with ccc==0 and quick check "yes". |
| 1168 | */ |
| 1169 | const UChar *prevBoundary=src; |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1170 | UChar32 minNoMaybeCP=minCompNoMaybeCP; |
| 1171 | if(limit==NULL) { |
| 1172 | UErrorCode errorCode=U_ZERO_ERROR; |
| 1173 | src=copyLowPrefixFromNulTerminated(src, minNoMaybeCP, NULL, errorCode); |
| 1174 | if(prevBoundary<src) { |
| 1175 | // Set prevBoundary to the last character in the prefix. |
| 1176 | prevBoundary=src-1; |
| 1177 | } |
| 1178 | limit=u_strchr(src, 0); |
| 1179 | } |
| 1180 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1181 | const UChar *prevSrc; |
| 1182 | UChar32 c=0; |
| 1183 | uint16_t norm16=0; |
| 1184 | uint8_t prevCC=0; |
| 1185 | |
| 1186 | for(;;) { |
| 1187 | // count code units below the minimum or with irrelevant data for the quick check |
| 1188 | for(prevSrc=src;;) { |
| 1189 | if(src==limit) { |
| 1190 | return src; |
| 1191 | } |
| 1192 | if( (c=*src)<minNoMaybeCP || |
| 1193 | isCompYesAndZeroCC(norm16=UTRIE2_GET16_FROM_U16_SINGLE_LEAD(normTrie, c)) |
| 1194 | ) { |
| 1195 | ++src; |
| 1196 | } else if(!U16_IS_SURROGATE(c)) { |
| 1197 | break; |
| 1198 | } else { |
| 1199 | UChar c2; |
| 1200 | if(U16_IS_SURROGATE_LEAD(c)) { |
| 1201 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])) { |
| 1202 | c=U16_GET_SUPPLEMENTARY(c, c2); |
| 1203 | } |
| 1204 | } else /* trail surrogate */ { |
| 1205 | if(prevSrc<src && U16_IS_LEAD(c2=*(src-1))) { |
| 1206 | --src; |
| 1207 | c=U16_GET_SUPPLEMENTARY(c2, c); |
| 1208 | } |
| 1209 | } |
| 1210 | if(isCompYesAndZeroCC(norm16=getNorm16(c))) { |
| 1211 | src+=U16_LENGTH(c); |
| 1212 | } else { |
| 1213 | break; |
| 1214 | } |
| 1215 | } |
| 1216 | } |
| 1217 | if(src!=prevSrc) { |
| 1218 | // Set prevBoundary to the last character in the quick check loop. |
| 1219 | prevBoundary=src-1; |
| 1220 | if( U16_IS_TRAIL(*prevBoundary) && prevSrc<prevBoundary && |
| 1221 | U16_IS_LEAD(*(prevBoundary-1)) |
| 1222 | ) { |
| 1223 | --prevBoundary; |
| 1224 | } |
| 1225 | prevCC=0; |
| 1226 | // The start of the current character (c). |
| 1227 | prevSrc=src; |
| 1228 | } |
| 1229 | |
| 1230 | src+=U16_LENGTH(c); |
| 1231 | /* |
| 1232 | * isCompYesAndZeroCC(norm16) is false, that is, norm16>=minNoNo. |
| 1233 | * c is either a "noNo" (has a mapping) or a "maybeYes" (combines backward) |
| 1234 | * or has ccc!=0. |
| 1235 | */ |
| 1236 | if(isMaybeOrNonZeroCC(norm16)) { |
| 1237 | uint8_t cc=getCCFromYesOrMaybe(norm16); |
| 1238 | if( onlyContiguous && // FCC |
| 1239 | cc!=0 && |
| 1240 | prevCC==0 && |
| 1241 | prevBoundary<prevSrc && |
| 1242 | // prevCC==0 && prevBoundary<prevSrc tell us that |
| 1243 | // [prevBoundary..prevSrc[ (which is exactly one character under these conditions) |
| 1244 | // passed the quick check "yes && ccc==0" test. |
| 1245 | // Check whether the last character was a "yesYes" or a "yesNo". |
| 1246 | // If a "yesNo", then we get its trailing ccc from its |
| 1247 | // mapping and check for canonical order. |
| 1248 | // All other cases are ok. |
| 1249 | getTrailCCFromCompYesAndZeroCC(prevBoundary, prevSrc)>cc |
| 1250 | ) { |
| 1251 | // Fails FCD test. |
| 1252 | } else if(prevCC<=cc || cc==0) { |
| 1253 | prevCC=cc; |
| 1254 | if(norm16<MIN_YES_YES_WITH_CC) { |
| 1255 | if(pQCResult!=NULL) { |
| 1256 | *pQCResult=UNORM_MAYBE; |
| 1257 | } else { |
| 1258 | return prevBoundary; |
| 1259 | } |
| 1260 | } |
| 1261 | continue; |
| 1262 | } |
| 1263 | } |
| 1264 | if(pQCResult!=NULL) { |
| 1265 | *pQCResult=UNORM_NO; |
| 1266 | } |
| 1267 | return prevBoundary; |
| 1268 | } |
| 1269 | } |
| 1270 | |
| 1271 | void Normalizer2Impl::composeAndAppend(const UChar *src, const UChar *limit, |
| 1272 | UBool doCompose, |
| 1273 | UBool onlyContiguous, |
| 1274 | ReorderingBuffer &buffer, |
| 1275 | UErrorCode &errorCode) const { |
| 1276 | if(!buffer.isEmpty()) { |
| 1277 | const UChar *firstStarterInSrc=findNextCompBoundary(src, limit); |
| 1278 | if(src!=firstStarterInSrc) { |
| 1279 | const UChar *lastStarterInDest=findPreviousCompBoundary(buffer.getStart(), |
| 1280 | buffer.getLimit()); |
| 1281 | UnicodeString middle(lastStarterInDest, |
| 1282 | (int32_t)(buffer.getLimit()-lastStarterInDest)); |
| 1283 | buffer.removeSuffix((int32_t)(buffer.getLimit()-lastStarterInDest)); |
| 1284 | middle.append(src, (int32_t)(firstStarterInSrc-src)); |
| 1285 | const UChar *middleStart=middle.getBuffer(); |
| 1286 | compose(middleStart, middleStart+middle.length(), onlyContiguous, |
| 1287 | TRUE, buffer, errorCode); |
| 1288 | if(U_FAILURE(errorCode)) { |
| 1289 | return; |
| 1290 | } |
| 1291 | src=firstStarterInSrc; |
| 1292 | } |
| 1293 | } |
| 1294 | if(doCompose) { |
| 1295 | compose(src, limit, onlyContiguous, TRUE, buffer, errorCode); |
| 1296 | } else { |
| 1297 | buffer.appendZeroCC(src, limit, errorCode); |
| 1298 | } |
| 1299 | } |
| 1300 | |
| 1301 | /** |
| 1302 | * Does c have a composition boundary before it? |
| 1303 | * True if its decomposition begins with a character that has |
| 1304 | * ccc=0 && NFC_QC=Yes (isCompYesAndZeroCC()). |
| 1305 | * As a shortcut, this is true if c itself has ccc=0 && NFC_QC=Yes |
| 1306 | * (isCompYesAndZeroCC()) so we need not decompose. |
| 1307 | */ |
| 1308 | UBool Normalizer2Impl::hasCompBoundaryBefore(UChar32 c, uint16_t norm16) const { |
| 1309 | for(;;) { |
| 1310 | if(isCompYesAndZeroCC(norm16)) { |
| 1311 | return TRUE; |
| 1312 | } else if(isMaybeOrNonZeroCC(norm16)) { |
| 1313 | return FALSE; |
| 1314 | } else if(isDecompNoAlgorithmic(norm16)) { |
| 1315 | c=mapAlgorithmic(c, norm16); |
| 1316 | norm16=getNorm16(c); |
| 1317 | } else { |
| 1318 | // c decomposes, get everything from the variable-length extra data |
| 1319 | const uint16_t *mapping=getMapping(norm16); |
| 1320 | uint16_t firstUnit=*mapping++; |
| 1321 | if((firstUnit&MAPPING_LENGTH_MASK)==0) { |
| 1322 | return FALSE; |
| 1323 | } |
| 1324 | if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD) && (*mapping++&0xff00)) { |
| 1325 | return FALSE; // non-zero leadCC |
| 1326 | } |
| 1327 | int32_t i=0; |
| 1328 | UChar32 c; |
| 1329 | U16_NEXT_UNSAFE(mapping, i, c); |
| 1330 | return isCompYesAndZeroCC(getNorm16(c)); |
| 1331 | } |
| 1332 | } |
| 1333 | } |
| 1334 | |
| 1335 | UBool Normalizer2Impl::hasCompBoundaryAfter(UChar32 c, UBool onlyContiguous, UBool testInert) const { |
| 1336 | for(;;) { |
| 1337 | uint16_t norm16=getNorm16(c); |
| 1338 | if(isInert(norm16)) { |
| 1339 | return TRUE; |
| 1340 | } else if(norm16<=minYesNo) { |
| 1341 | // Hangul LVT (==minYesNo) has a boundary after it. |
| 1342 | // Hangul LV and non-inert yesYes characters combine forward. |
| 1343 | return isHangul(norm16) && !Hangul::isHangulWithoutJamoT((UChar)c); |
| 1344 | } else if(norm16>= (testInert ? minNoNo : minMaybeYes)) { |
| 1345 | return FALSE; |
| 1346 | } else if(isDecompNoAlgorithmic(norm16)) { |
| 1347 | c=mapAlgorithmic(c, norm16); |
| 1348 | } else { |
| 1349 | // c decomposes, get everything from the variable-length extra data. |
| 1350 | // If testInert, then c must be a yesNo character which has lccc=0, |
| 1351 | // otherwise it could be a noNo. |
| 1352 | const uint16_t *mapping=getMapping(norm16); |
| 1353 | uint16_t firstUnit=*mapping; |
| 1354 | // TRUE if |
| 1355 | // c is not deleted, and |
| 1356 | // it and its decomposition do not combine forward, and it has a starter, and |
| 1357 | // if FCC then trailCC<=1 |
| 1358 | return |
| 1359 | (firstUnit&MAPPING_LENGTH_MASK)!=0 && |
| 1360 | (firstUnit&(MAPPING_PLUS_COMPOSITION_LIST|MAPPING_NO_COMP_BOUNDARY_AFTER))==0 && |
| 1361 | (!onlyContiguous || firstUnit<=0x1ff); |
| 1362 | } |
| 1363 | } |
| 1364 | } |
| 1365 | |
| 1366 | const UChar *Normalizer2Impl::findPreviousCompBoundary(const UChar *start, const UChar *p) const { |
| 1367 | BackwardUTrie2StringIterator iter(normTrie, start, p); |
| 1368 | uint16_t norm16; |
| 1369 | do { |
| 1370 | norm16=iter.previous16(); |
| 1371 | } while(!hasCompBoundaryBefore(iter.codePoint, norm16)); |
| 1372 | // We could also test hasCompBoundaryAfter() and return iter.codePointLimit, |
| 1373 | // but that's probably not worth the extra cost. |
| 1374 | return iter.codePointStart; |
| 1375 | } |
| 1376 | |
| 1377 | const UChar *Normalizer2Impl::findNextCompBoundary(const UChar *p, const UChar *limit) const { |
| 1378 | ForwardUTrie2StringIterator iter(normTrie, p, limit); |
| 1379 | uint16_t norm16; |
| 1380 | do { |
| 1381 | norm16=iter.next16(); |
| 1382 | } while(!hasCompBoundaryBefore(iter.codePoint, norm16)); |
| 1383 | return iter.codePointStart; |
| 1384 | } |
| 1385 | |
| 1386 | class FCDTrieSingleton : public UTrie2Singleton { |
| 1387 | public: |
| 1388 | FCDTrieSingleton(SimpleSingleton &s, Normalizer2Impl &ni, UErrorCode &ec) : |
| 1389 | UTrie2Singleton(s), impl(ni), errorCode(ec) {} |
| 1390 | UTrie2 *getInstance(UErrorCode &errorCode) { |
| 1391 | return UTrie2Singleton::getInstance(createInstance, this, errorCode); |
| 1392 | } |
| 1393 | static void *createInstance(const void *context, UErrorCode &errorCode); |
| 1394 | UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { |
| 1395 | if(value!=0) { |
| 1396 | impl.setFCD16FromNorm16(start, end, (uint16_t)value, newFCDTrie, errorCode); |
| 1397 | } |
| 1398 | return U_SUCCESS(errorCode); |
| 1399 | } |
| 1400 | |
| 1401 | Normalizer2Impl &impl; |
| 1402 | UTrie2 *newFCDTrie; |
| 1403 | UErrorCode &errorCode; |
| 1404 | }; |
| 1405 | |
| 1406 | U_CDECL_BEGIN |
| 1407 | |
| 1408 | // Set the FCD value for a range of same-norm16 characters. |
| 1409 | static UBool U_CALLCONV |
| 1410 | enumRangeHandler(const void *context, UChar32 start, UChar32 end, uint32_t value) { |
| 1411 | return ((FCDTrieSingleton *)context)->rangeHandler(start, end, value); |
| 1412 | } |
| 1413 | |
| 1414 | // Collect (OR together) the FCD values for a range of supplementary characters, |
| 1415 | // for their lead surrogate code unit. |
| 1416 | static UBool U_CALLCONV |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1417 | enumRangeOrValue(const void *context, UChar32 /*start*/, UChar32 /*end*/, uint32_t value) { |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1418 | *((uint32_t *)context)|=value; |
| 1419 | return TRUE; |
| 1420 | } |
| 1421 | |
| 1422 | U_CDECL_END |
| 1423 | |
| 1424 | void *FCDTrieSingleton::createInstance(const void *context, UErrorCode &errorCode) { |
| 1425 | FCDTrieSingleton *me=(FCDTrieSingleton *)context; |
| 1426 | me->newFCDTrie=utrie2_open(0, 0, &errorCode); |
| 1427 | if(U_SUCCESS(errorCode)) { |
| 1428 | utrie2_enum(me->impl.getNormTrie(), NULL, enumRangeHandler, me); |
| 1429 | for(UChar lead=0xd800; lead<0xdc00; ++lead) { |
| 1430 | uint32_t oredValue=utrie2_get32(me->newFCDTrie, lead); |
| 1431 | utrie2_enumForLeadSurrogate(me->newFCDTrie, lead, NULL, enumRangeOrValue, &oredValue); |
| 1432 | if(oredValue!=0) { |
| 1433 | // Set a "bad" value for makeFCD() to break the quick check loop |
| 1434 | // and look up the value for the supplementary code point. |
| 1435 | // If there is any lccc, then set the worst-case lccc of 1. |
| 1436 | // The ORed-together value's tccc is already the worst case. |
| 1437 | if(oredValue>0xff) { |
| 1438 | oredValue=0x100|(oredValue&0xff); |
| 1439 | } |
| 1440 | utrie2_set32ForLeadSurrogateCodeUnit(me->newFCDTrie, lead, oredValue, &errorCode); |
| 1441 | } |
| 1442 | } |
| 1443 | utrie2_freeze(me->newFCDTrie, UTRIE2_16_VALUE_BITS, &errorCode); |
| 1444 | if(U_SUCCESS(errorCode)) { |
| 1445 | return me->newFCDTrie; |
| 1446 | } |
| 1447 | } |
| 1448 | utrie2_close(me->newFCDTrie); |
| 1449 | return NULL; |
| 1450 | } |
| 1451 | |
| 1452 | void Normalizer2Impl::setFCD16FromNorm16(UChar32 start, UChar32 end, uint16_t norm16, |
| 1453 | UTrie2 *newFCDTrie, UErrorCode &errorCode) const { |
| 1454 | // Only loops for 1:1 algorithmic mappings. |
| 1455 | for(;;) { |
| 1456 | if(norm16>=MIN_NORMAL_MAYBE_YES) { |
| 1457 | norm16&=0xff; |
| 1458 | norm16|=norm16<<8; |
| 1459 | } else if(norm16<=minYesNo || minMaybeYes<=norm16) { |
| 1460 | // no decomposition or Hangul syllable, all zeros |
| 1461 | break; |
| 1462 | } else if(limitNoNo<=norm16) { |
| 1463 | int32_t delta=norm16-(minMaybeYes-MAX_DELTA-1); |
| 1464 | if(start==end) { |
| 1465 | start+=delta; |
| 1466 | norm16=getNorm16(start); |
| 1467 | } else { |
| 1468 | // the same delta leads from different original characters to different mappings |
| 1469 | do { |
| 1470 | UChar32 c=start+delta; |
| 1471 | setFCD16FromNorm16(c, c, getNorm16(c), newFCDTrie, errorCode); |
| 1472 | } while(++start<=end); |
| 1473 | break; |
| 1474 | } |
| 1475 | } else { |
| 1476 | // c decomposes, get everything from the variable-length extra data |
| 1477 | const uint16_t *mapping=getMapping(norm16); |
| 1478 | uint16_t firstUnit=*mapping; |
| 1479 | if((firstUnit&MAPPING_LENGTH_MASK)==0) { |
| 1480 | // A character that is deleted (maps to an empty string) must |
| 1481 | // get the worst-case lccc and tccc values because arbitrary |
| 1482 | // characters on both sides will become adjacent. |
| 1483 | norm16=0x1ff; |
| 1484 | } else { |
| 1485 | if(firstUnit&MAPPING_HAS_CCC_LCCC_WORD) { |
| 1486 | norm16=mapping[1]&0xff00; // lccc |
| 1487 | } else { |
| 1488 | norm16=0; |
| 1489 | } |
| 1490 | norm16|=firstUnit>>8; // tccc |
| 1491 | } |
| 1492 | } |
| 1493 | utrie2_setRange32(newFCDTrie, start, end, norm16, TRUE, &errorCode); |
| 1494 | break; |
| 1495 | } |
| 1496 | } |
| 1497 | |
| 1498 | const UTrie2 *Normalizer2Impl::getFCDTrie(UErrorCode &errorCode) const { |
| 1499 | // Logically const: Synchronized instantiation. |
| 1500 | Normalizer2Impl *me=const_cast<Normalizer2Impl *>(this); |
| 1501 | return FCDTrieSingleton(me->fcdTrieSingleton, *me, errorCode).getInstance(errorCode); |
| 1502 | } |
| 1503 | |
| 1504 | // Dual functionality: |
| 1505 | // buffer!=NULL: normalize |
| 1506 | // buffer==NULL: isNormalized/quickCheck/spanQuickCheckYes |
| 1507 | const UChar * |
| 1508 | Normalizer2Impl::makeFCD(const UChar *src, const UChar *limit, |
| 1509 | ReorderingBuffer *buffer, |
| 1510 | UErrorCode &errorCode) const { |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1511 | // Tracks the last FCD-safe boundary, before lccc=0 or after properly-ordered tccc<=1. |
| 1512 | // Similar to the prevBoundary in the compose() implementation. |
| 1513 | const UChar *prevBoundary=src; |
| 1514 | int32_t prevFCD16=0; |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1515 | if(limit==NULL) { |
| 1516 | src=copyLowPrefixFromNulTerminated(src, MIN_CCC_LCCC_CP, buffer, errorCode); |
| 1517 | if(U_FAILURE(errorCode)) { |
| 1518 | return src; |
| 1519 | } |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1520 | if(prevBoundary<src) { |
| 1521 | prevBoundary=src; |
| 1522 | // We know that the previous character's lccc==0. |
| 1523 | // Fetching the fcd16 value was deferred for this below-U+0300 code point. |
| 1524 | prevFCD16=getFCD16FromSingleLead(*(src-1)); |
| 1525 | if(prevFCD16>1) { |
| 1526 | --prevBoundary; |
| 1527 | } |
| 1528 | } |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1529 | limit=u_strchr(src, 0); |
| 1530 | } |
| 1531 | |
| 1532 | // Note: In this function we use buffer->appendZeroCC() because we track |
| 1533 | // the lead and trail combining classes here, rather than leaving it to |
| 1534 | // the ReorderingBuffer. |
| 1535 | // The exception is the call to decomposeShort() which uses the buffer |
| 1536 | // in the normal way. |
| 1537 | |
| 1538 | const UTrie2 *trie=fcdTrie(); |
| 1539 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1540 | const UChar *prevSrc; |
| 1541 | UChar32 c=0; |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1542 | uint16_t fcd16=0; |
| 1543 | |
| 1544 | for(;;) { |
| 1545 | // count code units with lccc==0 |
| 1546 | for(prevSrc=src; src!=limit;) { |
| 1547 | if((c=*src)<MIN_CCC_LCCC_CP) { |
| 1548 | prevFCD16=~c; |
| 1549 | ++src; |
| 1550 | } else if((fcd16=UTRIE2_GET16_FROM_U16_SINGLE_LEAD(trie, c))<=0xff) { |
| 1551 | prevFCD16=fcd16; |
| 1552 | ++src; |
| 1553 | } else if(!U16_IS_SURROGATE(c)) { |
| 1554 | break; |
| 1555 | } else { |
| 1556 | UChar c2; |
| 1557 | if(U16_IS_SURROGATE_LEAD(c)) { |
| 1558 | if((src+1)!=limit && U16_IS_TRAIL(c2=src[1])) { |
| 1559 | c=U16_GET_SUPPLEMENTARY(c, c2); |
| 1560 | } |
| 1561 | } else /* trail surrogate */ { |
| 1562 | if(prevSrc<src && U16_IS_LEAD(c2=*(src-1))) { |
| 1563 | --src; |
| 1564 | c=U16_GET_SUPPLEMENTARY(c2, c); |
| 1565 | } |
| 1566 | } |
| 1567 | if((fcd16=getFCD16(c))<=0xff) { |
| 1568 | prevFCD16=fcd16; |
| 1569 | src+=U16_LENGTH(c); |
| 1570 | } else { |
| 1571 | break; |
| 1572 | } |
| 1573 | } |
| 1574 | } |
| 1575 | // copy these code units all at once |
| 1576 | if(src!=prevSrc) { |
| 1577 | if(buffer!=NULL && !buffer->appendZeroCC(prevSrc, src, errorCode)) { |
| 1578 | break; |
| 1579 | } |
| 1580 | if(src==limit) { |
| 1581 | break; |
| 1582 | } |
| 1583 | prevBoundary=src; |
| 1584 | // We know that the previous character's lccc==0. |
| 1585 | if(prevFCD16<0) { |
| 1586 | // Fetching the fcd16 value was deferred for this below-U+0300 code point. |
| 1587 | prevFCD16=getFCD16FromSingleLead((UChar)~prevFCD16); |
| 1588 | if(prevFCD16>1) { |
| 1589 | --prevBoundary; |
| 1590 | } |
| 1591 | } else { |
| 1592 | const UChar *p=src-1; |
| 1593 | if(U16_IS_TRAIL(*p) && prevSrc<p && U16_IS_LEAD(*(p-1))) { |
| 1594 | --p; |
| 1595 | // Need to fetch the previous character's FCD value because |
| 1596 | // prevFCD16 was just for the trail surrogate code point. |
| 1597 | prevFCD16=getFCD16FromSurrogatePair(p[0], p[1]); |
| 1598 | // Still known to have lccc==0 because its lead surrogate unit had lccc==0. |
| 1599 | } |
| 1600 | if(prevFCD16>1) { |
| 1601 | prevBoundary=p; |
| 1602 | } |
| 1603 | } |
| 1604 | // The start of the current character (c). |
| 1605 | prevSrc=src; |
| 1606 | } else if(src==limit) { |
| 1607 | break; |
| 1608 | } |
| 1609 | |
| 1610 | src+=U16_LENGTH(c); |
| 1611 | // The current character (c) at [prevSrc..src[ has a non-zero lead combining class. |
| 1612 | // Check for proper order, and decompose locally if necessary. |
| 1613 | if((prevFCD16&0xff)<=(fcd16>>8)) { |
| 1614 | // proper order: prev tccc <= current lccc |
| 1615 | if((fcd16&0xff)<=1) { |
| 1616 | prevBoundary=src; |
| 1617 | } |
| 1618 | if(buffer!=NULL && !buffer->appendZeroCC(c, errorCode)) { |
| 1619 | break; |
| 1620 | } |
| 1621 | prevFCD16=fcd16; |
| 1622 | continue; |
| 1623 | } else if(buffer==NULL) { |
| 1624 | return prevBoundary; // quick check "no" |
| 1625 | } else { |
| 1626 | /* |
| 1627 | * Back out the part of the source that we copied or appended |
| 1628 | * already but is now going to be decomposed. |
| 1629 | * prevSrc is set to after what was copied/appended. |
| 1630 | */ |
| 1631 | buffer->removeSuffix((int32_t)(prevSrc-prevBoundary)); |
| 1632 | /* |
| 1633 | * Find the part of the source that needs to be decomposed, |
| 1634 | * up to the next safe boundary. |
| 1635 | */ |
| 1636 | src=findNextFCDBoundary(src, limit); |
| 1637 | /* |
| 1638 | * The source text does not fulfill the conditions for FCD. |
| 1639 | * Decompose and reorder a limited piece of the text. |
| 1640 | */ |
| 1641 | if(!decomposeShort(prevBoundary, src, *buffer, errorCode)) { |
| 1642 | break; |
| 1643 | } |
| 1644 | prevBoundary=src; |
| 1645 | prevFCD16=0; |
| 1646 | } |
| 1647 | } |
| 1648 | return src; |
| 1649 | } |
| 1650 | |
| 1651 | void Normalizer2Impl::makeFCDAndAppend(const UChar *src, const UChar *limit, |
| 1652 | UBool doMakeFCD, |
| 1653 | ReorderingBuffer &buffer, |
| 1654 | UErrorCode &errorCode) const { |
| 1655 | if(!buffer.isEmpty()) { |
| 1656 | const UChar *firstBoundaryInSrc=findNextFCDBoundary(src, limit); |
| 1657 | if(src!=firstBoundaryInSrc) { |
| 1658 | const UChar *lastBoundaryInDest=findPreviousFCDBoundary(buffer.getStart(), |
| 1659 | buffer.getLimit()); |
| 1660 | UnicodeString middle(lastBoundaryInDest, |
| 1661 | (int32_t)(buffer.getLimit()-lastBoundaryInDest)); |
| 1662 | buffer.removeSuffix((int32_t)(buffer.getLimit()-lastBoundaryInDest)); |
| 1663 | middle.append(src, (int32_t)(firstBoundaryInSrc-src)); |
| 1664 | const UChar *middleStart=middle.getBuffer(); |
| 1665 | makeFCD(middleStart, middleStart+middle.length(), &buffer, errorCode); |
| 1666 | if(U_FAILURE(errorCode)) { |
| 1667 | return; |
| 1668 | } |
| 1669 | src=firstBoundaryInSrc; |
| 1670 | } |
| 1671 | } |
| 1672 | if(doMakeFCD) { |
| 1673 | makeFCD(src, limit, &buffer, errorCode); |
| 1674 | } else { |
| 1675 | buffer.appendZeroCC(src, limit, errorCode); |
| 1676 | } |
| 1677 | } |
| 1678 | |
| 1679 | const UChar *Normalizer2Impl::findPreviousFCDBoundary(const UChar *start, const UChar *p) const { |
| 1680 | BackwardUTrie2StringIterator iter(fcdTrie(), start, p); |
| 1681 | uint16_t fcd16; |
| 1682 | do { |
| 1683 | fcd16=iter.previous16(); |
| 1684 | } while(fcd16>0xff); |
| 1685 | return iter.codePointStart; |
| 1686 | } |
| 1687 | |
| 1688 | const UChar *Normalizer2Impl::findNextFCDBoundary(const UChar *p, const UChar *limit) const { |
| 1689 | ForwardUTrie2StringIterator iter(fcdTrie(), p, limit); |
| 1690 | uint16_t fcd16; |
| 1691 | do { |
| 1692 | fcd16=iter.next16(); |
| 1693 | } while(fcd16>0xff); |
| 1694 | return iter.codePointStart; |
| 1695 | } |
| 1696 | |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1697 | // CanonicalIterator data -------------------------------------------------- *** |
| 1698 | |
| 1699 | CanonIterData::CanonIterData(UErrorCode &errorCode) : |
| 1700 | trie(utrie2_open(0, 0, &errorCode)), |
| 1701 | canonStartSets(uhash_deleteUObject, NULL, errorCode) {} |
| 1702 | |
| 1703 | CanonIterData::~CanonIterData() { |
| 1704 | utrie2_close(trie); |
| 1705 | } |
| 1706 | |
| 1707 | void CanonIterData::addToStartSet(UChar32 origin, UChar32 decompLead, UErrorCode &errorCode) { |
| 1708 | uint32_t canonValue=utrie2_get32(trie, decompLead); |
| 1709 | if((canonValue&(CANON_HAS_SET|CANON_VALUE_MASK))==0 && origin!=0) { |
| 1710 | // origin is the first character whose decomposition starts with |
| 1711 | // the character for which we are setting the value. |
| 1712 | utrie2_set32(trie, decompLead, canonValue|origin, &errorCode); |
| 1713 | } else { |
| 1714 | // origin is not the first character, or it is U+0000. |
| 1715 | UnicodeSet *set; |
| 1716 | if((canonValue&CANON_HAS_SET)==0) { |
| 1717 | set=new UnicodeSet; |
| 1718 | if(set==NULL) { |
| 1719 | errorCode=U_MEMORY_ALLOCATION_ERROR; |
| 1720 | return; |
| 1721 | } |
| 1722 | UChar32 firstOrigin=(UChar32)(canonValue&CANON_VALUE_MASK); |
| 1723 | canonValue=(canonValue&~CANON_VALUE_MASK)|CANON_HAS_SET|(uint32_t)canonStartSets.size(); |
| 1724 | utrie2_set32(trie, decompLead, canonValue, &errorCode); |
| 1725 | canonStartSets.addElement(set, errorCode); |
| 1726 | if(firstOrigin!=0) { |
| 1727 | set->add(firstOrigin); |
| 1728 | } |
| 1729 | } else { |
| 1730 | set=(UnicodeSet *)canonStartSets[(int32_t)(canonValue&CANON_VALUE_MASK)]; |
| 1731 | } |
| 1732 | set->add(origin); |
| 1733 | } |
| 1734 | } |
| 1735 | |
| 1736 | class CanonIterDataSingleton { |
| 1737 | public: |
| 1738 | CanonIterDataSingleton(SimpleSingleton &s, Normalizer2Impl &ni, UErrorCode &ec) : |
| 1739 | singleton(s), impl(ni), errorCode(ec) {} |
| 1740 | CanonIterData *getInstance(UErrorCode &errorCode) { |
| 1741 | void *duplicate; |
| 1742 | CanonIterData *instance= |
| 1743 | (CanonIterData *)singleton.getInstance(createInstance, this, duplicate, errorCode); |
| 1744 | delete (CanonIterData *)duplicate; |
| 1745 | return instance; |
| 1746 | } |
| 1747 | static void *createInstance(const void *context, UErrorCode &errorCode); |
| 1748 | UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { |
| 1749 | if(value!=0) { |
| 1750 | impl.makeCanonIterDataFromNorm16(start, end, (uint16_t)value, *newData, errorCode); |
| 1751 | } |
| 1752 | return U_SUCCESS(errorCode); |
| 1753 | } |
| 1754 | |
| 1755 | private: |
| 1756 | SimpleSingleton &singleton; |
| 1757 | Normalizer2Impl &impl; |
| 1758 | CanonIterData *newData; |
| 1759 | UErrorCode &errorCode; |
| 1760 | }; |
| 1761 | |
| 1762 | U_CDECL_BEGIN |
| 1763 | |
| 1764 | // Call Normalizer2Impl::makeCanonIterDataFromNorm16() for a range of same-norm16 characters. |
| 1765 | static UBool U_CALLCONV |
| 1766 | enumCIDRangeHandler(const void *context, UChar32 start, UChar32 end, uint32_t value) { |
| 1767 | return ((CanonIterDataSingleton *)context)->rangeHandler(start, end, value); |
| 1768 | } |
| 1769 | |
| 1770 | U_CDECL_END |
| 1771 | |
| 1772 | void *CanonIterDataSingleton::createInstance(const void *context, UErrorCode &errorCode) { |
| 1773 | CanonIterDataSingleton *me=(CanonIterDataSingleton *)context; |
| 1774 | me->newData=new CanonIterData(errorCode); |
| 1775 | if(me->newData==NULL) { |
| 1776 | errorCode=U_MEMORY_ALLOCATION_ERROR; |
| 1777 | return NULL; |
| 1778 | } |
| 1779 | if(U_SUCCESS(errorCode)) { |
| 1780 | utrie2_enum(me->impl.getNormTrie(), NULL, enumCIDRangeHandler, me); |
| 1781 | utrie2_freeze(me->newData->trie, UTRIE2_32_VALUE_BITS, &errorCode); |
| 1782 | if(U_SUCCESS(errorCode)) { |
| 1783 | return me->newData; |
| 1784 | } |
| 1785 | } |
| 1786 | delete me->newData; |
| 1787 | return NULL; |
| 1788 | } |
| 1789 | |
| 1790 | void Normalizer2Impl::makeCanonIterDataFromNorm16(UChar32 start, UChar32 end, uint16_t norm16, |
| 1791 | CanonIterData &newData, |
| 1792 | UErrorCode &errorCode) const { |
| 1793 | if(norm16==0 || (minYesNo<=norm16 && norm16<minNoNo)) { |
| 1794 | // Inert, or 2-way mapping (including Hangul syllable). |
| 1795 | // We do not write a canonStartSet for any yesNo character. |
| 1796 | // Composites from 2-way mappings are added at runtime from the |
| 1797 | // starter's compositions list, and the other characters in |
| 1798 | // 2-way mappings get CANON_NOT_SEGMENT_STARTER set because they are |
| 1799 | // "maybe" characters. |
| 1800 | return; |
| 1801 | } |
| 1802 | for(UChar32 c=start; c<=end; ++c) { |
| 1803 | uint32_t oldValue=utrie2_get32(newData.trie, c); |
| 1804 | uint32_t newValue=oldValue; |
| 1805 | if(norm16>=minMaybeYes) { |
| 1806 | // not a segment starter if it occurs in a decomposition or has cc!=0 |
| 1807 | newValue|=CANON_NOT_SEGMENT_STARTER; |
| 1808 | if(norm16<MIN_NORMAL_MAYBE_YES) { |
| 1809 | newValue|=CANON_HAS_COMPOSITIONS; |
| 1810 | } |
| 1811 | } else if(norm16<minYesNo) { |
| 1812 | newValue|=CANON_HAS_COMPOSITIONS; |
| 1813 | } else { |
| 1814 | // c has a one-way decomposition |
| 1815 | UChar32 c2=c; |
| 1816 | uint16_t norm16_2=norm16; |
| 1817 | while(limitNoNo<=norm16_2 && norm16_2<minMaybeYes) { |
| 1818 | c2=mapAlgorithmic(c2, norm16_2); |
| 1819 | norm16_2=getNorm16(c2); |
| 1820 | } |
| 1821 | if(minYesNo<=norm16_2 && norm16_2<limitNoNo) { |
| 1822 | // c decomposes, get everything from the variable-length extra data |
| 1823 | const uint16_t *mapping=getMapping(norm16_2); |
| 1824 | uint16_t firstUnit=*mapping++; |
| 1825 | int32_t length=firstUnit&MAPPING_LENGTH_MASK; |
| 1826 | if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { |
| 1827 | if(c==c2 && (*mapping&0xff)!=0) { |
| 1828 | newValue|=CANON_NOT_SEGMENT_STARTER; // original c has cc!=0 |
| 1829 | } |
| 1830 | ++mapping; |
| 1831 | } |
| 1832 | // Skip empty mappings (no characters in the decomposition). |
| 1833 | if(length!=0) { |
| 1834 | // add c to first code point's start set |
| 1835 | int32_t i=0; |
| 1836 | U16_NEXT_UNSAFE(mapping, i, c2); |
| 1837 | newData.addToStartSet(c, c2, errorCode); |
| 1838 | // Set CANON_NOT_SEGMENT_STARTER for each remaining code point of a |
| 1839 | // one-way mapping. A 2-way mapping is possible here after |
| 1840 | // intermediate algorithmic mapping. |
| 1841 | if(norm16_2>=minNoNo) { |
| 1842 | while(i<length) { |
| 1843 | U16_NEXT_UNSAFE(mapping, i, c2); |
| 1844 | uint32_t c2Value=utrie2_get32(newData.trie, c2); |
| 1845 | if((c2Value&CANON_NOT_SEGMENT_STARTER)==0) { |
| 1846 | utrie2_set32(newData.trie, c2, c2Value|CANON_NOT_SEGMENT_STARTER, |
| 1847 | &errorCode); |
| 1848 | } |
| 1849 | } |
| 1850 | } |
| 1851 | } |
| 1852 | } else { |
| 1853 | // c decomposed to c2 algorithmically; c has cc==0 |
| 1854 | newData.addToStartSet(c, c2, errorCode); |
| 1855 | } |
| 1856 | } |
| 1857 | if(newValue!=oldValue) { |
| 1858 | utrie2_set32(newData.trie, c, newValue, &errorCode); |
| 1859 | } |
| 1860 | } |
| 1861 | } |
| 1862 | |
| 1863 | UBool Normalizer2Impl::ensureCanonIterData(UErrorCode &errorCode) const { |
| 1864 | // Logically const: Synchronized instantiation. |
| 1865 | Normalizer2Impl *me=const_cast<Normalizer2Impl *>(this); |
| 1866 | CanonIterDataSingleton(me->canonIterDataSingleton, *me, errorCode).getInstance(errorCode); |
| 1867 | return U_SUCCESS(errorCode); |
| 1868 | } |
| 1869 | |
| 1870 | int32_t Normalizer2Impl::getCanonValue(UChar32 c) const { |
| 1871 | return (int32_t)utrie2_get32(((CanonIterData *)canonIterDataSingleton.fInstance)->trie, c); |
| 1872 | } |
| 1873 | |
| 1874 | const UnicodeSet &Normalizer2Impl::getCanonStartSet(int32_t n) const { |
| 1875 | return *(const UnicodeSet *)( |
| 1876 | ((CanonIterData *)canonIterDataSingleton.fInstance)->canonStartSets[n]); |
| 1877 | } |
| 1878 | |
| 1879 | UBool Normalizer2Impl::isCanonSegmentStarter(UChar32 c) const { |
| 1880 | return getCanonValue(c)>=0; |
| 1881 | } |
| 1882 | |
| 1883 | UBool Normalizer2Impl::getCanonStartSet(UChar32 c, UnicodeSet &set) const { |
| 1884 | int32_t canonValue=getCanonValue(c)&~CANON_NOT_SEGMENT_STARTER; |
| 1885 | if(canonValue==0) { |
| 1886 | return FALSE; |
| 1887 | } |
| 1888 | set.clear(); |
| 1889 | int32_t value=canonValue&CANON_VALUE_MASK; |
| 1890 | if((canonValue&CANON_HAS_SET)!=0) { |
| 1891 | set.addAll(getCanonStartSet(value)); |
| 1892 | } else if(value!=0) { |
| 1893 | set.add(value); |
| 1894 | } |
| 1895 | if((canonValue&CANON_HAS_COMPOSITIONS)!=0) { |
| 1896 | uint16_t norm16=getNorm16(c); |
| 1897 | if(norm16==JAMO_L) { |
| 1898 | UChar32 syllable= |
| 1899 | (UChar32)(Hangul::HANGUL_BASE+(c-Hangul::JAMO_L_BASE)*Hangul::JAMO_VT_COUNT); |
| 1900 | set.add(syllable, syllable+Hangul::JAMO_VT_COUNT-1); |
| 1901 | } else { |
| 1902 | addComposites(getCompositionsList(norm16), set); |
| 1903 | } |
| 1904 | } |
| 1905 | return TRUE; |
| 1906 | } |
| 1907 | |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1908 | U_NAMESPACE_END |
| 1909 | |
| 1910 | // Normalizer2 data swapping ----------------------------------------------- *** |
| 1911 | |
| 1912 | U_NAMESPACE_USE |
| 1913 | |
| 1914 | U_CAPI int32_t U_EXPORT2 |
| 1915 | unorm2_swap(const UDataSwapper *ds, |
| 1916 | const void *inData, int32_t length, void *outData, |
| 1917 | UErrorCode *pErrorCode) { |
| 1918 | const UDataInfo *pInfo; |
| 1919 | int32_t headerSize; |
| 1920 | |
| 1921 | const uint8_t *inBytes; |
| 1922 | uint8_t *outBytes; |
| 1923 | |
| 1924 | const int32_t *inIndexes; |
| 1925 | int32_t indexes[Normalizer2Impl::IX_MIN_MAYBE_YES+1]; |
| 1926 | |
| 1927 | int32_t i, offset, nextOffset, size; |
| 1928 | |
| 1929 | /* udata_swapDataHeader checks the arguments */ |
| 1930 | headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode); |
| 1931 | if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 1932 | return 0; |
| 1933 | } |
| 1934 | |
| 1935 | /* check data format and format version */ |
| 1936 | pInfo=(const UDataInfo *)((const char *)inData+4); |
| 1937 | if(!( |
| 1938 | pInfo->dataFormat[0]==0x4e && /* dataFormat="Nrm2" */ |
| 1939 | pInfo->dataFormat[1]==0x72 && |
| 1940 | pInfo->dataFormat[2]==0x6d && |
| 1941 | pInfo->dataFormat[3]==0x32 && |
| 1942 | pInfo->formatVersion[0]==1 |
| 1943 | )) { |
| 1944 | udata_printError(ds, "unorm2_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized as Normalizer2 data\n", |
| 1945 | pInfo->dataFormat[0], pInfo->dataFormat[1], |
| 1946 | pInfo->dataFormat[2], pInfo->dataFormat[3], |
| 1947 | pInfo->formatVersion[0]); |
| 1948 | *pErrorCode=U_UNSUPPORTED_ERROR; |
| 1949 | return 0; |
| 1950 | } |
| 1951 | |
| 1952 | inBytes=(const uint8_t *)inData+headerSize; |
| 1953 | outBytes=(uint8_t *)outData+headerSize; |
| 1954 | |
| 1955 | inIndexes=(const int32_t *)inBytes; |
| 1956 | |
| 1957 | if(length>=0) { |
| 1958 | length-=headerSize; |
claireho | 27f6547 | 2011-06-09 11:11:49 -0700 | [diff] [blame^] | 1959 | if(length<(int32_t)sizeof(indexes)) { |
claireho | 50294ea | 2010-05-03 15:44:48 -0700 | [diff] [blame] | 1960 | udata_printError(ds, "unorm2_swap(): too few bytes (%d after header) for Normalizer2 data\n", |
| 1961 | length); |
| 1962 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
| 1963 | return 0; |
| 1964 | } |
| 1965 | } |
| 1966 | |
| 1967 | /* read the first few indexes */ |
| 1968 | for(i=0; i<=Normalizer2Impl::IX_MIN_MAYBE_YES; ++i) { |
| 1969 | indexes[i]=udata_readInt32(ds, inIndexes[i]); |
| 1970 | } |
| 1971 | |
| 1972 | /* get the total length of the data */ |
| 1973 | size=indexes[Normalizer2Impl::IX_TOTAL_SIZE]; |
| 1974 | |
| 1975 | if(length>=0) { |
| 1976 | if(length<size) { |
| 1977 | udata_printError(ds, "unorm2_swap(): too few bytes (%d after header) for all of Normalizer2 data\n", |
| 1978 | length); |
| 1979 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
| 1980 | return 0; |
| 1981 | } |
| 1982 | |
| 1983 | /* copy the data for inaccessible bytes */ |
| 1984 | if(inBytes!=outBytes) { |
| 1985 | uprv_memcpy(outBytes, inBytes, size); |
| 1986 | } |
| 1987 | |
| 1988 | offset=0; |
| 1989 | |
| 1990 | /* swap the int32_t indexes[] */ |
| 1991 | nextOffset=indexes[Normalizer2Impl::IX_NORM_TRIE_OFFSET]; |
| 1992 | ds->swapArray32(ds, inBytes, nextOffset-offset, outBytes, pErrorCode); |
| 1993 | offset=nextOffset; |
| 1994 | |
| 1995 | /* swap the UTrie2 */ |
| 1996 | nextOffset=indexes[Normalizer2Impl::IX_EXTRA_DATA_OFFSET]; |
| 1997 | utrie2_swap(ds, inBytes+offset, nextOffset-offset, outBytes+offset, pErrorCode); |
| 1998 | offset=nextOffset; |
| 1999 | |
| 2000 | /* swap the uint16_t extraData[] */ |
| 2001 | nextOffset=indexes[Normalizer2Impl::IX_EXTRA_DATA_OFFSET+1]; |
| 2002 | ds->swapArray16(ds, inBytes+offset, nextOffset-offset, outBytes+offset, pErrorCode); |
| 2003 | offset=nextOffset; |
| 2004 | |
| 2005 | U_ASSERT(offset==size); |
| 2006 | } |
| 2007 | |
| 2008 | return headerSize+size; |
| 2009 | } |
| 2010 | |
| 2011 | #endif // !UCONFIG_NO_NORMALIZATION |