1 /* 2 ****************************************************************************** 3 * 4 * Copyright (C) 1999-2008, International Business Machines 5 * Corporation and others. All Rights Reserved. 6 * 7 ****************************************************************************** 8 * 9 * 10 * ucnv_io.c: 11 * initializes global variables and defines functions pertaining to converter 12 * name resolution aspect of the conversion code. 13 * 14 * new implementation: 15 * 16 * created on: 1999nov22 17 * created by: Markus W. Scherer 18 * 19 * Use the binary cnvalias.icu (created from convrtrs.txt) to work 20 * with aliases for converter names. 21 * 22 * Date Name Description 23 * 11/22/1999 markus Created 24 * 06/28/2002 grhoten Major overhaul of the converter alias design. 25 * Now an alias can map to different converters 26 * depending on the specified standard. 27 ******************************************************************************* 28 */ 29 30 #include "unicode/utypes.h" 31 32 #if !UCONFIG_NO_CONVERSION 33 34 #include "unicode/ucnv.h" 35 #include "unicode/udata.h" 36 37 #include "umutex.h" 38 #include "uarrsort.h" 39 #include "udataswp.h" 40 #include "cstring.h" 41 #include "cmemory.h" 42 #include "ucnv_io.h" 43 #include "uenumimp.h" 44 #include "ucln_cmn.h" 45 46 /* Format of cnvalias.icu ----------------------------------------------------- 47 * 48 * cnvalias.icu is a binary, memory-mappable form of convrtrs.txt. 49 * This binary form contains several tables. All indexes are to uint16_t 50 * units, and not to the bytes (uint8_t units). Addressing everything on 51 * 16-bit boundaries allows us to store more information with small index 52 * numbers, which are also 16-bit in size. The majority of the table (except 53 * the string table) are 16-bit numbers. 54 * 55 * First there is the size of the Table of Contents (TOC). The TOC 56 * entries contain the size of each section. In order to find the offset 57 * you just need to sum up the previous offsets. 58 * The TOC length and entries are an array of uint32_t values. 59 * The first section after the TOC starts immediately after the TOC. 60 * 61 * 1) This section contains a list of converters. This list contains indexes 62 * into the string table for the converter name. The index of this list is 63 * also used by other sections, which are mentioned later on. 64 * This list is not sorted. 65 * 66 * 2) This section contains a list of tags. This list contains indexes 67 * into the string table for the tag name. The index of this list is 68 * also used by other sections, which are mentioned later on. 69 * This list is in priority order of standards. 70 * 71 * 3) This section contains a list of sorted unique aliases. This 72 * list contains indexes into the string table for the alias name. The 73 * index of this list is also used by other sections, like the 4th section. 74 * The index for the 3rd and 4th section is used to get the 75 * alias -> converter name mapping. Section 3 and 4 form a two column table. 76 * Some of the most significant bits of each index may contain other 77 * information (see findConverter for details). 78 * 79 * 4) This section contains a list of mapped converter names. Consider this 80 * as a table that maps the 3rd section to the 1st section. This list contains 81 * indexes into the 1st section. The index of this list is the same index in 82 * the 3rd section. There is also some extra information in the high bits of 83 * each converter index in this table. Currently it's only used to say that 84 * an alias mapped to this converter is ambiguous. See UCNV_CONVERTER_INDEX_MASK 85 * and UCNV_AMBIGUOUS_ALIAS_MAP_BIT for more information. This section is 86 * the predigested form of the 5th section so that an alias lookup can be fast. 87 * 88 * 5) This section contains a 2D array with indexes to the 6th section. This 89 * section is the full form of all alias mappings. The column index is the 90 * index into the converter list (column header). The row index is the index 91 * to tag list (row header). This 2D array is the top part a 3D array. The 92 * third dimension is in the 6th section. 93 * 94 * 6) This is blob of variable length arrays. Each array starts with a size, 95 * and is followed by indexes to alias names in the string table. This is 96 * the third dimension to the section 5. No other section should be referencing 97 * this section. 98 * 99 * 7) Starting in ICU 3.6, this can be a UConverterAliasOptions struct. Its 100 * presence indicates that a section 9 exists. UConverterAliasOptions specifies 101 * what type of string normalization is used among other potential things in the 102 * future. 103 * 104 * 8) This is the string table. All strings are indexed on an even address. 105 * There are two reasons for this. First many chip architectures locate strings 106 * faster on even address boundaries. Second, since all indexes are 16-bit 107 * numbers, this string table can be 128KB in size instead of 64KB when we 108 * only have strings starting on an even address. 109 * 110 * 9) When present this is a set of prenormalized strings from section 8. This 111 * table contains normalized strings with the dashes and spaces stripped out, 112 * and all strings lowercased. In the future, the options in section 7 may state 113 * other types of normalization. 114 * 115 * Here is the concept of section 5 and 6. It's a 3D cube. Each tag 116 * has a unique alias among all converters. That same alias can 117 * be mentioned in other standards on different converters, 118 * but only one alias per tag can be unique. 119 * 120 * 121 * Converter Names (Usually in TR22 form) 122 * -------------------------------------------. 123 * T / /| 124 * a / / | 125 * g / / | 126 * s / / | 127 * / / | 128 * ------------------------------------------/ | 129 * A | | | 130 * l | | | 131 * i | | / 132 * a | | / 133 * s | | / 134 * e | | / 135 * s | |/ 136 * ------------------------------------------- 137 * 138 * 139 * 140 * Here is what it really looks like. It's like swiss cheese. 141 * There are holes. Some converters aren't recognized by 142 * a standard, or they are really old converters that the 143 * standard doesn't recognize anymore. 144 * 145 * Converter Names (Usually in TR22 form) 146 * -------------------------------------------. 147 * T /##########################################/| 148 * a / # # /# 149 * g / # ## ## ### # ### ### ### #/ 150 * s / # ##### #### ## ## #/# 151 * / ### # # ## # # # ### # # #/## 152 * ------------------------------------------/# # 153 * A |### # # ## # # # ### # # #|# # 154 * l |# # # # # ## # #|# # 155 * i |# # # # # # #|# 156 * a |# #|# 157 * s | #|# 158 * e 159 * s 160 * 161 */ 162 163 /** 164 * Used by the UEnumeration API 165 */ 166 typedef struct UAliasContext { 167 uint32_t listOffset; 168 uint32_t listIdx; 169 } UAliasContext; 170 171 static const char DATA_NAME[] = "cnvalias"; 172 static const char DATA_TYPE[] = "icu"; 173 174 static UDataMemory *gAliasData=NULL; 175 176 enum { 177 tocLengthIndex=0, 178 converterListIndex=1, 179 tagListIndex=2, 180 aliasListIndex=3, 181 untaggedConvArrayIndex=4, 182 taggedAliasArrayIndex=5, 183 taggedAliasListsIndex=6, 184 tableOptionsIndex=7, 185 stringTableIndex=8, 186 normalizedStringTableIndex=9, 187 offsetsCount, /* length of the swapper's temporary offsets[] */ 188 minTocLength=8 /* min. tocLength in the file, does not count the tocLengthIndex! */ 189 }; 190 191 static const UConverterAliasOptions defaultTableOptions = { 192 UCNV_IO_UNNORMALIZED, 193 0 /* containsCnvOptionInfo */ 194 }; 195 static UConverterAlias gMainTable; 196 197 #define GET_STRING(idx) (const char *)(gMainTable.stringTable + (idx)) 198 #define GET_NORMALIZED_STRING(idx) (const char *)(gMainTable.normalizedStringTable + (idx)) 199 200 static UBool U_CALLCONV 201 isAcceptable(void *context, 202 const char *type, const char *name, 203 const UDataInfo *pInfo) { 204 return (UBool)( 205 pInfo->size>=20 && 206 pInfo->isBigEndian==U_IS_BIG_ENDIAN && 207 pInfo->charsetFamily==U_CHARSET_FAMILY && 208 pInfo->dataFormat[0]==0x43 && /* dataFormat="CvAl" */ 209 pInfo->dataFormat[1]==0x76 && 210 pInfo->dataFormat[2]==0x41 && 211 pInfo->dataFormat[3]==0x6c && 212 pInfo->formatVersion[0]==3); 213 } 214 215 static UBool U_CALLCONV ucnv_io_cleanup(void) 216 { 217 if (gAliasData) { 218 udata_close(gAliasData); 219 gAliasData = NULL; 220 } 221 222 uprv_memset(&gMainTable, 0, sizeof(gMainTable)); 223 224 return TRUE; /* Everything was cleaned up */ 225 } 226 227 static UBool 228 haveAliasData(UErrorCode *pErrorCode) { 229 int needInit; 230 231 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { 232 return FALSE; 233 } 234 235 UMTX_CHECK(NULL, (gAliasData==NULL), needInit); 236 237 /* load converter alias data from file if necessary */ 238 if (needInit) { 239 UDataMemory *data; 240 const uint16_t *table; 241 const uint32_t *sectionSizes; 242 uint32_t tableStart; 243 uint32_t currOffset; 244 245 data = udata_openChoice(NULL, DATA_TYPE, DATA_NAME, isAcceptable, NULL, pErrorCode); 246 if(U_FAILURE(*pErrorCode)) { 247 return FALSE; 248 } 249 250 sectionSizes = (const uint32_t *)udata_getMemory(data); 251 table = (const uint16_t *)sectionSizes; 252 253 tableStart = sectionSizes[0]; 254 if (tableStart < minTocLength) { 255 *pErrorCode = U_INVALID_FORMAT_ERROR; 256 udata_close(data); 257 return FALSE; 258 } 259 260 umtx_lock(NULL); 261 if(gAliasData==NULL) { 262 gMainTable.converterListSize = sectionSizes[1]; 263 gMainTable.tagListSize = sectionSizes[2]; 264 gMainTable.aliasListSize = sectionSizes[3]; 265 gMainTable.untaggedConvArraySize = sectionSizes[4]; 266 gMainTable.taggedAliasArraySize = sectionSizes[5]; 267 gMainTable.taggedAliasListsSize = sectionSizes[6]; 268 gMainTable.optionTableSize = sectionSizes[7]; 269 gMainTable.stringTableSize = sectionSizes[8]; 270 271 if (tableStart > 8) { 272 gMainTable.normalizedStringTableSize = sectionSizes[9]; 273 } 274 275 currOffset = tableStart * (sizeof(uint32_t)/sizeof(uint16_t)) + (sizeof(uint32_t)/sizeof(uint16_t)); 276 gMainTable.converterList = table + currOffset; 277 278 currOffset += gMainTable.converterListSize; 279 gMainTable.tagList = table + currOffset; 280 281 currOffset += gMainTable.tagListSize; 282 gMainTable.aliasList = table + currOffset; 283 284 currOffset += gMainTable.aliasListSize; 285 gMainTable.untaggedConvArray = table + currOffset; 286 287 currOffset += gMainTable.untaggedConvArraySize; 288 gMainTable.taggedAliasArray = table + currOffset; 289 290 /* aliasLists is a 1's based array, but it has a padding character */ 291 currOffset += gMainTable.taggedAliasArraySize; 292 gMainTable.taggedAliasLists = table + currOffset; 293 294 currOffset += gMainTable.taggedAliasListsSize; 295 if (gMainTable.optionTableSize > 0 296 && ((const UConverterAliasOptions *)(table + currOffset))->stringNormalizationType < UCNV_IO_NORM_TYPE_COUNT) 297 { 298 /* Faster table */ 299 gMainTable.optionTable = (const UConverterAliasOptions *)(table + currOffset); 300 } 301 else { 302 /* Smaller table, or I can't handle this normalization mode! 303 Use the original slower table lookup. */ 304 gMainTable.optionTable = &defaultTableOptions; 305 } 306 307 currOffset += gMainTable.optionTableSize; 308 gMainTable.stringTable = table + currOffset; 309 310 currOffset += gMainTable.stringTableSize; 311 gMainTable.normalizedStringTable = ((gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED) 312 ? gMainTable.stringTable : (table + currOffset)); 313 314 ucln_common_registerCleanup(UCLN_COMMON_UCNV_IO, ucnv_io_cleanup); 315 316 gAliasData = data; 317 data=NULL; 318 } 319 umtx_unlock(NULL); 320 321 /* if a different thread set it first, then close the extra data */ 322 if(data!=NULL) { 323 udata_close(data); /* NULL if it was set correctly */ 324 } 325 } 326 327 return TRUE; 328 } 329 330 static U_INLINE UBool 331 isAlias(const char *alias, UErrorCode *pErrorCode) { 332 if(alias==NULL) { 333 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; 334 return FALSE; 335 } 336 return (UBool)(*alias!=0); 337 } 338 339 static uint32_t getTagNumber(const char *tagname) { 340 if (gMainTable.tagList) { 341 uint32_t tagNum; 342 for (tagNum = 0; tagNum < gMainTable.tagListSize; tagNum++) { 343 if (!uprv_stricmp(GET_STRING(gMainTable.tagList[tagNum]), tagname)) { 344 return tagNum; 345 } 346 } 347 } 348 349 return UINT32_MAX; 350 } 351 352 /* character types relevant for ucnv_compareNames() */ 353 enum { 354 IGNORE, 355 ZERO, 356 NONZERO, 357 MINLETTER /* any values from here on are lowercase letter mappings */ 358 }; 359 360 /* character types for ASCII 00..7F */ 361 static const uint8_t asciiTypes[128] = { 362 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 363 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 364 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 365 ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0, 366 0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 367 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0, 368 0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 369 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0 370 }; 371 372 #define GET_ASCII_TYPE(c) ((int8_t)(c) >= 0 ? asciiTypes[(uint8_t)c] : (uint8_t)IGNORE) 373 374 /* character types for EBCDIC 80..FF */ 375 static const uint8_t ebcdicTypes[128] = { 376 0, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0, 377 0, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0, 378 0, 0, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0, 379 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 380 0, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0, 381 0, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0, 382 0, 0, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0, 383 ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0 384 }; 385 386 #define GET_EBCDIC_TYPE(c) ((int8_t)(c) < 0 ? ebcdicTypes[(c)&0x7f] : (uint8_t)IGNORE) 387 388 #if U_CHARSET_FAMILY==U_ASCII_FAMILY 389 # define GET_CHAR_TYPE(c) GET_ASCII_TYPE(c) 390 #elif U_CHARSET_FAMILY==U_EBCDIC_FAMILY 391 # define GET_CHAR_TYPE(c) GET_EBCDIC_TYPE(c) 392 #else 393 # error U_CHARSET_FAMILY is not valid 394 #endif 395 396 /* @see ucnv_compareNames */ 397 U_CFUNC char * U_EXPORT2 398 ucnv_io_stripASCIIForCompare(char *dst, const char *name) { 399 char *dstItr = dst; 400 uint8_t type, nextType; 401 char c1; 402 UBool afterDigit = FALSE; 403 404 while ((c1 = *name++) != 0) { 405 type = GET_ASCII_TYPE(c1); 406 switch (type) { 407 case IGNORE: 408 afterDigit = FALSE; 409 continue; /* ignore all but letters and digits */ 410 case ZERO: 411 if (!afterDigit) { 412 nextType = GET_ASCII_TYPE(*name); 413 if (nextType == ZERO || nextType == NONZERO) { 414 continue; /* ignore leading zero before another digit */ 415 } 416 } 417 break; 418 case NONZERO: 419 afterDigit = TRUE; 420 break; 421 default: 422 c1 = (char)type; /* lowercased letter */ 423 afterDigit = FALSE; 424 break; 425 } 426 *dstItr++ = c1; 427 } 428 *dstItr = 0; 429 return dst; 430 } 431 432 U_CFUNC char * U_EXPORT2 433 ucnv_io_stripEBCDICForCompare(char *dst, const char *name) { 434 char *dstItr = dst; 435 uint8_t type, nextType; 436 char c1; 437 UBool afterDigit = FALSE; 438 439 while ((c1 = *name++) != 0) { 440 type = GET_EBCDIC_TYPE(c1); 441 switch (type) { 442 case IGNORE: 443 afterDigit = FALSE; 444 continue; /* ignore all but letters and digits */ 445 case ZERO: 446 if (!afterDigit) { 447 nextType = GET_EBCDIC_TYPE(*name); 448 if (nextType == ZERO || nextType == NONZERO) { 449 continue; /* ignore leading zero before another digit */ 450 } 451 } 452 break; 453 case NONZERO: 454 afterDigit = TRUE; 455 break; 456 default: 457 c1 = (char)type; /* lowercased letter */ 458 afterDigit = FALSE; 459 break; 460 } 461 *dstItr++ = c1; 462 } 463 *dstItr = 0; 464 return dst; 465 } 466 467 /** 468 * Do a fuzzy compare of two converter/alias names. 469 * The comparison is case-insensitive, ignores leading zeroes if they are not 470 * followed by further digits, and ignores all but letters and digits. 471 * Thus the strings "UTF-8", "utf_8", "u*T@f08" and "Utf 8" are exactly equivalent. 472 * See section 1.4, Charset Alias Matching in Unicode Technical Standard #22 473 * at http://www.unicode.org/reports/tr22/ 474 * 475 * This is a symmetrical (commutative) operation; order of arguments 476 * is insignificant. This is an important property for sorting the 477 * list (when the list is preprocessed into binary form) and for 478 * performing binary searches on it at run time. 479 * 480 * @param name1 a converter name or alias, zero-terminated 481 * @param name2 a converter name or alias, zero-terminated 482 * @return 0 if the names match, or a negative value if the name1 483 * lexically precedes name2, or a positive value if the name1 484 * lexically follows name2. 485 * 486 * @see ucnv_io_stripForCompare 487 */ 488 U_CAPI int U_EXPORT2 489 ucnv_compareNames(const char *name1, const char *name2) { 490 int rc; 491 uint8_t type, nextType; 492 char c1, c2; 493 UBool afterDigit1 = FALSE, afterDigit2 = FALSE; 494 495 for (;;) { 496 while ((c1 = *name1++) != 0) { 497 type = GET_CHAR_TYPE(c1); 498 switch (type) { 499 case IGNORE: 500 afterDigit1 = FALSE; 501 continue; /* ignore all but letters and digits */ 502 case ZERO: 503 if (!afterDigit1) { 504 nextType = GET_CHAR_TYPE(*name1); 505 if (nextType == ZERO || nextType == NONZERO) { 506 continue; /* ignore leading zero before another digit */ 507 } 508 } 509 break; 510 case NONZERO: 511 afterDigit1 = TRUE; 512 break; 513 default: 514 c1 = (char)type; /* lowercased letter */ 515 afterDigit1 = FALSE; 516 break; 517 } 518 break; /* deliver c1 */ 519 } 520 while ((c2 = *name2++) != 0) { 521 type = GET_CHAR_TYPE(c2); 522 switch (type) { 523 case IGNORE: 524 afterDigit2 = FALSE; 525 continue; /* ignore all but letters and digits */ 526 case ZERO: 527 if (!afterDigit2) { 528 nextType = GET_CHAR_TYPE(*name2); 529 if (nextType == ZERO || nextType == NONZERO) { 530 continue; /* ignore leading zero before another digit */ 531 } 532 } 533 break; 534 case NONZERO: 535 afterDigit2 = TRUE; 536 break; 537 default: 538 c2 = (char)type; /* lowercased letter */ 539 afterDigit2 = FALSE; 540 break; 541 } 542 break; /* deliver c2 */ 543 } 544 545 /* If we reach the ends of both strings then they match */ 546 if ((c1|c2)==0) { 547 return 0; 548 } 549 550 /* Case-insensitive comparison */ 551 rc = (int)(unsigned char)c1 - (int)(unsigned char)c2; 552 if (rc != 0) { 553 return rc; 554 } 555 } 556 } 557 558 /* 559 * search for an alias 560 * return the converter number index for gConverterList 561 */ 562 static U_INLINE uint32_t 563 findConverter(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) { 564 uint32_t mid, start, limit; 565 uint32_t lastMid; 566 int result; 567 int isUnnormalized = (gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED); 568 char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH]; 569 570 if (!isUnnormalized) { 571 if (uprv_strlen(alias) >= UCNV_MAX_CONVERTER_NAME_LENGTH) { 572 *pErrorCode = U_BUFFER_OVERFLOW_ERROR; 573 return UINT32_MAX; 574 } 575 576 /* Lower case and remove ignoreable characters. */ 577 ucnv_io_stripForCompare(strippedName, alias); 578 alias = strippedName; 579 } 580 581 /* do a binary search for the alias */ 582 start = 0; 583 limit = gMainTable.untaggedConvArraySize; 584 mid = limit; 585 lastMid = UINT32_MAX; 586 587 for (;;) { 588 mid = (uint32_t)((start + limit) / 2); 589 if (lastMid == mid) { /* Have we moved? */ 590 break; /* We haven't moved, and it wasn't found. */ 591 } 592 lastMid = mid; 593 if (isUnnormalized) { 594 result = ucnv_compareNames(alias, GET_STRING(gMainTable.aliasList[mid])); 595 } 596 else { 597 result = uprv_strcmp(alias, GET_NORMALIZED_STRING(gMainTable.aliasList[mid])); 598 } 599 600 if (result < 0) { 601 limit = mid; 602 } else if (result > 0) { 603 start = mid; 604 } else { 605 /* Since the gencnval tool folds duplicates into one entry, 606 * this alias in gAliasList is unique, but different standards 607 * may map an alias to different converters. 608 */ 609 if (gMainTable.untaggedConvArray[mid] & UCNV_AMBIGUOUS_ALIAS_MAP_BIT) { 610 *pErrorCode = U_AMBIGUOUS_ALIAS_WARNING; 611 } 612 /* State whether the canonical converter name contains an option. 613 This information is contained in this list in order to maintain backward & forward compatibility. */ 614 if (containsOption) { 615 UBool containsCnvOptionInfo = (UBool)gMainTable.optionTable->containsCnvOptionInfo; 616 *containsOption = (UBool)((containsCnvOptionInfo 617 && ((gMainTable.untaggedConvArray[mid] & UCNV_CONTAINS_OPTION_BIT) != 0)) 618 || !containsCnvOptionInfo); 619 } 620 return gMainTable.untaggedConvArray[mid] & UCNV_CONVERTER_INDEX_MASK; 621 } 622 } 623 624 return UINT32_MAX; 625 } 626 627 /* 628 * Is this alias in this list? 629 * alias and listOffset should be non-NULL. 630 */ 631 static U_INLINE UBool 632 isAliasInList(const char *alias, uint32_t listOffset) { 633 if (listOffset) { 634 uint32_t currAlias; 635 uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; 636 /* +1 to skip listCount */ 637 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; 638 for (currAlias = 0; currAlias < listCount; currAlias++) { 639 if (currList[currAlias] 640 && ucnv_compareNames(alias, GET_STRING(currList[currAlias]))==0) 641 { 642 return TRUE; 643 } 644 } 645 } 646 return FALSE; 647 } 648 649 /* 650 * Search for an standard name of an alias (what is the default name 651 * that this standard uses?) 652 * return the listOffset for gTaggedAliasLists. If it's 0, 653 * the it couldn't be found, but the parameters are valid. 654 */ 655 static uint32_t 656 findTaggedAliasListsOffset(const char *alias, const char *standard, UErrorCode *pErrorCode) { 657 uint32_t idx; 658 uint32_t listOffset; 659 uint32_t convNum; 660 UErrorCode myErr = U_ZERO_ERROR; 661 uint32_t tagNum = getTagNumber(standard); 662 663 /* Make a quick guess. Hopefully they used a TR22 canonical alias. */ 664 convNum = findConverter(alias, NULL, &myErr); 665 if (myErr != U_ZERO_ERROR) { 666 *pErrorCode = myErr; 667 } 668 669 if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) { 670 listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; 671 if (listOffset && gMainTable.taggedAliasLists[listOffset + 1]) { 672 return listOffset; 673 } 674 if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { 675 /* Uh Oh! They used an ambiguous alias. 676 We have to search the whole swiss cheese starting 677 at the highest standard affinity. 678 This may take a while. 679 */ 680 for (idx = 0; idx < gMainTable.taggedAliasArraySize; idx++) { 681 listOffset = gMainTable.taggedAliasArray[idx]; 682 if (listOffset && isAliasInList(alias, listOffset)) { 683 uint32_t currTagNum = idx/gMainTable.converterListSize; 684 uint32_t currConvNum = (idx - currTagNum*gMainTable.converterListSize); 685 uint32_t tempListOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + currConvNum]; 686 if (tempListOffset && gMainTable.taggedAliasLists[tempListOffset + 1]) { 687 return tempListOffset; 688 } 689 /* else keep on looking */ 690 /* We could speed this up by starting on the next row 691 because an alias is unique per row, right now. 692 This would change if alias versioning appears. */ 693 } 694 } 695 /* The standard doesn't know about the alias */ 696 } 697 /* else no default name */ 698 return 0; 699 } 700 /* else converter or tag not found */ 701 702 return UINT32_MAX; 703 } 704 705 /* Return the canonical name */ 706 static uint32_t 707 findTaggedConverterNum(const char *alias, const char *standard, UErrorCode *pErrorCode) { 708 uint32_t idx; 709 uint32_t listOffset; 710 uint32_t convNum; 711 UErrorCode myErr = U_ZERO_ERROR; 712 uint32_t tagNum = getTagNumber(standard); 713 714 /* Make a quick guess. Hopefully they used a TR22 canonical alias. */ 715 convNum = findConverter(alias, NULL, &myErr); 716 if (myErr != U_ZERO_ERROR) { 717 *pErrorCode = myErr; 718 } 719 720 if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) { 721 listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; 722 if (listOffset && isAliasInList(alias, listOffset)) { 723 return convNum; 724 } 725 if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { 726 /* Uh Oh! They used an ambiguous alias. 727 We have to search one slice of the swiss cheese. 728 We search only in the requested tag, not the whole thing. 729 This may take a while. 730 */ 731 uint32_t convStart = (tagNum)*gMainTable.converterListSize; 732 uint32_t convLimit = (tagNum+1)*gMainTable.converterListSize; 733 for (idx = convStart; idx < convLimit; idx++) { 734 listOffset = gMainTable.taggedAliasArray[idx]; 735 if (listOffset && isAliasInList(alias, listOffset)) { 736 return idx-convStart; 737 } 738 } 739 /* The standard doesn't know about the alias */ 740 } 741 /* else no canonical name */ 742 } 743 /* else converter or tag not found */ 744 745 return UINT32_MAX; 746 } 747 748 749 750 U_CFUNC const char * 751 ucnv_io_getConverterName(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) { 752 if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 753 uint32_t convNum = findConverter(alias, containsOption, pErrorCode); 754 if (convNum < gMainTable.converterListSize) { 755 return GET_STRING(gMainTable.converterList[convNum]); 756 } 757 /* else converter not found */ 758 } 759 return NULL; 760 } 761 762 static int32_t U_CALLCONV 763 ucnv_io_countStandardAliases(UEnumeration *enumerator, UErrorCode *pErrorCode) { 764 int32_t value = 0; 765 UAliasContext *myContext = (UAliasContext *)(enumerator->context); 766 uint32_t listOffset = myContext->listOffset; 767 768 if (listOffset) { 769 value = gMainTable.taggedAliasLists[listOffset]; 770 } 771 return value; 772 } 773 774 static const char* U_CALLCONV 775 ucnv_io_nextStandardAliases(UEnumeration *enumerator, 776 int32_t* resultLength, 777 UErrorCode *pErrorCode) 778 { 779 UAliasContext *myContext = (UAliasContext *)(enumerator->context); 780 uint32_t listOffset = myContext->listOffset; 781 782 if (listOffset) { 783 uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; 784 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; 785 786 if (myContext->listIdx < listCount) { 787 const char *myStr = GET_STRING(currList[myContext->listIdx++]); 788 if (resultLength) { 789 *resultLength = (int32_t)uprv_strlen(myStr); 790 } 791 return myStr; 792 } 793 } 794 /* Either we accessed a zero length list, or we enumerated too far. */ 795 if (resultLength) { 796 *resultLength = 0; 797 } 798 return NULL; 799 } 800 801 static void U_CALLCONV 802 ucnv_io_resetStandardAliases(UEnumeration *enumerator, UErrorCode *pErrorCode) { 803 ((UAliasContext *)(enumerator->context))->listIdx = 0; 804 } 805 806 static void U_CALLCONV 807 ucnv_io_closeUEnumeration(UEnumeration *enumerator) { 808 uprv_free(enumerator->context); 809 uprv_free(enumerator); 810 } 811 812 /* Enumerate the aliases for the specified converter and standard tag */ 813 static const UEnumeration gEnumAliases = { 814 NULL, 815 NULL, 816 ucnv_io_closeUEnumeration, 817 ucnv_io_countStandardAliases, 818 uenum_unextDefault, 819 ucnv_io_nextStandardAliases, 820 ucnv_io_resetStandardAliases 821 }; 822 823 U_CAPI UEnumeration * U_EXPORT2 824 ucnv_openStandardNames(const char *convName, 825 const char *standard, 826 UErrorCode *pErrorCode) 827 { 828 UEnumeration *myEnum = NULL; 829 if (haveAliasData(pErrorCode) && isAlias(convName, pErrorCode)) { 830 uint32_t listOffset = findTaggedAliasListsOffset(convName, standard, pErrorCode); 831 832 /* When listOffset == 0, we want to acknowledge that the 833 converter name and standard are okay, but there 834 is nothing to enumerate. */ 835 if (listOffset < gMainTable.taggedAliasListsSize) { 836 UAliasContext *myContext; 837 838 myEnum = uprv_malloc(sizeof(UEnumeration)); 839 if (myEnum == NULL) { 840 *pErrorCode = U_MEMORY_ALLOCATION_ERROR; 841 return NULL; 842 } 843 uprv_memcpy(myEnum, &gEnumAliases, sizeof(UEnumeration)); 844 myContext = uprv_malloc(sizeof(UAliasContext)); 845 if (myContext == NULL) { 846 *pErrorCode = U_MEMORY_ALLOCATION_ERROR; 847 uprv_free(myEnum); 848 return NULL; 849 } 850 myContext->listOffset = listOffset; 851 myContext->listIdx = 0; 852 myEnum->context = myContext; 853 } 854 /* else converter or tag not found */ 855 } 856 return myEnum; 857 } 858 859 static uint16_t 860 ucnv_io_countAliases(const char *alias, UErrorCode *pErrorCode) { 861 if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 862 uint32_t convNum = findConverter(alias, NULL, pErrorCode); 863 if (convNum < gMainTable.converterListSize) { 864 /* tagListNum - 1 is the ALL tag */ 865 int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; 866 867 if (listOffset) { 868 return gMainTable.taggedAliasLists[listOffset]; 869 } 870 /* else this shouldn't happen. internal program error */ 871 } 872 /* else converter not found */ 873 } 874 return 0; 875 } 876 877 static uint16_t 878 ucnv_io_getAliases(const char *alias, uint16_t start, const char **aliases, UErrorCode *pErrorCode) { 879 if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 880 uint32_t currAlias; 881 uint32_t convNum = findConverter(alias, NULL, pErrorCode); 882 if (convNum < gMainTable.converterListSize) { 883 /* tagListNum - 1 is the ALL tag */ 884 int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; 885 886 if (listOffset) { 887 uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; 888 /* +1 to skip listCount */ 889 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; 890 891 for (currAlias = start; currAlias < listCount; currAlias++) { 892 aliases[currAlias] = GET_STRING(currList[currAlias]); 893 } 894 } 895 /* else this shouldn't happen. internal program error */ 896 } 897 /* else converter not found */ 898 } 899 return 0; 900 } 901 902 static const char * 903 ucnv_io_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode) { 904 if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 905 uint32_t convNum = findConverter(alias, NULL, pErrorCode); 906 if (convNum < gMainTable.converterListSize) { 907 /* tagListNum - 1 is the ALL tag */ 908 int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; 909 910 if (listOffset) { 911 uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; 912 /* +1 to skip listCount */ 913 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; 914 915 if (n < listCount) { 916 return GET_STRING(currList[n]); 917 } 918 *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; 919 } 920 /* else this shouldn't happen. internal program error */ 921 } 922 /* else converter not found */ 923 } 924 return NULL; 925 } 926 927 static uint16_t 928 ucnv_io_countStandards(UErrorCode *pErrorCode) { 929 if (haveAliasData(pErrorCode)) { 930 /* Don't include the empty list */ 931 return (uint16_t)(gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS); 932 } 933 934 return 0; 935 } 936 937 U_CAPI const char * U_EXPORT2 938 ucnv_getStandard(uint16_t n, UErrorCode *pErrorCode) { 939 if (haveAliasData(pErrorCode)) { 940 if (n < gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) { 941 return GET_STRING(gMainTable.tagList[n]); 942 } 943 *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; 944 } 945 946 return NULL; 947 } 948 949 U_CAPI const char * U_EXPORT2 950 ucnv_getStandardName(const char *alias, const char *standard, UErrorCode *pErrorCode) { 951 if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 952 uint32_t listOffset = findTaggedAliasListsOffset(alias, standard, pErrorCode); 953 954 if (0 < listOffset && listOffset < gMainTable.taggedAliasListsSize) { 955 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; 956 957 /* Get the preferred name from this list */ 958 if (currList[0]) { 959 return GET_STRING(currList[0]); 960 } 961 /* else someone screwed up the alias table. */ 962 /* *pErrorCode = U_INVALID_FORMAT_ERROR */ 963 } 964 } 965 966 return NULL; 967 } 968 969 U_CAPI uint16_t U_EXPORT2 970 ucnv_countAliases(const char *alias, UErrorCode *pErrorCode) 971 { 972 return ucnv_io_countAliases(alias, pErrorCode); 973 } 974 975 976 U_CAPI const char* U_EXPORT2 977 ucnv_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode) 978 { 979 return ucnv_io_getAlias(alias, n, pErrorCode); 980 } 981 982 U_CAPI void U_EXPORT2 983 ucnv_getAliases(const char *alias, const char **aliases, UErrorCode *pErrorCode) 984 { 985 ucnv_io_getAliases(alias, 0, aliases, pErrorCode); 986 } 987 988 U_CAPI uint16_t U_EXPORT2 989 ucnv_countStandards(void) 990 { 991 UErrorCode err = U_ZERO_ERROR; 992 return ucnv_io_countStandards(&err); 993 } 994 995 U_CAPI const char * U_EXPORT2 996 ucnv_getCanonicalName(const char *alias, const char *standard, UErrorCode *pErrorCode) { 997 if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { 998 uint32_t convNum = findTaggedConverterNum(alias, standard, pErrorCode); 999 1000 if (convNum < gMainTable.converterListSize) { 1001 return GET_STRING(gMainTable.converterList[convNum]); 1002 } 1003 } 1004 1005 return NULL; 1006 } 1007 1008 static int32_t U_CALLCONV 1009 ucnv_io_countAllConverters(UEnumeration *enumerator, UErrorCode *pErrorCode) { 1010 return gMainTable.converterListSize; 1011 } 1012 1013 static const char* U_CALLCONV 1014 ucnv_io_nextAllConverters(UEnumeration *enumerator, 1015 int32_t* resultLength, 1016 UErrorCode *pErrorCode) 1017 { 1018 uint16_t *myContext = (uint16_t *)(enumerator->context); 1019 1020 if (*myContext < gMainTable.converterListSize) { 1021 const char *myStr = GET_STRING(gMainTable.converterList[(*myContext)++]); 1022 if (resultLength) { 1023 *resultLength = (int32_t)uprv_strlen(myStr); 1024 } 1025 return myStr; 1026 } 1027 /* Either we accessed a zero length list, or we enumerated too far. */ 1028 if (resultLength) { 1029 *resultLength = 0; 1030 } 1031 return NULL; 1032 } 1033 1034 static void U_CALLCONV 1035 ucnv_io_resetAllConverters(UEnumeration *enumerator, UErrorCode *pErrorCode) { 1036 *((uint16_t *)(enumerator->context)) = 0; 1037 } 1038 1039 static const UEnumeration gEnumAllConverters = { 1040 NULL, 1041 NULL, 1042 ucnv_io_closeUEnumeration, 1043 ucnv_io_countAllConverters, 1044 uenum_unextDefault, 1045 ucnv_io_nextAllConverters, 1046 ucnv_io_resetAllConverters 1047 }; 1048 1049 U_CAPI UEnumeration * U_EXPORT2 1050 ucnv_openAllNames(UErrorCode *pErrorCode) { 1051 UEnumeration *myEnum = NULL; 1052 if (haveAliasData(pErrorCode)) { 1053 uint16_t *myContext; 1054 1055 myEnum = uprv_malloc(sizeof(UEnumeration)); 1056 if (myEnum == NULL) { 1057 *pErrorCode = U_MEMORY_ALLOCATION_ERROR; 1058 return NULL; 1059 } 1060 uprv_memcpy(myEnum, &gEnumAllConverters, sizeof(UEnumeration)); 1061 myContext = uprv_malloc(sizeof(uint16_t)); 1062 if (myContext == NULL) { 1063 *pErrorCode = U_MEMORY_ALLOCATION_ERROR; 1064 uprv_free(myEnum); 1065 return NULL; 1066 } 1067 *myContext = 0; 1068 myEnum->context = myContext; 1069 } 1070 return myEnum; 1071 } 1072 1073 U_CFUNC uint16_t 1074 ucnv_io_countKnownConverters(UErrorCode *pErrorCode) { 1075 if (haveAliasData(pErrorCode)) { 1076 return (uint16_t)gMainTable.converterListSize; 1077 } 1078 return 0; 1079 } 1080 1081 /* alias table swapping ----------------------------------------------------- */ 1082 1083 typedef char * U_CALLCONV StripForCompareFn(char *dst, const char *name); 1084 1085 /* 1086 * row of a temporary array 1087 * 1088 * gets platform-endian charset string indexes and sorting indexes; 1089 * after sorting this array by strings, the actual arrays are permutated 1090 * according to the sorting indexes 1091 */ 1092 typedef struct TempRow { 1093 uint16_t strIndex, sortIndex; 1094 } TempRow; 1095 1096 typedef struct TempAliasTable { 1097 const char *chars; 1098 TempRow *rows; 1099 uint16_t *resort; 1100 StripForCompareFn *stripForCompare; 1101 } TempAliasTable; 1102 1103 enum { 1104 STACK_ROW_CAPACITY=500 1105 }; 1106 1107 static int32_t 1108 io_compareRows(const void *context, const void *left, const void *right) { 1109 char strippedLeft[UCNV_MAX_CONVERTER_NAME_LENGTH], 1110 strippedRight[UCNV_MAX_CONVERTER_NAME_LENGTH]; 1111 1112 TempAliasTable *tempTable=(TempAliasTable *)context; 1113 const char *chars=tempTable->chars; 1114 1115 return (int32_t)uprv_strcmp(tempTable->stripForCompare(strippedLeft, chars+2*((const TempRow *)left)->strIndex), 1116 tempTable->stripForCompare(strippedRight, chars+2*((const TempRow *)right)->strIndex)); 1117 } 1118 1119 U_CAPI int32_t U_EXPORT2 1120 ucnv_swapAliases(const UDataSwapper *ds, 1121 const void *inData, int32_t length, void *outData, 1122 UErrorCode *pErrorCode) { 1123 const UDataInfo *pInfo; 1124 int32_t headerSize; 1125 1126 const uint16_t *inTable; 1127 const uint32_t *inSectionSizes; 1128 uint32_t toc[offsetsCount]; 1129 uint32_t offsets[offsetsCount]; /* 16-bit-addressed offsets from inTable/outTable */ 1130 uint32_t i, count, tocLength, topOffset; 1131 1132 TempRow rows[STACK_ROW_CAPACITY]; 1133 uint16_t resort[STACK_ROW_CAPACITY]; 1134 TempAliasTable tempTable; 1135 1136 /* udata_swapDataHeader checks the arguments */ 1137 headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode); 1138 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { 1139 return 0; 1140 } 1141 1142 /* check data format and format version */ 1143 pInfo=(const UDataInfo *)((const char *)inData+4); 1144 if(!( 1145 pInfo->dataFormat[0]==0x43 && /* dataFormat="CvAl" */ 1146 pInfo->dataFormat[1]==0x76 && 1147 pInfo->dataFormat[2]==0x41 && 1148 pInfo->dataFormat[3]==0x6c && 1149 pInfo->formatVersion[0]==3 1150 )) { 1151 udata_printError(ds, "ucnv_swapAliases(): data format %02x.%02x.%02x.%02x (format version %02x) is not an alias table\n", 1152 pInfo->dataFormat[0], pInfo->dataFormat[1], 1153 pInfo->dataFormat[2], pInfo->dataFormat[3], 1154 pInfo->formatVersion[0]); 1155 *pErrorCode=U_UNSUPPORTED_ERROR; 1156 return 0; 1157 } 1158 1159 /* an alias table must contain at least the table of contents array */ 1160 if(length>=0 && (length-headerSize)<4*(1+minTocLength)) { 1161 udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n", 1162 length-headerSize); 1163 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; 1164 return 0; 1165 } 1166 1167 inSectionSizes=(const uint32_t *)((const char *)inData+headerSize); 1168 inTable=(const uint16_t *)inSectionSizes; 1169 uprv_memset(toc, 0, sizeof(toc)); 1170 toc[tocLengthIndex]=tocLength=ds->readUInt32(inSectionSizes[tocLengthIndex]); 1171 if(tocLength<minTocLength || offsetsCount<=tocLength) { 1172 udata_printError(ds, "ucnv_swapAliases(): table of contents contains unsupported number of sections (%u sections)\n", tocLength); 1173 *pErrorCode=U_INVALID_FORMAT_ERROR; 1174 return 0; 1175 } 1176 1177 /* read the known part of the table of contents */ 1178 for(i=converterListIndex; i<=tocLength; ++i) { 1179 toc[i]=ds->readUInt32(inSectionSizes[i]); 1180 } 1181 1182 /* compute offsets */ 1183 uprv_memset(offsets, 0, sizeof(offsets)); 1184 offsets[converterListIndex]=2*(1+tocLength); /* count two 16-bit units per toc entry */ 1185 for(i=tagListIndex; i<=tocLength; ++i) { 1186 offsets[i]=offsets[i-1]+toc[i-1]; 1187 } 1188 1189 /* compute the overall size of the after-header data, in numbers of 16-bit units */ 1190 topOffset=offsets[i-1]+toc[i-1]; 1191 1192 if(length>=0) { 1193 uint16_t *outTable; 1194 const uint16_t *p, *p2; 1195 uint16_t *q, *q2; 1196 uint16_t oldIndex; 1197 1198 if((length-headerSize)<(2*(int32_t)topOffset)) { 1199 udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n", 1200 length-headerSize); 1201 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; 1202 return 0; 1203 } 1204 1205 outTable=(uint16_t *)((char *)outData+headerSize); 1206 1207 /* swap the entire table of contents */ 1208 ds->swapArray32(ds, inTable, 4*(1+tocLength), outTable, pErrorCode); 1209 1210 /* swap unormalized strings & normalized strings */ 1211 ds->swapInvChars(ds, inTable+offsets[stringTableIndex], 2*(int32_t)(toc[stringTableIndex]+toc[normalizedStringTableIndex]), 1212 outTable+offsets[stringTableIndex], pErrorCode); 1213 if(U_FAILURE(*pErrorCode)) { 1214 udata_printError(ds, "ucnv_swapAliases().swapInvChars(charset names) failed\n"); 1215 return 0; 1216 } 1217 1218 if(ds->inCharset==ds->outCharset) { 1219 /* no need to sort, just swap all 16-bit values together */ 1220 ds->swapArray16(ds, 1221 inTable+offsets[converterListIndex], 1222 2*(int32_t)(offsets[stringTableIndex]-offsets[converterListIndex]), 1223 outTable+offsets[converterListIndex], 1224 pErrorCode); 1225 } else { 1226 /* allocate the temporary table for sorting */ 1227 count=toc[aliasListIndex]; 1228 1229 tempTable.chars=(const char *)(outTable+offsets[stringTableIndex]); /* sort by outCharset */ 1230 1231 if(count<=STACK_ROW_CAPACITY) { 1232 tempTable.rows=rows; 1233 tempTable.resort=resort; 1234 } else { 1235 tempTable.rows=(TempRow *)uprv_malloc(count*sizeof(TempRow)+count*2); 1236 if(tempTable.rows==NULL) { 1237 udata_printError(ds, "ucnv_swapAliases(): unable to allocate memory for sorting tables (max length: %u)\n", 1238 count); 1239 *pErrorCode=U_MEMORY_ALLOCATION_ERROR; 1240 return 0; 1241 } 1242 tempTable.resort=(uint16_t *)(tempTable.rows+count); 1243 } 1244 1245 if(ds->outCharset==U_ASCII_FAMILY) { 1246 tempTable.stripForCompare=ucnv_io_stripASCIIForCompare; 1247 } else /* U_EBCDIC_FAMILY */ { 1248 tempTable.stripForCompare=ucnv_io_stripEBCDICForCompare; 1249 } 1250 1251 /* 1252 * Sort unique aliases+mapped names. 1253 * 1254 * We need to sort the list again by outCharset strings because they 1255 * sort differently for different charset families. 1256 * First we set up a temporary table with the string indexes and 1257 * sorting indexes and sort that. 1258 * Then we permutate and copy/swap the actual values. 1259 */ 1260 p=inTable+offsets[aliasListIndex]; 1261 q=outTable+offsets[aliasListIndex]; 1262 1263 p2=inTable+offsets[untaggedConvArrayIndex]; 1264 q2=outTable+offsets[untaggedConvArrayIndex]; 1265 1266 for(i=0; i<count; ++i) { 1267 tempTable.rows[i].strIndex=ds->readUInt16(p[i]); 1268 tempTable.rows[i].sortIndex=(uint16_t)i; 1269 } 1270 1271 uprv_sortArray(tempTable.rows, (int32_t)count, sizeof(TempRow), 1272 io_compareRows, &tempTable, 1273 FALSE, pErrorCode); 1274 1275 if(U_SUCCESS(*pErrorCode)) { 1276 /* copy/swap/permutate items */ 1277 if(p!=q) { 1278 for(i=0; i<count; ++i) { 1279 oldIndex=tempTable.rows[i].sortIndex; 1280 ds->swapArray16(ds, p+oldIndex, 2, q+i, pErrorCode); 1281 ds->swapArray16(ds, p2+oldIndex, 2, q2+i, pErrorCode); 1282 } 1283 } else { 1284 /* 1285 * If we swap in-place, then the permutation must use another 1286 * temporary array (tempTable.resort) 1287 * before the results are copied to the outBundle. 1288 */ 1289 uint16_t *r=tempTable.resort; 1290 1291 for(i=0; i<count; ++i) { 1292 oldIndex=tempTable.rows[i].sortIndex; 1293 ds->swapArray16(ds, p+oldIndex, 2, r+i, pErrorCode); 1294 } 1295 uprv_memcpy(q, r, 2*count); 1296 1297 for(i=0; i<count; ++i) { 1298 oldIndex=tempTable.rows[i].sortIndex; 1299 ds->swapArray16(ds, p2+oldIndex, 2, r+i, pErrorCode); 1300 } 1301 uprv_memcpy(q2, r, 2*count); 1302 } 1303 } 1304 1305 if(tempTable.rows!=rows) { 1306 uprv_free(tempTable.rows); 1307 } 1308 1309 if(U_FAILURE(*pErrorCode)) { 1310 udata_printError(ds, "ucnv_swapAliases().uprv_sortArray(%u items) failed\n", 1311 count); 1312 return 0; 1313 } 1314 1315 /* swap remaining 16-bit values */ 1316 ds->swapArray16(ds, 1317 inTable+offsets[converterListIndex], 1318 2*(int32_t)(offsets[aliasListIndex]-offsets[converterListIndex]), 1319 outTable+offsets[converterListIndex], 1320 pErrorCode); 1321 ds->swapArray16(ds, 1322 inTable+offsets[taggedAliasArrayIndex], 1323 2*(int32_t)(offsets[stringTableIndex]-offsets[taggedAliasArrayIndex]), 1324 outTable+offsets[taggedAliasArrayIndex], 1325 pErrorCode); 1326 } 1327 } 1328 1329 return headerSize+2*(int32_t)topOffset; 1330 } 1331 1332 #endif 1333 1334 /* 1335 * Hey, Emacs, please set the following: 1336 * 1337 * Local Variables: 1338 * indent-tabs-mode: nil 1339 * End: 1340 * 1341 */ 1342
