1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ****************************************************************************** |
5 | * |
6 | * Copyright (C) 1999-2015, International Business Machines |
7 | * Corporation and others. All Rights Reserved. |
8 | * |
9 | ****************************************************************************** |
10 | * |
11 | * |
12 | * ucnv_io.cpp: |
13 | * initializes global variables and defines functions pertaining to converter |
14 | * name resolution aspect of the conversion code. |
15 | * |
16 | * new implementation: |
17 | * |
18 | * created on: 1999nov22 |
19 | * created by: Markus W. Scherer |
20 | * |
21 | * Use the binary cnvalias.icu (created from convrtrs.txt) to work |
22 | * with aliases for converter names. |
23 | * |
24 | * Date Name Description |
25 | * 11/22/1999 markus Created |
26 | * 06/28/2002 grhoten Major overhaul of the converter alias design. |
27 | * Now an alias can map to different converters |
28 | * depending on the specified standard. |
29 | ******************************************************************************* |
30 | */ |
31 | |
32 | #include "unicode/utypes.h" |
33 | |
34 | #if !UCONFIG_NO_CONVERSION |
35 | |
36 | #include "unicode/ucnv.h" |
37 | #include "unicode/udata.h" |
38 | |
39 | #include "umutex.h" |
40 | #include "uarrsort.h" |
41 | #include "uassert.h" |
42 | #include "udataswp.h" |
43 | #include "cstring.h" |
44 | #include "cmemory.h" |
45 | #include "ucnv_io.h" |
46 | #include "uenumimp.h" |
47 | #include "ucln_cmn.h" |
48 | |
49 | /* Format of cnvalias.icu ----------------------------------------------------- |
50 | * |
51 | * cnvalias.icu is a binary, memory-mappable form of convrtrs.txt. |
52 | * This binary form contains several tables. All indexes are to uint16_t |
53 | * units, and not to the bytes (uint8_t units). Addressing everything on |
54 | * 16-bit boundaries allows us to store more information with small index |
55 | * numbers, which are also 16-bit in size. The majority of the table (except |
56 | * the string table) are 16-bit numbers. |
57 | * |
58 | * First there is the size of the Table of Contents (TOC). The TOC |
59 | * entries contain the size of each section. In order to find the offset |
60 | * you just need to sum up the previous offsets. |
61 | * The TOC length and entries are an array of uint32_t values. |
62 | * The first section after the TOC starts immediately after the TOC. |
63 | * |
64 | * 1) This section contains a list of converters. This list contains indexes |
65 | * into the string table for the converter name. The index of this list is |
66 | * also used by other sections, which are mentioned later on. |
67 | * This list is not sorted. |
68 | * |
69 | * 2) This section contains a list of tags. This list contains indexes |
70 | * into the string table for the tag name. The index of this list is |
71 | * also used by other sections, which are mentioned later on. |
72 | * This list is in priority order of standards. |
73 | * |
74 | * 3) This section contains a list of sorted unique aliases. This |
75 | * list contains indexes into the string table for the alias name. The |
76 | * index of this list is also used by other sections, like the 4th section. |
77 | * The index for the 3rd and 4th section is used to get the |
78 | * alias -> converter name mapping. Section 3 and 4 form a two column table. |
79 | * Some of the most significant bits of each index may contain other |
80 | * information (see findConverter for details). |
81 | * |
82 | * 4) This section contains a list of mapped converter names. Consider this |
83 | * as a table that maps the 3rd section to the 1st section. This list contains |
84 | * indexes into the 1st section. The index of this list is the same index in |
85 | * the 3rd section. There is also some extra information in the high bits of |
86 | * each converter index in this table. Currently it's only used to say that |
87 | * an alias mapped to this converter is ambiguous. See UCNV_CONVERTER_INDEX_MASK |
88 | * and UCNV_AMBIGUOUS_ALIAS_MAP_BIT for more information. This section is |
89 | * the predigested form of the 5th section so that an alias lookup can be fast. |
90 | * |
91 | * 5) This section contains a 2D array with indexes to the 6th section. This |
92 | * section is the full form of all alias mappings. The column index is the |
93 | * index into the converter list (column header). The row index is the index |
94 | * to tag list (row header). This 2D array is the top part a 3D array. The |
95 | * third dimension is in the 6th section. |
96 | * |
97 | * 6) This is blob of variable length arrays. Each array starts with a size, |
98 | * and is followed by indexes to alias names in the string table. This is |
99 | * the third dimension to the section 5. No other section should be referencing |
100 | * this section. |
101 | * |
102 | * 7) Starting in ICU 3.6, this can be a UConverterAliasOptions struct. Its |
103 | * presence indicates that a section 9 exists. UConverterAliasOptions specifies |
104 | * what type of string normalization is used among other potential things in the |
105 | * future. |
106 | * |
107 | * 8) This is the string table. All strings are indexed on an even address. |
108 | * There are two reasons for this. First many chip architectures locate strings |
109 | * faster on even address boundaries. Second, since all indexes are 16-bit |
110 | * numbers, this string table can be 128KB in size instead of 64KB when we |
111 | * only have strings starting on an even address. |
112 | * |
113 | * 9) When present this is a set of prenormalized strings from section 8. This |
114 | * table contains normalized strings with the dashes and spaces stripped out, |
115 | * and all strings lowercased. In the future, the options in section 7 may state |
116 | * other types of normalization. |
117 | * |
118 | * Here is the concept of section 5 and 6. It's a 3D cube. Each tag |
119 | * has a unique alias among all converters. That same alias can |
120 | * be mentioned in other standards on different converters, |
121 | * but only one alias per tag can be unique. |
122 | * |
123 | * |
124 | * Converter Names (Usually in TR22 form) |
125 | * -------------------------------------------. |
126 | * T / /| |
127 | * a / / | |
128 | * g / / | |
129 | * s / / | |
130 | * / / | |
131 | * ------------------------------------------/ | |
132 | * A | | | |
133 | * l | | | |
134 | * i | | / |
135 | * a | | / |
136 | * s | | / |
137 | * e | | / |
138 | * s | |/ |
139 | * ------------------------------------------- |
140 | * |
141 | * |
142 | * |
143 | * Here is what it really looks like. It's like swiss cheese. |
144 | * There are holes. Some converters aren't recognized by |
145 | * a standard, or they are really old converters that the |
146 | * standard doesn't recognize anymore. |
147 | * |
148 | * Converter Names (Usually in TR22 form) |
149 | * -------------------------------------------. |
150 | * T /##########################################/| |
151 | * a / # # /# |
152 | * g / # ## ## ### # ### ### ### #/ |
153 | * s / # ##### #### ## ## #/# |
154 | * / ### # # ## # # # ### # # #/## |
155 | * ------------------------------------------/# # |
156 | * A |### # # ## # # # ### # # #|# # |
157 | * l |# # # # # ## # #|# # |
158 | * i |# # # # # # #|# |
159 | * a |# #|# |
160 | * s | #|# |
161 | * e |
162 | * s |
163 | * |
164 | */ |
165 | |
166 | /** |
167 | * Used by the UEnumeration API |
168 | */ |
169 | typedef struct UAliasContext { |
170 | uint32_t listOffset; |
171 | uint32_t listIdx; |
172 | } UAliasContext; |
173 | |
174 | static const char DATA_NAME[] = "cnvalias" ; |
175 | static const char DATA_TYPE[] = "icu" ; |
176 | |
177 | static UDataMemory *gAliasData=nullptr; |
178 | static icu::UInitOnce gAliasDataInitOnce {}; |
179 | |
180 | enum { |
181 | tocLengthIndex=0, |
182 | converterListIndex=1, |
183 | tagListIndex=2, |
184 | aliasListIndex=3, |
185 | untaggedConvArrayIndex=4, |
186 | taggedAliasArrayIndex=5, |
187 | taggedAliasListsIndex=6, |
188 | tableOptionsIndex=7, |
189 | stringTableIndex=8, |
190 | normalizedStringTableIndex=9, |
191 | offsetsCount, /* length of the swapper's temporary offsets[] */ |
192 | minTocLength=8 /* min. tocLength in the file, does not count the tocLengthIndex! */ |
193 | }; |
194 | |
195 | static const UConverterAliasOptions defaultTableOptions = { |
196 | UCNV_IO_UNNORMALIZED, |
197 | 0 /* containsCnvOptionInfo */ |
198 | }; |
199 | static UConverterAlias gMainTable; |
200 | |
201 | #define GET_STRING(idx) (const char *)(gMainTable.stringTable + (idx)) |
202 | #define GET_NORMALIZED_STRING(idx) (const char *)(gMainTable.normalizedStringTable + (idx)) |
203 | |
204 | static UBool U_CALLCONV |
205 | isAcceptable(void * /*context*/, |
206 | const char * /*type*/, const char * /*name*/, |
207 | const UDataInfo *pInfo) { |
208 | return (UBool)( |
209 | pInfo->size>=20 && |
210 | pInfo->isBigEndian==U_IS_BIG_ENDIAN && |
211 | pInfo->charsetFamily==U_CHARSET_FAMILY && |
212 | pInfo->dataFormat[0]==0x43 && /* dataFormat="CvAl" */ |
213 | pInfo->dataFormat[1]==0x76 && |
214 | pInfo->dataFormat[2]==0x41 && |
215 | pInfo->dataFormat[3]==0x6c && |
216 | pInfo->formatVersion[0]==3); |
217 | } |
218 | |
219 | static UBool U_CALLCONV ucnv_io_cleanup() |
220 | { |
221 | if (gAliasData) { |
222 | udata_close(gAliasData); |
223 | gAliasData = nullptr; |
224 | } |
225 | gAliasDataInitOnce.reset(); |
226 | |
227 | uprv_memset(&gMainTable, 0, sizeof(gMainTable)); |
228 | |
229 | return true; /* Everything was cleaned up */ |
230 | } |
231 | |
232 | static void U_CALLCONV initAliasData(UErrorCode &errCode) { |
233 | UDataMemory *data; |
234 | const uint16_t *table; |
235 | const uint32_t *sectionSizes; |
236 | uint32_t tableStart; |
237 | uint32_t currOffset; |
238 | |
239 | ucln_common_registerCleanup(UCLN_COMMON_UCNV_IO, ucnv_io_cleanup); |
240 | |
241 | U_ASSERT(gAliasData == nullptr); |
242 | data = udata_openChoice(nullptr, DATA_TYPE, DATA_NAME, isAcceptable, nullptr, &errCode); |
243 | if(U_FAILURE(errCode)) { |
244 | return; |
245 | } |
246 | |
247 | sectionSizes = (const uint32_t *)udata_getMemory(data); |
248 | table = (const uint16_t *)sectionSizes; |
249 | |
250 | tableStart = sectionSizes[0]; |
251 | if (tableStart < minTocLength) { |
252 | errCode = U_INVALID_FORMAT_ERROR; |
253 | udata_close(data); |
254 | return; |
255 | } |
256 | gAliasData = data; |
257 | |
258 | gMainTable.converterListSize = sectionSizes[1]; |
259 | gMainTable.tagListSize = sectionSizes[2]; |
260 | gMainTable.aliasListSize = sectionSizes[3]; |
261 | gMainTable.untaggedConvArraySize = sectionSizes[4]; |
262 | gMainTable.taggedAliasArraySize = sectionSizes[5]; |
263 | gMainTable.taggedAliasListsSize = sectionSizes[6]; |
264 | gMainTable.optionTableSize = sectionSizes[7]; |
265 | gMainTable.stringTableSize = sectionSizes[8]; |
266 | |
267 | if (tableStart > 8) { |
268 | gMainTable.normalizedStringTableSize = sectionSizes[9]; |
269 | } |
270 | |
271 | currOffset = tableStart * (sizeof(uint32_t)/sizeof(uint16_t)) + (sizeof(uint32_t)/sizeof(uint16_t)); |
272 | gMainTable.converterList = table + currOffset; |
273 | |
274 | currOffset += gMainTable.converterListSize; |
275 | gMainTable.tagList = table + currOffset; |
276 | |
277 | currOffset += gMainTable.tagListSize; |
278 | gMainTable.aliasList = table + currOffset; |
279 | |
280 | currOffset += gMainTable.aliasListSize; |
281 | gMainTable.untaggedConvArray = table + currOffset; |
282 | |
283 | currOffset += gMainTable.untaggedConvArraySize; |
284 | gMainTable.taggedAliasArray = table + currOffset; |
285 | |
286 | /* aliasLists is a 1's based array, but it has a padding character */ |
287 | currOffset += gMainTable.taggedAliasArraySize; |
288 | gMainTable.taggedAliasLists = table + currOffset; |
289 | |
290 | currOffset += gMainTable.taggedAliasListsSize; |
291 | if (gMainTable.optionTableSize > 0 |
292 | && ((const UConverterAliasOptions *)(table + currOffset))->stringNormalizationType < UCNV_IO_NORM_TYPE_COUNT) |
293 | { |
294 | /* Faster table */ |
295 | gMainTable.optionTable = (const UConverterAliasOptions *)(table + currOffset); |
296 | } |
297 | else { |
298 | /* Smaller table, or I can't handle this normalization mode! |
299 | Use the original slower table lookup. */ |
300 | gMainTable.optionTable = &defaultTableOptions; |
301 | } |
302 | |
303 | currOffset += gMainTable.optionTableSize; |
304 | gMainTable.stringTable = table + currOffset; |
305 | |
306 | currOffset += gMainTable.stringTableSize; |
307 | gMainTable.normalizedStringTable = ((gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED) |
308 | ? gMainTable.stringTable : (table + currOffset)); |
309 | } |
310 | |
311 | |
312 | static UBool |
313 | haveAliasData(UErrorCode *pErrorCode) { |
314 | umtx_initOnce(gAliasDataInitOnce, &initAliasData, *pErrorCode); |
315 | return U_SUCCESS(*pErrorCode); |
316 | } |
317 | |
318 | static inline UBool |
319 | isAlias(const char *alias, UErrorCode *pErrorCode) { |
320 | if(alias==nullptr) { |
321 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
322 | return false; |
323 | } |
324 | return (UBool)(*alias!=0); |
325 | } |
326 | |
327 | static uint32_t getTagNumber(const char *tagname) { |
328 | if (gMainTable.tagList) { |
329 | uint32_t tagNum; |
330 | for (tagNum = 0; tagNum < gMainTable.tagListSize; tagNum++) { |
331 | if (!uprv_stricmp(GET_STRING(gMainTable.tagList[tagNum]), tagname)) { |
332 | return tagNum; |
333 | } |
334 | } |
335 | } |
336 | |
337 | return UINT32_MAX; |
338 | } |
339 | |
340 | /* character types relevant for ucnv_compareNames() */ |
341 | enum { |
342 | UIGNORE, |
343 | ZERO, |
344 | NONZERO, |
345 | MINLETTER /* any values from here on are lowercase letter mappings */ |
346 | }; |
347 | |
348 | /* character types for ASCII 00..7F */ |
349 | static const uint8_t asciiTypes[128] = { |
350 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
351 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
352 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
353 | ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0, |
354 | 0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, |
355 | 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0, |
356 | 0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, |
357 | 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0 |
358 | }; |
359 | |
360 | #define GET_ASCII_TYPE(c) ((int8_t)(c) >= 0 ? asciiTypes[(uint8_t)c] : (uint8_t)UIGNORE) |
361 | |
362 | /* character types for EBCDIC 80..FF */ |
363 | static const uint8_t ebcdicTypes[128] = { |
364 | 0, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0, |
365 | 0, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0, |
366 | 0, 0, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0, |
367 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
368 | 0, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0, |
369 | 0, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0, |
370 | 0, 0, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0, |
371 | ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0 |
372 | }; |
373 | |
374 | #define GET_EBCDIC_TYPE(c) ((int8_t)(c) < 0 ? ebcdicTypes[(c)&0x7f] : (uint8_t)UIGNORE) |
375 | |
376 | #if U_CHARSET_FAMILY==U_ASCII_FAMILY |
377 | # define GET_CHAR_TYPE(c) GET_ASCII_TYPE(c) |
378 | #elif U_CHARSET_FAMILY==U_EBCDIC_FAMILY |
379 | # define GET_CHAR_TYPE(c) GET_EBCDIC_TYPE(c) |
380 | #else |
381 | # error U_CHARSET_FAMILY is not valid |
382 | #endif |
383 | |
384 | |
385 | /* @see ucnv_compareNames */ |
386 | U_CAPI char * U_CALLCONV |
387 | ucnv_io_stripASCIIForCompare(char *dst, const char *name) { |
388 | char *dstItr = dst; |
389 | uint8_t type, nextType; |
390 | char c1; |
391 | UBool afterDigit = false; |
392 | |
393 | while ((c1 = *name++) != 0) { |
394 | type = GET_ASCII_TYPE(c1); |
395 | switch (type) { |
396 | case UIGNORE: |
397 | afterDigit = false; |
398 | continue; /* ignore all but letters and digits */ |
399 | case ZERO: |
400 | if (!afterDigit) { |
401 | nextType = GET_ASCII_TYPE(*name); |
402 | if (nextType == ZERO || nextType == NONZERO) { |
403 | continue; /* ignore leading zero before another digit */ |
404 | } |
405 | } |
406 | break; |
407 | case NONZERO: |
408 | afterDigit = true; |
409 | break; |
410 | default: |
411 | c1 = (char)type; /* lowercased letter */ |
412 | afterDigit = false; |
413 | break; |
414 | } |
415 | *dstItr++ = c1; |
416 | } |
417 | *dstItr = 0; |
418 | return dst; |
419 | } |
420 | |
421 | U_CAPI char * U_CALLCONV |
422 | ucnv_io_stripEBCDICForCompare(char *dst, const char *name) { |
423 | char *dstItr = dst; |
424 | uint8_t type, nextType; |
425 | char c1; |
426 | UBool afterDigit = false; |
427 | |
428 | while ((c1 = *name++) != 0) { |
429 | type = GET_EBCDIC_TYPE(c1); |
430 | switch (type) { |
431 | case UIGNORE: |
432 | afterDigit = false; |
433 | continue; /* ignore all but letters and digits */ |
434 | case ZERO: |
435 | if (!afterDigit) { |
436 | nextType = GET_EBCDIC_TYPE(*name); |
437 | if (nextType == ZERO || nextType == NONZERO) { |
438 | continue; /* ignore leading zero before another digit */ |
439 | } |
440 | } |
441 | break; |
442 | case NONZERO: |
443 | afterDigit = true; |
444 | break; |
445 | default: |
446 | c1 = (char)type; /* lowercased letter */ |
447 | afterDigit = false; |
448 | break; |
449 | } |
450 | *dstItr++ = c1; |
451 | } |
452 | *dstItr = 0; |
453 | return dst; |
454 | } |
455 | |
456 | /** |
457 | * Do a fuzzy compare of two converter/alias names. |
458 | * The comparison is case-insensitive, ignores leading zeroes if they are not |
459 | * followed by further digits, and ignores all but letters and digits. |
460 | * Thus the strings "UTF-8", "utf_8", "u*T@f08" and "Utf 8" are exactly equivalent. |
461 | * See section 1.4, Charset Alias Matching in Unicode Technical Standard #22 |
462 | * at http://www.unicode.org/reports/tr22/ |
463 | * |
464 | * This is a symmetrical (commutative) operation; order of arguments |
465 | * is insignificant. This is an important property for sorting the |
466 | * list (when the list is preprocessed into binary form) and for |
467 | * performing binary searches on it at run time. |
468 | * |
469 | * @param name1 a converter name or alias, zero-terminated |
470 | * @param name2 a converter name or alias, zero-terminated |
471 | * @return 0 if the names match, or a negative value if the name1 |
472 | * lexically precedes name2, or a positive value if the name1 |
473 | * lexically follows name2. |
474 | * |
475 | * @see ucnv_io_stripForCompare |
476 | */ |
477 | U_CAPI int U_EXPORT2 |
478 | ucnv_compareNames(const char *name1, const char *name2) { |
479 | int rc; |
480 | uint8_t type, nextType; |
481 | char c1, c2; |
482 | UBool afterDigit1 = false, afterDigit2 = false; |
483 | |
484 | for (;;) { |
485 | while ((c1 = *name1++) != 0) { |
486 | type = GET_CHAR_TYPE(c1); |
487 | switch (type) { |
488 | case UIGNORE: |
489 | afterDigit1 = false; |
490 | continue; /* ignore all but letters and digits */ |
491 | case ZERO: |
492 | if (!afterDigit1) { |
493 | nextType = GET_CHAR_TYPE(*name1); |
494 | if (nextType == ZERO || nextType == NONZERO) { |
495 | continue; /* ignore leading zero before another digit */ |
496 | } |
497 | } |
498 | break; |
499 | case NONZERO: |
500 | afterDigit1 = true; |
501 | break; |
502 | default: |
503 | c1 = (char)type; /* lowercased letter */ |
504 | afterDigit1 = false; |
505 | break; |
506 | } |
507 | break; /* deliver c1 */ |
508 | } |
509 | while ((c2 = *name2++) != 0) { |
510 | type = GET_CHAR_TYPE(c2); |
511 | switch (type) { |
512 | case UIGNORE: |
513 | afterDigit2 = false; |
514 | continue; /* ignore all but letters and digits */ |
515 | case ZERO: |
516 | if (!afterDigit2) { |
517 | nextType = GET_CHAR_TYPE(*name2); |
518 | if (nextType == ZERO || nextType == NONZERO) { |
519 | continue; /* ignore leading zero before another digit */ |
520 | } |
521 | } |
522 | break; |
523 | case NONZERO: |
524 | afterDigit2 = true; |
525 | break; |
526 | default: |
527 | c2 = (char)type; /* lowercased letter */ |
528 | afterDigit2 = false; |
529 | break; |
530 | } |
531 | break; /* deliver c2 */ |
532 | } |
533 | |
534 | /* If we reach the ends of both strings then they match */ |
535 | if ((c1|c2)==0) { |
536 | return 0; |
537 | } |
538 | |
539 | /* Case-insensitive comparison */ |
540 | rc = (int)(unsigned char)c1 - (int)(unsigned char)c2; |
541 | if (rc != 0) { |
542 | return rc; |
543 | } |
544 | } |
545 | } |
546 | |
547 | /* |
548 | * search for an alias |
549 | * return the converter number index for gConverterList |
550 | */ |
551 | static inline uint32_t |
552 | findConverter(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) { |
553 | uint32_t mid, start, limit; |
554 | uint32_t lastMid; |
555 | int result; |
556 | int isUnnormalized = (gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED); |
557 | char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH]; |
558 | |
559 | if (!isUnnormalized) { |
560 | if (uprv_strlen(alias) >= UCNV_MAX_CONVERTER_NAME_LENGTH) { |
561 | *pErrorCode = U_BUFFER_OVERFLOW_ERROR; |
562 | return UINT32_MAX; |
563 | } |
564 | |
565 | /* Lower case and remove ignoreable characters. */ |
566 | ucnv_io_stripForCompare(strippedName, alias); |
567 | alias = strippedName; |
568 | } |
569 | |
570 | /* do a binary search for the alias */ |
571 | start = 0; |
572 | limit = gMainTable.untaggedConvArraySize; |
573 | mid = limit; |
574 | lastMid = UINT32_MAX; |
575 | |
576 | for (;;) { |
577 | mid = (uint32_t)((start + limit) / 2); |
578 | if (lastMid == mid) { /* Have we moved? */ |
579 | break; /* We haven't moved, and it wasn't found. */ |
580 | } |
581 | lastMid = mid; |
582 | if (isUnnormalized) { |
583 | result = ucnv_compareNames(alias, GET_STRING(gMainTable.aliasList[mid])); |
584 | } |
585 | else { |
586 | result = uprv_strcmp(alias, GET_NORMALIZED_STRING(gMainTable.aliasList[mid])); |
587 | } |
588 | |
589 | if (result < 0) { |
590 | limit = mid; |
591 | } else if (result > 0) { |
592 | start = mid; |
593 | } else { |
594 | /* Since the gencnval tool folds duplicates into one entry, |
595 | * this alias in gAliasList is unique, but different standards |
596 | * may map an alias to different converters. |
597 | */ |
598 | if (gMainTable.untaggedConvArray[mid] & UCNV_AMBIGUOUS_ALIAS_MAP_BIT) { |
599 | *pErrorCode = U_AMBIGUOUS_ALIAS_WARNING; |
600 | } |
601 | /* State whether the canonical converter name contains an option. |
602 | This information is contained in this list in order to maintain backward & forward compatibility. */ |
603 | if (containsOption) { |
604 | UBool containsCnvOptionInfo = (UBool)gMainTable.optionTable->containsCnvOptionInfo; |
605 | *containsOption = (UBool)((containsCnvOptionInfo |
606 | && ((gMainTable.untaggedConvArray[mid] & UCNV_CONTAINS_OPTION_BIT) != 0)) |
607 | || !containsCnvOptionInfo); |
608 | } |
609 | return gMainTable.untaggedConvArray[mid] & UCNV_CONVERTER_INDEX_MASK; |
610 | } |
611 | } |
612 | |
613 | return UINT32_MAX; |
614 | } |
615 | |
616 | /* |
617 | * Is this alias in this list? |
618 | * alias and listOffset should be non-nullptr. |
619 | */ |
620 | static inline UBool |
621 | isAliasInList(const char *alias, uint32_t listOffset) { |
622 | if (listOffset) { |
623 | uint32_t currAlias; |
624 | uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; |
625 | /* +1 to skip listCount */ |
626 | const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; |
627 | for (currAlias = 0; currAlias < listCount; currAlias++) { |
628 | if (currList[currAlias] |
629 | && ucnv_compareNames(alias, GET_STRING(currList[currAlias]))==0) |
630 | { |
631 | return true; |
632 | } |
633 | } |
634 | } |
635 | return false; |
636 | } |
637 | |
638 | /* |
639 | * Search for an standard name of an alias (what is the default name |
640 | * that this standard uses?) |
641 | * return the listOffset for gTaggedAliasLists. If it's 0, |
642 | * the it couldn't be found, but the parameters are valid. |
643 | */ |
644 | static uint32_t |
645 | findTaggedAliasListsOffset(const char *alias, const char *standard, UErrorCode *pErrorCode) { |
646 | uint32_t idx; |
647 | uint32_t listOffset; |
648 | uint32_t convNum; |
649 | UErrorCode myErr = U_ZERO_ERROR; |
650 | uint32_t tagNum = getTagNumber(standard); |
651 | |
652 | /* Make a quick guess. Hopefully they used a TR22 canonical alias. */ |
653 | convNum = findConverter(alias, nullptr, &myErr); |
654 | if (myErr != U_ZERO_ERROR) { |
655 | *pErrorCode = myErr; |
656 | } |
657 | |
658 | if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) { |
659 | listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; |
660 | if (listOffset && gMainTable.taggedAliasLists[listOffset + 1]) { |
661 | return listOffset; |
662 | } |
663 | if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { |
664 | /* Uh Oh! They used an ambiguous alias. |
665 | We have to search the whole swiss cheese starting |
666 | at the highest standard affinity. |
667 | This may take a while. |
668 | */ |
669 | for (idx = 0; idx < gMainTable.taggedAliasArraySize; idx++) { |
670 | listOffset = gMainTable.taggedAliasArray[idx]; |
671 | if (listOffset && isAliasInList(alias, listOffset)) { |
672 | uint32_t currTagNum = idx/gMainTable.converterListSize; |
673 | uint32_t currConvNum = (idx - currTagNum*gMainTable.converterListSize); |
674 | uint32_t tempListOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + currConvNum]; |
675 | if (tempListOffset && gMainTable.taggedAliasLists[tempListOffset + 1]) { |
676 | return tempListOffset; |
677 | } |
678 | /* else keep on looking */ |
679 | /* We could speed this up by starting on the next row |
680 | because an alias is unique per row, right now. |
681 | This would change if alias versioning appears. */ |
682 | } |
683 | } |
684 | /* The standard doesn't know about the alias */ |
685 | } |
686 | /* else no default name */ |
687 | return 0; |
688 | } |
689 | /* else converter or tag not found */ |
690 | |
691 | return UINT32_MAX; |
692 | } |
693 | |
694 | /* Return the canonical name */ |
695 | static uint32_t |
696 | findTaggedConverterNum(const char *alias, const char *standard, UErrorCode *pErrorCode) { |
697 | uint32_t idx; |
698 | uint32_t listOffset; |
699 | uint32_t convNum; |
700 | UErrorCode myErr = U_ZERO_ERROR; |
701 | uint32_t tagNum = getTagNumber(standard); |
702 | |
703 | /* Make a quick guess. Hopefully they used a TR22 canonical alias. */ |
704 | convNum = findConverter(alias, nullptr, &myErr); |
705 | if (myErr != U_ZERO_ERROR) { |
706 | *pErrorCode = myErr; |
707 | } |
708 | |
709 | if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) { |
710 | listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; |
711 | if (listOffset && isAliasInList(alias, listOffset)) { |
712 | return convNum; |
713 | } |
714 | if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { |
715 | /* Uh Oh! They used an ambiguous alias. |
716 | We have to search one slice of the swiss cheese. |
717 | We search only in the requested tag, not the whole thing. |
718 | This may take a while. |
719 | */ |
720 | uint32_t convStart = (tagNum)*gMainTable.converterListSize; |
721 | uint32_t convLimit = (tagNum+1)*gMainTable.converterListSize; |
722 | for (idx = convStart; idx < convLimit; idx++) { |
723 | listOffset = gMainTable.taggedAliasArray[idx]; |
724 | if (listOffset && isAliasInList(alias, listOffset)) { |
725 | return idx-convStart; |
726 | } |
727 | } |
728 | /* The standard doesn't know about the alias */ |
729 | } |
730 | /* else no canonical name */ |
731 | } |
732 | /* else converter or tag not found */ |
733 | |
734 | return UINT32_MAX; |
735 | } |
736 | |
737 | U_CAPI const char * |
738 | ucnv_io_getConverterName(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) { |
739 | const char *aliasTmp = alias; |
740 | int32_t i = 0; |
741 | for (i = 0; i < 2; i++) { |
742 | if (i == 1) { |
743 | /* |
744 | * After the first unsuccess converter lookup, check to see if |
745 | * the name begins with 'x-'. If it does, strip it off and try |
746 | * again. This behaviour is similar to how ICU4J does it. |
747 | */ |
748 | if (aliasTmp[0] == 'x' && aliasTmp[1] == '-') { |
749 | aliasTmp = aliasTmp+2; |
750 | } else { |
751 | break; |
752 | } |
753 | } |
754 | if(haveAliasData(pErrorCode) && isAlias(aliasTmp, pErrorCode)) { |
755 | uint32_t convNum = findConverter(aliasTmp, containsOption, pErrorCode); |
756 | if (convNum < gMainTable.converterListSize) { |
757 | return GET_STRING(gMainTable.converterList[convNum]); |
758 | } |
759 | /* else converter not found */ |
760 | } else { |
761 | break; |
762 | } |
763 | } |
764 | |
765 | return nullptr; |
766 | } |
767 | |
768 | U_CDECL_BEGIN |
769 | |
770 | |
771 | static int32_t U_CALLCONV |
772 | ucnv_io_countStandardAliases(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) { |
773 | int32_t value = 0; |
774 | UAliasContext *myContext = (UAliasContext *)(enumerator->context); |
775 | uint32_t listOffset = myContext->listOffset; |
776 | |
777 | if (listOffset) { |
778 | value = gMainTable.taggedAliasLists[listOffset]; |
779 | } |
780 | return value; |
781 | } |
782 | |
783 | static const char * U_CALLCONV |
784 | ucnv_io_nextStandardAliases(UEnumeration *enumerator, |
785 | int32_t* resultLength, |
786 | UErrorCode * /*pErrorCode*/) |
787 | { |
788 | UAliasContext *myContext = (UAliasContext *)(enumerator->context); |
789 | uint32_t listOffset = myContext->listOffset; |
790 | |
791 | if (listOffset) { |
792 | uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; |
793 | const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; |
794 | |
795 | if (myContext->listIdx < listCount) { |
796 | const char *myStr = GET_STRING(currList[myContext->listIdx++]); |
797 | if (resultLength) { |
798 | *resultLength = (int32_t)uprv_strlen(myStr); |
799 | } |
800 | return myStr; |
801 | } |
802 | } |
803 | /* Either we accessed a zero length list, or we enumerated too far. */ |
804 | if (resultLength) { |
805 | *resultLength = 0; |
806 | } |
807 | return nullptr; |
808 | } |
809 | |
810 | static void U_CALLCONV |
811 | ucnv_io_resetStandardAliases(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) { |
812 | ((UAliasContext *)(enumerator->context))->listIdx = 0; |
813 | } |
814 | |
815 | static void U_CALLCONV |
816 | ucnv_io_closeUEnumeration(UEnumeration *enumerator) { |
817 | uprv_free(enumerator->context); |
818 | uprv_free(enumerator); |
819 | } |
820 | |
821 | U_CDECL_END |
822 | |
823 | /* Enumerate the aliases for the specified converter and standard tag */ |
824 | static const UEnumeration gEnumAliases = { |
825 | nullptr, |
826 | nullptr, |
827 | ucnv_io_closeUEnumeration, |
828 | ucnv_io_countStandardAliases, |
829 | uenum_unextDefault, |
830 | ucnv_io_nextStandardAliases, |
831 | ucnv_io_resetStandardAliases |
832 | }; |
833 | |
834 | U_CAPI UEnumeration * U_EXPORT2 |
835 | ucnv_openStandardNames(const char *convName, |
836 | const char *standard, |
837 | UErrorCode *pErrorCode) |
838 | { |
839 | UEnumeration *myEnum = nullptr; |
840 | if (haveAliasData(pErrorCode) && isAlias(convName, pErrorCode)) { |
841 | uint32_t listOffset = findTaggedAliasListsOffset(convName, standard, pErrorCode); |
842 | |
843 | /* When listOffset == 0, we want to acknowledge that the |
844 | converter name and standard are okay, but there |
845 | is nothing to enumerate. */ |
846 | if (listOffset < gMainTable.taggedAliasListsSize) { |
847 | UAliasContext *myContext; |
848 | |
849 | myEnum = static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration))); |
850 | if (myEnum == nullptr) { |
851 | *pErrorCode = U_MEMORY_ALLOCATION_ERROR; |
852 | return nullptr; |
853 | } |
854 | uprv_memcpy(myEnum, &gEnumAliases, sizeof(UEnumeration)); |
855 | myContext = static_cast<UAliasContext *>(uprv_malloc(sizeof(UAliasContext))); |
856 | if (myContext == nullptr) { |
857 | *pErrorCode = U_MEMORY_ALLOCATION_ERROR; |
858 | uprv_free(myEnum); |
859 | return nullptr; |
860 | } |
861 | myContext->listOffset = listOffset; |
862 | myContext->listIdx = 0; |
863 | myEnum->context = myContext; |
864 | } |
865 | /* else converter or tag not found */ |
866 | } |
867 | return myEnum; |
868 | } |
869 | |
870 | static uint16_t |
871 | ucnv_io_countAliases(const char *alias, UErrorCode *pErrorCode) { |
872 | if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { |
873 | uint32_t convNum = findConverter(alias, nullptr, pErrorCode); |
874 | if (convNum < gMainTable.converterListSize) { |
875 | /* tagListNum - 1 is the ALL tag */ |
876 | int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; |
877 | |
878 | if (listOffset) { |
879 | return gMainTable.taggedAliasLists[listOffset]; |
880 | } |
881 | /* else this shouldn't happen. internal program error */ |
882 | } |
883 | /* else converter not found */ |
884 | } |
885 | return 0; |
886 | } |
887 | |
888 | static uint16_t |
889 | ucnv_io_getAliases(const char *alias, uint16_t start, const char **aliases, UErrorCode *pErrorCode) { |
890 | if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { |
891 | uint32_t currAlias; |
892 | uint32_t convNum = findConverter(alias, nullptr, pErrorCode); |
893 | if (convNum < gMainTable.converterListSize) { |
894 | /* tagListNum - 1 is the ALL tag */ |
895 | int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; |
896 | |
897 | if (listOffset) { |
898 | uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; |
899 | /* +1 to skip listCount */ |
900 | const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; |
901 | |
902 | for (currAlias = start; currAlias < listCount; currAlias++) { |
903 | aliases[currAlias] = GET_STRING(currList[currAlias]); |
904 | } |
905 | } |
906 | /* else this shouldn't happen. internal program error */ |
907 | } |
908 | /* else converter not found */ |
909 | } |
910 | return 0; |
911 | } |
912 | |
913 | static const char * |
914 | ucnv_io_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode) { |
915 | if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { |
916 | uint32_t convNum = findConverter(alias, nullptr, pErrorCode); |
917 | if (convNum < gMainTable.converterListSize) { |
918 | /* tagListNum - 1 is the ALL tag */ |
919 | int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum]; |
920 | |
921 | if (listOffset) { |
922 | uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; |
923 | /* +1 to skip listCount */ |
924 | const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; |
925 | |
926 | if (n < listCount) { |
927 | return GET_STRING(currList[n]); |
928 | } |
929 | *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; |
930 | } |
931 | /* else this shouldn't happen. internal program error */ |
932 | } |
933 | /* else converter not found */ |
934 | } |
935 | return nullptr; |
936 | } |
937 | |
938 | static uint16_t |
939 | ucnv_io_countStandards(UErrorCode *pErrorCode) { |
940 | if (haveAliasData(pErrorCode)) { |
941 | /* Don't include the empty list */ |
942 | return (uint16_t)(gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS); |
943 | } |
944 | |
945 | return 0; |
946 | } |
947 | |
948 | U_CAPI const char * U_EXPORT2 |
949 | ucnv_getStandard(uint16_t n, UErrorCode *pErrorCode) { |
950 | if (haveAliasData(pErrorCode)) { |
951 | if (n < gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) { |
952 | return GET_STRING(gMainTable.tagList[n]); |
953 | } |
954 | *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR; |
955 | } |
956 | |
957 | return nullptr; |
958 | } |
959 | |
960 | U_CAPI const char * U_EXPORT2 |
961 | ucnv_getStandardName(const char *alias, const char *standard, UErrorCode *pErrorCode) { |
962 | if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { |
963 | uint32_t listOffset = findTaggedAliasListsOffset(alias, standard, pErrorCode); |
964 | |
965 | if (0 < listOffset && listOffset < gMainTable.taggedAliasListsSize) { |
966 | const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; |
967 | |
968 | /* Get the preferred name from this list */ |
969 | if (currList[0]) { |
970 | return GET_STRING(currList[0]); |
971 | } |
972 | /* else someone screwed up the alias table. */ |
973 | /* *pErrorCode = U_INVALID_FORMAT_ERROR */ |
974 | } |
975 | } |
976 | |
977 | return nullptr; |
978 | } |
979 | |
980 | U_CAPI uint16_t U_EXPORT2 |
981 | ucnv_countAliases(const char *alias, UErrorCode *pErrorCode) |
982 | { |
983 | return ucnv_io_countAliases(alias, pErrorCode); |
984 | } |
985 | |
986 | |
987 | U_CAPI const char* U_EXPORT2 |
988 | ucnv_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode) |
989 | { |
990 | return ucnv_io_getAlias(alias, n, pErrorCode); |
991 | } |
992 | |
993 | U_CAPI void U_EXPORT2 |
994 | ucnv_getAliases(const char *alias, const char **aliases, UErrorCode *pErrorCode) |
995 | { |
996 | ucnv_io_getAliases(alias, 0, aliases, pErrorCode); |
997 | } |
998 | |
999 | U_CAPI uint16_t U_EXPORT2 |
1000 | ucnv_countStandards() |
1001 | { |
1002 | UErrorCode err = U_ZERO_ERROR; |
1003 | return ucnv_io_countStandards(&err); |
1004 | } |
1005 | |
1006 | U_CAPI const char * U_EXPORT2 |
1007 | ucnv_getCanonicalName(const char *alias, const char *standard, UErrorCode *pErrorCode) { |
1008 | if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) { |
1009 | uint32_t convNum = findTaggedConverterNum(alias, standard, pErrorCode); |
1010 | |
1011 | if (convNum < gMainTable.converterListSize) { |
1012 | return GET_STRING(gMainTable.converterList[convNum]); |
1013 | } |
1014 | } |
1015 | |
1016 | return nullptr; |
1017 | } |
1018 | |
1019 | U_CDECL_BEGIN |
1020 | |
1021 | |
1022 | static int32_t U_CALLCONV |
1023 | ucnv_io_countAllConverters(UEnumeration * /*enumerator*/, UErrorCode * /*pErrorCode*/) { |
1024 | return gMainTable.converterListSize; |
1025 | } |
1026 | |
1027 | static const char * U_CALLCONV |
1028 | ucnv_io_nextAllConverters(UEnumeration *enumerator, |
1029 | int32_t* resultLength, |
1030 | UErrorCode * /*pErrorCode*/) |
1031 | { |
1032 | uint16_t *myContext = (uint16_t *)(enumerator->context); |
1033 | |
1034 | if (*myContext < gMainTable.converterListSize) { |
1035 | const char *myStr = GET_STRING(gMainTable.converterList[(*myContext)++]); |
1036 | if (resultLength) { |
1037 | *resultLength = (int32_t)uprv_strlen(myStr); |
1038 | } |
1039 | return myStr; |
1040 | } |
1041 | /* Either we accessed a zero length list, or we enumerated too far. */ |
1042 | if (resultLength) { |
1043 | *resultLength = 0; |
1044 | } |
1045 | return nullptr; |
1046 | } |
1047 | |
1048 | static void U_CALLCONV |
1049 | ucnv_io_resetAllConverters(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) { |
1050 | *((uint16_t *)(enumerator->context)) = 0; |
1051 | } |
1052 | U_CDECL_END |
1053 | static const UEnumeration gEnumAllConverters = { |
1054 | nullptr, |
1055 | nullptr, |
1056 | ucnv_io_closeUEnumeration, |
1057 | ucnv_io_countAllConverters, |
1058 | uenum_unextDefault, |
1059 | ucnv_io_nextAllConverters, |
1060 | ucnv_io_resetAllConverters |
1061 | }; |
1062 | |
1063 | U_CAPI UEnumeration * U_EXPORT2 |
1064 | ucnv_openAllNames(UErrorCode *pErrorCode) { |
1065 | UEnumeration *myEnum = nullptr; |
1066 | if (haveAliasData(pErrorCode)) { |
1067 | uint16_t *myContext; |
1068 | |
1069 | myEnum = static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration))); |
1070 | if (myEnum == nullptr) { |
1071 | *pErrorCode = U_MEMORY_ALLOCATION_ERROR; |
1072 | return nullptr; |
1073 | } |
1074 | uprv_memcpy(myEnum, &gEnumAllConverters, sizeof(UEnumeration)); |
1075 | myContext = static_cast<uint16_t *>(uprv_malloc(sizeof(uint16_t))); |
1076 | if (myContext == nullptr) { |
1077 | *pErrorCode = U_MEMORY_ALLOCATION_ERROR; |
1078 | uprv_free(myEnum); |
1079 | return nullptr; |
1080 | } |
1081 | *myContext = 0; |
1082 | myEnum->context = myContext; |
1083 | } |
1084 | return myEnum; |
1085 | } |
1086 | |
1087 | U_CAPI uint16_t |
1088 | ucnv_io_countKnownConverters(UErrorCode *pErrorCode) { |
1089 | if (haveAliasData(pErrorCode)) { |
1090 | return (uint16_t)gMainTable.converterListSize; |
1091 | } |
1092 | return 0; |
1093 | } |
1094 | |
1095 | /* alias table swapping ----------------------------------------------------- */ |
1096 | |
1097 | U_CDECL_BEGIN |
1098 | |
1099 | typedef char * U_CALLCONV StripForCompareFn(char *dst, const char *name); |
1100 | U_CDECL_END |
1101 | |
1102 | |
1103 | /* |
1104 | * row of a temporary array |
1105 | * |
1106 | * gets platform-endian charset string indexes and sorting indexes; |
1107 | * after sorting this array by strings, the actual arrays are permutated |
1108 | * according to the sorting indexes |
1109 | */ |
1110 | typedef struct TempRow { |
1111 | uint16_t strIndex, sortIndex; |
1112 | } TempRow; |
1113 | |
1114 | typedef struct TempAliasTable { |
1115 | const char *chars; |
1116 | TempRow *rows; |
1117 | uint16_t *resort; |
1118 | StripForCompareFn *stripForCompare; |
1119 | } TempAliasTable; |
1120 | |
1121 | enum { |
1122 | STACK_ROW_CAPACITY=500 |
1123 | }; |
1124 | |
1125 | static int32_t U_CALLCONV |
1126 | io_compareRows(const void *context, const void *left, const void *right) { |
1127 | char strippedLeft[UCNV_MAX_CONVERTER_NAME_LENGTH], |
1128 | strippedRight[UCNV_MAX_CONVERTER_NAME_LENGTH]; |
1129 | |
1130 | TempAliasTable *tempTable=(TempAliasTable *)context; |
1131 | const char *chars=tempTable->chars; |
1132 | |
1133 | return (int32_t)uprv_strcmp(tempTable->stripForCompare(strippedLeft, chars+2*((const TempRow *)left)->strIndex), |
1134 | tempTable->stripForCompare(strippedRight, chars+2*((const TempRow *)right)->strIndex)); |
1135 | } |
1136 | |
1137 | U_CAPI int32_t U_EXPORT2 |
1138 | ucnv_swapAliases(const UDataSwapper *ds, |
1139 | const void *inData, int32_t length, void *outData, |
1140 | UErrorCode *pErrorCode) { |
1141 | const UDataInfo *pInfo; |
1142 | int32_t headerSize; |
1143 | |
1144 | const uint16_t *inTable; |
1145 | const uint32_t *inSectionSizes; |
1146 | uint32_t toc[offsetsCount]; |
1147 | uint32_t offsets[offsetsCount]; /* 16-bit-addressed offsets from inTable/outTable */ |
1148 | uint32_t i, count, tocLength, topOffset; |
1149 | |
1150 | TempRow rows[STACK_ROW_CAPACITY]; |
1151 | uint16_t resort[STACK_ROW_CAPACITY]; |
1152 | TempAliasTable tempTable; |
1153 | |
1154 | /* udata_swapDataHeader checks the arguments */ |
1155 | headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode); |
1156 | if(pErrorCode==nullptr || U_FAILURE(*pErrorCode)) { |
1157 | return 0; |
1158 | } |
1159 | |
1160 | /* check data format and format version */ |
1161 | pInfo=(const UDataInfo *)((const char *)inData+4); |
1162 | if(!( |
1163 | pInfo->dataFormat[0]==0x43 && /* dataFormat="CvAl" */ |
1164 | pInfo->dataFormat[1]==0x76 && |
1165 | pInfo->dataFormat[2]==0x41 && |
1166 | pInfo->dataFormat[3]==0x6c && |
1167 | pInfo->formatVersion[0]==3 |
1168 | )) { |
1169 | udata_printError(ds, "ucnv_swapAliases(): data format %02x.%02x.%02x.%02x (format version %02x) is not an alias table\n" , |
1170 | pInfo->dataFormat[0], pInfo->dataFormat[1], |
1171 | pInfo->dataFormat[2], pInfo->dataFormat[3], |
1172 | pInfo->formatVersion[0]); |
1173 | *pErrorCode=U_UNSUPPORTED_ERROR; |
1174 | return 0; |
1175 | } |
1176 | |
1177 | /* an alias table must contain at least the table of contents array */ |
1178 | if(length>=0 && (length-headerSize)<4*(1+minTocLength)) { |
1179 | udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n" , |
1180 | length-headerSize); |
1181 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1182 | return 0; |
1183 | } |
1184 | |
1185 | inSectionSizes=(const uint32_t *)((const char *)inData+headerSize); |
1186 | inTable=(const uint16_t *)inSectionSizes; |
1187 | uprv_memset(toc, 0, sizeof(toc)); |
1188 | toc[tocLengthIndex]=tocLength=ds->readUInt32(inSectionSizes[tocLengthIndex]); |
1189 | if(tocLength<minTocLength || offsetsCount<=tocLength) { |
1190 | udata_printError(ds, "ucnv_swapAliases(): table of contents contains unsupported number of sections (%u sections)\n" , tocLength); |
1191 | *pErrorCode=U_INVALID_FORMAT_ERROR; |
1192 | return 0; |
1193 | } |
1194 | |
1195 | /* read the known part of the table of contents */ |
1196 | for(i=converterListIndex; i<=tocLength; ++i) { |
1197 | toc[i]=ds->readUInt32(inSectionSizes[i]); |
1198 | } |
1199 | |
1200 | /* compute offsets */ |
1201 | uprv_memset(offsets, 0, sizeof(offsets)); |
1202 | offsets[converterListIndex]=2*(1+tocLength); /* count two 16-bit units per toc entry */ |
1203 | for(i=tagListIndex; i<=tocLength; ++i) { |
1204 | offsets[i]=offsets[i-1]+toc[i-1]; |
1205 | } |
1206 | |
1207 | /* compute the overall size of the after-header data, in numbers of 16-bit units */ |
1208 | topOffset=offsets[i-1]+toc[i-1]; |
1209 | |
1210 | if(length>=0) { |
1211 | uint16_t *outTable; |
1212 | const uint16_t *p, *p2; |
1213 | uint16_t *q, *q2; |
1214 | uint16_t oldIndex; |
1215 | |
1216 | if((length-headerSize)<(2*(int32_t)topOffset)) { |
1217 | udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n" , |
1218 | length-headerSize); |
1219 | *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; |
1220 | return 0; |
1221 | } |
1222 | |
1223 | outTable=(uint16_t *)((char *)outData+headerSize); |
1224 | |
1225 | /* swap the entire table of contents */ |
1226 | ds->swapArray32(ds, inTable, 4*(1+tocLength), outTable, pErrorCode); |
1227 | |
1228 | /* swap unormalized strings & normalized strings */ |
1229 | ds->swapInvChars(ds, inTable+offsets[stringTableIndex], 2*(int32_t)(toc[stringTableIndex]+toc[normalizedStringTableIndex]), |
1230 | outTable+offsets[stringTableIndex], pErrorCode); |
1231 | if(U_FAILURE(*pErrorCode)) { |
1232 | udata_printError(ds, "ucnv_swapAliases().swapInvChars(charset names) failed\n" ); |
1233 | return 0; |
1234 | } |
1235 | |
1236 | if(ds->inCharset==ds->outCharset) { |
1237 | /* no need to sort, just swap all 16-bit values together */ |
1238 | ds->swapArray16(ds, |
1239 | inTable+offsets[converterListIndex], |
1240 | 2*(int32_t)(offsets[stringTableIndex]-offsets[converterListIndex]), |
1241 | outTable+offsets[converterListIndex], |
1242 | pErrorCode); |
1243 | } else { |
1244 | /* allocate the temporary table for sorting */ |
1245 | count=toc[aliasListIndex]; |
1246 | |
1247 | tempTable.chars=(const char *)(outTable+offsets[stringTableIndex]); /* sort by outCharset */ |
1248 | |
1249 | if(count<=STACK_ROW_CAPACITY) { |
1250 | tempTable.rows=rows; |
1251 | tempTable.resort=resort; |
1252 | } else { |
1253 | tempTable.rows=(TempRow *)uprv_malloc(count*sizeof(TempRow)+count*2); |
1254 | if(tempTable.rows==nullptr) { |
1255 | udata_printError(ds, "ucnv_swapAliases(): unable to allocate memory for sorting tables (max length: %u)\n" , |
1256 | count); |
1257 | *pErrorCode=U_MEMORY_ALLOCATION_ERROR; |
1258 | return 0; |
1259 | } |
1260 | tempTable.resort=(uint16_t *)(tempTable.rows+count); |
1261 | } |
1262 | |
1263 | if(ds->outCharset==U_ASCII_FAMILY) { |
1264 | tempTable.stripForCompare=ucnv_io_stripASCIIForCompare; |
1265 | } else /* U_EBCDIC_FAMILY */ { |
1266 | tempTable.stripForCompare=ucnv_io_stripEBCDICForCompare; |
1267 | } |
1268 | |
1269 | /* |
1270 | * Sort unique aliases+mapped names. |
1271 | * |
1272 | * We need to sort the list again by outCharset strings because they |
1273 | * sort differently for different charset families. |
1274 | * First we set up a temporary table with the string indexes and |
1275 | * sorting indexes and sort that. |
1276 | * Then we permutate and copy/swap the actual values. |
1277 | */ |
1278 | p=inTable+offsets[aliasListIndex]; |
1279 | q=outTable+offsets[aliasListIndex]; |
1280 | |
1281 | p2=inTable+offsets[untaggedConvArrayIndex]; |
1282 | q2=outTable+offsets[untaggedConvArrayIndex]; |
1283 | |
1284 | for(i=0; i<count; ++i) { |
1285 | tempTable.rows[i].strIndex=ds->readUInt16(p[i]); |
1286 | tempTable.rows[i].sortIndex=(uint16_t)i; |
1287 | } |
1288 | |
1289 | uprv_sortArray(tempTable.rows, (int32_t)count, sizeof(TempRow), |
1290 | io_compareRows, &tempTable, |
1291 | false, pErrorCode); |
1292 | |
1293 | if(U_SUCCESS(*pErrorCode)) { |
1294 | /* copy/swap/permutate items */ |
1295 | if(p!=q) { |
1296 | for(i=0; i<count; ++i) { |
1297 | oldIndex=tempTable.rows[i].sortIndex; |
1298 | ds->swapArray16(ds, p+oldIndex, 2, q+i, pErrorCode); |
1299 | ds->swapArray16(ds, p2+oldIndex, 2, q2+i, pErrorCode); |
1300 | } |
1301 | } else { |
1302 | /* |
1303 | * If we swap in-place, then the permutation must use another |
1304 | * temporary array (tempTable.resort) |
1305 | * before the results are copied to the outBundle. |
1306 | */ |
1307 | uint16_t *r=tempTable.resort; |
1308 | |
1309 | for(i=0; i<count; ++i) { |
1310 | oldIndex=tempTable.rows[i].sortIndex; |
1311 | ds->swapArray16(ds, p+oldIndex, 2, r+i, pErrorCode); |
1312 | } |
1313 | uprv_memcpy(q, r, 2*(size_t)count); |
1314 | |
1315 | for(i=0; i<count; ++i) { |
1316 | oldIndex=tempTable.rows[i].sortIndex; |
1317 | ds->swapArray16(ds, p2+oldIndex, 2, r+i, pErrorCode); |
1318 | } |
1319 | uprv_memcpy(q2, r, 2*(size_t)count); |
1320 | } |
1321 | } |
1322 | |
1323 | if(tempTable.rows!=rows) { |
1324 | uprv_free(tempTable.rows); |
1325 | } |
1326 | |
1327 | if(U_FAILURE(*pErrorCode)) { |
1328 | udata_printError(ds, "ucnv_swapAliases().uprv_sortArray(%u items) failed\n" , |
1329 | count); |
1330 | return 0; |
1331 | } |
1332 | |
1333 | /* swap remaining 16-bit values */ |
1334 | ds->swapArray16(ds, |
1335 | inTable+offsets[converterListIndex], |
1336 | 2*(int32_t)(offsets[aliasListIndex]-offsets[converterListIndex]), |
1337 | outTable+offsets[converterListIndex], |
1338 | pErrorCode); |
1339 | ds->swapArray16(ds, |
1340 | inTable+offsets[taggedAliasArrayIndex], |
1341 | 2*(int32_t)(offsets[stringTableIndex]-offsets[taggedAliasArrayIndex]), |
1342 | outTable+offsets[taggedAliasArrayIndex], |
1343 | pErrorCode); |
1344 | } |
1345 | } |
1346 | |
1347 | return headerSize+2*(int32_t)topOffset; |
1348 | } |
1349 | |
1350 | #endif |
1351 | |
1352 | |
1353 | /* |
1354 | * Hey, Emacs, please set the following: |
1355 | * |
1356 | * Local Variables: |
1357 | * indent-tabs-mode: nil |
1358 | * End: |
1359 | * |
1360 | */ |
1361 | |