1 | // © 2019 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | |
4 | // loclikelysubtags.cpp |
5 | // created: 2019may08 Markus W. Scherer |
6 | |
7 | #include <utility> |
8 | #include "unicode/utypes.h" |
9 | #include "unicode/bytestrie.h" |
10 | #include "unicode/localpointer.h" |
11 | #include "unicode/locid.h" |
12 | #include "unicode/uobject.h" |
13 | #include "unicode/ures.h" |
14 | #include "charstr.h" |
15 | #include "cstring.h" |
16 | #include "loclikelysubtags.h" |
17 | #include "lsr.h" |
18 | #include "uassert.h" |
19 | #include "ucln_cmn.h" |
20 | #include "uhash.h" |
21 | #include "uinvchar.h" |
22 | #include "umutex.h" |
23 | #include "uniquecharstr.h" |
24 | #include "uresdata.h" |
25 | #include "uresimp.h" |
26 | |
27 | U_NAMESPACE_BEGIN |
28 | |
29 | namespace { |
30 | |
31 | constexpr char PSEUDO_ACCENTS_PREFIX = '\''; // -XA, -PSACCENT |
32 | constexpr char PSEUDO_BIDI_PREFIX = '+'; // -XB, -PSBIDI |
33 | constexpr char PSEUDO_CRACKED_PREFIX = ','; // -XC, -PSCRACK |
34 | |
35 | } // namespace |
36 | |
37 | LocaleDistanceData::LocaleDistanceData(LocaleDistanceData &&data) : |
38 | distanceTrieBytes(data.distanceTrieBytes), |
39 | regionToPartitions(data.regionToPartitions), |
40 | partitions(data.partitions), |
41 | paradigms(data.paradigms), paradigmsLength(data.paradigmsLength), |
42 | distances(data.distances) { |
43 | data.partitions = nullptr; |
44 | data.paradigms = nullptr; |
45 | } |
46 | |
47 | LocaleDistanceData::~LocaleDistanceData() { |
48 | uprv_free(partitions); |
49 | delete[] paradigms; |
50 | } |
51 | |
52 | // TODO(ICU-20777): Rename to just LikelySubtagsData. |
53 | struct XLikelySubtagsData { |
54 | UResourceBundle *langInfoBundle = nullptr; |
55 | UniqueCharStrings strings; |
56 | CharStringMap languageAliases; |
57 | CharStringMap regionAliases; |
58 | const uint8_t *trieBytes = nullptr; |
59 | LSR *lsrs = nullptr; |
60 | int32_t lsrsLength = 0; |
61 | |
62 | LocaleDistanceData distanceData; |
63 | |
64 | XLikelySubtagsData(UErrorCode &errorCode) : strings(errorCode) {} |
65 | |
66 | ~XLikelySubtagsData() { |
67 | ures_close(langInfoBundle); |
68 | delete[] lsrs; |
69 | } |
70 | |
71 | void load(UErrorCode &errorCode) { |
72 | langInfoBundle = ures_openDirect(nullptr, "langInfo" , &errorCode); |
73 | if (U_FAILURE(errorCode)) { return; } |
74 | StackUResourceBundle stackTempBundle; |
75 | ResourceDataValue value; |
76 | ures_getValueWithFallback(langInfoBundle, "likely" , stackTempBundle.getAlias(), |
77 | value, errorCode); |
78 | ResourceTable likelyTable = value.getTable(errorCode); |
79 | if (U_FAILURE(errorCode)) { return; } |
80 | |
81 | // Read all strings in the resource bundle and convert them to invariant char *. |
82 | LocalMemory<int32_t> languageIndexes, regionIndexes, lsrSubtagIndexes; |
83 | int32_t languagesLength = 0, regionsLength = 0, lsrSubtagsLength = 0; |
84 | if (!readStrings(likelyTable, "languageAliases" , value, |
85 | languageIndexes, languagesLength, errorCode) || |
86 | !readStrings(likelyTable, "regionAliases" , value, |
87 | regionIndexes, regionsLength, errorCode) || |
88 | !readStrings(likelyTable, "lsrs" , value, |
89 | lsrSubtagIndexes,lsrSubtagsLength, errorCode)) { |
90 | return; |
91 | } |
92 | if ((languagesLength & 1) != 0 || |
93 | (regionsLength & 1) != 0 || |
94 | (lsrSubtagsLength % 3) != 0) { |
95 | errorCode = U_INVALID_FORMAT_ERROR; |
96 | return; |
97 | } |
98 | if (lsrSubtagsLength == 0) { |
99 | errorCode = U_MISSING_RESOURCE_ERROR; |
100 | return; |
101 | } |
102 | |
103 | if (!likelyTable.findValue("trie" , value)) { |
104 | errorCode = U_MISSING_RESOURCE_ERROR; |
105 | return; |
106 | } |
107 | int32_t length; |
108 | trieBytes = value.getBinary(length, errorCode); |
109 | if (U_FAILURE(errorCode)) { return; } |
110 | |
111 | // Also read distance/matcher data if available, |
112 | // to open & keep only one resource bundle pointer |
113 | // and to use one single UniqueCharStrings. |
114 | UErrorCode matchErrorCode = U_ZERO_ERROR; |
115 | ures_getValueWithFallback(langInfoBundle, "match" , stackTempBundle.getAlias(), |
116 | value, matchErrorCode); |
117 | LocalMemory<int32_t> partitionIndexes, paradigmSubtagIndexes; |
118 | int32_t partitionsLength = 0, paradigmSubtagsLength = 0; |
119 | if (U_SUCCESS(matchErrorCode)) { |
120 | ResourceTable matchTable = value.getTable(errorCode); |
121 | if (U_FAILURE(errorCode)) { return; } |
122 | |
123 | if (matchTable.findValue("trie" , value)) { |
124 | distanceData.distanceTrieBytes = value.getBinary(length, errorCode); |
125 | if (U_FAILURE(errorCode)) { return; } |
126 | } |
127 | |
128 | if (matchTable.findValue("regionToPartitions" , value)) { |
129 | distanceData.regionToPartitions = value.getBinary(length, errorCode); |
130 | if (U_FAILURE(errorCode)) { return; } |
131 | if (length < LSR::REGION_INDEX_LIMIT) { |
132 | errorCode = U_INVALID_FORMAT_ERROR; |
133 | return; |
134 | } |
135 | } |
136 | |
137 | if (!readStrings(matchTable, "partitions" , value, |
138 | partitionIndexes, partitionsLength, errorCode) || |
139 | !readStrings(matchTable, "paradigms" , value, |
140 | paradigmSubtagIndexes, paradigmSubtagsLength, errorCode)) { |
141 | return; |
142 | } |
143 | if ((paradigmSubtagsLength % 3) != 0) { |
144 | errorCode = U_INVALID_FORMAT_ERROR; |
145 | return; |
146 | } |
147 | |
148 | if (matchTable.findValue("distances" , value)) { |
149 | distanceData.distances = value.getIntVector(length, errorCode); |
150 | if (U_FAILURE(errorCode)) { return; } |
151 | if (length < 4) { // LocaleDistance IX_LIMIT |
152 | errorCode = U_INVALID_FORMAT_ERROR; |
153 | return; |
154 | } |
155 | } |
156 | } else if (matchErrorCode == U_MISSING_RESOURCE_ERROR) { |
157 | // ok for likely subtags |
158 | } else { // error other than missing resource |
159 | errorCode = matchErrorCode; |
160 | return; |
161 | } |
162 | |
163 | // Fetch & store invariant-character versions of strings |
164 | // only after we have collected and de-duplicated all of them. |
165 | strings.freeze(); |
166 | |
167 | languageAliases = CharStringMap(languagesLength / 2, errorCode); |
168 | for (int32_t i = 0; i < languagesLength; i += 2) { |
169 | languageAliases.put(strings.get(languageIndexes[i]), |
170 | strings.get(languageIndexes[i + 1]), errorCode); |
171 | } |
172 | |
173 | regionAliases = CharStringMap(regionsLength / 2, errorCode); |
174 | for (int32_t i = 0; i < regionsLength; i += 2) { |
175 | regionAliases.put(strings.get(regionIndexes[i]), |
176 | strings.get(regionIndexes[i + 1]), errorCode); |
177 | } |
178 | if (U_FAILURE(errorCode)) { return; } |
179 | |
180 | lsrsLength = lsrSubtagsLength / 3; |
181 | lsrs = new LSR[lsrsLength]; |
182 | if (lsrs == nullptr) { |
183 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
184 | return; |
185 | } |
186 | for (int32_t i = 0, j = 0; i < lsrSubtagsLength; i += 3, ++j) { |
187 | lsrs[j] = LSR(strings.get(lsrSubtagIndexes[i]), |
188 | strings.get(lsrSubtagIndexes[i + 1]), |
189 | strings.get(lsrSubtagIndexes[i + 2]), |
190 | LSR::IMPLICIT_LSR); |
191 | } |
192 | |
193 | if (partitionsLength > 0) { |
194 | distanceData.partitions = static_cast<const char **>( |
195 | uprv_malloc(partitionsLength * sizeof(const char *))); |
196 | if (distanceData.partitions == nullptr) { |
197 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
198 | return; |
199 | } |
200 | for (int32_t i = 0; i < partitionsLength; ++i) { |
201 | distanceData.partitions[i] = strings.get(partitionIndexes[i]); |
202 | } |
203 | } |
204 | |
205 | if (paradigmSubtagsLength > 0) { |
206 | distanceData.paradigmsLength = paradigmSubtagsLength / 3; |
207 | LSR *paradigms = new LSR[distanceData.paradigmsLength]; |
208 | if (paradigms == nullptr) { |
209 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
210 | return; |
211 | } |
212 | for (int32_t i = 0, j = 0; i < paradigmSubtagsLength; i += 3, ++j) { |
213 | paradigms[j] = LSR(strings.get(paradigmSubtagIndexes[i]), |
214 | strings.get(paradigmSubtagIndexes[i + 1]), |
215 | strings.get(paradigmSubtagIndexes[i + 2]), |
216 | LSR::DONT_CARE_FLAGS); |
217 | } |
218 | distanceData.paradigms = paradigms; |
219 | } |
220 | } |
221 | |
222 | private: |
223 | bool readStrings(const ResourceTable &table, const char *key, ResourceValue &value, |
224 | LocalMemory<int32_t> &indexes, int32_t &length, UErrorCode &errorCode) { |
225 | if (table.findValue(key, value)) { |
226 | ResourceArray stringArray = value.getArray(errorCode); |
227 | if (U_FAILURE(errorCode)) { return false; } |
228 | length = stringArray.getSize(); |
229 | if (length == 0) { return true; } |
230 | int32_t *rawIndexes = indexes.allocateInsteadAndCopy(length); |
231 | if (rawIndexes == nullptr) { |
232 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
233 | return false; |
234 | } |
235 | for (int i = 0; i < length; ++i) { |
236 | stringArray.getValue(i, value); // returns true because i < length |
237 | rawIndexes[i] = strings.add(value.getUnicodeString(errorCode), errorCode); |
238 | if (U_FAILURE(errorCode)) { return false; } |
239 | } |
240 | } |
241 | return true; |
242 | } |
243 | }; |
244 | |
245 | namespace { |
246 | |
247 | XLikelySubtags *gLikelySubtags = nullptr; |
248 | UInitOnce gInitOnce {}; |
249 | |
250 | UBool U_CALLCONV cleanup() { |
251 | delete gLikelySubtags; |
252 | gLikelySubtags = nullptr; |
253 | gInitOnce.reset(); |
254 | return true; |
255 | } |
256 | |
257 | } // namespace |
258 | |
259 | void U_CALLCONV XLikelySubtags::initLikelySubtags(UErrorCode &errorCode) { |
260 | // This function is invoked only via umtx_initOnce(). |
261 | U_ASSERT(gLikelySubtags == nullptr); |
262 | XLikelySubtagsData data(errorCode); |
263 | data.load(errorCode); |
264 | if (U_FAILURE(errorCode)) { return; } |
265 | gLikelySubtags = new XLikelySubtags(data); |
266 | if (gLikelySubtags == nullptr) { |
267 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
268 | return; |
269 | } |
270 | ucln_common_registerCleanup(UCLN_COMMON_LIKELY_SUBTAGS, cleanup); |
271 | } |
272 | |
273 | const XLikelySubtags *XLikelySubtags::getSingleton(UErrorCode &errorCode) { |
274 | if (U_FAILURE(errorCode)) { return nullptr; } |
275 | umtx_initOnce(gInitOnce, &XLikelySubtags::initLikelySubtags, errorCode); |
276 | return gLikelySubtags; |
277 | } |
278 | |
279 | XLikelySubtags::XLikelySubtags(XLikelySubtagsData &data) : |
280 | langInfoBundle(data.langInfoBundle), |
281 | strings(data.strings.orphanCharStrings()), |
282 | languageAliases(std::move(data.languageAliases)), |
283 | regionAliases(std::move(data.regionAliases)), |
284 | trie(data.trieBytes), |
285 | lsrs(data.lsrs), |
286 | #if U_DEBUG |
287 | lsrsLength(data.lsrsLength), |
288 | #endif |
289 | distanceData(std::move(data.distanceData)) { |
290 | data.langInfoBundle = nullptr; |
291 | data.lsrs = nullptr; |
292 | |
293 | // Cache the result of looking up language="und" encoded as "*", and "und-Zzzz" ("**"). |
294 | UStringTrieResult result = trie.next(u'*'); |
295 | U_ASSERT(USTRINGTRIE_HAS_NEXT(result)); |
296 | trieUndState = trie.getState64(); |
297 | result = trie.next(u'*'); |
298 | U_ASSERT(USTRINGTRIE_HAS_NEXT(result)); |
299 | trieUndZzzzState = trie.getState64(); |
300 | result = trie.next(u'*'); |
301 | U_ASSERT(USTRINGTRIE_HAS_VALUE(result)); |
302 | defaultLsrIndex = trie.getValue(); |
303 | trie.reset(); |
304 | |
305 | for (char16_t c = u'a'; c <= u'z'; ++c) { |
306 | result = trie.next(c); |
307 | if (result == USTRINGTRIE_NO_VALUE) { |
308 | trieFirstLetterStates[c - u'a'] = trie.getState64(); |
309 | } |
310 | trie.reset(); |
311 | } |
312 | } |
313 | |
314 | XLikelySubtags::~XLikelySubtags() { |
315 | ures_close(langInfoBundle); |
316 | delete strings; |
317 | delete[] lsrs; |
318 | } |
319 | |
320 | LSR XLikelySubtags::makeMaximizedLsrFrom(const Locale &locale, UErrorCode &errorCode) const { |
321 | const char *name = locale.getName(); |
322 | if (uprv_isAtSign(name[0]) && name[1] == 'x' && name[2] == '=') { // name.startsWith("@x=") |
323 | // Private use language tag x-subtag-subtag... which CLDR changes to |
324 | // und-x-subtag-subtag... |
325 | return LSR(name, "" , "" , LSR::EXPLICIT_LSR); |
326 | } |
327 | return makeMaximizedLsr(locale.getLanguage(), locale.getScript(), locale.getCountry(), |
328 | locale.getVariant(), errorCode); |
329 | } |
330 | |
331 | namespace { |
332 | |
333 | const char *getCanonical(const CharStringMap &aliases, const char *alias) { |
334 | const char *canonical = aliases.get(alias); |
335 | return canonical == nullptr ? alias : canonical; |
336 | } |
337 | |
338 | } // namespace |
339 | |
340 | LSR XLikelySubtags::makeMaximizedLsr(const char *language, const char *script, const char *region, |
341 | const char *variant, UErrorCode &errorCode) const { |
342 | // Handle pseudolocales like en-XA, ar-XB, fr-PSCRACK. |
343 | // They should match only themselves, |
344 | // not other locales with what looks like the same language and script subtags. |
345 | char c1; |
346 | if (region[0] == 'X' && (c1 = region[1]) != 0 && region[2] == 0) { |
347 | switch (c1) { |
348 | case 'A': |
349 | return LSR(PSEUDO_ACCENTS_PREFIX, language, script, region, |
350 | LSR::EXPLICIT_LSR, errorCode); |
351 | case 'B': |
352 | return LSR(PSEUDO_BIDI_PREFIX, language, script, region, |
353 | LSR::EXPLICIT_LSR, errorCode); |
354 | case 'C': |
355 | return LSR(PSEUDO_CRACKED_PREFIX, language, script, region, |
356 | LSR::EXPLICIT_LSR, errorCode); |
357 | default: // normal locale |
358 | break; |
359 | } |
360 | } |
361 | |
362 | if (variant[0] == 'P' && variant[1] == 'S') { |
363 | int32_t lsrFlags = *region == 0 ? |
364 | LSR::EXPLICIT_LANGUAGE | LSR::EXPLICIT_SCRIPT : LSR::EXPLICIT_LSR; |
365 | if (uprv_strcmp(variant, "PSACCENT" ) == 0) { |
366 | return LSR(PSEUDO_ACCENTS_PREFIX, language, script, |
367 | *region == 0 ? "XA" : region, lsrFlags, errorCode); |
368 | } else if (uprv_strcmp(variant, "PSBIDI" ) == 0) { |
369 | return LSR(PSEUDO_BIDI_PREFIX, language, script, |
370 | *region == 0 ? "XB" : region, lsrFlags, errorCode); |
371 | } else if (uprv_strcmp(variant, "PSCRACK" ) == 0) { |
372 | return LSR(PSEUDO_CRACKED_PREFIX, language, script, |
373 | *region == 0 ? "XC" : region, lsrFlags, errorCode); |
374 | } |
375 | // else normal locale |
376 | } |
377 | |
378 | language = getCanonical(languageAliases, language); |
379 | // (We have no script mappings.) |
380 | region = getCanonical(regionAliases, region); |
381 | return maximize(language, script, region); |
382 | } |
383 | |
384 | LSR XLikelySubtags::maximize(const char *language, const char *script, const char *region) const { |
385 | if (uprv_strcmp(language, "und" ) == 0) { |
386 | language = "" ; |
387 | } |
388 | if (uprv_strcmp(script, "Zzzz" ) == 0) { |
389 | script = "" ; |
390 | } |
391 | if (uprv_strcmp(region, "ZZ" ) == 0) { |
392 | region = "" ; |
393 | } |
394 | if (*script != 0 && *region != 0 && *language != 0) { |
395 | return LSR(language, script, region, LSR::EXPLICIT_LSR); // already maximized |
396 | } |
397 | |
398 | uint32_t retainOldMask = 0; |
399 | BytesTrie iter(trie); |
400 | uint64_t state; |
401 | int32_t value; |
402 | // Small optimization: Array lookup for first language letter. |
403 | int32_t c0; |
404 | if (0 <= (c0 = uprv_lowerOrdinal(language[0])) && c0 <= 25 && |
405 | language[1] != 0 && // language.length() >= 2 |
406 | (state = trieFirstLetterStates[c0]) != 0) { |
407 | value = trieNext(iter.resetToState64(state), language, 1); |
408 | } else { |
409 | value = trieNext(iter, language, 0); |
410 | } |
411 | if (value >= 0) { |
412 | if (*language != 0) { |
413 | retainOldMask |= 4; |
414 | } |
415 | state = iter.getState64(); |
416 | } else { |
417 | retainOldMask |= 4; |
418 | iter.resetToState64(trieUndState); // "und" ("*") |
419 | state = 0; |
420 | } |
421 | |
422 | if (value > 0) { |
423 | // Intermediate or final value from just language. |
424 | if (value == SKIP_SCRIPT) { |
425 | value = 0; |
426 | } |
427 | if (*script != 0) { |
428 | retainOldMask |= 2; |
429 | } |
430 | } else { |
431 | value = trieNext(iter, script, 0); |
432 | if (value >= 0) { |
433 | if (*script != 0) { |
434 | retainOldMask |= 2; |
435 | } |
436 | state = iter.getState64(); |
437 | } else { |
438 | retainOldMask |= 2; |
439 | if (state == 0) { |
440 | iter.resetToState64(trieUndZzzzState); // "und-Zzzz" ("**") |
441 | } else { |
442 | iter.resetToState64(state); |
443 | value = trieNext(iter, "" , 0); |
444 | U_ASSERT(value >= 0); |
445 | state = iter.getState64(); |
446 | } |
447 | } |
448 | } |
449 | |
450 | if (value > 0) { |
451 | // Final value from just language or language+script. |
452 | if (*region != 0) { |
453 | retainOldMask |= 1; |
454 | } |
455 | } else { |
456 | value = trieNext(iter, region, 0); |
457 | if (value >= 0) { |
458 | if (*region != 0) { |
459 | retainOldMask |= 1; |
460 | } |
461 | } else { |
462 | retainOldMask |= 1; |
463 | if (state == 0) { |
464 | value = defaultLsrIndex; |
465 | } else { |
466 | iter.resetToState64(state); |
467 | value = trieNext(iter, "" , 0); |
468 | U_ASSERT(value > 0); |
469 | } |
470 | } |
471 | } |
472 | U_ASSERT(value < lsrsLength); |
473 | const LSR &result = lsrs[value]; |
474 | |
475 | if (*language == 0) { |
476 | language = "und" ; |
477 | } |
478 | |
479 | if (retainOldMask == 0) { |
480 | // Quickly return a copy of the lookup-result LSR |
481 | // without new allocation of the subtags. |
482 | return LSR(result.language, result.script, result.region, result.flags); |
483 | } |
484 | if ((retainOldMask & 4) == 0) { |
485 | language = result.language; |
486 | } |
487 | if ((retainOldMask & 2) == 0) { |
488 | script = result.script; |
489 | } |
490 | if ((retainOldMask & 1) == 0) { |
491 | region = result.region; |
492 | } |
493 | // retainOldMask flags = LSR explicit-subtag flags |
494 | return LSR(language, script, region, retainOldMask); |
495 | } |
496 | |
497 | int32_t XLikelySubtags::compareLikely(const LSR &lsr, const LSR &other, int32_t likelyInfo) const { |
498 | // If likelyInfo >= 0: |
499 | // likelyInfo bit 1 is set if the previous comparison with lsr |
500 | // was for equal language and script. |
501 | // Otherwise the scripts differed. |
502 | if (uprv_strcmp(lsr.language, other.language) != 0) { |
503 | return 0xfffffffc; // negative, lsr not better than other |
504 | } |
505 | if (uprv_strcmp(lsr.script, other.script) != 0) { |
506 | int32_t index; |
507 | if (likelyInfo >= 0 && (likelyInfo & 2) == 0) { |
508 | index = likelyInfo >> 2; |
509 | } else { |
510 | index = getLikelyIndex(lsr.language, "" ); |
511 | likelyInfo = index << 2; |
512 | } |
513 | const LSR &likely = lsrs[index]; |
514 | if (uprv_strcmp(lsr.script, likely.script) == 0) { |
515 | return likelyInfo | 1; |
516 | } else { |
517 | return likelyInfo & ~1; |
518 | } |
519 | } |
520 | if (uprv_strcmp(lsr.region, other.region) != 0) { |
521 | int32_t index; |
522 | if (likelyInfo >= 0 && (likelyInfo & 2) != 0) { |
523 | index = likelyInfo >> 2; |
524 | } else { |
525 | index = getLikelyIndex(lsr.language, lsr.region); |
526 | likelyInfo = (index << 2) | 2; |
527 | } |
528 | const LSR &likely = lsrs[index]; |
529 | if (uprv_strcmp(lsr.region, likely.region) == 0) { |
530 | return likelyInfo | 1; |
531 | } else { |
532 | return likelyInfo & ~1; |
533 | } |
534 | } |
535 | return likelyInfo & ~1; // lsr not better than other |
536 | } |
537 | |
538 | // Subset of maximize(). |
539 | int32_t XLikelySubtags::getLikelyIndex(const char *language, const char *script) const { |
540 | if (uprv_strcmp(language, "und" ) == 0) { |
541 | language = "" ; |
542 | } |
543 | if (uprv_strcmp(script, "Zzzz" ) == 0) { |
544 | script = "" ; |
545 | } |
546 | |
547 | BytesTrie iter(trie); |
548 | uint64_t state; |
549 | int32_t value; |
550 | // Small optimization: Array lookup for first language letter. |
551 | int32_t c0; |
552 | if (0 <= (c0 = uprv_lowerOrdinal(language[0])) && c0 <= 25 && |
553 | language[1] != 0 && // language.length() >= 2 |
554 | (state = trieFirstLetterStates[c0]) != 0) { |
555 | value = trieNext(iter.resetToState64(state), language, 1); |
556 | } else { |
557 | value = trieNext(iter, language, 0); |
558 | } |
559 | if (value >= 0) { |
560 | state = iter.getState64(); |
561 | } else { |
562 | iter.resetToState64(trieUndState); // "und" ("*") |
563 | state = 0; |
564 | } |
565 | |
566 | if (value > 0) { |
567 | // Intermediate or final value from just language. |
568 | if (value == SKIP_SCRIPT) { |
569 | value = 0; |
570 | } |
571 | } else { |
572 | value = trieNext(iter, script, 0); |
573 | if (value >= 0) { |
574 | state = iter.getState64(); |
575 | } else { |
576 | if (state == 0) { |
577 | iter.resetToState64(trieUndZzzzState); // "und-Zzzz" ("**") |
578 | } else { |
579 | iter.resetToState64(state); |
580 | value = trieNext(iter, "" , 0); |
581 | U_ASSERT(value >= 0); |
582 | state = iter.getState64(); |
583 | } |
584 | } |
585 | } |
586 | |
587 | if (value > 0) { |
588 | // Final value from just language or language+script. |
589 | } else { |
590 | value = trieNext(iter, "" , 0); |
591 | U_ASSERT(value > 0); |
592 | } |
593 | U_ASSERT(value < lsrsLength); |
594 | return value; |
595 | } |
596 | |
597 | int32_t XLikelySubtags::trieNext(BytesTrie &iter, const char *s, int32_t i) { |
598 | UStringTrieResult result; |
599 | uint8_t c; |
600 | if ((c = s[i]) == 0) { |
601 | result = iter.next(u'*'); |
602 | } else { |
603 | for (;;) { |
604 | c = uprv_invCharToAscii(c); |
605 | // EBCDIC: If s[i] is not an invariant character, |
606 | // then c is now 0 and will simply not match anything, which is harmless. |
607 | uint8_t next = s[++i]; |
608 | if (next != 0) { |
609 | if (!USTRINGTRIE_HAS_NEXT(iter.next(c))) { |
610 | return -1; |
611 | } |
612 | } else { |
613 | // last character of this subtag |
614 | result = iter.next(c | 0x80); |
615 | break; |
616 | } |
617 | c = next; |
618 | } |
619 | } |
620 | switch (result) { |
621 | case USTRINGTRIE_NO_MATCH: return -1; |
622 | case USTRINGTRIE_NO_VALUE: return 0; |
623 | case USTRINGTRIE_INTERMEDIATE_VALUE: |
624 | U_ASSERT(iter.getValue() == SKIP_SCRIPT); |
625 | return SKIP_SCRIPT; |
626 | case USTRINGTRIE_FINAL_VALUE: return iter.getValue(); |
627 | default: return -1; |
628 | } |
629 | } |
630 | |
631 | // TODO(ICU-20777): Switch Locale/uloc_ likely-subtags API from the old code |
632 | // in loclikely.cpp to this new code, including activating this |
633 | // minimizeSubtags() function. The LocaleMatcher does not minimize. |
634 | #if 0 |
635 | LSR XLikelySubtags::minimizeSubtags(const char *languageIn, const char *scriptIn, |
636 | const char *regionIn, ULocale.Minimize fieldToFavor, |
637 | UErrorCode &errorCode) const { |
638 | LSR result = maximize(languageIn, scriptIn, regionIn); |
639 | |
640 | // We could try just a series of checks, like: |
641 | // LSR result2 = addLikelySubtags(languageIn, "", ""); |
642 | // if result.equals(result2) return result2; |
643 | // However, we can optimize 2 of the cases: |
644 | // (languageIn, "", "") |
645 | // (languageIn, "", regionIn) |
646 | |
647 | // value00 = lookup(result.language, "", "") |
648 | BytesTrie iter = new BytesTrie(trie); |
649 | int value = trieNext(iter, result.language, 0); |
650 | U_ASSERT(value >= 0); |
651 | if (value == 0) { |
652 | value = trieNext(iter, "" , 0); |
653 | U_ASSERT(value >= 0); |
654 | if (value == 0) { |
655 | value = trieNext(iter, "" , 0); |
656 | } |
657 | } |
658 | U_ASSERT(value > 0); |
659 | LSR value00 = lsrs[value]; |
660 | boolean favorRegionOk = false; |
661 | if (result.script.equals(value00.script)) { //script is default |
662 | if (result.region.equals(value00.region)) { |
663 | return new LSR(result.language, "" , "" , LSR.DONT_CARE_FLAGS); |
664 | } else if (fieldToFavor == ULocale.Minimize.FAVOR_REGION) { |
665 | return new LSR(result.language, "" , result.region, LSR.DONT_CARE_FLAGS); |
666 | } else { |
667 | favorRegionOk = true; |
668 | } |
669 | } |
670 | |
671 | // The last case is not as easy to optimize. |
672 | // Maybe do later, but for now use the straightforward code. |
673 | LSR result2 = maximize(languageIn, scriptIn, "" ); |
674 | if (result2.equals(result)) { |
675 | return new LSR(result.language, result.script, "" , LSR.DONT_CARE_FLAGS); |
676 | } else if (favorRegionOk) { |
677 | return new LSR(result.language, "" , result.region, LSR.DONT_CARE_FLAGS); |
678 | } |
679 | return result; |
680 | } |
681 | #endif |
682 | |
683 | U_NAMESPACE_END |
684 | |