1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ******************************************************************************* |
5 | * Copyright (C) 1996-2015, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ******************************************************************************* |
8 | * rulebasedcollator.cpp |
9 | * |
10 | * (replaced the former tblcoll.cpp) |
11 | * |
12 | * created on: 2012feb14 with new and old collation code |
13 | * created by: Markus W. Scherer |
14 | */ |
15 | |
16 | #include "unicode/utypes.h" |
17 | |
18 | #if !UCONFIG_NO_COLLATION |
19 | |
20 | #include "unicode/coll.h" |
21 | #include "unicode/coleitr.h" |
22 | #include "unicode/localpointer.h" |
23 | #include "unicode/locid.h" |
24 | #include "unicode/sortkey.h" |
25 | #include "unicode/tblcoll.h" |
26 | #include "unicode/ucol.h" |
27 | #include "unicode/uiter.h" |
28 | #include "unicode/uloc.h" |
29 | #include "unicode/uniset.h" |
30 | #include "unicode/unistr.h" |
31 | #include "unicode/usetiter.h" |
32 | #include "unicode/utf8.h" |
33 | #include "unicode/uversion.h" |
34 | #include "bocsu.h" |
35 | #include "charstr.h" |
36 | #include "cmemory.h" |
37 | #include "collation.h" |
38 | #include "collationcompare.h" |
39 | #include "collationdata.h" |
40 | #include "collationdatareader.h" |
41 | #include "collationfastlatin.h" |
42 | #include "collationiterator.h" |
43 | #include "collationkeys.h" |
44 | #include "collationroot.h" |
45 | #include "collationsets.h" |
46 | #include "collationsettings.h" |
47 | #include "collationtailoring.h" |
48 | #include "cstring.h" |
49 | #include "uassert.h" |
50 | #include "ucol_imp.h" |
51 | #include "uhash.h" |
52 | #include "uitercollationiterator.h" |
53 | #include "ustr_imp.h" |
54 | #include "utf16collationiterator.h" |
55 | #include "utf8collationiterator.h" |
56 | #include "uvectr64.h" |
57 | |
58 | U_NAMESPACE_BEGIN |
59 | |
60 | namespace { |
61 | |
62 | class FixedSortKeyByteSink : public SortKeyByteSink { |
63 | public: |
64 | FixedSortKeyByteSink(char *dest, int32_t destCapacity) |
65 | : SortKeyByteSink(dest, destCapacity) {} |
66 | virtual ~FixedSortKeyByteSink(); |
67 | |
68 | private: |
69 | virtual void AppendBeyondCapacity(const char *bytes, int32_t n, int32_t length); |
70 | virtual UBool Resize(int32_t appendCapacity, int32_t length); |
71 | }; |
72 | |
73 | FixedSortKeyByteSink::~FixedSortKeyByteSink() {} |
74 | |
75 | void |
76 | FixedSortKeyByteSink::AppendBeyondCapacity(const char *bytes, int32_t /*n*/, int32_t length) { |
77 | // buffer_ != NULL && bytes != NULL && n > 0 && appended_ > capacity_ |
78 | // Fill the buffer completely. |
79 | int32_t available = capacity_ - length; |
80 | if (available > 0) { |
81 | uprv_memcpy(buffer_ + length, bytes, available); |
82 | } |
83 | } |
84 | |
85 | UBool |
86 | FixedSortKeyByteSink::Resize(int32_t /*appendCapacity*/, int32_t /*length*/) { |
87 | return FALSE; |
88 | } |
89 | |
90 | } // namespace |
91 | |
92 | // Not in an anonymous namespace, so that it can be a friend of CollationKey. |
93 | class CollationKeyByteSink : public SortKeyByteSink { |
94 | public: |
95 | CollationKeyByteSink(CollationKey &key) |
96 | : SortKeyByteSink(reinterpret_cast<char *>(key.getBytes()), key.getCapacity()), |
97 | key_(key) {} |
98 | virtual ~CollationKeyByteSink(); |
99 | |
100 | private: |
101 | virtual void AppendBeyondCapacity(const char *bytes, int32_t n, int32_t length); |
102 | virtual UBool Resize(int32_t appendCapacity, int32_t length); |
103 | |
104 | CollationKey &key_; |
105 | }; |
106 | |
107 | CollationKeyByteSink::~CollationKeyByteSink() {} |
108 | |
109 | void |
110 | CollationKeyByteSink::AppendBeyondCapacity(const char *bytes, int32_t n, int32_t length) { |
111 | // buffer_ != NULL && bytes != NULL && n > 0 && appended_ > capacity_ |
112 | if (Resize(n, length)) { |
113 | uprv_memcpy(buffer_ + length, bytes, n); |
114 | } |
115 | } |
116 | |
117 | UBool |
118 | CollationKeyByteSink::Resize(int32_t appendCapacity, int32_t length) { |
119 | if (buffer_ == NULL) { |
120 | return FALSE; // allocation failed before already |
121 | } |
122 | int32_t newCapacity = 2 * capacity_; |
123 | int32_t altCapacity = length + 2 * appendCapacity; |
124 | if (newCapacity < altCapacity) { |
125 | newCapacity = altCapacity; |
126 | } |
127 | if (newCapacity < 200) { |
128 | newCapacity = 200; |
129 | } |
130 | uint8_t *newBuffer = key_.reallocate(newCapacity, length); |
131 | if (newBuffer == NULL) { |
132 | SetNotOk(); |
133 | return FALSE; |
134 | } |
135 | buffer_ = reinterpret_cast<char *>(newBuffer); |
136 | capacity_ = newCapacity; |
137 | return TRUE; |
138 | } |
139 | |
140 | RuleBasedCollator::RuleBasedCollator(const RuleBasedCollator &other) |
141 | : Collator(other), |
142 | data(other.data), |
143 | settings(other.settings), |
144 | tailoring(other.tailoring), |
145 | cacheEntry(other.cacheEntry), |
146 | validLocale(other.validLocale), |
147 | explicitlySetAttributes(other.explicitlySetAttributes), |
148 | actualLocaleIsSameAsValid(other.actualLocaleIsSameAsValid) { |
149 | settings->addRef(); |
150 | cacheEntry->addRef(); |
151 | } |
152 | |
153 | RuleBasedCollator::RuleBasedCollator(const uint8_t *bin, int32_t length, |
154 | const RuleBasedCollator *base, UErrorCode &errorCode) |
155 | : data(NULL), |
156 | settings(NULL), |
157 | tailoring(NULL), |
158 | cacheEntry(NULL), |
159 | validLocale("" ), |
160 | explicitlySetAttributes(0), |
161 | actualLocaleIsSameAsValid(FALSE) { |
162 | if(U_FAILURE(errorCode)) { return; } |
163 | if(bin == NULL || length == 0 || base == NULL) { |
164 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
165 | return; |
166 | } |
167 | const CollationTailoring *root = CollationRoot::getRoot(errorCode); |
168 | if(U_FAILURE(errorCode)) { return; } |
169 | if(base->tailoring != root) { |
170 | errorCode = U_UNSUPPORTED_ERROR; |
171 | return; |
172 | } |
173 | LocalPointer<CollationTailoring> t(new CollationTailoring(base->tailoring->settings)); |
174 | if(t.isNull() || t->isBogus()) { |
175 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
176 | return; |
177 | } |
178 | CollationDataReader::read(base->tailoring, bin, length, *t, errorCode); |
179 | if(U_FAILURE(errorCode)) { return; } |
180 | t->actualLocale.setToBogus(); |
181 | adoptTailoring(t.orphan(), errorCode); |
182 | } |
183 | |
184 | RuleBasedCollator::RuleBasedCollator(const CollationCacheEntry *entry) |
185 | : data(entry->tailoring->data), |
186 | settings(entry->tailoring->settings), |
187 | tailoring(entry->tailoring), |
188 | cacheEntry(entry), |
189 | validLocale(entry->validLocale), |
190 | explicitlySetAttributes(0), |
191 | actualLocaleIsSameAsValid(FALSE) { |
192 | settings->addRef(); |
193 | cacheEntry->addRef(); |
194 | } |
195 | |
196 | RuleBasedCollator::~RuleBasedCollator() { |
197 | SharedObject::clearPtr(settings); |
198 | SharedObject::clearPtr(cacheEntry); |
199 | } |
200 | |
201 | void |
202 | RuleBasedCollator::adoptTailoring(CollationTailoring *t, UErrorCode &errorCode) { |
203 | if(U_FAILURE(errorCode)) { |
204 | t->deleteIfZeroRefCount(); |
205 | return; |
206 | } |
207 | U_ASSERT(settings == NULL && data == NULL && tailoring == NULL && cacheEntry == NULL); |
208 | cacheEntry = new CollationCacheEntry(t->actualLocale, t); |
209 | if(cacheEntry == NULL) { |
210 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
211 | t->deleteIfZeroRefCount(); |
212 | return; |
213 | } |
214 | data = t->data; |
215 | settings = t->settings; |
216 | settings->addRef(); |
217 | tailoring = t; |
218 | cacheEntry->addRef(); |
219 | validLocale = t->actualLocale; |
220 | actualLocaleIsSameAsValid = FALSE; |
221 | } |
222 | |
223 | RuleBasedCollator * |
224 | RuleBasedCollator::clone() const { |
225 | return new RuleBasedCollator(*this); |
226 | } |
227 | |
228 | RuleBasedCollator &RuleBasedCollator::operator=(const RuleBasedCollator &other) { |
229 | if(this == &other) { return *this; } |
230 | SharedObject::copyPtr(other.settings, settings); |
231 | tailoring = other.tailoring; |
232 | SharedObject::copyPtr(other.cacheEntry, cacheEntry); |
233 | data = tailoring->data; |
234 | validLocale = other.validLocale; |
235 | explicitlySetAttributes = other.explicitlySetAttributes; |
236 | actualLocaleIsSameAsValid = other.actualLocaleIsSameAsValid; |
237 | return *this; |
238 | } |
239 | |
240 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(RuleBasedCollator) |
241 | |
242 | UBool |
243 | RuleBasedCollator::operator==(const Collator& other) const { |
244 | if(this == &other) { return TRUE; } |
245 | if(!Collator::operator==(other)) { return FALSE; } |
246 | const RuleBasedCollator &o = static_cast<const RuleBasedCollator &>(other); |
247 | if(*settings != *o.settings) { return FALSE; } |
248 | if(data == o.data) { return TRUE; } |
249 | UBool thisIsRoot = data->base == NULL; |
250 | UBool otherIsRoot = o.data->base == NULL; |
251 | U_ASSERT(!thisIsRoot || !otherIsRoot); // otherwise their data pointers should be == |
252 | if(thisIsRoot != otherIsRoot) { return FALSE; } |
253 | if((thisIsRoot || !tailoring->rules.isEmpty()) && |
254 | (otherIsRoot || !o.tailoring->rules.isEmpty())) { |
255 | // Shortcut: If both collators have valid rule strings, then compare those. |
256 | if(tailoring->rules == o.tailoring->rules) { return TRUE; } |
257 | } |
258 | // Different rule strings can result in the same or equivalent tailoring. |
259 | // The rule strings are optional in ICU resource bundles, although included by default. |
260 | // cloneBinary() drops the rule string. |
261 | UErrorCode errorCode = U_ZERO_ERROR; |
262 | LocalPointer<UnicodeSet> thisTailored(getTailoredSet(errorCode)); |
263 | LocalPointer<UnicodeSet> otherTailored(o.getTailoredSet(errorCode)); |
264 | if(U_FAILURE(errorCode)) { return FALSE; } |
265 | if(*thisTailored != *otherTailored) { return FALSE; } |
266 | // For completeness, we should compare all of the mappings; |
267 | // or we should create a list of strings, sort it with one collator, |
268 | // and check if both collators compare adjacent strings the same |
269 | // (order & strength, down to quaternary); or similar. |
270 | // Testing equality of collators seems unusual. |
271 | return TRUE; |
272 | } |
273 | |
274 | int32_t |
275 | RuleBasedCollator::hashCode() const { |
276 | int32_t h = settings->hashCode(); |
277 | if(data->base == NULL) { return h; } // root collator |
278 | // Do not rely on the rule string, see comments in operator==(). |
279 | UErrorCode errorCode = U_ZERO_ERROR; |
280 | LocalPointer<UnicodeSet> set(getTailoredSet(errorCode)); |
281 | if(U_FAILURE(errorCode)) { return 0; } |
282 | UnicodeSetIterator iter(*set); |
283 | while(iter.next() && !iter.isString()) { |
284 | h ^= data->getCE32(iter.getCodepoint()); |
285 | } |
286 | return h; |
287 | } |
288 | |
289 | void |
290 | RuleBasedCollator::setLocales(const Locale &requested, const Locale &valid, |
291 | const Locale &actual) { |
292 | if(actual == tailoring->actualLocale) { |
293 | actualLocaleIsSameAsValid = FALSE; |
294 | } else { |
295 | U_ASSERT(actual == valid); |
296 | actualLocaleIsSameAsValid = TRUE; |
297 | } |
298 | // Do not modify tailoring.actualLocale: |
299 | // We cannot be sure that that would be thread-safe. |
300 | validLocale = valid; |
301 | (void)requested; // Ignore, see also ticket #10477. |
302 | } |
303 | |
304 | Locale |
305 | RuleBasedCollator::getLocale(ULocDataLocaleType type, UErrorCode& errorCode) const { |
306 | if(U_FAILURE(errorCode)) { |
307 | return Locale::getRoot(); |
308 | } |
309 | switch(type) { |
310 | case ULOC_ACTUAL_LOCALE: |
311 | return actualLocaleIsSameAsValid ? validLocale : tailoring->actualLocale; |
312 | case ULOC_VALID_LOCALE: |
313 | return validLocale; |
314 | case ULOC_REQUESTED_LOCALE: |
315 | default: |
316 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
317 | return Locale::getRoot(); |
318 | } |
319 | } |
320 | |
321 | const char * |
322 | RuleBasedCollator::internalGetLocaleID(ULocDataLocaleType type, UErrorCode &errorCode) const { |
323 | if(U_FAILURE(errorCode)) { |
324 | return NULL; |
325 | } |
326 | const Locale *result; |
327 | switch(type) { |
328 | case ULOC_ACTUAL_LOCALE: |
329 | result = actualLocaleIsSameAsValid ? &validLocale : &tailoring->actualLocale; |
330 | break; |
331 | case ULOC_VALID_LOCALE: |
332 | result = &validLocale; |
333 | break; |
334 | case ULOC_REQUESTED_LOCALE: |
335 | default: |
336 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
337 | return NULL; |
338 | } |
339 | if(result->isBogus()) { return NULL; } |
340 | const char *id = result->getName(); |
341 | return id[0] == 0 ? "root" : id; |
342 | } |
343 | |
344 | const UnicodeString& |
345 | RuleBasedCollator::getRules() const { |
346 | return tailoring->rules; |
347 | } |
348 | |
349 | void |
350 | RuleBasedCollator::getRules(UColRuleOption delta, UnicodeString &buffer) const { |
351 | if(delta == UCOL_TAILORING_ONLY) { |
352 | buffer = tailoring->rules; |
353 | return; |
354 | } |
355 | // UCOL_FULL_RULES |
356 | buffer.remove(); |
357 | CollationLoader::appendRootRules(buffer); |
358 | buffer.append(tailoring->rules).getTerminatedBuffer(); |
359 | } |
360 | |
361 | void |
362 | RuleBasedCollator::getVersion(UVersionInfo version) const { |
363 | uprv_memcpy(version, tailoring->version, U_MAX_VERSION_LENGTH); |
364 | version[0] += (UCOL_RUNTIME_VERSION << 4) + (UCOL_RUNTIME_VERSION >> 4); |
365 | } |
366 | |
367 | UnicodeSet * |
368 | RuleBasedCollator::getTailoredSet(UErrorCode &errorCode) const { |
369 | if(U_FAILURE(errorCode)) { return NULL; } |
370 | UnicodeSet *tailored = new UnicodeSet(); |
371 | if(tailored == NULL) { |
372 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
373 | return NULL; |
374 | } |
375 | if(data->base != NULL) { |
376 | TailoredSet(tailored).forData(data, errorCode); |
377 | if(U_FAILURE(errorCode)) { |
378 | delete tailored; |
379 | return NULL; |
380 | } |
381 | } |
382 | return tailored; |
383 | } |
384 | |
385 | void |
386 | RuleBasedCollator::internalGetContractionsAndExpansions( |
387 | UnicodeSet *contractions, UnicodeSet *expansions, |
388 | UBool addPrefixes, UErrorCode &errorCode) const { |
389 | if(U_FAILURE(errorCode)) { return; } |
390 | if(contractions != NULL) { |
391 | contractions->clear(); |
392 | } |
393 | if(expansions != NULL) { |
394 | expansions->clear(); |
395 | } |
396 | ContractionsAndExpansions(contractions, expansions, NULL, addPrefixes).forData(data, errorCode); |
397 | } |
398 | |
399 | void |
400 | RuleBasedCollator::internalAddContractions(UChar32 c, UnicodeSet &set, UErrorCode &errorCode) const { |
401 | if(U_FAILURE(errorCode)) { return; } |
402 | ContractionsAndExpansions(&set, NULL, NULL, FALSE).forCodePoint(data, c, errorCode); |
403 | } |
404 | |
405 | const CollationSettings & |
406 | RuleBasedCollator::getDefaultSettings() const { |
407 | return *tailoring->settings; |
408 | } |
409 | |
410 | UColAttributeValue |
411 | RuleBasedCollator::getAttribute(UColAttribute attr, UErrorCode &errorCode) const { |
412 | if(U_FAILURE(errorCode)) { return UCOL_DEFAULT; } |
413 | int32_t option; |
414 | switch(attr) { |
415 | case UCOL_FRENCH_COLLATION: |
416 | option = CollationSettings::BACKWARD_SECONDARY; |
417 | break; |
418 | case UCOL_ALTERNATE_HANDLING: |
419 | return settings->getAlternateHandling(); |
420 | case UCOL_CASE_FIRST: |
421 | return settings->getCaseFirst(); |
422 | case UCOL_CASE_LEVEL: |
423 | option = CollationSettings::CASE_LEVEL; |
424 | break; |
425 | case UCOL_NORMALIZATION_MODE: |
426 | option = CollationSettings::CHECK_FCD; |
427 | break; |
428 | case UCOL_STRENGTH: |
429 | return (UColAttributeValue)settings->getStrength(); |
430 | case UCOL_HIRAGANA_QUATERNARY_MODE: |
431 | // Deprecated attribute, unsettable. |
432 | return UCOL_OFF; |
433 | case UCOL_NUMERIC_COLLATION: |
434 | option = CollationSettings::NUMERIC; |
435 | break; |
436 | default: |
437 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
438 | return UCOL_DEFAULT; |
439 | } |
440 | return ((settings->options & option) == 0) ? UCOL_OFF : UCOL_ON; |
441 | } |
442 | |
443 | void |
444 | RuleBasedCollator::setAttribute(UColAttribute attr, UColAttributeValue value, |
445 | UErrorCode &errorCode) { |
446 | UColAttributeValue oldValue = getAttribute(attr, errorCode); |
447 | if(U_FAILURE(errorCode)) { return; } |
448 | if(value == oldValue) { |
449 | setAttributeExplicitly(attr); |
450 | return; |
451 | } |
452 | const CollationSettings &defaultSettings = getDefaultSettings(); |
453 | if(settings == &defaultSettings) { |
454 | if(value == UCOL_DEFAULT) { |
455 | setAttributeDefault(attr); |
456 | return; |
457 | } |
458 | } |
459 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); |
460 | if(ownedSettings == NULL) { |
461 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
462 | return; |
463 | } |
464 | |
465 | switch(attr) { |
466 | case UCOL_FRENCH_COLLATION: |
467 | ownedSettings->setFlag(CollationSettings::BACKWARD_SECONDARY, value, |
468 | defaultSettings.options, errorCode); |
469 | break; |
470 | case UCOL_ALTERNATE_HANDLING: |
471 | ownedSettings->setAlternateHandling(value, defaultSettings.options, errorCode); |
472 | break; |
473 | case UCOL_CASE_FIRST: |
474 | ownedSettings->setCaseFirst(value, defaultSettings.options, errorCode); |
475 | break; |
476 | case UCOL_CASE_LEVEL: |
477 | ownedSettings->setFlag(CollationSettings::CASE_LEVEL, value, |
478 | defaultSettings.options, errorCode); |
479 | break; |
480 | case UCOL_NORMALIZATION_MODE: |
481 | ownedSettings->setFlag(CollationSettings::CHECK_FCD, value, |
482 | defaultSettings.options, errorCode); |
483 | break; |
484 | case UCOL_STRENGTH: |
485 | ownedSettings->setStrength(value, defaultSettings.options, errorCode); |
486 | break; |
487 | case UCOL_HIRAGANA_QUATERNARY_MODE: |
488 | // Deprecated attribute. Check for valid values but do not change anything. |
489 | if(value != UCOL_OFF && value != UCOL_ON && value != UCOL_DEFAULT) { |
490 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
491 | } |
492 | break; |
493 | case UCOL_NUMERIC_COLLATION: |
494 | ownedSettings->setFlag(CollationSettings::NUMERIC, value, defaultSettings.options, errorCode); |
495 | break; |
496 | default: |
497 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
498 | break; |
499 | } |
500 | if(U_FAILURE(errorCode)) { return; } |
501 | setFastLatinOptions(*ownedSettings); |
502 | if(value == UCOL_DEFAULT) { |
503 | setAttributeDefault(attr); |
504 | } else { |
505 | setAttributeExplicitly(attr); |
506 | } |
507 | } |
508 | |
509 | Collator & |
510 | RuleBasedCollator::setMaxVariable(UColReorderCode group, UErrorCode &errorCode) { |
511 | if(U_FAILURE(errorCode)) { return *this; } |
512 | // Convert the reorder code into a MaxVariable number, or UCOL_DEFAULT=-1. |
513 | int32_t value; |
514 | if(group == UCOL_REORDER_CODE_DEFAULT) { |
515 | value = UCOL_DEFAULT; |
516 | } else if(UCOL_REORDER_CODE_FIRST <= group && group <= UCOL_REORDER_CODE_CURRENCY) { |
517 | value = group - UCOL_REORDER_CODE_FIRST; |
518 | } else { |
519 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
520 | return *this; |
521 | } |
522 | CollationSettings::MaxVariable oldValue = settings->getMaxVariable(); |
523 | if(value == oldValue) { |
524 | setAttributeExplicitly(ATTR_VARIABLE_TOP); |
525 | return *this; |
526 | } |
527 | const CollationSettings &defaultSettings = getDefaultSettings(); |
528 | if(settings == &defaultSettings) { |
529 | if(value == UCOL_DEFAULT) { |
530 | setAttributeDefault(ATTR_VARIABLE_TOP); |
531 | return *this; |
532 | } |
533 | } |
534 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); |
535 | if(ownedSettings == NULL) { |
536 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
537 | return *this; |
538 | } |
539 | |
540 | if(group == UCOL_REORDER_CODE_DEFAULT) { |
541 | group = (UColReorderCode)(UCOL_REORDER_CODE_FIRST + defaultSettings.getMaxVariable()); |
542 | } |
543 | uint32_t varTop = data->getLastPrimaryForGroup(group); |
544 | U_ASSERT(varTop != 0); |
545 | ownedSettings->setMaxVariable(value, defaultSettings.options, errorCode); |
546 | if(U_FAILURE(errorCode)) { return *this; } |
547 | ownedSettings->variableTop = varTop; |
548 | setFastLatinOptions(*ownedSettings); |
549 | if(value == UCOL_DEFAULT) { |
550 | setAttributeDefault(ATTR_VARIABLE_TOP); |
551 | } else { |
552 | setAttributeExplicitly(ATTR_VARIABLE_TOP); |
553 | } |
554 | return *this; |
555 | } |
556 | |
557 | UColReorderCode |
558 | RuleBasedCollator::getMaxVariable() const { |
559 | return (UColReorderCode)(UCOL_REORDER_CODE_FIRST + settings->getMaxVariable()); |
560 | } |
561 | |
562 | uint32_t |
563 | RuleBasedCollator::getVariableTop(UErrorCode & /*errorCode*/) const { |
564 | return settings->variableTop; |
565 | } |
566 | |
567 | uint32_t |
568 | RuleBasedCollator::setVariableTop(const UChar *varTop, int32_t len, UErrorCode &errorCode) { |
569 | if(U_FAILURE(errorCode)) { return 0; } |
570 | if(varTop == NULL && len !=0) { |
571 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
572 | return 0; |
573 | } |
574 | if(len < 0) { len = u_strlen(varTop); } |
575 | if(len == 0) { |
576 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
577 | return 0; |
578 | } |
579 | UBool numeric = settings->isNumeric(); |
580 | int64_t ce1, ce2; |
581 | if(settings->dontCheckFCD()) { |
582 | UTF16CollationIterator ci(data, numeric, varTop, varTop, varTop + len); |
583 | ce1 = ci.nextCE(errorCode); |
584 | ce2 = ci.nextCE(errorCode); |
585 | } else { |
586 | FCDUTF16CollationIterator ci(data, numeric, varTop, varTop, varTop + len); |
587 | ce1 = ci.nextCE(errorCode); |
588 | ce2 = ci.nextCE(errorCode); |
589 | } |
590 | if(ce1 == Collation::NO_CE || ce2 != Collation::NO_CE) { |
591 | errorCode = U_CE_NOT_FOUND_ERROR; |
592 | return 0; |
593 | } |
594 | setVariableTop((uint32_t)(ce1 >> 32), errorCode); |
595 | return settings->variableTop; |
596 | } |
597 | |
598 | uint32_t |
599 | RuleBasedCollator::setVariableTop(const UnicodeString &varTop, UErrorCode &errorCode) { |
600 | return setVariableTop(varTop.getBuffer(), varTop.length(), errorCode); |
601 | } |
602 | |
603 | void |
604 | RuleBasedCollator::setVariableTop(uint32_t varTop, UErrorCode &errorCode) { |
605 | if(U_FAILURE(errorCode)) { return; } |
606 | if(varTop != settings->variableTop) { |
607 | // Pin the variable top to the end of the reordering group which contains it. |
608 | // Only a few special groups are supported. |
609 | int32_t group = data->getGroupForPrimary(varTop); |
610 | if(group < UCOL_REORDER_CODE_FIRST || UCOL_REORDER_CODE_CURRENCY < group) { |
611 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
612 | return; |
613 | } |
614 | uint32_t v = data->getLastPrimaryForGroup(group); |
615 | U_ASSERT(v != 0 && v >= varTop); |
616 | varTop = v; |
617 | if(varTop != settings->variableTop) { |
618 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); |
619 | if(ownedSettings == NULL) { |
620 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
621 | return; |
622 | } |
623 | ownedSettings->setMaxVariable(group - UCOL_REORDER_CODE_FIRST, |
624 | getDefaultSettings().options, errorCode); |
625 | if(U_FAILURE(errorCode)) { return; } |
626 | ownedSettings->variableTop = varTop; |
627 | setFastLatinOptions(*ownedSettings); |
628 | } |
629 | } |
630 | if(varTop == getDefaultSettings().variableTop) { |
631 | setAttributeDefault(ATTR_VARIABLE_TOP); |
632 | } else { |
633 | setAttributeExplicitly(ATTR_VARIABLE_TOP); |
634 | } |
635 | } |
636 | |
637 | int32_t |
638 | RuleBasedCollator::getReorderCodes(int32_t *dest, int32_t capacity, |
639 | UErrorCode &errorCode) const { |
640 | if(U_FAILURE(errorCode)) { return 0; } |
641 | if(capacity < 0 || (dest == NULL && capacity > 0)) { |
642 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
643 | return 0; |
644 | } |
645 | int32_t length = settings->reorderCodesLength; |
646 | if(length == 0) { return 0; } |
647 | if(length > capacity) { |
648 | errorCode = U_BUFFER_OVERFLOW_ERROR; |
649 | return length; |
650 | } |
651 | uprv_memcpy(dest, settings->reorderCodes, length * 4); |
652 | return length; |
653 | } |
654 | |
655 | void |
656 | RuleBasedCollator::setReorderCodes(const int32_t *reorderCodes, int32_t length, |
657 | UErrorCode &errorCode) { |
658 | if(U_FAILURE(errorCode)) { return; } |
659 | if(length < 0 || (reorderCodes == NULL && length > 0)) { |
660 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
661 | return; |
662 | } |
663 | if(length == 1 && reorderCodes[0] == UCOL_REORDER_CODE_NONE) { |
664 | length = 0; |
665 | } |
666 | if(length == settings->reorderCodesLength && |
667 | uprv_memcmp(reorderCodes, settings->reorderCodes, length * 4) == 0) { |
668 | return; |
669 | } |
670 | const CollationSettings &defaultSettings = getDefaultSettings(); |
671 | if(length == 1 && reorderCodes[0] == UCOL_REORDER_CODE_DEFAULT) { |
672 | if(settings != &defaultSettings) { |
673 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); |
674 | if(ownedSettings == NULL) { |
675 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
676 | return; |
677 | } |
678 | ownedSettings->copyReorderingFrom(defaultSettings, errorCode); |
679 | setFastLatinOptions(*ownedSettings); |
680 | } |
681 | return; |
682 | } |
683 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); |
684 | if(ownedSettings == NULL) { |
685 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
686 | return; |
687 | } |
688 | ownedSettings->setReordering(*data, reorderCodes, length, errorCode); |
689 | setFastLatinOptions(*ownedSettings); |
690 | } |
691 | |
692 | void |
693 | RuleBasedCollator::setFastLatinOptions(CollationSettings &ownedSettings) const { |
694 | ownedSettings.fastLatinOptions = CollationFastLatin::getOptions( |
695 | data, ownedSettings, |
696 | ownedSettings.fastLatinPrimaries, UPRV_LENGTHOF(ownedSettings.fastLatinPrimaries)); |
697 | } |
698 | |
699 | UCollationResult |
700 | RuleBasedCollator::compare(const UnicodeString &left, const UnicodeString &right, |
701 | UErrorCode &errorCode) const { |
702 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } |
703 | return doCompare(left.getBuffer(), left.length(), |
704 | right.getBuffer(), right.length(), errorCode); |
705 | } |
706 | |
707 | UCollationResult |
708 | RuleBasedCollator::compare(const UnicodeString &left, const UnicodeString &right, |
709 | int32_t length, UErrorCode &errorCode) const { |
710 | if(U_FAILURE(errorCode) || length == 0) { return UCOL_EQUAL; } |
711 | if(length < 0) { |
712 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
713 | return UCOL_EQUAL; |
714 | } |
715 | int32_t leftLength = left.length(); |
716 | int32_t rightLength = right.length(); |
717 | if(leftLength > length) { leftLength = length; } |
718 | if(rightLength > length) { rightLength = length; } |
719 | return doCompare(left.getBuffer(), leftLength, |
720 | right.getBuffer(), rightLength, errorCode); |
721 | } |
722 | |
723 | UCollationResult |
724 | RuleBasedCollator::compare(const UChar *left, int32_t leftLength, |
725 | const UChar *right, int32_t rightLength, |
726 | UErrorCode &errorCode) const { |
727 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } |
728 | if((left == NULL && leftLength != 0) || (right == NULL && rightLength != 0)) { |
729 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
730 | return UCOL_EQUAL; |
731 | } |
732 | // Make sure both or neither strings have a known length. |
733 | // We do not optimize for mixed length/termination. |
734 | if(leftLength >= 0) { |
735 | if(rightLength < 0) { rightLength = u_strlen(right); } |
736 | } else { |
737 | if(rightLength >= 0) { leftLength = u_strlen(left); } |
738 | } |
739 | return doCompare(left, leftLength, right, rightLength, errorCode); |
740 | } |
741 | |
742 | UCollationResult |
743 | RuleBasedCollator::compareUTF8(const StringPiece &left, const StringPiece &right, |
744 | UErrorCode &errorCode) const { |
745 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } |
746 | const uint8_t *leftBytes = reinterpret_cast<const uint8_t *>(left.data()); |
747 | const uint8_t *rightBytes = reinterpret_cast<const uint8_t *>(right.data()); |
748 | if((leftBytes == NULL && !left.empty()) || (rightBytes == NULL && !right.empty())) { |
749 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
750 | return UCOL_EQUAL; |
751 | } |
752 | return doCompare(leftBytes, left.length(), rightBytes, right.length(), errorCode); |
753 | } |
754 | |
755 | UCollationResult |
756 | RuleBasedCollator::internalCompareUTF8(const char *left, int32_t leftLength, |
757 | const char *right, int32_t rightLength, |
758 | UErrorCode &errorCode) const { |
759 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } |
760 | if((left == NULL && leftLength != 0) || (right == NULL && rightLength != 0)) { |
761 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
762 | return UCOL_EQUAL; |
763 | } |
764 | // Make sure both or neither strings have a known length. |
765 | // We do not optimize for mixed length/termination. |
766 | if(leftLength >= 0) { |
767 | if(rightLength < 0) { rightLength = static_cast<int32_t>(uprv_strlen(right)); } |
768 | } else { |
769 | if(rightLength >= 0) { leftLength = static_cast<int32_t>(uprv_strlen(left)); } |
770 | } |
771 | return doCompare(reinterpret_cast<const uint8_t *>(left), leftLength, |
772 | reinterpret_cast<const uint8_t *>(right), rightLength, errorCode); |
773 | } |
774 | |
775 | namespace { |
776 | |
777 | /** |
778 | * Abstract iterator for identical-level string comparisons. |
779 | * Returns FCD code points and handles temporary switching to NFD. |
780 | */ |
781 | class NFDIterator : public UObject { |
782 | public: |
783 | NFDIterator() : index(-1), length(0) {} |
784 | virtual ~NFDIterator() {} |
785 | /** |
786 | * Returns the next code point from the internal normalization buffer, |
787 | * or else the next text code point. |
788 | * Returns -1 at the end of the text. |
789 | */ |
790 | UChar32 nextCodePoint() { |
791 | if(index >= 0) { |
792 | if(index == length) { |
793 | index = -1; |
794 | } else { |
795 | UChar32 c; |
796 | U16_NEXT_UNSAFE(decomp, index, c); |
797 | return c; |
798 | } |
799 | } |
800 | return nextRawCodePoint(); |
801 | } |
802 | /** |
803 | * @param nfcImpl |
804 | * @param c the last code point returned by nextCodePoint() or nextDecomposedCodePoint() |
805 | * @return the first code point in c's decomposition, |
806 | * or c itself if it was decomposed already or if it does not decompose |
807 | */ |
808 | UChar32 nextDecomposedCodePoint(const Normalizer2Impl &nfcImpl, UChar32 c) { |
809 | if(index >= 0) { return c; } |
810 | decomp = nfcImpl.getDecomposition(c, buffer, length); |
811 | if(decomp == NULL) { return c; } |
812 | index = 0; |
813 | U16_NEXT_UNSAFE(decomp, index, c); |
814 | return c; |
815 | } |
816 | protected: |
817 | /** |
818 | * Returns the next text code point in FCD order. |
819 | * Returns -1 at the end of the text. |
820 | */ |
821 | virtual UChar32 () = 0; |
822 | private: |
823 | const UChar *decomp; |
824 | UChar buffer[4]; |
825 | int32_t index; |
826 | int32_t length; |
827 | }; |
828 | |
829 | class UTF16NFDIterator : public NFDIterator { |
830 | public: |
831 | UTF16NFDIterator(const UChar *text, const UChar *textLimit) : s(text), limit(textLimit) {} |
832 | protected: |
833 | virtual UChar32 () { |
834 | if(s == limit) { return U_SENTINEL; } |
835 | UChar32 c = *s++; |
836 | if(limit == NULL && c == 0) { |
837 | s = NULL; |
838 | return U_SENTINEL; |
839 | } |
840 | UChar trail; |
841 | if(U16_IS_LEAD(c) && s != limit && U16_IS_TRAIL(trail = *s)) { |
842 | ++s; |
843 | c = U16_GET_SUPPLEMENTARY(c, trail); |
844 | } |
845 | return c; |
846 | } |
847 | |
848 | const UChar *s; |
849 | const UChar *limit; |
850 | }; |
851 | |
852 | class FCDUTF16NFDIterator : public UTF16NFDIterator { |
853 | public: |
854 | FCDUTF16NFDIterator(const Normalizer2Impl &nfcImpl, const UChar *text, const UChar *textLimit) |
855 | : UTF16NFDIterator(NULL, NULL) { |
856 | UErrorCode errorCode = U_ZERO_ERROR; |
857 | const UChar *spanLimit = nfcImpl.makeFCD(text, textLimit, NULL, errorCode); |
858 | if(U_FAILURE(errorCode)) { return; } |
859 | if(spanLimit == textLimit || (textLimit == NULL && *spanLimit == 0)) { |
860 | s = text; |
861 | limit = spanLimit; |
862 | } else { |
863 | str.setTo(text, (int32_t)(spanLimit - text)); |
864 | { |
865 | ReorderingBuffer r_buffer(nfcImpl, str); |
866 | if(r_buffer.init(str.length(), errorCode)) { |
867 | nfcImpl.makeFCD(spanLimit, textLimit, &r_buffer, errorCode); |
868 | } |
869 | } |
870 | if(U_SUCCESS(errorCode)) { |
871 | s = str.getBuffer(); |
872 | limit = s + str.length(); |
873 | } |
874 | } |
875 | } |
876 | private: |
877 | UnicodeString str; |
878 | }; |
879 | |
880 | class UTF8NFDIterator : public NFDIterator { |
881 | public: |
882 | UTF8NFDIterator(const uint8_t *text, int32_t textLength) |
883 | : s(text), pos(0), length(textLength) {} |
884 | protected: |
885 | virtual UChar32 () { |
886 | if(pos == length || (s[pos] == 0 && length < 0)) { return U_SENTINEL; } |
887 | UChar32 c; |
888 | U8_NEXT_OR_FFFD(s, pos, length, c); |
889 | return c; |
890 | } |
891 | |
892 | const uint8_t *s; |
893 | int32_t pos; |
894 | int32_t length; |
895 | }; |
896 | |
897 | class FCDUTF8NFDIterator : public NFDIterator { |
898 | public: |
899 | FCDUTF8NFDIterator(const CollationData *data, const uint8_t *text, int32_t textLength) |
900 | : u8ci(data, FALSE, text, 0, textLength) {} |
901 | protected: |
902 | virtual UChar32 () { |
903 | UErrorCode errorCode = U_ZERO_ERROR; |
904 | return u8ci.nextCodePoint(errorCode); |
905 | } |
906 | private: |
907 | FCDUTF8CollationIterator u8ci; |
908 | }; |
909 | |
910 | class UIterNFDIterator : public NFDIterator { |
911 | public: |
912 | UIterNFDIterator(UCharIterator &it) : iter(it) {} |
913 | protected: |
914 | virtual UChar32 () { |
915 | return uiter_next32(&iter); |
916 | } |
917 | private: |
918 | UCharIterator &iter; |
919 | }; |
920 | |
921 | class FCDUIterNFDIterator : public NFDIterator { |
922 | public: |
923 | FCDUIterNFDIterator(const CollationData *data, UCharIterator &it, int32_t startIndex) |
924 | : uici(data, FALSE, it, startIndex) {} |
925 | protected: |
926 | virtual UChar32 () { |
927 | UErrorCode errorCode = U_ZERO_ERROR; |
928 | return uici.nextCodePoint(errorCode); |
929 | } |
930 | private: |
931 | FCDUIterCollationIterator uici; |
932 | }; |
933 | |
934 | UCollationResult compareNFDIter(const Normalizer2Impl &nfcImpl, |
935 | NFDIterator &left, NFDIterator &right) { |
936 | for(;;) { |
937 | // Fetch the next FCD code point from each string. |
938 | UChar32 leftCp = left.nextCodePoint(); |
939 | UChar32 rightCp = right.nextCodePoint(); |
940 | if(leftCp == rightCp) { |
941 | if(leftCp < 0) { break; } |
942 | continue; |
943 | } |
944 | // If they are different, then decompose each and compare again. |
945 | if(leftCp < 0) { |
946 | leftCp = -2; // end of string |
947 | } else if(leftCp == 0xfffe) { |
948 | leftCp = -1; // U+FFFE: merge separator |
949 | } else { |
950 | leftCp = left.nextDecomposedCodePoint(nfcImpl, leftCp); |
951 | } |
952 | if(rightCp < 0) { |
953 | rightCp = -2; // end of string |
954 | } else if(rightCp == 0xfffe) { |
955 | rightCp = -1; // U+FFFE: merge separator |
956 | } else { |
957 | rightCp = right.nextDecomposedCodePoint(nfcImpl, rightCp); |
958 | } |
959 | if(leftCp < rightCp) { return UCOL_LESS; } |
960 | if(leftCp > rightCp) { return UCOL_GREATER; } |
961 | } |
962 | return UCOL_EQUAL; |
963 | } |
964 | |
965 | } // namespace |
966 | |
967 | UCollationResult |
968 | RuleBasedCollator::doCompare(const UChar *left, int32_t leftLength, |
969 | const UChar *right, int32_t rightLength, |
970 | UErrorCode &errorCode) const { |
971 | // U_FAILURE(errorCode) checked by caller. |
972 | if(left == right && leftLength == rightLength) { |
973 | return UCOL_EQUAL; |
974 | } |
975 | |
976 | // Identical-prefix test. |
977 | const UChar *leftLimit; |
978 | const UChar *rightLimit; |
979 | int32_t equalPrefixLength = 0; |
980 | if(leftLength < 0) { |
981 | leftLimit = NULL; |
982 | rightLimit = NULL; |
983 | UChar c; |
984 | while((c = left[equalPrefixLength]) == right[equalPrefixLength]) { |
985 | if(c == 0) { return UCOL_EQUAL; } |
986 | ++equalPrefixLength; |
987 | } |
988 | } else { |
989 | leftLimit = left + leftLength; |
990 | rightLimit = right + rightLength; |
991 | for(;;) { |
992 | if(equalPrefixLength == leftLength) { |
993 | if(equalPrefixLength == rightLength) { return UCOL_EQUAL; } |
994 | break; |
995 | } else if(equalPrefixLength == rightLength || |
996 | left[equalPrefixLength] != right[equalPrefixLength]) { |
997 | break; |
998 | } |
999 | ++equalPrefixLength; |
1000 | } |
1001 | } |
1002 | |
1003 | UBool numeric = settings->isNumeric(); |
1004 | if(equalPrefixLength > 0) { |
1005 | if((equalPrefixLength != leftLength && |
1006 | data->isUnsafeBackward(left[equalPrefixLength], numeric)) || |
1007 | (equalPrefixLength != rightLength && |
1008 | data->isUnsafeBackward(right[equalPrefixLength], numeric))) { |
1009 | // Identical prefix: Back up to the start of a contraction or reordering sequence. |
1010 | while(--equalPrefixLength > 0 && |
1011 | data->isUnsafeBackward(left[equalPrefixLength], numeric)) {} |
1012 | } |
1013 | // Notes: |
1014 | // - A longer string can compare equal to a prefix of it if only ignorables follow. |
1015 | // - With a backward level, a longer string can compare less-than a prefix of it. |
1016 | |
1017 | // Pass the actual start of each string into the CollationIterators, |
1018 | // plus the equalPrefixLength position, |
1019 | // so that prefix matches back into the equal prefix work. |
1020 | } |
1021 | |
1022 | int32_t result; |
1023 | int32_t fastLatinOptions = settings->fastLatinOptions; |
1024 | if(fastLatinOptions >= 0 && |
1025 | (equalPrefixLength == leftLength || |
1026 | left[equalPrefixLength] <= CollationFastLatin::LATIN_MAX) && |
1027 | (equalPrefixLength == rightLength || |
1028 | right[equalPrefixLength] <= CollationFastLatin::LATIN_MAX)) { |
1029 | if(leftLength >= 0) { |
1030 | result = CollationFastLatin::compareUTF16(data->fastLatinTable, |
1031 | settings->fastLatinPrimaries, |
1032 | fastLatinOptions, |
1033 | left + equalPrefixLength, |
1034 | leftLength - equalPrefixLength, |
1035 | right + equalPrefixLength, |
1036 | rightLength - equalPrefixLength); |
1037 | } else { |
1038 | result = CollationFastLatin::compareUTF16(data->fastLatinTable, |
1039 | settings->fastLatinPrimaries, |
1040 | fastLatinOptions, |
1041 | left + equalPrefixLength, -1, |
1042 | right + equalPrefixLength, -1); |
1043 | } |
1044 | } else { |
1045 | result = CollationFastLatin::BAIL_OUT_RESULT; |
1046 | } |
1047 | |
1048 | if(result == CollationFastLatin::BAIL_OUT_RESULT) { |
1049 | if(settings->dontCheckFCD()) { |
1050 | UTF16CollationIterator leftIter(data, numeric, |
1051 | left, left + equalPrefixLength, leftLimit); |
1052 | UTF16CollationIterator rightIter(data, numeric, |
1053 | right, right + equalPrefixLength, rightLimit); |
1054 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1055 | } else { |
1056 | FCDUTF16CollationIterator leftIter(data, numeric, |
1057 | left, left + equalPrefixLength, leftLimit); |
1058 | FCDUTF16CollationIterator rightIter(data, numeric, |
1059 | right, right + equalPrefixLength, rightLimit); |
1060 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1061 | } |
1062 | } |
1063 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { |
1064 | return (UCollationResult)result; |
1065 | } |
1066 | |
1067 | // Note: If NUL-terminated, we could get the actual limits from the iterators now. |
1068 | // That would complicate the iterators a bit, NUL-terminated strings are only a C convenience, |
1069 | // and the benefit seems unlikely to be measurable. |
1070 | |
1071 | // Compare identical level. |
1072 | const Normalizer2Impl &nfcImpl = data->nfcImpl; |
1073 | left += equalPrefixLength; |
1074 | right += equalPrefixLength; |
1075 | if(settings->dontCheckFCD()) { |
1076 | UTF16NFDIterator leftIter(left, leftLimit); |
1077 | UTF16NFDIterator rightIter(right, rightLimit); |
1078 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1079 | } else { |
1080 | FCDUTF16NFDIterator leftIter(nfcImpl, left, leftLimit); |
1081 | FCDUTF16NFDIterator rightIter(nfcImpl, right, rightLimit); |
1082 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1083 | } |
1084 | } |
1085 | |
1086 | UCollationResult |
1087 | RuleBasedCollator::doCompare(const uint8_t *left, int32_t leftLength, |
1088 | const uint8_t *right, int32_t rightLength, |
1089 | UErrorCode &errorCode) const { |
1090 | // U_FAILURE(errorCode) checked by caller. |
1091 | if(left == right && leftLength == rightLength) { |
1092 | return UCOL_EQUAL; |
1093 | } |
1094 | |
1095 | // Identical-prefix test. |
1096 | int32_t equalPrefixLength = 0; |
1097 | if(leftLength < 0) { |
1098 | uint8_t c; |
1099 | while((c = left[equalPrefixLength]) == right[equalPrefixLength]) { |
1100 | if(c == 0) { return UCOL_EQUAL; } |
1101 | ++equalPrefixLength; |
1102 | } |
1103 | } else { |
1104 | for(;;) { |
1105 | if(equalPrefixLength == leftLength) { |
1106 | if(equalPrefixLength == rightLength) { return UCOL_EQUAL; } |
1107 | break; |
1108 | } else if(equalPrefixLength == rightLength || |
1109 | left[equalPrefixLength] != right[equalPrefixLength]) { |
1110 | break; |
1111 | } |
1112 | ++equalPrefixLength; |
1113 | } |
1114 | } |
1115 | // Back up to the start of a partially-equal code point. |
1116 | if(equalPrefixLength > 0 && |
1117 | ((equalPrefixLength != leftLength && U8_IS_TRAIL(left[equalPrefixLength])) || |
1118 | (equalPrefixLength != rightLength && U8_IS_TRAIL(right[equalPrefixLength])))) { |
1119 | while(--equalPrefixLength > 0 && U8_IS_TRAIL(left[equalPrefixLength])) {} |
1120 | } |
1121 | |
1122 | UBool numeric = settings->isNumeric(); |
1123 | if(equalPrefixLength > 0) { |
1124 | UBool unsafe = FALSE; |
1125 | if(equalPrefixLength != leftLength) { |
1126 | int32_t i = equalPrefixLength; |
1127 | UChar32 c; |
1128 | U8_NEXT_OR_FFFD(left, i, leftLength, c); |
1129 | unsafe = data->isUnsafeBackward(c, numeric); |
1130 | } |
1131 | if(!unsafe && equalPrefixLength != rightLength) { |
1132 | int32_t i = equalPrefixLength; |
1133 | UChar32 c; |
1134 | U8_NEXT_OR_FFFD(right, i, rightLength, c); |
1135 | unsafe = data->isUnsafeBackward(c, numeric); |
1136 | } |
1137 | if(unsafe) { |
1138 | // Identical prefix: Back up to the start of a contraction or reordering sequence. |
1139 | UChar32 c; |
1140 | do { |
1141 | U8_PREV_OR_FFFD(left, 0, equalPrefixLength, c); |
1142 | } while(equalPrefixLength > 0 && data->isUnsafeBackward(c, numeric)); |
1143 | } |
1144 | // See the notes in the UTF-16 version. |
1145 | |
1146 | // Pass the actual start of each string into the CollationIterators, |
1147 | // plus the equalPrefixLength position, |
1148 | // so that prefix matches back into the equal prefix work. |
1149 | } |
1150 | |
1151 | int32_t result; |
1152 | int32_t fastLatinOptions = settings->fastLatinOptions; |
1153 | if(fastLatinOptions >= 0 && |
1154 | (equalPrefixLength == leftLength || |
1155 | left[equalPrefixLength] <= CollationFastLatin::LATIN_MAX_UTF8_LEAD) && |
1156 | (equalPrefixLength == rightLength || |
1157 | right[equalPrefixLength] <= CollationFastLatin::LATIN_MAX_UTF8_LEAD)) { |
1158 | if(leftLength >= 0) { |
1159 | result = CollationFastLatin::compareUTF8(data->fastLatinTable, |
1160 | settings->fastLatinPrimaries, |
1161 | fastLatinOptions, |
1162 | left + equalPrefixLength, |
1163 | leftLength - equalPrefixLength, |
1164 | right + equalPrefixLength, |
1165 | rightLength - equalPrefixLength); |
1166 | } else { |
1167 | result = CollationFastLatin::compareUTF8(data->fastLatinTable, |
1168 | settings->fastLatinPrimaries, |
1169 | fastLatinOptions, |
1170 | left + equalPrefixLength, -1, |
1171 | right + equalPrefixLength, -1); |
1172 | } |
1173 | } else { |
1174 | result = CollationFastLatin::BAIL_OUT_RESULT; |
1175 | } |
1176 | |
1177 | if(result == CollationFastLatin::BAIL_OUT_RESULT) { |
1178 | if(settings->dontCheckFCD()) { |
1179 | UTF8CollationIterator leftIter(data, numeric, left, equalPrefixLength, leftLength); |
1180 | UTF8CollationIterator rightIter(data, numeric, right, equalPrefixLength, rightLength); |
1181 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1182 | } else { |
1183 | FCDUTF8CollationIterator leftIter(data, numeric, left, equalPrefixLength, leftLength); |
1184 | FCDUTF8CollationIterator rightIter(data, numeric, right, equalPrefixLength, rightLength); |
1185 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1186 | } |
1187 | } |
1188 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { |
1189 | return (UCollationResult)result; |
1190 | } |
1191 | |
1192 | // Note: If NUL-terminated, we could get the actual limits from the iterators now. |
1193 | // That would complicate the iterators a bit, NUL-terminated strings are only a C convenience, |
1194 | // and the benefit seems unlikely to be measurable. |
1195 | |
1196 | // Compare identical level. |
1197 | const Normalizer2Impl &nfcImpl = data->nfcImpl; |
1198 | left += equalPrefixLength; |
1199 | right += equalPrefixLength; |
1200 | if(leftLength > 0) { |
1201 | leftLength -= equalPrefixLength; |
1202 | rightLength -= equalPrefixLength; |
1203 | } |
1204 | if(settings->dontCheckFCD()) { |
1205 | UTF8NFDIterator leftIter(left, leftLength); |
1206 | UTF8NFDIterator rightIter(right, rightLength); |
1207 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1208 | } else { |
1209 | FCDUTF8NFDIterator leftIter(data, left, leftLength); |
1210 | FCDUTF8NFDIterator rightIter(data, right, rightLength); |
1211 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1212 | } |
1213 | } |
1214 | |
1215 | UCollationResult |
1216 | RuleBasedCollator::compare(UCharIterator &left, UCharIterator &right, |
1217 | UErrorCode &errorCode) const { |
1218 | if(U_FAILURE(errorCode) || &left == &right) { return UCOL_EQUAL; } |
1219 | UBool numeric = settings->isNumeric(); |
1220 | |
1221 | // Identical-prefix test. |
1222 | int32_t equalPrefixLength = 0; |
1223 | { |
1224 | UChar32 leftUnit; |
1225 | UChar32 rightUnit; |
1226 | while((leftUnit = left.next(&left)) == (rightUnit = right.next(&right))) { |
1227 | if(leftUnit < 0) { return UCOL_EQUAL; } |
1228 | ++equalPrefixLength; |
1229 | } |
1230 | |
1231 | // Back out the code units that differed, for the real collation comparison. |
1232 | if(leftUnit >= 0) { left.previous(&left); } |
1233 | if(rightUnit >= 0) { right.previous(&right); } |
1234 | |
1235 | if(equalPrefixLength > 0) { |
1236 | if((leftUnit >= 0 && data->isUnsafeBackward(leftUnit, numeric)) || |
1237 | (rightUnit >= 0 && data->isUnsafeBackward(rightUnit, numeric))) { |
1238 | // Identical prefix: Back up to the start of a contraction or reordering sequence. |
1239 | do { |
1240 | --equalPrefixLength; |
1241 | leftUnit = left.previous(&left); |
1242 | right.previous(&right); |
1243 | } while(equalPrefixLength > 0 && data->isUnsafeBackward(leftUnit, numeric)); |
1244 | } |
1245 | // See the notes in the UTF-16 version. |
1246 | } |
1247 | } |
1248 | |
1249 | UCollationResult result; |
1250 | if(settings->dontCheckFCD()) { |
1251 | UIterCollationIterator leftIter(data, numeric, left); |
1252 | UIterCollationIterator rightIter(data, numeric, right); |
1253 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1254 | } else { |
1255 | FCDUIterCollationIterator leftIter(data, numeric, left, equalPrefixLength); |
1256 | FCDUIterCollationIterator rightIter(data, numeric, right, equalPrefixLength); |
1257 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); |
1258 | } |
1259 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { |
1260 | return result; |
1261 | } |
1262 | |
1263 | // Compare identical level. |
1264 | left.move(&left, equalPrefixLength, UITER_ZERO); |
1265 | right.move(&right, equalPrefixLength, UITER_ZERO); |
1266 | const Normalizer2Impl &nfcImpl = data->nfcImpl; |
1267 | if(settings->dontCheckFCD()) { |
1268 | UIterNFDIterator leftIter(left); |
1269 | UIterNFDIterator rightIter(right); |
1270 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1271 | } else { |
1272 | FCDUIterNFDIterator leftIter(data, left, equalPrefixLength); |
1273 | FCDUIterNFDIterator rightIter(data, right, equalPrefixLength); |
1274 | return compareNFDIter(nfcImpl, leftIter, rightIter); |
1275 | } |
1276 | } |
1277 | |
1278 | CollationKey & |
1279 | RuleBasedCollator::getCollationKey(const UnicodeString &s, CollationKey &key, |
1280 | UErrorCode &errorCode) const { |
1281 | return getCollationKey(s.getBuffer(), s.length(), key, errorCode); |
1282 | } |
1283 | |
1284 | CollationKey & |
1285 | RuleBasedCollator::getCollationKey(const UChar *s, int32_t length, CollationKey& key, |
1286 | UErrorCode &errorCode) const { |
1287 | if(U_FAILURE(errorCode)) { |
1288 | return key.setToBogus(); |
1289 | } |
1290 | if(s == NULL && length != 0) { |
1291 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
1292 | return key.setToBogus(); |
1293 | } |
1294 | key.reset(); // resets the "bogus" state |
1295 | CollationKeyByteSink sink(key); |
1296 | writeSortKey(s, length, sink, errorCode); |
1297 | if(U_FAILURE(errorCode)) { |
1298 | key.setToBogus(); |
1299 | } else if(key.isBogus()) { |
1300 | errorCode = U_MEMORY_ALLOCATION_ERROR; |
1301 | } else { |
1302 | key.setLength(sink.NumberOfBytesAppended()); |
1303 | } |
1304 | return key; |
1305 | } |
1306 | |
1307 | int32_t |
1308 | RuleBasedCollator::getSortKey(const UnicodeString &s, |
1309 | uint8_t *dest, int32_t capacity) const { |
1310 | return getSortKey(s.getBuffer(), s.length(), dest, capacity); |
1311 | } |
1312 | |
1313 | int32_t |
1314 | RuleBasedCollator::getSortKey(const UChar *s, int32_t length, |
1315 | uint8_t *dest, int32_t capacity) const { |
1316 | if((s == NULL && length != 0) || capacity < 0 || (dest == NULL && capacity > 0)) { |
1317 | return 0; |
1318 | } |
1319 | uint8_t noDest[1] = { 0 }; |
1320 | if(dest == NULL) { |
1321 | // Distinguish pure preflighting from an allocation error. |
1322 | dest = noDest; |
1323 | capacity = 0; |
1324 | } |
1325 | FixedSortKeyByteSink sink(reinterpret_cast<char *>(dest), capacity); |
1326 | UErrorCode errorCode = U_ZERO_ERROR; |
1327 | writeSortKey(s, length, sink, errorCode); |
1328 | return U_SUCCESS(errorCode) ? sink.NumberOfBytesAppended() : 0; |
1329 | } |
1330 | |
1331 | void |
1332 | RuleBasedCollator::writeSortKey(const UChar *s, int32_t length, |
1333 | SortKeyByteSink &sink, UErrorCode &errorCode) const { |
1334 | if(U_FAILURE(errorCode)) { return; } |
1335 | const UChar *limit = (length >= 0) ? s + length : NULL; |
1336 | UBool numeric = settings->isNumeric(); |
1337 | CollationKeys::LevelCallback callback; |
1338 | if(settings->dontCheckFCD()) { |
1339 | UTF16CollationIterator iter(data, numeric, s, s, limit); |
1340 | CollationKeys::writeSortKeyUpToQuaternary(iter, data->compressibleBytes, *settings, |
1341 | sink, Collation::PRIMARY_LEVEL, |
1342 | callback, TRUE, errorCode); |
1343 | } else { |
1344 | FCDUTF16CollationIterator iter(data, numeric, s, s, limit); |
1345 | CollationKeys::writeSortKeyUpToQuaternary(iter, data->compressibleBytes, *settings, |
1346 | sink, Collation::PRIMARY_LEVEL, |
1347 | callback, TRUE, errorCode); |
1348 | } |
1349 | if(settings->getStrength() == UCOL_IDENTICAL) { |
1350 | writeIdenticalLevel(s, limit, sink, errorCode); |
1351 | } |
1352 | static const char terminator = 0; // TERMINATOR_BYTE |
1353 | sink.Append(&terminator, 1); |
1354 | } |
1355 | |
1356 | void |
1357 | RuleBasedCollator::writeIdenticalLevel(const UChar *s, const UChar *limit, |
1358 | SortKeyByteSink &sink, UErrorCode &errorCode) const { |
1359 | // NFD quick check |
1360 | const UChar *nfdQCYesLimit = data->nfcImpl.decompose(s, limit, NULL, errorCode); |
1361 | if(U_FAILURE(errorCode)) { return; } |
1362 | sink.Append(Collation::LEVEL_SEPARATOR_BYTE); |
1363 | UChar32 prev = 0; |
1364 | if(nfdQCYesLimit != s) { |
1365 | prev = u_writeIdenticalLevelRun(prev, s, (int32_t)(nfdQCYesLimit - s), sink); |
1366 | } |
1367 | // Is there non-NFD text? |
1368 | int32_t destLengthEstimate; |
1369 | if(limit != NULL) { |
1370 | if(nfdQCYesLimit == limit) { return; } |
1371 | destLengthEstimate = (int32_t)(limit - nfdQCYesLimit); |
1372 | } else { |
1373 | // s is NUL-terminated |
1374 | if(*nfdQCYesLimit == 0) { return; } |
1375 | destLengthEstimate = -1; |
1376 | } |
1377 | UnicodeString nfd; |
1378 | data->nfcImpl.decompose(nfdQCYesLimit, limit, nfd, destLengthEstimate, errorCode); |
1379 | u_writeIdenticalLevelRun(prev, nfd.getBuffer(), nfd.length(), sink); |
1380 | } |
1381 | |
1382 | namespace { |
1383 | |
1384 | /** |
1385 | * internalNextSortKeyPart() calls CollationKeys::writeSortKeyUpToQuaternary() |
1386 | * with an instance of this callback class. |
1387 | * When another level is about to be written, the callback |
1388 | * records the level and the number of bytes that will be written until |
1389 | * the sink (which is actually a FixedSortKeyByteSink) fills up. |
1390 | * |
1391 | * When internalNextSortKeyPart() is called again, it restarts with the last level |
1392 | * and ignores as many bytes as were written previously for that level. |
1393 | */ |
1394 | class PartLevelCallback : public CollationKeys::LevelCallback { |
1395 | public: |
1396 | PartLevelCallback(const SortKeyByteSink &s) |
1397 | : sink(s), level(Collation::PRIMARY_LEVEL) { |
1398 | levelCapacity = sink.GetRemainingCapacity(); |
1399 | } |
1400 | virtual ~PartLevelCallback() {} |
1401 | virtual UBool needToWrite(Collation::Level l) { |
1402 | if(!sink.Overflowed()) { |
1403 | // Remember a level that will be at least partially written. |
1404 | level = l; |
1405 | levelCapacity = sink.GetRemainingCapacity(); |
1406 | return TRUE; |
1407 | } else { |
1408 | return FALSE; |
1409 | } |
1410 | } |
1411 | Collation::Level getLevel() const { return level; } |
1412 | int32_t getLevelCapacity() const { return levelCapacity; } |
1413 | |
1414 | private: |
1415 | const SortKeyByteSink &sink; |
1416 | Collation::Level level; |
1417 | int32_t levelCapacity; |
1418 | }; |
1419 | |
1420 | } // namespace |
1421 | |
1422 | int32_t |
1423 | RuleBasedCollator::internalNextSortKeyPart(UCharIterator *iter, uint32_t state[2], |
1424 | uint8_t *dest, int32_t count, UErrorCode &errorCode) const { |
1425 | if(U_FAILURE(errorCode)) { return 0; } |
1426 | if(iter == NULL || state == NULL || count < 0 || (count > 0 && dest == NULL)) { |
1427 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
1428 | return 0; |
1429 | } |
1430 | if(count == 0) { return 0; } |
1431 | |
1432 | FixedSortKeyByteSink sink(reinterpret_cast<char *>(dest), count); |
1433 | sink.IgnoreBytes((int32_t)state[1]); |
1434 | iter->move(iter, 0, UITER_START); |
1435 | |
1436 | Collation::Level level = (Collation::Level)state[0]; |
1437 | if(level <= Collation::QUATERNARY_LEVEL) { |
1438 | UBool numeric = settings->isNumeric(); |
1439 | PartLevelCallback callback(sink); |
1440 | if(settings->dontCheckFCD()) { |
1441 | UIterCollationIterator ci(data, numeric, *iter); |
1442 | CollationKeys::writeSortKeyUpToQuaternary(ci, data->compressibleBytes, *settings, |
1443 | sink, level, callback, FALSE, errorCode); |
1444 | } else { |
1445 | FCDUIterCollationIterator ci(data, numeric, *iter, 0); |
1446 | CollationKeys::writeSortKeyUpToQuaternary(ci, data->compressibleBytes, *settings, |
1447 | sink, level, callback, FALSE, errorCode); |
1448 | } |
1449 | if(U_FAILURE(errorCode)) { return 0; } |
1450 | if(sink.NumberOfBytesAppended() > count) { |
1451 | state[0] = (uint32_t)callback.getLevel(); |
1452 | state[1] = (uint32_t)callback.getLevelCapacity(); |
1453 | return count; |
1454 | } |
1455 | // All of the normal levels are done. |
1456 | if(settings->getStrength() == UCOL_IDENTICAL) { |
1457 | level = Collation::IDENTICAL_LEVEL; |
1458 | iter->move(iter, 0, UITER_START); |
1459 | } |
1460 | // else fall through to setting ZERO_LEVEL |
1461 | } |
1462 | |
1463 | if(level == Collation::IDENTICAL_LEVEL) { |
1464 | int32_t levelCapacity = sink.GetRemainingCapacity(); |
1465 | UnicodeString s; |
1466 | for(;;) { |
1467 | UChar32 c = iter->next(iter); |
1468 | if(c < 0) { break; } |
1469 | s.append((UChar)c); |
1470 | } |
1471 | const UChar *sArray = s.getBuffer(); |
1472 | writeIdenticalLevel(sArray, sArray + s.length(), sink, errorCode); |
1473 | if(U_FAILURE(errorCode)) { return 0; } |
1474 | if(sink.NumberOfBytesAppended() > count) { |
1475 | state[0] = (uint32_t)level; |
1476 | state[1] = (uint32_t)levelCapacity; |
1477 | return count; |
1478 | } |
1479 | } |
1480 | |
1481 | // ZERO_LEVEL: Fill the remainder of dest with 00 bytes. |
1482 | state[0] = (uint32_t)Collation::ZERO_LEVEL; |
1483 | state[1] = 0; |
1484 | int32_t length = sink.NumberOfBytesAppended(); |
1485 | int32_t i = length; |
1486 | while(i < count) { dest[i++] = 0; } |
1487 | return length; |
1488 | } |
1489 | |
1490 | void |
1491 | RuleBasedCollator::internalGetCEs(const UnicodeString &str, UVector64 &ces, |
1492 | UErrorCode &errorCode) const { |
1493 | if(U_FAILURE(errorCode)) { return; } |
1494 | const UChar *s = str.getBuffer(); |
1495 | const UChar *limit = s + str.length(); |
1496 | UBool numeric = settings->isNumeric(); |
1497 | if(settings->dontCheckFCD()) { |
1498 | UTF16CollationIterator iter(data, numeric, s, s, limit); |
1499 | int64_t ce; |
1500 | while((ce = iter.nextCE(errorCode)) != Collation::NO_CE) { |
1501 | ces.addElement(ce, errorCode); |
1502 | } |
1503 | } else { |
1504 | FCDUTF16CollationIterator iter(data, numeric, s, s, limit); |
1505 | int64_t ce; |
1506 | while((ce = iter.nextCE(errorCode)) != Collation::NO_CE) { |
1507 | ces.addElement(ce, errorCode); |
1508 | } |
1509 | } |
1510 | } |
1511 | |
1512 | namespace { |
1513 | |
1514 | void appendSubtag(CharString &s, char letter, const char *subtag, int32_t length, |
1515 | UErrorCode &errorCode) { |
1516 | if(U_FAILURE(errorCode) || length == 0) { return; } |
1517 | if(!s.isEmpty()) { |
1518 | s.append('_', errorCode); |
1519 | } |
1520 | s.append(letter, errorCode); |
1521 | for(int32_t i = 0; i < length; ++i) { |
1522 | s.append(uprv_toupper(subtag[i]), errorCode); |
1523 | } |
1524 | } |
1525 | |
1526 | void appendAttribute(CharString &s, char letter, UColAttributeValue value, |
1527 | UErrorCode &errorCode) { |
1528 | if(U_FAILURE(errorCode)) { return; } |
1529 | if(!s.isEmpty()) { |
1530 | s.append('_', errorCode); |
1531 | } |
1532 | static const char *valueChars = "1234...........IXO..SN..LU......" ; |
1533 | s.append(letter, errorCode); |
1534 | s.append(valueChars[value], errorCode); |
1535 | } |
1536 | |
1537 | } // namespace |
1538 | |
1539 | int32_t |
1540 | RuleBasedCollator::internalGetShortDefinitionString(const char *locale, |
1541 | char *buffer, int32_t capacity, |
1542 | UErrorCode &errorCode) const { |
1543 | if(U_FAILURE(errorCode)) { return 0; } |
1544 | if(buffer == NULL ? capacity != 0 : capacity < 0) { |
1545 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; |
1546 | return 0; |
1547 | } |
1548 | if(locale == NULL) { |
1549 | locale = internalGetLocaleID(ULOC_VALID_LOCALE, errorCode); |
1550 | } |
1551 | |
1552 | char resultLocale[ULOC_FULLNAME_CAPACITY + 1]; |
1553 | int32_t length = ucol_getFunctionalEquivalent(resultLocale, ULOC_FULLNAME_CAPACITY, |
1554 | "collation" , locale, |
1555 | NULL, &errorCode); |
1556 | if(U_FAILURE(errorCode)) { return 0; } |
1557 | resultLocale[length] = 0; |
1558 | |
1559 | // Append items in alphabetic order of their short definition letters. |
1560 | CharString result; |
1561 | char subtag[ULOC_KEYWORD_AND_VALUES_CAPACITY]; |
1562 | |
1563 | if(attributeHasBeenSetExplicitly(UCOL_ALTERNATE_HANDLING)) { |
1564 | appendAttribute(result, 'A', getAttribute(UCOL_ALTERNATE_HANDLING, errorCode), errorCode); |
1565 | } |
1566 | // ATTR_VARIABLE_TOP not supported because 'B' was broken. |
1567 | // See ICU tickets #10372 and #10386. |
1568 | if(attributeHasBeenSetExplicitly(UCOL_CASE_FIRST)) { |
1569 | appendAttribute(result, 'C', getAttribute(UCOL_CASE_FIRST, errorCode), errorCode); |
1570 | } |
1571 | if(attributeHasBeenSetExplicitly(UCOL_NUMERIC_COLLATION)) { |
1572 | appendAttribute(result, 'D', getAttribute(UCOL_NUMERIC_COLLATION, errorCode), errorCode); |
1573 | } |
1574 | if(attributeHasBeenSetExplicitly(UCOL_CASE_LEVEL)) { |
1575 | appendAttribute(result, 'E', getAttribute(UCOL_CASE_LEVEL, errorCode), errorCode); |
1576 | } |
1577 | if(attributeHasBeenSetExplicitly(UCOL_FRENCH_COLLATION)) { |
1578 | appendAttribute(result, 'F', getAttribute(UCOL_FRENCH_COLLATION, errorCode), errorCode); |
1579 | } |
1580 | // Note: UCOL_HIRAGANA_QUATERNARY_MODE is deprecated and never changes away from default. |
1581 | length = uloc_getKeywordValue(resultLocale, "collation" , subtag, UPRV_LENGTHOF(subtag), &errorCode); |
1582 | appendSubtag(result, 'K', subtag, length, errorCode); |
1583 | length = uloc_getLanguage(resultLocale, subtag, UPRV_LENGTHOF(subtag), &errorCode); |
1584 | if (length == 0) { |
1585 | appendSubtag(result, 'L', "root" , 4, errorCode); |
1586 | } else { |
1587 | appendSubtag(result, 'L', subtag, length, errorCode); |
1588 | } |
1589 | if(attributeHasBeenSetExplicitly(UCOL_NORMALIZATION_MODE)) { |
1590 | appendAttribute(result, 'N', getAttribute(UCOL_NORMALIZATION_MODE, errorCode), errorCode); |
1591 | } |
1592 | length = uloc_getCountry(resultLocale, subtag, UPRV_LENGTHOF(subtag), &errorCode); |
1593 | appendSubtag(result, 'R', subtag, length, errorCode); |
1594 | if(attributeHasBeenSetExplicitly(UCOL_STRENGTH)) { |
1595 | appendAttribute(result, 'S', getAttribute(UCOL_STRENGTH, errorCode), errorCode); |
1596 | } |
1597 | length = uloc_getVariant(resultLocale, subtag, UPRV_LENGTHOF(subtag), &errorCode); |
1598 | appendSubtag(result, 'V', subtag, length, errorCode); |
1599 | length = uloc_getScript(resultLocale, subtag, UPRV_LENGTHOF(subtag), &errorCode); |
1600 | appendSubtag(result, 'Z', subtag, length, errorCode); |
1601 | |
1602 | if(U_FAILURE(errorCode)) { return 0; } |
1603 | if(result.length() <= capacity) { |
1604 | uprv_memcpy(buffer, result.data(), result.length()); |
1605 | } |
1606 | return u_terminateChars(buffer, capacity, result.length(), &errorCode); |
1607 | } |
1608 | |
1609 | UBool |
1610 | RuleBasedCollator::isUnsafe(UChar32 c) const { |
1611 | return data->isUnsafeBackward(c, settings->isNumeric()); |
1612 | } |
1613 | |
1614 | void U_CALLCONV |
1615 | RuleBasedCollator::computeMaxExpansions(const CollationTailoring *t, UErrorCode &errorCode) { |
1616 | t->maxExpansions = CollationElementIterator::computeMaxExpansions(t->data, errorCode); |
1617 | } |
1618 | |
1619 | UBool |
1620 | RuleBasedCollator::initMaxExpansions(UErrorCode &errorCode) const { |
1621 | umtx_initOnce(tailoring->maxExpansionsInitOnce, computeMaxExpansions, tailoring, errorCode); |
1622 | return U_SUCCESS(errorCode); |
1623 | } |
1624 | |
1625 | CollationElementIterator * |
1626 | RuleBasedCollator::createCollationElementIterator(const UnicodeString& source) const { |
1627 | UErrorCode errorCode = U_ZERO_ERROR; |
1628 | if(!initMaxExpansions(errorCode)) { return NULL; } |
1629 | CollationElementIterator *cei = new CollationElementIterator(source, this, errorCode); |
1630 | if(U_FAILURE(errorCode)) { |
1631 | delete cei; |
1632 | return NULL; |
1633 | } |
1634 | return cei; |
1635 | } |
1636 | |
1637 | CollationElementIterator * |
1638 | RuleBasedCollator::createCollationElementIterator(const CharacterIterator& source) const { |
1639 | UErrorCode errorCode = U_ZERO_ERROR; |
1640 | if(!initMaxExpansions(errorCode)) { return NULL; } |
1641 | CollationElementIterator *cei = new CollationElementIterator(source, this, errorCode); |
1642 | if(U_FAILURE(errorCode)) { |
1643 | delete cei; |
1644 | return NULL; |
1645 | } |
1646 | return cei; |
1647 | } |
1648 | |
1649 | int32_t |
1650 | RuleBasedCollator::getMaxExpansion(int32_t order) const { |
1651 | UErrorCode errorCode = U_ZERO_ERROR; |
1652 | (void)initMaxExpansions(errorCode); |
1653 | return CollationElementIterator::getMaxExpansion(tailoring->maxExpansions, order); |
1654 | } |
1655 | |
1656 | U_NAMESPACE_END |
1657 | |
1658 | #endif // !UCONFIG_NO_COLLATION |
1659 | |