]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | ******************************************************************************* | |
3 | * Copyright (C) 1996-2014, International Business Machines | |
4 | * Corporation and others. All Rights Reserved. | |
5 | ******************************************************************************* | |
6 | * rulebasedcollator.cpp | |
7 | * | |
8 | * (replaced the former tblcoll.cpp) | |
9 | * | |
10 | * created on: 2012feb14 with new and old collation code | |
11 | * created by: Markus W. Scherer | |
12 | */ | |
13 | ||
14 | #include "unicode/utypes.h" | |
15 | ||
16 | #if !UCONFIG_NO_COLLATION | |
17 | ||
18 | #include "unicode/coll.h" | |
19 | #include "unicode/coleitr.h" | |
20 | #include "unicode/localpointer.h" | |
21 | #include "unicode/locid.h" | |
22 | #include "unicode/sortkey.h" | |
23 | #include "unicode/tblcoll.h" | |
24 | #include "unicode/ucol.h" | |
25 | #include "unicode/uiter.h" | |
26 | #include "unicode/uloc.h" | |
27 | #include "unicode/uniset.h" | |
28 | #include "unicode/unistr.h" | |
29 | #include "unicode/usetiter.h" | |
30 | #include "unicode/utf8.h" | |
31 | #include "unicode/uversion.h" | |
32 | #include "bocsu.h" | |
33 | #include "charstr.h" | |
34 | #include "cmemory.h" | |
35 | #include "collation.h" | |
36 | #include "collationcompare.h" | |
37 | #include "collationdata.h" | |
38 | #include "collationdatareader.h" | |
39 | #include "collationfastlatin.h" | |
40 | #include "collationiterator.h" | |
41 | #include "collationkeys.h" | |
42 | #include "collationroot.h" | |
43 | #include "collationsets.h" | |
44 | #include "collationsettings.h" | |
45 | #include "collationtailoring.h" | |
46 | #include "cstring.h" | |
47 | #include "uassert.h" | |
48 | #include "ucol_imp.h" | |
49 | #include "uhash.h" | |
50 | #include "uitercollationiterator.h" | |
51 | #include "ustr_imp.h" | |
52 | #include "utf16collationiterator.h" | |
53 | #include "utf8collationiterator.h" | |
54 | #include "uvectr64.h" | |
55 | ||
56 | #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) | |
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 | validLocale(other.validLocale), | |
146 | explicitlySetAttributes(other.explicitlySetAttributes), | |
147 | actualLocaleIsSameAsValid(other.actualLocaleIsSameAsValid) { | |
148 | settings->addRef(); | |
149 | tailoring->addRef(); | |
150 | } | |
151 | ||
152 | RuleBasedCollator::RuleBasedCollator(const uint8_t *bin, int32_t length, | |
153 | const RuleBasedCollator *base, UErrorCode &errorCode) | |
154 | : data(NULL), | |
155 | settings(NULL), | |
156 | tailoring(NULL), | |
157 | validLocale(""), | |
158 | explicitlySetAttributes(0), | |
159 | actualLocaleIsSameAsValid(FALSE) { | |
160 | if(U_FAILURE(errorCode)) { return; } | |
161 | if(bin == NULL || length <= 0 || base == NULL) { | |
162 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
163 | return; | |
164 | } | |
165 | const CollationTailoring *root = CollationRoot::getRoot(errorCode); | |
166 | if(U_FAILURE(errorCode)) { return; } | |
167 | if(base->tailoring != root) { | |
168 | errorCode = U_UNSUPPORTED_ERROR; | |
169 | return; | |
170 | } | |
171 | LocalPointer<CollationTailoring> t(new CollationTailoring(base->tailoring->settings)); | |
172 | if(t.isNull() || t->isBogus()) { | |
173 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
174 | return; | |
175 | } | |
176 | CollationDataReader::read(base->tailoring, bin, length, *t, errorCode); | |
177 | if(U_FAILURE(errorCode)) { return; } | |
178 | t->actualLocale.setToBogus(); | |
179 | adoptTailoring(t.orphan()); | |
180 | } | |
181 | ||
182 | RuleBasedCollator::RuleBasedCollator(const CollationTailoring *t, const Locale &vl) | |
183 | : data(t->data), | |
184 | settings(t->settings), | |
185 | tailoring(t), | |
186 | validLocale(vl), | |
187 | explicitlySetAttributes(0), | |
188 | actualLocaleIsSameAsValid(FALSE) { | |
189 | settings->addRef(); | |
190 | tailoring->addRef(); | |
191 | } | |
192 | ||
193 | RuleBasedCollator::~RuleBasedCollator() { | |
194 | SharedObject::clearPtr(settings); | |
195 | SharedObject::clearPtr(tailoring); | |
196 | } | |
197 | ||
198 | void | |
199 | RuleBasedCollator::adoptTailoring(CollationTailoring *t) { | |
200 | U_ASSERT(settings == NULL && data == NULL && tailoring == NULL); | |
201 | data = t->data; | |
202 | settings = t->settings; | |
203 | settings->addRef(); | |
204 | t->addRef(); | |
205 | tailoring = t; | |
206 | validLocale = t->actualLocale; | |
207 | actualLocaleIsSameAsValid = FALSE; | |
208 | } | |
209 | ||
210 | Collator * | |
211 | RuleBasedCollator::clone() const { | |
212 | return new RuleBasedCollator(*this); | |
213 | } | |
214 | ||
215 | RuleBasedCollator &RuleBasedCollator::operator=(const RuleBasedCollator &other) { | |
216 | if(this == &other) { return *this; } | |
217 | SharedObject::copyPtr(other.settings, settings); | |
218 | SharedObject::copyPtr(other.tailoring, tailoring); | |
219 | data = tailoring->data; | |
220 | validLocale = other.validLocale; | |
221 | explicitlySetAttributes = other.explicitlySetAttributes; | |
222 | actualLocaleIsSameAsValid = other.actualLocaleIsSameAsValid; | |
223 | return *this; | |
224 | } | |
225 | ||
226 | UOBJECT_DEFINE_RTTI_IMPLEMENTATION(RuleBasedCollator) | |
227 | ||
228 | UBool | |
229 | RuleBasedCollator::operator==(const Collator& other) const { | |
230 | if(this == &other) { return TRUE; } | |
231 | if(!Collator::operator==(other)) { return FALSE; } | |
232 | const RuleBasedCollator &o = static_cast<const RuleBasedCollator &>(other); | |
233 | if(*settings != *o.settings) { return FALSE; } | |
234 | if(data == o.data) { return TRUE; } | |
235 | UBool thisIsRoot = data->base == NULL; | |
236 | UBool otherIsRoot = o.data->base == NULL; | |
237 | U_ASSERT(!thisIsRoot || !otherIsRoot); // otherwise their data pointers should be == | |
238 | if(thisIsRoot != otherIsRoot) { return FALSE; } | |
239 | if((thisIsRoot || !tailoring->rules.isEmpty()) && | |
240 | (otherIsRoot || !o.tailoring->rules.isEmpty())) { | |
241 | // Shortcut: If both collators have valid rule strings, then compare those. | |
242 | if(tailoring->rules == o.tailoring->rules) { return TRUE; } | |
243 | } | |
244 | // Different rule strings can result in the same or equivalent tailoring. | |
245 | // The rule strings are optional in ICU resource bundles, although included by default. | |
246 | // cloneBinary() drops the rule string. | |
247 | UErrorCode errorCode = U_ZERO_ERROR; | |
248 | LocalPointer<UnicodeSet> thisTailored(getTailoredSet(errorCode)); | |
249 | LocalPointer<UnicodeSet> otherTailored(o.getTailoredSet(errorCode)); | |
250 | if(U_FAILURE(errorCode)) { return FALSE; } | |
251 | if(*thisTailored != *otherTailored) { return FALSE; } | |
252 | // For completeness, we should compare all of the mappings; | |
253 | // or we should create a list of strings, sort it with one collator, | |
254 | // and check if both collators compare adjacent strings the same | |
255 | // (order & strength, down to quaternary); or similar. | |
256 | // Testing equality of collators seems unusual. | |
257 | return TRUE; | |
258 | } | |
259 | ||
260 | int32_t | |
261 | RuleBasedCollator::hashCode() const { | |
262 | int32_t h = settings->hashCode(); | |
263 | if(data->base == NULL) { return h; } // root collator | |
264 | // Do not rely on the rule string, see comments in operator==(). | |
265 | UErrorCode errorCode = U_ZERO_ERROR; | |
266 | LocalPointer<UnicodeSet> set(getTailoredSet(errorCode)); | |
267 | if(U_FAILURE(errorCode)) { return 0; } | |
268 | UnicodeSetIterator iter(*set); | |
269 | while(iter.next() && !iter.isString()) { | |
270 | h ^= data->getCE32(iter.getCodepoint()); | |
271 | } | |
272 | return h; | |
273 | } | |
274 | ||
275 | void | |
276 | RuleBasedCollator::setLocales(const Locale &requested, const Locale &valid, | |
277 | const Locale &actual) { | |
278 | if(actual == tailoring->actualLocale) { | |
279 | actualLocaleIsSameAsValid = FALSE; | |
280 | } else { | |
281 | U_ASSERT(actual == valid); | |
282 | actualLocaleIsSameAsValid = TRUE; | |
283 | } | |
284 | // Do not modify tailoring.actualLocale: | |
285 | // We cannot be sure that that would be thread-safe. | |
286 | validLocale = valid; | |
287 | (void)requested; // Ignore, see also ticket #10477. | |
288 | } | |
289 | ||
290 | Locale | |
291 | RuleBasedCollator::getLocale(ULocDataLocaleType type, UErrorCode& errorCode) const { | |
292 | if(U_FAILURE(errorCode)) { | |
293 | return Locale::getRoot(); | |
294 | } | |
295 | switch(type) { | |
296 | case ULOC_ACTUAL_LOCALE: | |
297 | return actualLocaleIsSameAsValid ? validLocale : tailoring->actualLocale; | |
298 | case ULOC_VALID_LOCALE: | |
299 | case ULOC_REQUESTED_LOCALE: // TODO: Drop this, see ticket #10477. | |
300 | return validLocale; | |
301 | default: | |
302 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
303 | return Locale::getRoot(); | |
304 | } | |
305 | } | |
306 | ||
307 | const char * | |
308 | RuleBasedCollator::internalGetLocaleID(ULocDataLocaleType type, UErrorCode &errorCode) const { | |
309 | if(U_FAILURE(errorCode)) { | |
310 | return NULL; | |
311 | } | |
312 | const Locale *result; | |
313 | switch(type) { | |
314 | case ULOC_ACTUAL_LOCALE: | |
315 | result = actualLocaleIsSameAsValid ? &validLocale : &tailoring->actualLocale; | |
316 | break; | |
317 | case ULOC_VALID_LOCALE: | |
318 | case ULOC_REQUESTED_LOCALE: // TODO: Drop this, see ticket #10477. | |
319 | result = &validLocale; | |
320 | break; | |
321 | default: | |
322 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
323 | return NULL; | |
324 | } | |
325 | if(result->isBogus()) { return NULL; } | |
326 | const char *id = result->getName(); | |
327 | return id[0] == 0 ? "root" : id; | |
328 | } | |
329 | ||
330 | const UnicodeString& | |
331 | RuleBasedCollator::getRules() const { | |
332 | return tailoring->rules; | |
333 | } | |
334 | ||
335 | void | |
336 | RuleBasedCollator::getRules(UColRuleOption delta, UnicodeString &buffer) const { | |
337 | if(delta == UCOL_TAILORING_ONLY) { | |
338 | buffer = tailoring->rules; | |
339 | return; | |
340 | } | |
341 | // UCOL_FULL_RULES | |
342 | buffer.remove(); | |
343 | CollationLoader::appendRootRules(buffer); | |
344 | buffer.append(tailoring->rules).getTerminatedBuffer(); | |
345 | } | |
346 | ||
347 | void | |
348 | RuleBasedCollator::getVersion(UVersionInfo version) const { | |
349 | uprv_memcpy(version, tailoring->version, U_MAX_VERSION_LENGTH); | |
350 | version[0] += (UCOL_RUNTIME_VERSION << 4) + (UCOL_RUNTIME_VERSION >> 4); | |
351 | } | |
352 | ||
353 | UnicodeSet * | |
354 | RuleBasedCollator::getTailoredSet(UErrorCode &errorCode) const { | |
355 | if(U_FAILURE(errorCode)) { return NULL; } | |
356 | UnicodeSet *tailored = new UnicodeSet(); | |
357 | if(tailored == NULL) { | |
358 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
359 | return NULL; | |
360 | } | |
361 | if(data->base != NULL) { | |
362 | TailoredSet(tailored).forData(data, errorCode); | |
363 | if(U_FAILURE(errorCode)) { | |
364 | delete tailored; | |
365 | return NULL; | |
366 | } | |
367 | } | |
368 | return tailored; | |
369 | } | |
370 | ||
371 | void | |
372 | RuleBasedCollator::internalGetContractionsAndExpansions( | |
373 | UnicodeSet *contractions, UnicodeSet *expansions, | |
374 | UBool addPrefixes, UErrorCode &errorCode) const { | |
375 | if(U_FAILURE(errorCode)) { return; } | |
376 | if(contractions != NULL) { | |
377 | contractions->clear(); | |
378 | } | |
379 | if(expansions != NULL) { | |
380 | expansions->clear(); | |
381 | } | |
382 | ContractionsAndExpansions(contractions, expansions, NULL, addPrefixes).forData(data, errorCode); | |
383 | } | |
384 | ||
385 | void | |
386 | RuleBasedCollator::internalAddContractions(UChar32 c, UnicodeSet &set, UErrorCode &errorCode) const { | |
387 | if(U_FAILURE(errorCode)) { return; } | |
388 | ContractionsAndExpansions(&set, NULL, NULL, FALSE).forCodePoint(data, c, errorCode); | |
389 | } | |
390 | ||
391 | const CollationSettings & | |
392 | RuleBasedCollator::getDefaultSettings() const { | |
393 | return *tailoring->settings; | |
394 | } | |
395 | ||
396 | UColAttributeValue | |
397 | RuleBasedCollator::getAttribute(UColAttribute attr, UErrorCode &errorCode) const { | |
398 | if(U_FAILURE(errorCode)) { return UCOL_DEFAULT; } | |
399 | int32_t option; | |
400 | switch(attr) { | |
401 | case UCOL_FRENCH_COLLATION: | |
402 | option = CollationSettings::BACKWARD_SECONDARY; | |
403 | break; | |
404 | case UCOL_ALTERNATE_HANDLING: | |
405 | return settings->getAlternateHandling(); | |
406 | case UCOL_CASE_FIRST: | |
407 | return settings->getCaseFirst(); | |
408 | case UCOL_CASE_LEVEL: | |
409 | option = CollationSettings::CASE_LEVEL; | |
410 | break; | |
411 | case UCOL_NORMALIZATION_MODE: | |
412 | option = CollationSettings::CHECK_FCD; | |
413 | break; | |
414 | case UCOL_STRENGTH: | |
415 | return (UColAttributeValue)settings->getStrength(); | |
416 | case UCOL_HIRAGANA_QUATERNARY_MODE: | |
417 | // Deprecated attribute, unsettable. | |
418 | return UCOL_OFF; | |
419 | case UCOL_NUMERIC_COLLATION: | |
420 | option = CollationSettings::NUMERIC; | |
421 | break; | |
422 | default: | |
423 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
424 | return UCOL_DEFAULT; | |
425 | } | |
426 | return ((settings->options & option) == 0) ? UCOL_OFF : UCOL_ON; | |
427 | } | |
428 | ||
429 | void | |
430 | RuleBasedCollator::setAttribute(UColAttribute attr, UColAttributeValue value, | |
431 | UErrorCode &errorCode) { | |
432 | UColAttributeValue oldValue = getAttribute(attr, errorCode); | |
433 | if(U_FAILURE(errorCode)) { return; } | |
434 | if(value == oldValue) { | |
435 | setAttributeExplicitly(attr); | |
436 | return; | |
437 | } | |
438 | const CollationSettings &defaultSettings = getDefaultSettings(); | |
439 | if(settings == &defaultSettings) { | |
440 | if(value == UCOL_DEFAULT) { | |
441 | setAttributeDefault(attr); | |
442 | return; | |
443 | } | |
444 | } | |
445 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); | |
446 | if(ownedSettings == NULL) { | |
447 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
448 | return; | |
449 | } | |
450 | ||
451 | switch(attr) { | |
452 | case UCOL_FRENCH_COLLATION: | |
453 | ownedSettings->setFlag(CollationSettings::BACKWARD_SECONDARY, value, | |
454 | defaultSettings.options, errorCode); | |
455 | break; | |
456 | case UCOL_ALTERNATE_HANDLING: | |
457 | ownedSettings->setAlternateHandling(value, defaultSettings.options, errorCode); | |
458 | break; | |
459 | case UCOL_CASE_FIRST: | |
460 | ownedSettings->setCaseFirst(value, defaultSettings.options, errorCode); | |
461 | break; | |
462 | case UCOL_CASE_LEVEL: | |
463 | ownedSettings->setFlag(CollationSettings::CASE_LEVEL, value, | |
464 | defaultSettings.options, errorCode); | |
465 | break; | |
466 | case UCOL_NORMALIZATION_MODE: | |
467 | ownedSettings->setFlag(CollationSettings::CHECK_FCD, value, | |
468 | defaultSettings.options, errorCode); | |
469 | break; | |
470 | case UCOL_STRENGTH: | |
471 | ownedSettings->setStrength(value, defaultSettings.options, errorCode); | |
472 | break; | |
473 | case UCOL_HIRAGANA_QUATERNARY_MODE: | |
474 | // Deprecated attribute. Check for valid values but do not change anything. | |
475 | if(value != UCOL_OFF && value != UCOL_ON && value != UCOL_DEFAULT) { | |
476 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
477 | } | |
478 | break; | |
479 | case UCOL_NUMERIC_COLLATION: | |
480 | ownedSettings->setFlag(CollationSettings::NUMERIC, value, defaultSettings.options, errorCode); | |
481 | break; | |
482 | default: | |
483 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
484 | break; | |
485 | } | |
486 | if(U_FAILURE(errorCode)) { return; } | |
487 | setFastLatinOptions(*ownedSettings); | |
488 | if(value == UCOL_DEFAULT) { | |
489 | setAttributeDefault(attr); | |
490 | } else { | |
491 | setAttributeExplicitly(attr); | |
492 | } | |
493 | } | |
494 | ||
495 | Collator & | |
496 | RuleBasedCollator::setMaxVariable(UColReorderCode group, UErrorCode &errorCode) { | |
497 | if(U_FAILURE(errorCode)) { return *this; } | |
498 | // Convert the reorder code into a MaxVariable number, or UCOL_DEFAULT=-1. | |
499 | int32_t value; | |
500 | if(group == UCOL_REORDER_CODE_DEFAULT) { | |
501 | value = UCOL_DEFAULT; | |
502 | } else if(UCOL_REORDER_CODE_FIRST <= group && group <= UCOL_REORDER_CODE_CURRENCY) { | |
503 | value = group - UCOL_REORDER_CODE_FIRST; | |
504 | } else { | |
505 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
506 | return *this; | |
507 | } | |
508 | CollationSettings::MaxVariable oldValue = settings->getMaxVariable(); | |
509 | if(value == oldValue) { | |
510 | setAttributeExplicitly(ATTR_VARIABLE_TOP); | |
511 | return *this; | |
512 | } | |
513 | const CollationSettings &defaultSettings = getDefaultSettings(); | |
514 | if(settings == &defaultSettings) { | |
515 | if(value == UCOL_DEFAULT) { | |
516 | setAttributeDefault(ATTR_VARIABLE_TOP); | |
517 | return *this; | |
518 | } | |
519 | } | |
520 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); | |
521 | if(ownedSettings == NULL) { | |
522 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
523 | return *this; | |
524 | } | |
525 | ||
526 | if(group == UCOL_REORDER_CODE_DEFAULT) { | |
527 | group = (UColReorderCode)(UCOL_REORDER_CODE_FIRST + defaultSettings.getMaxVariable()); | |
528 | } | |
529 | uint32_t varTop = data->getLastPrimaryForGroup(group); | |
530 | U_ASSERT(varTop != 0); | |
531 | ownedSettings->setMaxVariable(value, defaultSettings.options, errorCode); | |
532 | if(U_FAILURE(errorCode)) { return *this; } | |
533 | ownedSettings->variableTop = varTop; | |
534 | setFastLatinOptions(*ownedSettings); | |
535 | if(value == UCOL_DEFAULT) { | |
536 | setAttributeDefault(ATTR_VARIABLE_TOP); | |
537 | } else { | |
538 | setAttributeExplicitly(ATTR_VARIABLE_TOP); | |
539 | } | |
540 | return *this; | |
541 | } | |
542 | ||
543 | UColReorderCode | |
544 | RuleBasedCollator::getMaxVariable() const { | |
545 | return (UColReorderCode)(UCOL_REORDER_CODE_FIRST + settings->getMaxVariable()); | |
546 | } | |
547 | ||
548 | uint32_t | |
549 | RuleBasedCollator::getVariableTop(UErrorCode & /*errorCode*/) const { | |
550 | return settings->variableTop; | |
551 | } | |
552 | ||
553 | uint32_t | |
554 | RuleBasedCollator::setVariableTop(const UChar *varTop, int32_t len, UErrorCode &errorCode) { | |
555 | if(U_FAILURE(errorCode)) { return 0; } | |
556 | if(varTop == NULL && len !=0) { | |
557 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
558 | return 0; | |
559 | } | |
560 | if(len < 0) { len = u_strlen(varTop); } | |
561 | if(len == 0) { | |
562 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
563 | return 0; | |
564 | } | |
565 | UBool numeric = settings->isNumeric(); | |
566 | int64_t ce1, ce2; | |
567 | if(settings->dontCheckFCD()) { | |
568 | UTF16CollationIterator ci(data, numeric, varTop, varTop, varTop + len); | |
569 | ce1 = ci.nextCE(errorCode); | |
570 | ce2 = ci.nextCE(errorCode); | |
571 | } else { | |
572 | FCDUTF16CollationIterator ci(data, numeric, varTop, varTop, varTop + len); | |
573 | ce1 = ci.nextCE(errorCode); | |
574 | ce2 = ci.nextCE(errorCode); | |
575 | } | |
576 | if(ce1 == Collation::NO_CE || ce2 != Collation::NO_CE) { | |
577 | errorCode = U_CE_NOT_FOUND_ERROR; | |
578 | return 0; | |
579 | } | |
580 | setVariableTop((uint32_t)(ce1 >> 32), errorCode); | |
581 | return settings->variableTop; | |
582 | } | |
583 | ||
584 | uint32_t | |
585 | RuleBasedCollator::setVariableTop(const UnicodeString &varTop, UErrorCode &errorCode) { | |
586 | return setVariableTop(varTop.getBuffer(), varTop.length(), errorCode); | |
587 | } | |
588 | ||
589 | void | |
590 | RuleBasedCollator::setVariableTop(uint32_t varTop, UErrorCode &errorCode) { | |
591 | if(U_FAILURE(errorCode)) { return; } | |
592 | if(varTop != settings->variableTop) { | |
593 | // Pin the variable top to the end of the reordering group which contains it. | |
594 | // Only a few special groups are supported. | |
595 | int32_t group = data->getGroupForPrimary(varTop); | |
596 | if(group < UCOL_REORDER_CODE_FIRST || UCOL_REORDER_CODE_CURRENCY < group) { | |
597 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
598 | return; | |
599 | } | |
600 | uint32_t v = data->getLastPrimaryForGroup(group); | |
601 | U_ASSERT(v != 0 && v >= varTop); | |
602 | varTop = v; | |
603 | if(varTop != settings->variableTop) { | |
604 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); | |
605 | if(ownedSettings == NULL) { | |
606 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
607 | return; | |
608 | } | |
609 | ownedSettings->setMaxVariable(group - UCOL_REORDER_CODE_FIRST, | |
610 | getDefaultSettings().options, errorCode); | |
611 | if(U_FAILURE(errorCode)) { return; } | |
612 | ownedSettings->variableTop = varTop; | |
613 | setFastLatinOptions(*ownedSettings); | |
614 | } | |
615 | } | |
616 | if(varTop == getDefaultSettings().variableTop) { | |
617 | setAttributeDefault(ATTR_VARIABLE_TOP); | |
618 | } else { | |
619 | setAttributeExplicitly(ATTR_VARIABLE_TOP); | |
620 | } | |
621 | } | |
622 | ||
623 | int32_t | |
624 | RuleBasedCollator::getReorderCodes(int32_t *dest, int32_t capacity, | |
625 | UErrorCode &errorCode) const { | |
626 | if(U_FAILURE(errorCode)) { return 0; } | |
627 | if(capacity < 0 || (dest == NULL && capacity > 0)) { | |
628 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
629 | return 0; | |
630 | } | |
631 | int32_t length = settings->reorderCodesLength; | |
632 | if(length == 0) { return 0; } | |
633 | if(length > capacity) { | |
634 | errorCode = U_BUFFER_OVERFLOW_ERROR; | |
635 | return length; | |
636 | } | |
637 | uprv_memcpy(dest, settings->reorderCodes, length * 4); | |
638 | return length; | |
639 | } | |
640 | ||
641 | void | |
642 | RuleBasedCollator::setReorderCodes(const int32_t *reorderCodes, int32_t length, | |
643 | UErrorCode &errorCode) { | |
644 | if(U_FAILURE(errorCode)) { return; } | |
645 | if(length < 0 || (reorderCodes == NULL && length > 0)) { | |
646 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
647 | return; | |
648 | } | |
649 | if(length == settings->reorderCodesLength && | |
650 | uprv_memcmp(reorderCodes, settings->reorderCodes, length * 4) == 0) { | |
651 | return; | |
652 | } | |
653 | const CollationSettings &defaultSettings = getDefaultSettings(); | |
654 | if(length == 1 && reorderCodes[0] == UCOL_REORDER_CODE_DEFAULT) { | |
655 | if(settings != &defaultSettings) { | |
656 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); | |
657 | if(ownedSettings == NULL) { | |
658 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
659 | return; | |
660 | } | |
661 | ownedSettings->aliasReordering(defaultSettings.reorderCodes, | |
662 | defaultSettings.reorderCodesLength, | |
663 | defaultSettings.reorderTable); | |
664 | setFastLatinOptions(*ownedSettings); | |
665 | } | |
666 | return; | |
667 | } | |
668 | CollationSettings *ownedSettings = SharedObject::copyOnWrite(settings); | |
669 | if(ownedSettings == NULL) { | |
670 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
671 | return; | |
672 | } | |
673 | if(length == 0) { | |
674 | ownedSettings->resetReordering(); | |
675 | } else { | |
676 | uint8_t reorderTable[256]; | |
677 | data->makeReorderTable(reorderCodes, length, reorderTable, errorCode); | |
678 | if(U_FAILURE(errorCode)) { return; } | |
679 | if(!ownedSettings->setReordering(reorderCodes, length, reorderTable)) { | |
680 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
681 | return; | |
682 | } | |
683 | } | |
684 | setFastLatinOptions(*ownedSettings); | |
685 | } | |
686 | ||
687 | void | |
688 | RuleBasedCollator::setFastLatinOptions(CollationSettings &ownedSettings) const { | |
689 | ownedSettings.fastLatinOptions = CollationFastLatin::getOptions( | |
690 | data, ownedSettings, | |
691 | ownedSettings.fastLatinPrimaries, LENGTHOF(ownedSettings.fastLatinPrimaries)); | |
692 | } | |
693 | ||
694 | UCollationResult | |
695 | RuleBasedCollator::compare(const UnicodeString &left, const UnicodeString &right, | |
696 | UErrorCode &errorCode) const { | |
697 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } | |
698 | return doCompare(left.getBuffer(), left.length(), | |
699 | right.getBuffer(), right.length(), errorCode); | |
700 | } | |
701 | ||
702 | UCollationResult | |
703 | RuleBasedCollator::compare(const UnicodeString &left, const UnicodeString &right, | |
704 | int32_t length, UErrorCode &errorCode) const { | |
705 | if(U_FAILURE(errorCode) || length == 0) { return UCOL_EQUAL; } | |
706 | if(length < 0) { | |
707 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
708 | return UCOL_EQUAL; | |
709 | } | |
710 | int32_t leftLength = left.length(); | |
711 | int32_t rightLength = right.length(); | |
712 | if(leftLength > length) { leftLength = length; } | |
713 | if(rightLength > length) { rightLength = length; } | |
714 | return doCompare(left.getBuffer(), leftLength, | |
715 | right.getBuffer(), rightLength, errorCode); | |
716 | } | |
717 | ||
718 | UCollationResult | |
719 | RuleBasedCollator::compare(const UChar *left, int32_t leftLength, | |
720 | const UChar *right, int32_t rightLength, | |
721 | UErrorCode &errorCode) const { | |
722 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } | |
723 | if((left == NULL && leftLength != 0) || (right == NULL && rightLength != 0)) { | |
724 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
725 | return UCOL_EQUAL; | |
726 | } | |
727 | // Make sure both or neither strings have a known length. | |
728 | // We do not optimize for mixed length/termination. | |
729 | if(leftLength >= 0) { | |
730 | if(rightLength < 0) { rightLength = u_strlen(right); } | |
731 | } else { | |
732 | if(rightLength >= 0) { leftLength = u_strlen(left); } | |
733 | } | |
734 | return doCompare(left, leftLength, right, rightLength, errorCode); | |
735 | } | |
736 | ||
737 | UCollationResult | |
738 | RuleBasedCollator::compareUTF8(const StringPiece &left, const StringPiece &right, | |
739 | UErrorCode &errorCode) const { | |
740 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } | |
741 | const uint8_t *leftBytes = reinterpret_cast<const uint8_t *>(left.data()); | |
742 | const uint8_t *rightBytes = reinterpret_cast<const uint8_t *>(right.data()); | |
743 | if((leftBytes == NULL && !left.empty()) || (rightBytes == NULL && !right.empty())) { | |
744 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
745 | return UCOL_EQUAL; | |
746 | } | |
747 | return doCompare(leftBytes, left.length(), rightBytes, right.length(), errorCode); | |
748 | } | |
749 | ||
750 | UCollationResult | |
751 | RuleBasedCollator::internalCompareUTF8(const char *left, int32_t leftLength, | |
752 | const char *right, int32_t rightLength, | |
753 | UErrorCode &errorCode) const { | |
754 | if(U_FAILURE(errorCode)) { return UCOL_EQUAL; } | |
755 | if((left == NULL && leftLength != 0) || (right == NULL && rightLength != 0)) { | |
756 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
757 | return UCOL_EQUAL; | |
758 | } | |
759 | // Make sure both or neither strings have a known length. | |
760 | // We do not optimize for mixed length/termination. | |
761 | if(leftLength >= 0) { | |
762 | if(rightLength < 0) { rightLength = uprv_strlen(right); } | |
763 | } else { | |
764 | if(rightLength >= 0) { leftLength = uprv_strlen(left); } | |
765 | } | |
766 | return doCompare(reinterpret_cast<const uint8_t *>(left), leftLength, | |
767 | reinterpret_cast<const uint8_t *>(right), rightLength, errorCode); | |
768 | } | |
769 | ||
770 | namespace { | |
771 | ||
772 | /** | |
773 | * Abstract iterator for identical-level string comparisons. | |
774 | * Returns FCD code points and handles temporary switching to NFD. | |
775 | */ | |
776 | class NFDIterator { | |
777 | public: | |
778 | NFDIterator() : index(-1), length(0) {} | |
779 | virtual ~NFDIterator() {} | |
780 | /** | |
781 | * Returns the next code point from the internal normalization buffer, | |
782 | * or else the next text code point. | |
783 | * Returns -1 at the end of the text. | |
784 | */ | |
785 | UChar32 nextCodePoint() { | |
786 | if(index >= 0) { | |
787 | if(index == length) { | |
788 | index = -1; | |
789 | } else { | |
790 | UChar32 c; | |
791 | U16_NEXT_UNSAFE(decomp, index, c); | |
792 | return c; | |
793 | } | |
794 | } | |
795 | return nextRawCodePoint(); | |
796 | } | |
797 | /** | |
798 | * @param nfcImpl | |
799 | * @param c the last code point returned by nextCodePoint() or nextDecomposedCodePoint() | |
800 | * @return the first code point in c's decomposition, | |
801 | * or c itself if it was decomposed already or if it does not decompose | |
802 | */ | |
803 | UChar32 nextDecomposedCodePoint(const Normalizer2Impl &nfcImpl, UChar32 c) { | |
804 | if(index >= 0) { return c; } | |
805 | decomp = nfcImpl.getDecomposition(c, buffer, length); | |
806 | if(decomp == NULL) { return c; } | |
807 | index = 0; | |
808 | U16_NEXT_UNSAFE(decomp, index, c); | |
809 | return c; | |
810 | } | |
811 | protected: | |
812 | /** | |
813 | * Returns the next text code point in FCD order. | |
814 | * Returns -1 at the end of the text. | |
815 | */ | |
816 | virtual UChar32 nextRawCodePoint() = 0; | |
817 | private: | |
818 | const UChar *decomp; | |
819 | UChar buffer[4]; | |
820 | int32_t index; | |
821 | int32_t length; | |
822 | }; | |
823 | ||
824 | class UTF16NFDIterator : public NFDIterator { | |
825 | public: | |
826 | UTF16NFDIterator(const UChar *text, const UChar *textLimit) : s(text), limit(textLimit) {} | |
827 | protected: | |
828 | virtual UChar32 nextRawCodePoint() { | |
829 | if(s == limit) { return U_SENTINEL; } | |
830 | UChar32 c = *s++; | |
831 | if(limit == NULL && c == 0) { | |
832 | s = NULL; | |
833 | return U_SENTINEL; | |
834 | } | |
835 | UChar trail; | |
836 | if(U16_IS_LEAD(c) && s != limit && U16_IS_TRAIL(trail = *s)) { | |
837 | ++s; | |
838 | c = U16_GET_SUPPLEMENTARY(c, trail); | |
839 | } | |
840 | return c; | |
841 | } | |
842 | ||
843 | const UChar *s; | |
844 | const UChar *limit; | |
845 | }; | |
846 | ||
847 | class FCDUTF16NFDIterator : public UTF16NFDIterator { | |
848 | public: | |
849 | FCDUTF16NFDIterator(const Normalizer2Impl &nfcImpl, const UChar *text, const UChar *textLimit) | |
850 | : UTF16NFDIterator(NULL, NULL) { | |
851 | UErrorCode errorCode = U_ZERO_ERROR; | |
852 | const UChar *spanLimit = nfcImpl.makeFCD(text, textLimit, NULL, errorCode); | |
853 | if(U_FAILURE(errorCode)) { return; } | |
854 | if(spanLimit == textLimit || (textLimit == NULL && *spanLimit == 0)) { | |
855 | s = text; | |
856 | limit = spanLimit; | |
857 | } else { | |
858 | str.setTo(text, (int32_t)(spanLimit - text)); | |
859 | { | |
860 | ReorderingBuffer buffer(nfcImpl, str); | |
861 | if(buffer.init(str.length(), errorCode)) { | |
862 | nfcImpl.makeFCD(spanLimit, textLimit, &buffer, errorCode); | |
863 | } | |
864 | } | |
865 | if(U_SUCCESS(errorCode)) { | |
866 | s = str.getBuffer(); | |
867 | limit = s + str.length(); | |
868 | } | |
869 | } | |
870 | } | |
871 | private: | |
872 | UnicodeString str; | |
873 | }; | |
874 | ||
875 | class UTF8NFDIterator : public NFDIterator { | |
876 | public: | |
877 | UTF8NFDIterator(const uint8_t *text, int32_t textLength) | |
878 | : s(text), pos(0), length(textLength) {} | |
879 | protected: | |
880 | virtual UChar32 nextRawCodePoint() { | |
881 | if(pos == length || (s[pos] == 0 && length < 0)) { return U_SENTINEL; } | |
882 | UChar32 c; | |
883 | U8_NEXT_OR_FFFD(s, pos, length, c); | |
884 | return c; | |
885 | } | |
886 | ||
887 | const uint8_t *s; | |
888 | int32_t pos; | |
889 | int32_t length; | |
890 | }; | |
891 | ||
892 | class FCDUTF8NFDIterator : public NFDIterator { | |
893 | public: | |
894 | FCDUTF8NFDIterator(const CollationData *data, const uint8_t *text, int32_t textLength) | |
895 | : u8ci(data, FALSE, text, 0, textLength) {} | |
896 | protected: | |
897 | virtual UChar32 nextRawCodePoint() { | |
898 | UErrorCode errorCode = U_ZERO_ERROR; | |
899 | return u8ci.nextCodePoint(errorCode); | |
900 | } | |
901 | private: | |
902 | FCDUTF8CollationIterator u8ci; | |
903 | }; | |
904 | ||
905 | class UIterNFDIterator : public NFDIterator { | |
906 | public: | |
907 | UIterNFDIterator(UCharIterator &it) : iter(it) {} | |
908 | protected: | |
909 | virtual UChar32 nextRawCodePoint() { | |
910 | return uiter_next32(&iter); | |
911 | } | |
912 | private: | |
913 | UCharIterator &iter; | |
914 | }; | |
915 | ||
916 | class FCDUIterNFDIterator : public NFDIterator { | |
917 | public: | |
918 | FCDUIterNFDIterator(const CollationData *data, UCharIterator &it, int32_t startIndex) | |
919 | : uici(data, FALSE, it, startIndex) {} | |
920 | protected: | |
921 | virtual UChar32 nextRawCodePoint() { | |
922 | UErrorCode errorCode = U_ZERO_ERROR; | |
923 | return uici.nextCodePoint(errorCode); | |
924 | } | |
925 | private: | |
926 | FCDUIterCollationIterator uici; | |
927 | }; | |
928 | ||
929 | UCollationResult compareNFDIter(const Normalizer2Impl &nfcImpl, | |
930 | NFDIterator &left, NFDIterator &right) { | |
931 | for(;;) { | |
932 | // Fetch the next FCD code point from each string. | |
933 | UChar32 leftCp = left.nextCodePoint(); | |
934 | UChar32 rightCp = right.nextCodePoint(); | |
935 | if(leftCp == rightCp) { | |
936 | if(leftCp < 0) { break; } | |
937 | continue; | |
938 | } | |
939 | // If they are different, then decompose each and compare again. | |
940 | if(leftCp < 0) { | |
941 | leftCp = -2; // end of string | |
942 | } else if(leftCp == 0xfffe) { | |
943 | leftCp = -1; // U+FFFE: merge separator | |
944 | } else { | |
945 | leftCp = left.nextDecomposedCodePoint(nfcImpl, leftCp); | |
946 | } | |
947 | if(rightCp < 0) { | |
948 | rightCp = -2; // end of string | |
949 | } else if(rightCp == 0xfffe) { | |
950 | rightCp = -1; // U+FFFE: merge separator | |
951 | } else { | |
952 | rightCp = right.nextDecomposedCodePoint(nfcImpl, rightCp); | |
953 | } | |
954 | if(leftCp < rightCp) { return UCOL_LESS; } | |
955 | if(leftCp > rightCp) { return UCOL_GREATER; } | |
956 | } | |
957 | return UCOL_EQUAL; | |
958 | } | |
959 | ||
960 | } // namespace | |
961 | ||
962 | UCollationResult | |
963 | RuleBasedCollator::doCompare(const UChar *left, int32_t leftLength, | |
964 | const UChar *right, int32_t rightLength, | |
965 | UErrorCode &errorCode) const { | |
966 | // U_FAILURE(errorCode) checked by caller. | |
967 | if(left == right && leftLength == rightLength) { | |
968 | return UCOL_EQUAL; | |
969 | } | |
970 | ||
971 | // Identical-prefix test. | |
972 | const UChar *leftLimit; | |
973 | const UChar *rightLimit; | |
974 | int32_t equalPrefixLength = 0; | |
975 | if(leftLength < 0) { | |
976 | leftLimit = NULL; | |
977 | rightLimit = NULL; | |
978 | UChar c; | |
979 | while((c = left[equalPrefixLength]) == right[equalPrefixLength]) { | |
980 | if(c == 0) { return UCOL_EQUAL; } | |
981 | ++equalPrefixLength; | |
982 | } | |
983 | } else { | |
984 | leftLimit = left + leftLength; | |
985 | rightLimit = right + rightLength; | |
986 | for(;;) { | |
987 | if(equalPrefixLength == leftLength) { | |
988 | if(equalPrefixLength == rightLength) { return UCOL_EQUAL; } | |
989 | break; | |
990 | } else if(equalPrefixLength == rightLength || | |
991 | left[equalPrefixLength] != right[equalPrefixLength]) { | |
992 | break; | |
993 | } | |
994 | ++equalPrefixLength; | |
995 | } | |
996 | } | |
997 | ||
998 | UBool numeric = settings->isNumeric(); | |
999 | if(equalPrefixLength > 0) { | |
1000 | if((equalPrefixLength != leftLength && | |
1001 | data->isUnsafeBackward(left[equalPrefixLength], numeric)) || | |
1002 | (equalPrefixLength != rightLength && | |
1003 | data->isUnsafeBackward(right[equalPrefixLength], numeric))) { | |
1004 | // Identical prefix: Back up to the start of a contraction or reordering sequence. | |
1005 | while(--equalPrefixLength > 0 && | |
1006 | data->isUnsafeBackward(left[equalPrefixLength], numeric)) {} | |
1007 | } | |
1008 | // Notes: | |
1009 | // - A longer string can compare equal to a prefix of it if only ignorables follow. | |
1010 | // - With a backward level, a longer string can compare less-than a prefix of it. | |
1011 | ||
1012 | // Pass the actual start of each string into the CollationIterators, | |
1013 | // plus the equalPrefixLength position, | |
1014 | // so that prefix matches back into the equal prefix work. | |
1015 | } | |
1016 | ||
1017 | int32_t result; | |
1018 | int32_t fastLatinOptions = settings->fastLatinOptions; | |
1019 | if(fastLatinOptions >= 0 && | |
1020 | (equalPrefixLength == leftLength || | |
1021 | left[equalPrefixLength] <= CollationFastLatin::LATIN_MAX) && | |
1022 | (equalPrefixLength == rightLength || | |
1023 | right[equalPrefixLength] <= CollationFastLatin::LATIN_MAX)) { | |
1024 | if(leftLength >= 0) { | |
1025 | result = CollationFastLatin::compareUTF16(data->fastLatinTable, | |
1026 | settings->fastLatinPrimaries, | |
1027 | fastLatinOptions, | |
1028 | left + equalPrefixLength, | |
1029 | leftLength - equalPrefixLength, | |
1030 | right + equalPrefixLength, | |
1031 | rightLength - equalPrefixLength); | |
1032 | } else { | |
1033 | result = CollationFastLatin::compareUTF16(data->fastLatinTable, | |
1034 | settings->fastLatinPrimaries, | |
1035 | fastLatinOptions, | |
1036 | left + equalPrefixLength, -1, | |
1037 | right + equalPrefixLength, -1); | |
1038 | } | |
1039 | } else { | |
1040 | result = CollationFastLatin::BAIL_OUT_RESULT; | |
1041 | } | |
1042 | ||
1043 | if(result == CollationFastLatin::BAIL_OUT_RESULT) { | |
1044 | if(settings->dontCheckFCD()) { | |
1045 | UTF16CollationIterator leftIter(data, numeric, | |
1046 | left, left + equalPrefixLength, leftLimit); | |
1047 | UTF16CollationIterator rightIter(data, numeric, | |
1048 | right, right + equalPrefixLength, rightLimit); | |
1049 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1050 | } else { | |
1051 | FCDUTF16CollationIterator leftIter(data, numeric, | |
1052 | left, left + equalPrefixLength, leftLimit); | |
1053 | FCDUTF16CollationIterator rightIter(data, numeric, | |
1054 | right, right + equalPrefixLength, rightLimit); | |
1055 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1056 | } | |
1057 | } | |
1058 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { | |
1059 | return (UCollationResult)result; | |
1060 | } | |
1061 | ||
1062 | // Note: If NUL-terminated, we could get the actual limits from the iterators now. | |
1063 | // That would complicate the iterators a bit, NUL-terminated strings are only a C convenience, | |
1064 | // and the benefit seems unlikely to be measurable. | |
1065 | ||
1066 | // Compare identical level. | |
1067 | const Normalizer2Impl &nfcImpl = data->nfcImpl; | |
1068 | left += equalPrefixLength; | |
1069 | right += equalPrefixLength; | |
1070 | if(settings->dontCheckFCD()) { | |
1071 | UTF16NFDIterator leftIter(left, leftLimit); | |
1072 | UTF16NFDIterator rightIter(right, rightLimit); | |
1073 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1074 | } else { | |
1075 | FCDUTF16NFDIterator leftIter(nfcImpl, left, leftLimit); | |
1076 | FCDUTF16NFDIterator rightIter(nfcImpl, right, rightLimit); | |
1077 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1078 | } | |
1079 | } | |
1080 | ||
1081 | UCollationResult | |
1082 | RuleBasedCollator::doCompare(const uint8_t *left, int32_t leftLength, | |
1083 | const uint8_t *right, int32_t rightLength, | |
1084 | UErrorCode &errorCode) const { | |
1085 | // U_FAILURE(errorCode) checked by caller. | |
1086 | if(left == right && leftLength == rightLength) { | |
1087 | return UCOL_EQUAL; | |
1088 | } | |
1089 | ||
1090 | // Identical-prefix test. | |
1091 | int32_t equalPrefixLength = 0; | |
1092 | if(leftLength < 0) { | |
1093 | uint8_t c; | |
1094 | while((c = left[equalPrefixLength]) == right[equalPrefixLength]) { | |
1095 | if(c == 0) { return UCOL_EQUAL; } | |
1096 | ++equalPrefixLength; | |
1097 | } | |
1098 | } else { | |
1099 | for(;;) { | |
1100 | if(equalPrefixLength == leftLength) { | |
1101 | if(equalPrefixLength == rightLength) { return UCOL_EQUAL; } | |
1102 | break; | |
1103 | } else if(equalPrefixLength == rightLength || | |
1104 | left[equalPrefixLength] != right[equalPrefixLength]) { | |
1105 | break; | |
1106 | } | |
1107 | ++equalPrefixLength; | |
1108 | } | |
1109 | } | |
1110 | // Back up to the start of a partially-equal code point. | |
1111 | if(equalPrefixLength > 0 && | |
1112 | ((equalPrefixLength != leftLength && U8_IS_TRAIL(left[equalPrefixLength])) || | |
1113 | (equalPrefixLength != rightLength && U8_IS_TRAIL(right[equalPrefixLength])))) { | |
1114 | while(--equalPrefixLength > 0 && U8_IS_TRAIL(left[equalPrefixLength])) {} | |
1115 | } | |
1116 | ||
1117 | UBool numeric = settings->isNumeric(); | |
1118 | if(equalPrefixLength > 0) { | |
1119 | UBool unsafe = FALSE; | |
1120 | if(equalPrefixLength != leftLength) { | |
1121 | int32_t i = equalPrefixLength; | |
1122 | UChar32 c; | |
1123 | U8_NEXT_OR_FFFD(left, i, leftLength, c); | |
1124 | unsafe = data->isUnsafeBackward(c, numeric); | |
1125 | } | |
1126 | if(!unsafe && equalPrefixLength != rightLength) { | |
1127 | int32_t i = equalPrefixLength; | |
1128 | UChar32 c; | |
1129 | U8_NEXT_OR_FFFD(right, i, rightLength, c); | |
1130 | unsafe = data->isUnsafeBackward(c, numeric); | |
1131 | } | |
1132 | if(unsafe) { | |
1133 | // Identical prefix: Back up to the start of a contraction or reordering sequence. | |
1134 | UChar32 c; | |
1135 | do { | |
1136 | U8_PREV_OR_FFFD(left, 0, equalPrefixLength, c); | |
1137 | } while(equalPrefixLength > 0 && data->isUnsafeBackward(c, numeric)); | |
1138 | } | |
1139 | // See the notes in the UTF-16 version. | |
1140 | ||
1141 | // Pass the actual start of each string into the CollationIterators, | |
1142 | // plus the equalPrefixLength position, | |
1143 | // so that prefix matches back into the equal prefix work. | |
1144 | } | |
1145 | ||
1146 | int32_t result; | |
1147 | int32_t fastLatinOptions = settings->fastLatinOptions; | |
1148 | if(fastLatinOptions >= 0 && | |
1149 | (equalPrefixLength == leftLength || | |
1150 | left[equalPrefixLength] <= CollationFastLatin::LATIN_MAX_UTF8_LEAD) && | |
1151 | (equalPrefixLength == rightLength || | |
1152 | right[equalPrefixLength] <= CollationFastLatin::LATIN_MAX_UTF8_LEAD)) { | |
1153 | if(leftLength >= 0) { | |
1154 | result = CollationFastLatin::compareUTF8(data->fastLatinTable, | |
1155 | settings->fastLatinPrimaries, | |
1156 | fastLatinOptions, | |
1157 | left + equalPrefixLength, | |
1158 | leftLength - equalPrefixLength, | |
1159 | right + equalPrefixLength, | |
1160 | rightLength - equalPrefixLength); | |
1161 | } else { | |
1162 | result = CollationFastLatin::compareUTF8(data->fastLatinTable, | |
1163 | settings->fastLatinPrimaries, | |
1164 | fastLatinOptions, | |
1165 | left + equalPrefixLength, -1, | |
1166 | right + equalPrefixLength, -1); | |
1167 | } | |
1168 | } else { | |
1169 | result = CollationFastLatin::BAIL_OUT_RESULT; | |
1170 | } | |
1171 | ||
1172 | if(result == CollationFastLatin::BAIL_OUT_RESULT) { | |
1173 | if(settings->dontCheckFCD()) { | |
1174 | UTF8CollationIterator leftIter(data, numeric, left, equalPrefixLength, leftLength); | |
1175 | UTF8CollationIterator rightIter(data, numeric, right, equalPrefixLength, rightLength); | |
1176 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1177 | } else { | |
1178 | FCDUTF8CollationIterator leftIter(data, numeric, left, equalPrefixLength, leftLength); | |
1179 | FCDUTF8CollationIterator rightIter(data, numeric, right, equalPrefixLength, rightLength); | |
1180 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1181 | } | |
1182 | } | |
1183 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { | |
1184 | return (UCollationResult)result; | |
1185 | } | |
1186 | ||
1187 | // Note: If NUL-terminated, we could get the actual limits from the iterators now. | |
1188 | // That would complicate the iterators a bit, NUL-terminated strings are only a C convenience, | |
1189 | // and the benefit seems unlikely to be measurable. | |
1190 | ||
1191 | // Compare identical level. | |
1192 | const Normalizer2Impl &nfcImpl = data->nfcImpl; | |
1193 | left += equalPrefixLength; | |
1194 | right += equalPrefixLength; | |
1195 | if(leftLength > 0) { | |
1196 | leftLength -= equalPrefixLength; | |
1197 | rightLength -= equalPrefixLength; | |
1198 | } | |
1199 | if(settings->dontCheckFCD()) { | |
1200 | UTF8NFDIterator leftIter(left, leftLength); | |
1201 | UTF8NFDIterator rightIter(right, rightLength); | |
1202 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1203 | } else { | |
1204 | FCDUTF8NFDIterator leftIter(data, left, leftLength); | |
1205 | FCDUTF8NFDIterator rightIter(data, right, rightLength); | |
1206 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1207 | } | |
1208 | } | |
1209 | ||
1210 | UCollationResult | |
1211 | RuleBasedCollator::compare(UCharIterator &left, UCharIterator &right, | |
1212 | UErrorCode &errorCode) const { | |
1213 | if(U_FAILURE(errorCode) || &left == &right) { return UCOL_EQUAL; } | |
1214 | UBool numeric = settings->isNumeric(); | |
1215 | ||
1216 | // Identical-prefix test. | |
1217 | int32_t equalPrefixLength = 0; | |
1218 | { | |
1219 | UChar32 leftUnit; | |
1220 | UChar32 rightUnit; | |
1221 | while((leftUnit = left.next(&left)) == (rightUnit = right.next(&right))) { | |
1222 | if(leftUnit < 0) { return UCOL_EQUAL; } | |
1223 | ++equalPrefixLength; | |
1224 | } | |
1225 | ||
1226 | // Back out the code units that differed, for the real collation comparison. | |
1227 | if(leftUnit >= 0) { left.previous(&left); } | |
1228 | if(rightUnit >= 0) { right.previous(&right); } | |
1229 | ||
1230 | if(equalPrefixLength > 0) { | |
1231 | if((leftUnit >= 0 && data->isUnsafeBackward(leftUnit, numeric)) || | |
1232 | (rightUnit >= 0 && data->isUnsafeBackward(rightUnit, numeric))) { | |
1233 | // Identical prefix: Back up to the start of a contraction or reordering sequence. | |
1234 | do { | |
1235 | --equalPrefixLength; | |
1236 | leftUnit = left.previous(&left); | |
1237 | right.previous(&right); | |
1238 | } while(equalPrefixLength > 0 && data->isUnsafeBackward(leftUnit, numeric)); | |
1239 | } | |
1240 | // See the notes in the UTF-16 version. | |
1241 | } | |
1242 | } | |
1243 | ||
1244 | UCollationResult result; | |
1245 | if(settings->dontCheckFCD()) { | |
1246 | UIterCollationIterator leftIter(data, numeric, left); | |
1247 | UIterCollationIterator rightIter(data, numeric, right); | |
1248 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1249 | } else { | |
1250 | FCDUIterCollationIterator leftIter(data, numeric, left, equalPrefixLength); | |
1251 | FCDUIterCollationIterator rightIter(data, numeric, right, equalPrefixLength); | |
1252 | result = CollationCompare::compareUpToQuaternary(leftIter, rightIter, *settings, errorCode); | |
1253 | } | |
1254 | if(result != UCOL_EQUAL || settings->getStrength() < UCOL_IDENTICAL || U_FAILURE(errorCode)) { | |
1255 | return result; | |
1256 | } | |
1257 | ||
1258 | // Compare identical level. | |
1259 | left.move(&left, equalPrefixLength, UITER_ZERO); | |
1260 | right.move(&right, equalPrefixLength, UITER_ZERO); | |
1261 | const Normalizer2Impl &nfcImpl = data->nfcImpl; | |
1262 | if(settings->dontCheckFCD()) { | |
1263 | UIterNFDIterator leftIter(left); | |
1264 | UIterNFDIterator rightIter(right); | |
1265 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1266 | } else { | |
1267 | FCDUIterNFDIterator leftIter(data, left, equalPrefixLength); | |
1268 | FCDUIterNFDIterator rightIter(data, right, equalPrefixLength); | |
1269 | return compareNFDIter(nfcImpl, leftIter, rightIter); | |
1270 | } | |
1271 | } | |
1272 | ||
1273 | CollationKey & | |
1274 | RuleBasedCollator::getCollationKey(const UnicodeString &s, CollationKey &key, | |
1275 | UErrorCode &errorCode) const { | |
1276 | return getCollationKey(s.getBuffer(), s.length(), key, errorCode); | |
1277 | } | |
1278 | ||
1279 | CollationKey & | |
1280 | RuleBasedCollator::getCollationKey(const UChar *s, int32_t length, CollationKey& key, | |
1281 | UErrorCode &errorCode) const { | |
1282 | if(U_FAILURE(errorCode)) { | |
1283 | return key.setToBogus(); | |
1284 | } | |
1285 | if(s == NULL && length != 0) { | |
1286 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
1287 | return key.setToBogus(); | |
1288 | } | |
1289 | key.reset(); // resets the "bogus" state | |
1290 | CollationKeyByteSink sink(key); | |
1291 | writeSortKey(s, length, sink, errorCode); | |
1292 | if(U_FAILURE(errorCode)) { | |
1293 | key.setToBogus(); | |
1294 | } else if(key.isBogus()) { | |
1295 | errorCode = U_MEMORY_ALLOCATION_ERROR; | |
1296 | } else { | |
1297 | key.setLength(sink.NumberOfBytesAppended()); | |
1298 | } | |
1299 | return key; | |
1300 | } | |
1301 | ||
1302 | int32_t | |
1303 | RuleBasedCollator::getSortKey(const UnicodeString &s, | |
1304 | uint8_t *dest, int32_t capacity) const { | |
1305 | return getSortKey(s.getBuffer(), s.length(), dest, capacity); | |
1306 | } | |
1307 | ||
1308 | int32_t | |
1309 | RuleBasedCollator::getSortKey(const UChar *s, int32_t length, | |
1310 | uint8_t *dest, int32_t capacity) const { | |
1311 | if((s == NULL && length != 0) || capacity < 0 || (dest == NULL && capacity > 0)) { | |
1312 | return 0; | |
1313 | } | |
1314 | uint8_t noDest[1] = { 0 }; | |
1315 | if(dest == NULL) { | |
1316 | // Distinguish pure preflighting from an allocation error. | |
1317 | dest = noDest; | |
1318 | capacity = 0; | |
1319 | } | |
1320 | FixedSortKeyByteSink sink(reinterpret_cast<char *>(dest), capacity); | |
1321 | UErrorCode errorCode = U_ZERO_ERROR; | |
1322 | writeSortKey(s, length, sink, errorCode); | |
1323 | return U_SUCCESS(errorCode) ? sink.NumberOfBytesAppended() : 0; | |
1324 | } | |
1325 | ||
1326 | void | |
1327 | RuleBasedCollator::writeSortKey(const UChar *s, int32_t length, | |
1328 | SortKeyByteSink &sink, UErrorCode &errorCode) const { | |
1329 | if(U_FAILURE(errorCode)) { return; } | |
1330 | const UChar *limit = (length >= 0) ? s + length : NULL; | |
1331 | UBool numeric = settings->isNumeric(); | |
1332 | CollationKeys::LevelCallback callback; | |
1333 | if(settings->dontCheckFCD()) { | |
1334 | UTF16CollationIterator iter(data, numeric, s, s, limit); | |
1335 | CollationKeys::writeSortKeyUpToQuaternary(iter, data->compressibleBytes, *settings, | |
1336 | sink, Collation::PRIMARY_LEVEL, | |
1337 | callback, TRUE, errorCode); | |
1338 | } else { | |
1339 | FCDUTF16CollationIterator iter(data, numeric, s, s, limit); | |
1340 | CollationKeys::writeSortKeyUpToQuaternary(iter, data->compressibleBytes, *settings, | |
1341 | sink, Collation::PRIMARY_LEVEL, | |
1342 | callback, TRUE, errorCode); | |
1343 | } | |
1344 | if(settings->getStrength() == UCOL_IDENTICAL) { | |
1345 | writeIdenticalLevel(s, limit, sink, errorCode); | |
1346 | } | |
1347 | static const char terminator = 0; // TERMINATOR_BYTE | |
1348 | sink.Append(&terminator, 1); | |
1349 | } | |
1350 | ||
1351 | void | |
1352 | RuleBasedCollator::writeIdenticalLevel(const UChar *s, const UChar *limit, | |
1353 | SortKeyByteSink &sink, UErrorCode &errorCode) const { | |
1354 | // NFD quick check | |
1355 | const UChar *nfdQCYesLimit = data->nfcImpl.decompose(s, limit, NULL, errorCode); | |
1356 | if(U_FAILURE(errorCode)) { return; } | |
1357 | sink.Append(Collation::LEVEL_SEPARATOR_BYTE); | |
1358 | UChar32 prev = 0; | |
1359 | if(nfdQCYesLimit != s) { | |
1360 | prev = u_writeIdenticalLevelRun(prev, s, (int32_t)(nfdQCYesLimit - s), sink); | |
1361 | } | |
1362 | // Is there non-NFD text? | |
1363 | int32_t destLengthEstimate; | |
1364 | if(limit != NULL) { | |
1365 | if(nfdQCYesLimit == limit) { return; } | |
1366 | destLengthEstimate = (int32_t)(limit - nfdQCYesLimit); | |
1367 | } else { | |
1368 | // s is NUL-terminated | |
1369 | if(*nfdQCYesLimit == 0) { return; } | |
1370 | destLengthEstimate = -1; | |
1371 | } | |
1372 | UnicodeString nfd; | |
1373 | data->nfcImpl.decompose(nfdQCYesLimit, limit, nfd, destLengthEstimate, errorCode); | |
1374 | u_writeIdenticalLevelRun(prev, nfd.getBuffer(), nfd.length(), sink); | |
1375 | } | |
1376 | ||
1377 | namespace { | |
1378 | ||
1379 | /** | |
1380 | * internalNextSortKeyPart() calls CollationKeys::writeSortKeyUpToQuaternary() | |
1381 | * with an instance of this callback class. | |
1382 | * When another level is about to be written, the callback | |
1383 | * records the level and the number of bytes that will be written until | |
1384 | * the sink (which is actually a FixedSortKeyByteSink) fills up. | |
1385 | * | |
1386 | * When internalNextSortKeyPart() is called again, it restarts with the last level | |
1387 | * and ignores as many bytes as were written previously for that level. | |
1388 | */ | |
1389 | class PartLevelCallback : public CollationKeys::LevelCallback { | |
1390 | public: | |
1391 | PartLevelCallback(const SortKeyByteSink &s) | |
1392 | : sink(s), level(Collation::PRIMARY_LEVEL) { | |
1393 | levelCapacity = sink.GetRemainingCapacity(); | |
1394 | } | |
1395 | virtual ~PartLevelCallback() {} | |
1396 | virtual UBool needToWrite(Collation::Level l) { | |
1397 | if(!sink.Overflowed()) { | |
1398 | // Remember a level that will be at least partially written. | |
1399 | level = l; | |
1400 | levelCapacity = sink.GetRemainingCapacity(); | |
1401 | return TRUE; | |
1402 | } else { | |
1403 | return FALSE; | |
1404 | } | |
1405 | } | |
1406 | Collation::Level getLevel() const { return level; } | |
1407 | int32_t getLevelCapacity() const { return levelCapacity; } | |
1408 | ||
1409 | private: | |
1410 | const SortKeyByteSink &sink; | |
1411 | Collation::Level level; | |
1412 | int32_t levelCapacity; | |
1413 | }; | |
1414 | ||
1415 | } // namespace | |
1416 | ||
1417 | int32_t | |
1418 | RuleBasedCollator::internalNextSortKeyPart(UCharIterator *iter, uint32_t state[2], | |
1419 | uint8_t *dest, int32_t count, UErrorCode &errorCode) const { | |
1420 | if(U_FAILURE(errorCode)) { return 0; } | |
1421 | if(iter == NULL || state == NULL || count < 0 || (count > 0 && dest == NULL)) { | |
1422 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
1423 | return 0; | |
1424 | } | |
1425 | if(count == 0) { return 0; } | |
1426 | ||
1427 | FixedSortKeyByteSink sink(reinterpret_cast<char *>(dest), count); | |
1428 | sink.IgnoreBytes((int32_t)state[1]); | |
1429 | iter->move(iter, 0, UITER_START); | |
1430 | ||
1431 | Collation::Level level = (Collation::Level)state[0]; | |
1432 | if(level <= Collation::QUATERNARY_LEVEL) { | |
1433 | UBool numeric = settings->isNumeric(); | |
1434 | PartLevelCallback callback(sink); | |
1435 | if(settings->dontCheckFCD()) { | |
1436 | UIterCollationIterator ci(data, numeric, *iter); | |
1437 | CollationKeys::writeSortKeyUpToQuaternary(ci, data->compressibleBytes, *settings, | |
1438 | sink, level, callback, FALSE, errorCode); | |
1439 | } else { | |
1440 | FCDUIterCollationIterator ci(data, numeric, *iter, 0); | |
1441 | CollationKeys::writeSortKeyUpToQuaternary(ci, data->compressibleBytes, *settings, | |
1442 | sink, level, callback, FALSE, errorCode); | |
1443 | } | |
1444 | if(U_FAILURE(errorCode)) { return 0; } | |
1445 | if(sink.NumberOfBytesAppended() > count) { | |
1446 | state[0] = (uint32_t)callback.getLevel(); | |
1447 | state[1] = (uint32_t)callback.getLevelCapacity(); | |
1448 | return count; | |
1449 | } | |
1450 | // All of the normal levels are done. | |
1451 | if(settings->getStrength() == UCOL_IDENTICAL) { | |
1452 | level = Collation::IDENTICAL_LEVEL; | |
1453 | iter->move(iter, 0, UITER_START); | |
1454 | } | |
1455 | // else fall through to setting ZERO_LEVEL | |
1456 | } | |
1457 | ||
1458 | if(level == Collation::IDENTICAL_LEVEL) { | |
1459 | int32_t levelCapacity = sink.GetRemainingCapacity(); | |
1460 | UnicodeString s; | |
1461 | for(;;) { | |
1462 | UChar32 c = iter->next(iter); | |
1463 | if(c < 0) { break; } | |
1464 | s.append((UChar)c); | |
1465 | } | |
1466 | const UChar *sArray = s.getBuffer(); | |
1467 | writeIdenticalLevel(sArray, sArray + s.length(), sink, errorCode); | |
1468 | if(U_FAILURE(errorCode)) { return 0; } | |
1469 | if(sink.NumberOfBytesAppended() > count) { | |
1470 | state[0] = (uint32_t)level; | |
1471 | state[1] = (uint32_t)levelCapacity; | |
1472 | return count; | |
1473 | } | |
1474 | } | |
1475 | ||
1476 | // ZERO_LEVEL: Fill the remainder of dest with 00 bytes. | |
1477 | state[0] = (uint32_t)Collation::ZERO_LEVEL; | |
1478 | state[1] = 0; | |
1479 | int32_t length = sink.NumberOfBytesAppended(); | |
1480 | int32_t i = length; | |
1481 | while(i < count) { dest[i++] = 0; } | |
1482 | return length; | |
1483 | } | |
1484 | ||
1485 | void | |
1486 | RuleBasedCollator::internalGetCEs(const UnicodeString &str, UVector64 &ces, | |
1487 | UErrorCode &errorCode) const { | |
1488 | if(U_FAILURE(errorCode)) { return; } | |
1489 | const UChar *s = str.getBuffer(); | |
1490 | const UChar *limit = s + str.length(); | |
1491 | UBool numeric = settings->isNumeric(); | |
1492 | if(settings->dontCheckFCD()) { | |
1493 | UTF16CollationIterator iter(data, numeric, s, s, limit); | |
1494 | int64_t ce; | |
1495 | while((ce = iter.nextCE(errorCode)) != Collation::NO_CE) { | |
1496 | ces.addElement(ce, errorCode); | |
1497 | } | |
1498 | } else { | |
1499 | FCDUTF16CollationIterator iter(data, numeric, s, s, limit); | |
1500 | int64_t ce; | |
1501 | while((ce = iter.nextCE(errorCode)) != Collation::NO_CE) { | |
1502 | ces.addElement(ce, errorCode); | |
1503 | } | |
1504 | } | |
1505 | } | |
1506 | ||
1507 | namespace { | |
1508 | ||
1509 | void appendSubtag(CharString &s, char letter, const char *subtag, int32_t length, | |
1510 | UErrorCode &errorCode) { | |
1511 | if(U_FAILURE(errorCode) || length == 0) { return; } | |
1512 | if(!s.isEmpty()) { | |
1513 | s.append('_', errorCode); | |
1514 | } | |
1515 | s.append(letter, errorCode); | |
1516 | for(int32_t i = 0; i < length; ++i) { | |
1517 | s.append(uprv_toupper(subtag[i]), errorCode); | |
1518 | } | |
1519 | } | |
1520 | ||
1521 | void appendAttribute(CharString &s, char letter, UColAttributeValue value, | |
1522 | UErrorCode &errorCode) { | |
1523 | if(U_FAILURE(errorCode)) { return; } | |
1524 | if(!s.isEmpty()) { | |
1525 | s.append('_', errorCode); | |
1526 | } | |
1527 | static const char *valueChars = "1234...........IXO..SN..LU......"; | |
1528 | s.append(letter, errorCode); | |
1529 | s.append(valueChars[value], errorCode); | |
1530 | } | |
1531 | ||
1532 | } // namespace | |
1533 | ||
1534 | int32_t | |
1535 | RuleBasedCollator::internalGetShortDefinitionString(const char *locale, | |
1536 | char *buffer, int32_t capacity, | |
1537 | UErrorCode &errorCode) const { | |
1538 | if(U_FAILURE(errorCode)) { return 0; } | |
1539 | if(buffer == NULL ? capacity != 0 : capacity < 0) { | |
1540 | errorCode = U_ILLEGAL_ARGUMENT_ERROR; | |
1541 | return 0; | |
1542 | } | |
1543 | if(locale == NULL) { | |
1544 | locale = internalGetLocaleID(ULOC_VALID_LOCALE, errorCode); | |
1545 | } | |
1546 | ||
1547 | char resultLocale[ULOC_FULLNAME_CAPACITY + 1]; | |
1548 | int32_t length = ucol_getFunctionalEquivalent(resultLocale, ULOC_FULLNAME_CAPACITY, | |
1549 | "collation", locale, | |
1550 | NULL, &errorCode); | |
1551 | if(U_FAILURE(errorCode)) { return 0; } | |
1552 | if(length == 0) { | |
1553 | uprv_strcpy(resultLocale, "root"); | |
1554 | } else { | |
1555 | resultLocale[length] = 0; | |
1556 | } | |
1557 | ||
1558 | // Append items in alphabetic order of their short definition letters. | |
1559 | CharString result; | |
1560 | char subtag[ULOC_KEYWORD_AND_VALUES_CAPACITY]; | |
1561 | ||
1562 | if(attributeHasBeenSetExplicitly(UCOL_ALTERNATE_HANDLING)) { | |
1563 | appendAttribute(result, 'A', getAttribute(UCOL_ALTERNATE_HANDLING, errorCode), errorCode); | |
1564 | } | |
1565 | // ATTR_VARIABLE_TOP not supported because 'B' was broken. | |
1566 | // See ICU tickets #10372 and #10386. | |
1567 | if(attributeHasBeenSetExplicitly(UCOL_CASE_FIRST)) { | |
1568 | appendAttribute(result, 'C', getAttribute(UCOL_CASE_FIRST, errorCode), errorCode); | |
1569 | } | |
1570 | if(attributeHasBeenSetExplicitly(UCOL_NUMERIC_COLLATION)) { | |
1571 | appendAttribute(result, 'D', getAttribute(UCOL_NUMERIC_COLLATION, errorCode), errorCode); | |
1572 | } | |
1573 | if(attributeHasBeenSetExplicitly(UCOL_CASE_LEVEL)) { | |
1574 | appendAttribute(result, 'E', getAttribute(UCOL_CASE_LEVEL, errorCode), errorCode); | |
1575 | } | |
1576 | if(attributeHasBeenSetExplicitly(UCOL_FRENCH_COLLATION)) { | |
1577 | appendAttribute(result, 'F', getAttribute(UCOL_FRENCH_COLLATION, errorCode), errorCode); | |
1578 | } | |
1579 | // Note: UCOL_HIRAGANA_QUATERNARY_MODE is deprecated and never changes away from default. | |
1580 | length = uloc_getKeywordValue(resultLocale, "collation", subtag, LENGTHOF(subtag), &errorCode); | |
1581 | appendSubtag(result, 'K', subtag, length, errorCode); | |
1582 | length = uloc_getLanguage(resultLocale, subtag, LENGTHOF(subtag), &errorCode); | |
1583 | appendSubtag(result, 'L', subtag, length, errorCode); | |
1584 | if(attributeHasBeenSetExplicitly(UCOL_NORMALIZATION_MODE)) { | |
1585 | appendAttribute(result, 'N', getAttribute(UCOL_NORMALIZATION_MODE, errorCode), errorCode); | |
1586 | } | |
1587 | length = uloc_getCountry(resultLocale, subtag, LENGTHOF(subtag), &errorCode); | |
1588 | appendSubtag(result, 'R', subtag, length, errorCode); | |
1589 | if(attributeHasBeenSetExplicitly(UCOL_STRENGTH)) { | |
1590 | appendAttribute(result, 'S', getAttribute(UCOL_STRENGTH, errorCode), errorCode); | |
1591 | } | |
1592 | length = uloc_getVariant(resultLocale, subtag, LENGTHOF(subtag), &errorCode); | |
1593 | appendSubtag(result, 'V', subtag, length, errorCode); | |
1594 | length = uloc_getScript(resultLocale, subtag, LENGTHOF(subtag), &errorCode); | |
1595 | appendSubtag(result, 'Z', subtag, length, errorCode); | |
1596 | ||
1597 | if(U_FAILURE(errorCode)) { return 0; } | |
1598 | if(result.length() <= capacity) { | |
1599 | uprv_memcpy(buffer, result.data(), result.length()); | |
1600 | } | |
1601 | return u_terminateChars(buffer, capacity, result.length(), &errorCode); | |
1602 | } | |
1603 | ||
1604 | UBool | |
1605 | RuleBasedCollator::isUnsafe(UChar32 c) const { | |
1606 | return data->isUnsafeBackward(c, settings->isNumeric()); | |
1607 | } | |
1608 | ||
1609 | void | |
1610 | RuleBasedCollator::computeMaxExpansions(const CollationTailoring *t, UErrorCode &errorCode) { | |
1611 | t->maxExpansions = CollationElementIterator::computeMaxExpansions(t->data, errorCode); | |
1612 | } | |
1613 | ||
1614 | UBool | |
1615 | RuleBasedCollator::initMaxExpansions(UErrorCode &errorCode) const { | |
1616 | umtx_initOnce(tailoring->maxExpansionsInitOnce, computeMaxExpansions, tailoring, errorCode); | |
1617 | return U_SUCCESS(errorCode); | |
1618 | } | |
1619 | ||
1620 | CollationElementIterator * | |
1621 | RuleBasedCollator::createCollationElementIterator(const UnicodeString& source) const { | |
1622 | UErrorCode errorCode = U_ZERO_ERROR; | |
1623 | if(!initMaxExpansions(errorCode)) { return NULL; } | |
1624 | CollationElementIterator *cei = new CollationElementIterator(source, this, errorCode); | |
1625 | if(U_FAILURE(errorCode)) { | |
1626 | delete cei; | |
1627 | return NULL; | |
1628 | } | |
1629 | return cei; | |
1630 | } | |
1631 | ||
1632 | CollationElementIterator * | |
1633 | RuleBasedCollator::createCollationElementIterator(const CharacterIterator& source) const { | |
1634 | UErrorCode errorCode = U_ZERO_ERROR; | |
1635 | if(!initMaxExpansions(errorCode)) { return NULL; } | |
1636 | CollationElementIterator *cei = new CollationElementIterator(source, this, errorCode); | |
1637 | if(U_FAILURE(errorCode)) { | |
1638 | delete cei; | |
1639 | return NULL; | |
1640 | } | |
1641 | return cei; | |
1642 | } | |
1643 | ||
1644 | int32_t | |
1645 | RuleBasedCollator::getMaxExpansion(int32_t order) const { | |
1646 | UErrorCode errorCode = U_ZERO_ERROR; | |
1647 | (void)initMaxExpansions(errorCode); | |
1648 | return CollationElementIterator::getMaxExpansion(tailoring->maxExpansions, order); | |
1649 | } | |
1650 | ||
1651 | U_NAMESPACE_END | |
1652 | ||
1653 | #endif // !UCONFIG_NO_COLLATION |