1 // © 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
3 /*
4 ************************************************************************************
5 * Copyright (C) 2006-2016, International Business Machines Corporation
6 * and others. All Rights Reserved.
7 ************************************************************************************
8 */
9
10 #include "unicode/utypes.h"
11
12 #if !UCONFIG_NO_BREAK_ITERATION
13
14 #include "unicode/uchar.h"
15 #include "unicode/uniset.h"
16 #include "unicode/chariter.h"
17 #include "unicode/ures.h"
18 #include "unicode/udata.h"
19 #include "unicode/putil.h"
20 #include "unicode/ustring.h"
21 #include "unicode/uscript.h"
22 #include "unicode/ucharstrie.h"
23 #include "unicode/bytestrie.h"
24 #include "unicode/rbbi.h"
25
26 #include "brkeng.h"
27 #include "cmemory.h"
28 #include "dictbe.h"
29 #include "lstmbe.h"
30 #include "charstr.h"
31 #include "dictionarydata.h"
32 #include "mutex.h"
33 #include "uvector.h"
34 #include "umutex.h"
35 #include "uresimp.h"
36 #include "ubrkimpl.h"
37
38 U_NAMESPACE_BEGIN
39
40 /*
41 ******************************************************************
42 */
43
LanguageBreakEngine()44 LanguageBreakEngine::LanguageBreakEngine() {
45 }
46
~LanguageBreakEngine()47 LanguageBreakEngine::~LanguageBreakEngine() {
48 }
49
50 /*
51 ******************************************************************
52 */
53
LanguageBreakFactory()54 LanguageBreakFactory::LanguageBreakFactory() {
55 }
56
~LanguageBreakFactory()57 LanguageBreakFactory::~LanguageBreakFactory() {
58 }
59
60 /*
61 ******************************************************************
62 */
63
UnhandledEngine(UErrorCode & status)64 UnhandledEngine::UnhandledEngine(UErrorCode &status) : fHandled(nullptr) {
65 (void)status;
66 }
67
~UnhandledEngine()68 UnhandledEngine::~UnhandledEngine() {
69 delete fHandled;
70 fHandled = nullptr;
71 }
72
73 UBool
handles(UChar32 c,const char * locale) const74 UnhandledEngine::handles(UChar32 c, const char* locale) const {
75 (void)locale; // Unused
76 return fHandled && fHandled->contains(c);
77 }
78
79 int32_t
findBreaks(UText * text,int32_t startPos,int32_t endPos,UVector32 &,UBool,UErrorCode & status) const80 UnhandledEngine::findBreaks( UText *text,
81 int32_t startPos,
82 int32_t endPos,
83 UVector32 &/*foundBreaks*/,
84 UBool /* isPhraseBreaking */,
85 UErrorCode &status) const {
86 if (U_FAILURE(status)) return 0;
87 utext_setNativeIndex(text, startPos);
88 UChar32 c = utext_current32(text);
89 while((int32_t)utext_getNativeIndex(text) < endPos && fHandled->contains(c)) {
90 utext_next32(text); // TODO: recast loop to work with post-increment operations.
91 c = utext_current32(text);
92 }
93 return 0;
94 }
95
96 void
handleCharacter(UChar32 c)97 UnhandledEngine::handleCharacter(UChar32 c) {
98 if (fHandled == nullptr) {
99 fHandled = new UnicodeSet();
100 if (fHandled == nullptr) {
101 return;
102 }
103 }
104 if (!fHandled->contains(c)) {
105 UErrorCode status = U_ZERO_ERROR;
106 // Apply the entire script of the character.
107 int32_t script = u_getIntPropertyValue(c, UCHAR_SCRIPT);
108 fHandled->applyIntPropertyValue(UCHAR_SCRIPT, script, status);
109 }
110 }
111
112 /*
113 ******************************************************************
114 */
115
ICULanguageBreakFactory(UErrorCode &)116 ICULanguageBreakFactory::ICULanguageBreakFactory(UErrorCode &/*status*/) {
117 fEngines = nullptr;
118 }
119
~ICULanguageBreakFactory()120 ICULanguageBreakFactory::~ICULanguageBreakFactory() {
121 delete fEngines;
122 }
123
ensureEngines(UErrorCode & status)124 void ICULanguageBreakFactory::ensureEngines(UErrorCode& status) {
125 static UMutex gBreakEngineMutex;
126 Mutex m(&gBreakEngineMutex);
127 if (fEngines == nullptr) {
128 LocalPointer<UStack> engines(new UStack(uprv_deleteUObject, nullptr, status), status);
129 if (U_SUCCESS(status)) {
130 fEngines = engines.orphan();
131 }
132 }
133 }
134
135 const LanguageBreakEngine *
getEngineFor(UChar32 c,const char * locale)136 ICULanguageBreakFactory::getEngineFor(UChar32 c, const char* locale) {
137 const LanguageBreakEngine *lbe = nullptr;
138 UErrorCode status = U_ZERO_ERROR;
139 ensureEngines(status);
140 if (U_FAILURE(status) ) {
141 // Note: no way to return error code to caller.
142 return nullptr;
143 }
144
145 static UMutex gBreakEngineMutex;
146 Mutex m(&gBreakEngineMutex);
147 int32_t i = fEngines->size();
148 while (--i >= 0) {
149 lbe = (const LanguageBreakEngine *)(fEngines->elementAt(i));
150 if (lbe != nullptr && lbe->handles(c, locale)) {
151 return lbe;
152 }
153 }
154
155 // We didn't find an engine. Create one.
156 lbe = loadEngineFor(c, locale);
157 if (lbe != nullptr) {
158 fEngines->push((void *)lbe, status);
159 }
160 return U_SUCCESS(status) ? lbe : nullptr;
161 }
162
163 const LanguageBreakEngine *
loadEngineFor(UChar32 c,const char *)164 ICULanguageBreakFactory::loadEngineFor(UChar32 c, const char*) {
165 UErrorCode status = U_ZERO_ERROR;
166 UScriptCode code = uscript_getScript(c, &status);
167 if (U_SUCCESS(status)) {
168 const LanguageBreakEngine *engine = nullptr;
169 // Try to use LSTM first
170 const LSTMData *data = CreateLSTMDataForScript(code, status);
171 if (U_SUCCESS(status)) {
172 if (data != nullptr) {
173 engine = CreateLSTMBreakEngine(code, data, status);
174 if (U_SUCCESS(status) && engine != nullptr) {
175 return engine;
176 }
177 if (engine != nullptr) {
178 delete engine;
179 engine = nullptr;
180 } else {
181 DeleteLSTMData(data);
182 }
183 }
184 }
185 status = U_ZERO_ERROR; // fallback to dictionary based
186 DictionaryMatcher *m = loadDictionaryMatcherFor(code);
187 if (m != nullptr) {
188 switch(code) {
189 case USCRIPT_THAI:
190 engine = new ThaiBreakEngine(m, status);
191 break;
192 case USCRIPT_LAO:
193 engine = new LaoBreakEngine(m, status);
194 break;
195 case USCRIPT_MYANMAR:
196 engine = new BurmeseBreakEngine(m, status);
197 break;
198 case USCRIPT_KHMER:
199 engine = new KhmerBreakEngine(m, status);
200 break;
201
202 #if !UCONFIG_NO_NORMALIZATION
203 // CJK not available w/o normalization
204 case USCRIPT_HANGUL:
205 engine = new CjkBreakEngine(m, kKorean, status);
206 break;
207
208 // use same BreakEngine and dictionary for both Chinese and Japanese
209 case USCRIPT_HIRAGANA:
210 case USCRIPT_KATAKANA:
211 case USCRIPT_HAN:
212 engine = new CjkBreakEngine(m, kChineseJapanese, status);
213 break;
214 #if 0
215 // TODO: Have to get some characters with script=common handled
216 // by CjkBreakEngine (e.g. U+309B). Simply subjecting
217 // them to CjkBreakEngine does not work. The engine has to
218 // special-case them.
219 case USCRIPT_COMMON:
220 {
221 UBlockCode block = ublock_getCode(code);
222 if (block == UBLOCK_HIRAGANA || block == UBLOCK_KATAKANA)
223 engine = new CjkBreakEngine(dict, kChineseJapanese, status);
224 break;
225 }
226 #endif
227 #endif
228
229 default:
230 break;
231 }
232 if (engine == nullptr) {
233 delete m;
234 }
235 else if (U_FAILURE(status)) {
236 delete engine;
237 engine = nullptr;
238 }
239 return engine;
240 }
241 }
242 return nullptr;
243 }
244
245 DictionaryMatcher *
loadDictionaryMatcherFor(UScriptCode script)246 ICULanguageBreakFactory::loadDictionaryMatcherFor(UScriptCode script) {
247 UErrorCode status = U_ZERO_ERROR;
248 // open root from brkitr tree.
249 UResourceBundle *b = ures_open(U_ICUDATA_BRKITR, "", &status);
250 b = ures_getByKeyWithFallback(b, "dictionaries", b, &status);
251 int32_t dictnlength = 0;
252 const char16_t *dictfname =
253 ures_getStringByKeyWithFallback(b, uscript_getShortName(script), &dictnlength, &status);
254 if (U_FAILURE(status)) {
255 ures_close(b);
256 return nullptr;
257 }
258 CharString dictnbuf;
259 CharString ext;
260 const char16_t *extStart = u_memrchr(dictfname, 0x002e, dictnlength); // last dot
261 if (extStart != nullptr) {
262 int32_t len = (int32_t)(extStart - dictfname);
263 ext.appendInvariantChars(UnicodeString(false, extStart + 1, dictnlength - len - 1), status);
264 dictnlength = len;
265 }
266 dictnbuf.appendInvariantChars(UnicodeString(false, dictfname, dictnlength), status);
267 ures_close(b);
268
269 UDataMemory *file = udata_open(U_ICUDATA_BRKITR, ext.data(), dictnbuf.data(), &status);
270 if (U_SUCCESS(status)) {
271 // build trie
272 const uint8_t *data = (const uint8_t *)udata_getMemory(file);
273 const int32_t *indexes = (const int32_t *)data;
274 const int32_t offset = indexes[DictionaryData::IX_STRING_TRIE_OFFSET];
275 const int32_t trieType = indexes[DictionaryData::IX_TRIE_TYPE] & DictionaryData::TRIE_TYPE_MASK;
276 DictionaryMatcher *m = nullptr;
277 if (trieType == DictionaryData::TRIE_TYPE_BYTES) {
278 const int32_t transform = indexes[DictionaryData::IX_TRANSFORM];
279 const char *characters = (const char *)(data + offset);
280 m = new BytesDictionaryMatcher(characters, transform, file);
281 }
282 else if (trieType == DictionaryData::TRIE_TYPE_UCHARS) {
283 const char16_t *characters = (const char16_t *)(data + offset);
284 m = new UCharsDictionaryMatcher(characters, file);
285 }
286 if (m == nullptr) {
287 // no matcher exists to take ownership - either we are an invalid
288 // type or memory allocation failed
289 udata_close(file);
290 }
291 return m;
292 } else if (dictfname != nullptr) {
293 // we don't have a dictionary matcher.
294 // returning nullptr here will cause us to fail to find a dictionary break engine, as expected
295 status = U_ZERO_ERROR;
296 return nullptr;
297 }
298 return nullptr;
299 }
300
301
addExternalEngine(ExternalBreakEngine * external,UErrorCode & status)302 void ICULanguageBreakFactory::addExternalEngine(
303 ExternalBreakEngine* external, UErrorCode& status) {
304 LocalPointer<ExternalBreakEngine> engine(external, status);
305 ensureEngines(status);
306 LocalPointer<BreakEngineWrapper> wrapper(
307 new BreakEngineWrapper(engine.orphan(), status), status);
308 static UMutex gBreakEngineMutex;
309 Mutex m(&gBreakEngineMutex);
310 fEngines->push(wrapper.getAlias(), status);
311 wrapper.orphan();
312 }
313
BreakEngineWrapper(ExternalBreakEngine * engine,UErrorCode & status)314 BreakEngineWrapper::BreakEngineWrapper(
315 ExternalBreakEngine* engine, UErrorCode &status) : delegate(engine, status) {
316 }
317
~BreakEngineWrapper()318 BreakEngineWrapper::~BreakEngineWrapper() {
319 }
320
handles(UChar32 c,const char * locale) const321 UBool BreakEngineWrapper::handles(UChar32 c, const char* locale) const {
322 return delegate->isFor(c, locale);
323 }
324
findBreaks(UText * text,int32_t startPos,int32_t endPos,UVector32 & foundBreaks,UBool,UErrorCode & status) const325 int32_t BreakEngineWrapper::findBreaks(
326 UText *text,
327 int32_t startPos,
328 int32_t endPos,
329 UVector32 &foundBreaks,
330 UBool /* isPhraseBreaking */,
331 UErrorCode &status) const {
332 if (U_FAILURE(status)) return 0;
333 int32_t result = 0;
334
335 // Find the span of characters included in the set.
336 // The span to break begins at the current position in the text, and
337 // extends towards the start or end of the text, depending on 'reverse'.
338
339 utext_setNativeIndex(text, startPos);
340 int32_t start = (int32_t)utext_getNativeIndex(text);
341 int32_t current;
342 int32_t rangeStart;
343 int32_t rangeEnd;
344 UChar32 c = utext_current32(text);
345 while((current = (int32_t)utext_getNativeIndex(text)) < endPos && delegate->handles(c)) {
346 utext_next32(text); // TODO: recast loop for postincrement
347 c = utext_current32(text);
348 }
349 rangeStart = start;
350 rangeEnd = current;
351 int32_t beforeSize = foundBreaks.size();
352 int32_t additionalCapacity = rangeEnd - rangeStart + 1;
353 // enlarge to contains (rangeEnd-rangeStart+1) more items
354 foundBreaks.ensureCapacity(beforeSize+additionalCapacity, status);
355 if (U_FAILURE(status)) return 0;
356 foundBreaks.setSize(beforeSize + beforeSize+additionalCapacity);
357 result = delegate->fillBreaks(text, rangeStart, rangeEnd, foundBreaks.getBuffer()+beforeSize,
358 additionalCapacity, status);
359 if (U_FAILURE(status)) return 0;
360 foundBreaks.setSize(beforeSize + result);
361 utext_setNativeIndex(text, current);
362 return result;
363 }
364
365 U_NAMESPACE_END
366
367 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
368