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
25 #include "brkeng.h"
26 #include "cmemory.h"
27 #include "dictbe.h"
28 #include "lstmbe.h"
29 #include "charstr.h"
30 #include "dictionarydata.h"
31 #include "mutex.h"
32 #include "uvector.h"
33 #include "umutex.h"
34 #include "uresimp.h"
35 #include "ubrkimpl.h"
36
37 U_NAMESPACE_BEGIN
38
39 /*
40 ******************************************************************
41 */
42
LanguageBreakEngine()43 LanguageBreakEngine::LanguageBreakEngine() {
44 }
45
~LanguageBreakEngine()46 LanguageBreakEngine::~LanguageBreakEngine() {
47 }
48
49 /*
50 ******************************************************************
51 */
52
LanguageBreakFactory()53 LanguageBreakFactory::LanguageBreakFactory() {
54 }
55
~LanguageBreakFactory()56 LanguageBreakFactory::~LanguageBreakFactory() {
57 }
58
59 /*
60 ******************************************************************
61 */
62
UnhandledEngine(UErrorCode & status)63 UnhandledEngine::UnhandledEngine(UErrorCode &status) : fHandled(nullptr) {
64 (void)status;
65 }
66
~UnhandledEngine()67 UnhandledEngine::~UnhandledEngine() {
68 delete fHandled;
69 fHandled = nullptr;
70 }
71
72 UBool
handles(UChar32 c) const73 UnhandledEngine::handles(UChar32 c) const {
74 return fHandled && fHandled->contains(c);
75 }
76
77 int32_t
findBreaks(UText * text,int32_t,int32_t endPos,UVector32 &,UErrorCode & status) const78 UnhandledEngine::findBreaks( UText *text,
79 int32_t /* startPos */,
80 int32_t endPos,
81 UVector32 &/*foundBreaks*/,
82 UErrorCode &status) const {
83 if (U_FAILURE(status)) return 0;
84 UChar32 c = utext_current32(text);
85 while((int32_t)utext_getNativeIndex(text) < endPos && fHandled->contains(c)) {
86 utext_next32(text); // TODO: recast loop to work with post-increment operations.
87 c = utext_current32(text);
88 }
89 return 0;
90 }
91
92 void
handleCharacter(UChar32 c)93 UnhandledEngine::handleCharacter(UChar32 c) {
94 if (fHandled == nullptr) {
95 fHandled = new UnicodeSet();
96 if (fHandled == nullptr) {
97 return;
98 }
99 }
100 if (!fHandled->contains(c)) {
101 UErrorCode status = U_ZERO_ERROR;
102 // Apply the entire script of the character.
103 int32_t script = u_getIntPropertyValue(c, UCHAR_SCRIPT);
104 fHandled->applyIntPropertyValue(UCHAR_SCRIPT, script, status);
105 }
106 }
107
108 /*
109 ******************************************************************
110 */
111
ICULanguageBreakFactory(UErrorCode &)112 ICULanguageBreakFactory::ICULanguageBreakFactory(UErrorCode &/*status*/) {
113 fEngines = 0;
114 }
115
~ICULanguageBreakFactory()116 ICULanguageBreakFactory::~ICULanguageBreakFactory() {
117 if (fEngines != 0) {
118 delete fEngines;
119 }
120 }
121
122 U_NAMESPACE_END
123 U_CDECL_BEGIN
_deleteEngine(void * obj)124 static void U_CALLCONV _deleteEngine(void *obj) {
125 delete (const icu::LanguageBreakEngine *) obj;
126 }
127 U_CDECL_END
128 U_NAMESPACE_BEGIN
129
130 const LanguageBreakEngine *
getEngineFor(UChar32 c)131 ICULanguageBreakFactory::getEngineFor(UChar32 c) {
132 const LanguageBreakEngine *lbe = NULL;
133 UErrorCode status = U_ZERO_ERROR;
134
135 static UMutex gBreakEngineMutex;
136 Mutex m(&gBreakEngineMutex);
137
138 if (fEngines == nullptr) {
139 LocalPointer<UStack> engines(new UStack(_deleteEngine, nullptr, status), status);
140 if (U_FAILURE(status) ) {
141 // Note: no way to return error code to caller.
142 return nullptr;
143 }
144 fEngines = engines.orphan();
145 } else {
146 int32_t i = fEngines->size();
147 while (--i >= 0) {
148 lbe = (const LanguageBreakEngine *)(fEngines->elementAt(i));
149 if (lbe != NULL && lbe->handles(c)) {
150 return lbe;
151 }
152 }
153 }
154
155 // We didn't find an engine. Create one.
156 lbe = loadEngineFor(c);
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)164 ICULanguageBreakFactory::loadEngineFor(UChar32 c) {
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 != NULL) {
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 == NULL) {
233 delete m;
234 }
235 else if (U_FAILURE(status)) {
236 delete engine;
237 engine = NULL;
238 }
239 return engine;
240 }
241 }
242 return NULL;
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 UChar *dictfname =
253 ures_getStringByKeyWithFallback(b, uscript_getShortName(script), &dictnlength, &status);
254 if (U_FAILURE(status)) {
255 ures_close(b);
256 return NULL;
257 }
258 CharString dictnbuf;
259 CharString ext;
260 const UChar *extStart = u_memrchr(dictfname, 0x002e, dictnlength); // last dot
261 if (extStart != NULL) {
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 = NULL;
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 UChar *characters = (const UChar *)(data + offset);
284 m = new UCharsDictionaryMatcher(characters, file);
285 }
286 if (m == NULL) {
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 != NULL) {
293 // we don't have a dictionary matcher.
294 // returning NULL here will cause us to fail to find a dictionary break engine, as expected
295 status = U_ZERO_ERROR;
296 return NULL;
297 }
298 return NULL;
299 }
300
301 U_NAMESPACE_END
302
303 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
304