]>
Commit | Line | Data |
---|---|---|
729e4ab9 A |
1 | /* |
2 | ******************************************************************************* | |
3 | * | |
b331163b | 4 | * Copyright (C) 2009-2014, International Business Machines |
729e4ab9 A |
5 | * Corporation and others. All Rights Reserved. |
6 | * | |
7 | ******************************************************************************* | |
8 | * file name: n2builder.cpp | |
9 | * encoding: US-ASCII | |
10 | * tab size: 8 (not used) | |
11 | * indentation:4 | |
12 | * | |
13 | * created on: 2009nov25 | |
14 | * created by: Markus W. Scherer | |
15 | * | |
16 | * Builds Normalizer2 data and writes a binary .nrm file. | |
17 | * For the file format see source/common/normalizer2impl.h. | |
18 | */ | |
19 | ||
20 | #include "unicode/utypes.h" | |
21 | #include "n2builder.h" | |
22 | ||
23 | #include <stdio.h> | |
24 | #include <stdlib.h> | |
25 | #include <string.h> | |
26 | #if U_HAVE_STD_STRING | |
27 | #include <vector> | |
28 | #endif | |
29 | #include "unicode/errorcode.h" | |
30 | #include "unicode/localpointer.h" | |
31 | #include "unicode/putil.h" | |
32 | #include "unicode/udata.h" | |
33 | #include "unicode/uniset.h" | |
34 | #include "unicode/unistr.h" | |
35 | #include "unicode/ustring.h" | |
b331163b | 36 | #include "charstr.h" |
729e4ab9 A |
37 | #include "hash.h" |
38 | #include "normalizer2impl.h" | |
39 | #include "toolutil.h" | |
40 | #include "unewdata.h" | |
41 | #include "utrie2.h" | |
42 | #include "uvectr32.h" | |
b331163b | 43 | #include "writesrc.h" |
729e4ab9 A |
44 | |
45 | #if !UCONFIG_NO_NORMALIZATION | |
46 | ||
47 | /* UDataInfo cf. udata.h */ | |
48 | static UDataInfo dataInfo={ | |
49 | sizeof(UDataInfo), | |
50 | 0, | |
51 | ||
52 | U_IS_BIG_ENDIAN, | |
53 | U_CHARSET_FAMILY, | |
54 | U_SIZEOF_UCHAR, | |
55 | 0, | |
56 | ||
57 | { 0x4e, 0x72, 0x6d, 0x32 }, /* dataFormat="Nrm2" */ | |
4388f060 | 58 | { 2, 0, 0, 0 }, /* formatVersion */ |
729e4ab9 A |
59 | { 5, 2, 0, 0 } /* dataVersion (Unicode version) */ |
60 | }; | |
61 | ||
62 | U_NAMESPACE_BEGIN | |
63 | ||
64 | class HangulIterator { | |
65 | public: | |
66 | struct Range { | |
67 | UChar32 start, limit; | |
68 | uint16_t norm16; | |
69 | }; | |
70 | ||
71 | HangulIterator() : rangeIndex(0) {} | |
72 | const Range *nextRange() { | |
b331163b | 73 | if(rangeIndex<UPRV_LENGTHOF(ranges)) { |
729e4ab9 A |
74 | return ranges+rangeIndex++; |
75 | } else { | |
76 | return NULL; | |
77 | } | |
78 | } | |
79 | void reset() { rangeIndex=0; } | |
80 | private: | |
81 | static const Range ranges[4]; | |
82 | int32_t rangeIndex; | |
83 | }; | |
84 | ||
85 | const HangulIterator::Range HangulIterator::ranges[4]={ | |
86 | { Hangul::JAMO_L_BASE, Hangul::JAMO_L_BASE+Hangul::JAMO_L_COUNT, 1 }, | |
87 | { Hangul::JAMO_V_BASE, Hangul::JAMO_V_BASE+Hangul::JAMO_V_COUNT, Normalizer2Impl::JAMO_VT }, | |
88 | // JAMO_T_BASE+1: not U+11A7 | |
89 | { Hangul::JAMO_T_BASE+1, Hangul::JAMO_T_BASE+Hangul::JAMO_T_COUNT, Normalizer2Impl::JAMO_VT }, | |
90 | { Hangul::HANGUL_BASE, Hangul::HANGUL_BASE+Hangul::HANGUL_COUNT, 0 }, // will become minYesNo | |
91 | }; | |
92 | ||
93 | struct CompositionPair { | |
94 | CompositionPair(UChar32 t, UChar32 c) : trail(t), composite(c) {} | |
95 | UChar32 trail, composite; | |
96 | }; | |
97 | ||
98 | struct Norm { | |
99 | enum MappingType { NONE, REMOVED, ROUND_TRIP, ONE_WAY }; | |
100 | ||
101 | UBool hasMapping() const { return mappingType>REMOVED; } | |
102 | ||
103 | // Requires hasMapping() and well-formed mapping. | |
104 | void setMappingCP() { | |
105 | UChar32 c; | |
106 | if(!mapping->isEmpty() && mapping->length()==U16_LENGTH(c=mapping->char32At(0))) { | |
107 | mappingCP=c; | |
108 | } else { | |
109 | mappingCP=U_SENTINEL; | |
110 | } | |
111 | } | |
112 | ||
113 | const CompositionPair *getCompositionPairs(int32_t &length) const { | |
114 | if(compositions==NULL) { | |
115 | length=0; | |
116 | return NULL; | |
117 | } else { | |
118 | length=compositions->size()/2; | |
119 | return reinterpret_cast<const CompositionPair *>(compositions->getBuffer()); | |
120 | } | |
121 | } | |
122 | ||
123 | UnicodeString *mapping; | |
4388f060 | 124 | UnicodeString *rawMapping; // non-NULL if the mapping is further decomposed |
729e4ab9 A |
125 | UChar32 mappingCP; // >=0 if mapping to 1 code point |
126 | int32_t mappingPhase; | |
127 | MappingType mappingType; | |
128 | ||
129 | UVector32 *compositions; // (trail, composite) pairs | |
130 | uint8_t cc; | |
131 | UBool combinesBack; | |
132 | UBool hasNoCompBoundaryAfter; | |
133 | ||
134 | enum OffsetType { | |
4388f060 A |
135 | OFFSET_NONE, |
136 | // Composition for back-combining character. Allowed, but not normally used. | |
137 | OFFSET_MAYBE_YES, | |
138 | // Composition for a starter that does not have a decomposition mapping. | |
139 | OFFSET_YES_YES, | |
140 | // Round-trip mapping & composition for a starter. | |
141 | OFFSET_YES_NO_MAPPING_AND_COMPOSITION, | |
142 | // Round-trip mapping for a starter that itself does not combine-forward. | |
143 | OFFSET_YES_NO_MAPPING_ONLY, | |
144 | // One-way mapping. | |
145 | OFFSET_NO_NO, | |
146 | // Delta for an algorithmic one-way mapping. | |
729e4ab9 A |
147 | OFFSET_DELTA |
148 | }; | |
149 | enum { OFFSET_SHIFT=4, OFFSET_MASK=(1<<OFFSET_SHIFT)-1 }; | |
150 | int32_t offset; | |
151 | }; | |
152 | ||
153 | class Normalizer2DBEnumerator { | |
154 | public: | |
155 | Normalizer2DBEnumerator(Normalizer2DataBuilder &b) : builder(b) {} | |
156 | virtual ~Normalizer2DBEnumerator() {} | |
157 | virtual UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) = 0; | |
158 | Normalizer2DBEnumerator *ptr() { return this; } | |
159 | protected: | |
160 | Normalizer2DataBuilder &builder; | |
161 | }; | |
162 | ||
163 | U_CDECL_BEGIN | |
164 | ||
165 | static UBool U_CALLCONV | |
166 | enumRangeHandler(const void *context, UChar32 start, UChar32 end, uint32_t value) { | |
167 | return ((Normalizer2DBEnumerator *)context)->rangeHandler(start, end, value); | |
168 | } | |
169 | ||
170 | U_CDECL_END | |
171 | ||
172 | Normalizer2DataBuilder::Normalizer2DataBuilder(UErrorCode &errorCode) : | |
b331163b A |
173 | phase(0), overrideHandling(OVERRIDE_PREVIOUS), optimization(OPTIMIZE_NORMAL), |
174 | norm16TrieLength(0) { | |
729e4ab9 A |
175 | memset(unicodeVersion, 0, sizeof(unicodeVersion)); |
176 | normTrie=utrie2_open(0, 0, &errorCode); | |
177 | normMem=utm_open("gennorm2 normalization structs", 10000, 0x110100, sizeof(Norm)); | |
178 | norms=allocNorm(); // unused Norm struct at index 0 | |
179 | memset(indexes, 0, sizeof(indexes)); | |
4388f060 | 180 | memset(smallFCD, 0, sizeof(smallFCD)); |
729e4ab9 A |
181 | } |
182 | ||
183 | Normalizer2DataBuilder::~Normalizer2DataBuilder() { | |
184 | utrie2_close(normTrie); | |
185 | int32_t normsLength=utm_countItems(normMem); | |
186 | for(int32_t i=1; i<normsLength; ++i) { | |
187 | delete norms[i].mapping; | |
4388f060 | 188 | delete norms[i].rawMapping; |
729e4ab9 A |
189 | delete norms[i].compositions; |
190 | } | |
191 | utm_close(normMem); | |
192 | utrie2_close(norm16Trie); | |
193 | } | |
194 | ||
195 | void | |
196 | Normalizer2DataBuilder::setUnicodeVersion(const char *v) { | |
4388f060 A |
197 | UVersionInfo nullVersion={ 0, 0, 0, 0 }; |
198 | UVersionInfo version; | |
199 | u_versionFromString(version, v); | |
200 | if( 0!=memcmp(version, unicodeVersion, U_MAX_VERSION_LENGTH) && | |
201 | 0!=memcmp(nullVersion, unicodeVersion, U_MAX_VERSION_LENGTH) | |
202 | ) { | |
203 | char buffer[U_MAX_VERSION_STRING_LENGTH]; | |
204 | u_versionToString(unicodeVersion, buffer); | |
205 | fprintf(stderr, "gennorm2 error: multiple inconsistent Unicode version numbers %s vs. %s\n", | |
206 | buffer, v); | |
207 | exit(U_ILLEGAL_ARGUMENT_ERROR); | |
208 | } | |
209 | memcpy(unicodeVersion, version, U_MAX_VERSION_LENGTH); | |
729e4ab9 A |
210 | } |
211 | ||
212 | Norm *Normalizer2DataBuilder::allocNorm() { | |
213 | Norm *p=(Norm *)utm_alloc(normMem); | |
214 | norms=(Norm *)utm_getStart(normMem); // in case it got reallocated | |
215 | return p; | |
216 | } | |
217 | ||
218 | /* get an existing Norm unit */ | |
219 | Norm *Normalizer2DataBuilder::getNorm(UChar32 c) { | |
220 | uint32_t i=utrie2_get32(normTrie, c); | |
221 | if(i==0) { | |
222 | return NULL; | |
223 | } | |
224 | return norms+i; | |
225 | } | |
226 | ||
227 | const Norm &Normalizer2DataBuilder::getNormRef(UChar32 c) const { | |
228 | return norms[utrie2_get32(normTrie, c)]; | |
229 | } | |
230 | ||
231 | /* | |
232 | * get or create a Norm unit; | |
233 | * get or create the intermediate trie entries for it as well | |
234 | */ | |
235 | Norm *Normalizer2DataBuilder::createNorm(UChar32 c) { | |
236 | uint32_t i=utrie2_get32(normTrie, c); | |
237 | if(i!=0) { | |
238 | return norms+i; | |
239 | } else { | |
240 | /* allocate Norm */ | |
241 | Norm *p=allocNorm(); | |
242 | IcuToolErrorCode errorCode("gennorm2/createNorm()"); | |
243 | utrie2_set32(normTrie, c, (uint32_t)(p-norms), errorCode); | |
244 | return p; | |
245 | } | |
246 | } | |
247 | ||
248 | Norm *Normalizer2DataBuilder::checkNormForMapping(Norm *p, UChar32 c) { | |
249 | if(p!=NULL) { | |
250 | if(p->mappingType!=Norm::NONE) { | |
251 | if( overrideHandling==OVERRIDE_NONE || | |
252 | (overrideHandling==OVERRIDE_PREVIOUS && p->mappingPhase==phase) | |
253 | ) { | |
254 | fprintf(stderr, | |
255 | "error in gennorm2 phase %d: " | |
256 | "not permitted to override mapping for U+%04lX from phase %d\n", | |
257 | (int)phase, (long)c, (int)p->mappingPhase); | |
258 | exit(U_INVALID_FORMAT_ERROR); | |
259 | } | |
260 | delete p->mapping; | |
261 | p->mapping=NULL; | |
262 | } | |
263 | p->mappingPhase=phase; | |
264 | } | |
265 | return p; | |
266 | } | |
267 | ||
268 | void Normalizer2DataBuilder::setOverrideHandling(OverrideHandling oh) { | |
269 | overrideHandling=oh; | |
270 | ++phase; | |
271 | } | |
272 | ||
273 | void Normalizer2DataBuilder::setCC(UChar32 c, uint8_t cc) { | |
274 | createNorm(c)->cc=cc; | |
275 | } | |
276 | ||
277 | uint8_t Normalizer2DataBuilder::getCC(UChar32 c) const { | |
278 | return getNormRef(c).cc; | |
279 | } | |
280 | ||
281 | static UBool isWellFormed(const UnicodeString &s) { | |
282 | UErrorCode errorCode=U_ZERO_ERROR; | |
283 | u_strToUTF8(NULL, 0, NULL, s.getBuffer(), s.length(), &errorCode); | |
284 | return U_SUCCESS(errorCode) || errorCode==U_BUFFER_OVERFLOW_ERROR; | |
285 | } | |
286 | ||
287 | void Normalizer2DataBuilder::setOneWayMapping(UChar32 c, const UnicodeString &m) { | |
288 | if(!isWellFormed(m)) { | |
289 | fprintf(stderr, | |
290 | "error in gennorm2 phase %d: " | |
291 | "illegal one-way mapping from U+%04lX to malformed string\n", | |
292 | (int)phase, (long)c); | |
293 | exit(U_INVALID_FORMAT_ERROR); | |
294 | } | |
295 | Norm *p=checkNormForMapping(createNorm(c), c); | |
296 | p->mapping=new UnicodeString(m); | |
297 | p->mappingType=Norm::ONE_WAY; | |
298 | p->setMappingCP(); | |
299 | } | |
300 | ||
301 | void Normalizer2DataBuilder::setRoundTripMapping(UChar32 c, const UnicodeString &m) { | |
302 | if(U_IS_SURROGATE(c)) { | |
303 | fprintf(stderr, | |
304 | "error in gennorm2 phase %d: " | |
305 | "illegal round-trip mapping from surrogate code point U+%04lX\n", | |
306 | (int)phase, (long)c); | |
307 | exit(U_INVALID_FORMAT_ERROR); | |
308 | } | |
309 | if(!isWellFormed(m)) { | |
310 | fprintf(stderr, | |
311 | "error in gennorm2 phase %d: " | |
312 | "illegal round-trip mapping from U+%04lX to malformed string\n", | |
313 | (int)phase, (long)c); | |
314 | exit(U_INVALID_FORMAT_ERROR); | |
315 | } | |
316 | int32_t numCP=u_countChar32(m.getBuffer(), m.length()); | |
317 | if(numCP!=2) { | |
318 | fprintf(stderr, | |
319 | "error in gennorm2 phase %d: " | |
320 | "illegal round-trip mapping from U+%04lX to %d!=2 code points\n", | |
321 | (int)phase, (long)c, (int)numCP); | |
322 | exit(U_INVALID_FORMAT_ERROR); | |
323 | } | |
324 | Norm *p=checkNormForMapping(createNorm(c), c); | |
325 | p->mapping=new UnicodeString(m); | |
326 | p->mappingType=Norm::ROUND_TRIP; | |
327 | p->mappingCP=U_SENTINEL; | |
328 | } | |
329 | ||
330 | void Normalizer2DataBuilder::removeMapping(UChar32 c) { | |
331 | Norm *p=checkNormForMapping(getNorm(c), c); | |
332 | if(p!=NULL) { | |
333 | p->mappingType=Norm::REMOVED; | |
334 | } | |
335 | } | |
336 | ||
337 | class CompositionBuilder : public Normalizer2DBEnumerator { | |
338 | public: | |
339 | CompositionBuilder(Normalizer2DataBuilder &b) : Normalizer2DBEnumerator(b) {} | |
340 | virtual UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { | |
341 | builder.addComposition(start, end, value); | |
342 | return TRUE; | |
343 | } | |
344 | }; | |
345 | ||
346 | void | |
347 | Normalizer2DataBuilder::addComposition(UChar32 start, UChar32 end, uint32_t value) { | |
348 | if(norms[value].mappingType==Norm::ROUND_TRIP) { | |
349 | if(start!=end) { | |
350 | fprintf(stderr, | |
351 | "gennorm2 error: same round-trip mapping for " | |
352 | "more than 1 code point U+%04lX..U+%04lX\n", | |
353 | (long)start, (long)end); | |
354 | exit(U_INVALID_FORMAT_ERROR); | |
355 | } | |
356 | if(norms[value].cc!=0) { | |
357 | fprintf(stderr, | |
358 | "gennorm2 error: " | |
359 | "U+%04lX has a round-trip mapping and ccc!=0, " | |
360 | "not possible in Unicode normalization\n", | |
361 | (long)start); | |
362 | exit(U_INVALID_FORMAT_ERROR); | |
363 | } | |
364 | // setRoundTripMapping() ensured that there are exactly two code points. | |
365 | const UnicodeString &m=*norms[value].mapping; | |
366 | UChar32 lead=m.char32At(0); | |
367 | UChar32 trail=m.char32At(m.length()-1); | |
368 | if(getCC(lead)!=0) { | |
369 | fprintf(stderr, | |
370 | "gennorm2 error: " | |
371 | "U+%04lX's round-trip mapping's starter U+%04lX has ccc!=0, " | |
372 | "not possible in Unicode normalization\n", | |
373 | (long)start, (long)lead); | |
374 | exit(U_INVALID_FORMAT_ERROR); | |
375 | } | |
376 | // Flag for trailing character. | |
377 | createNorm(trail)->combinesBack=TRUE; | |
378 | // Insert (trail, composite) pair into compositions list for the lead character. | |
379 | IcuToolErrorCode errorCode("gennorm2/addComposition()"); | |
380 | Norm *leadNorm=createNorm(lead); | |
381 | UVector32 *compositions=leadNorm->compositions; | |
382 | int32_t i; | |
383 | if(compositions==NULL) { | |
384 | compositions=leadNorm->compositions=new UVector32(errorCode); | |
385 | i=0; // "insert" the first pair at index 0 | |
386 | } else { | |
387 | // Insertion sort, and check for duplicate trail characters. | |
388 | int32_t length; | |
389 | const CompositionPair *pairs=leadNorm->getCompositionPairs(length); | |
390 | for(i=0; i<length; ++i) { | |
391 | if(trail==pairs[i].trail) { | |
392 | fprintf(stderr, | |
393 | "gennorm2 error: same round-trip mapping for " | |
394 | "more than 1 code point (e.g., U+%04lX) to U+%04lX + U+%04lX\n", | |
395 | (long)start, (long)lead, (long)trail); | |
396 | exit(U_INVALID_FORMAT_ERROR); | |
397 | } | |
398 | if(trail<pairs[i].trail) { | |
399 | break; | |
400 | } | |
401 | } | |
402 | } | |
403 | compositions->insertElementAt(trail, 2*i, errorCode); | |
404 | compositions->insertElementAt(start, 2*i+1, errorCode); | |
405 | } | |
406 | } | |
407 | ||
408 | UBool Normalizer2DataBuilder::combinesWithCCBetween(const Norm &norm, | |
409 | uint8_t lowCC, uint8_t highCC) const { | |
410 | if((highCC-lowCC)>=2) { | |
411 | int32_t length; | |
412 | const CompositionPair *pairs=norm.getCompositionPairs(length); | |
413 | for(int32_t i=0; i<length; ++i) { | |
414 | uint8_t trailCC=getCC(pairs[i].trail); | |
415 | if(lowCC<trailCC && trailCC<highCC) { | |
416 | return TRUE; | |
417 | } | |
418 | } | |
419 | } | |
420 | return FALSE; | |
421 | } | |
422 | ||
423 | UChar32 Normalizer2DataBuilder::combine(const Norm &norm, UChar32 trail) const { | |
424 | int32_t length; | |
425 | const CompositionPair *pairs=norm.getCompositionPairs(length); | |
426 | for(int32_t i=0; i<length; ++i) { | |
427 | if(trail==pairs[i].trail) { | |
428 | return pairs[i].composite; | |
429 | } | |
430 | if(trail<pairs[i].trail) { | |
431 | break; | |
432 | } | |
433 | } | |
434 | return U_SENTINEL; | |
435 | } | |
436 | ||
437 | class Decomposer : public Normalizer2DBEnumerator { | |
438 | public: | |
439 | Decomposer(Normalizer2DataBuilder &b) : Normalizer2DBEnumerator(b), didDecompose(FALSE) {} | |
440 | virtual UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { | |
441 | didDecompose|=builder.decompose(start, end, value); | |
442 | return TRUE; | |
443 | } | |
444 | UBool didDecompose; | |
445 | }; | |
446 | ||
447 | UBool | |
448 | Normalizer2DataBuilder::decompose(UChar32 start, UChar32 end, uint32_t value) { | |
449 | if(norms[value].hasMapping()) { | |
4388f060 A |
450 | Norm &norm=norms[value]; |
451 | const UnicodeString &m=*norm.mapping; | |
729e4ab9 A |
452 | UnicodeString *decomposed=NULL; |
453 | const UChar *s=m.getBuffer(); | |
454 | int32_t length=m.length(); | |
455 | int32_t prev, i=0; | |
456 | UChar32 c; | |
457 | while(i<length) { | |
458 | prev=i; | |
459 | U16_NEXT(s, i, length, c); | |
460 | if(start<=c && c<=end) { | |
461 | fprintf(stderr, | |
462 | "gennorm2 error: U+%04lX maps to itself directly or indirectly\n", | |
463 | (long)c); | |
464 | exit(U_INVALID_FORMAT_ERROR); | |
465 | } | |
466 | const Norm &cNorm=getNormRef(c); | |
467 | if(cNorm.hasMapping()) { | |
4388f060 | 468 | if(norm.mappingType==Norm::ROUND_TRIP) { |
729e4ab9 A |
469 | if(prev==0) { |
470 | if(cNorm.mappingType!=Norm::ROUND_TRIP) { | |
471 | fprintf(stderr, | |
472 | "gennorm2 error: " | |
473 | "U+%04lX's round-trip mapping's starter " | |
474 | "U+%04lX one-way-decomposes, " | |
475 | "not possible in Unicode normalization\n", | |
476 | (long)start, (long)c); | |
477 | exit(U_INVALID_FORMAT_ERROR); | |
478 | } | |
479 | uint8_t myTrailCC=getCC(m.char32At(i)); | |
480 | UChar32 cTrailChar=cNorm.mapping->char32At(cNorm.mapping->length()-1); | |
481 | uint8_t cTrailCC=getCC(cTrailChar); | |
482 | if(cTrailCC>myTrailCC) { | |
483 | fprintf(stderr, | |
484 | "gennorm2 error: " | |
485 | "U+%04lX's round-trip mapping's starter " | |
486 | "U+%04lX decomposes and the " | |
487 | "inner/earlier tccc=%hu > outer/following tccc=%hu, " | |
488 | "not possible in Unicode normalization\n", | |
489 | (long)start, (long)c, | |
490 | (short)cTrailCC, (short)myTrailCC); | |
491 | exit(U_INVALID_FORMAT_ERROR); | |
492 | } | |
493 | } else { | |
494 | fprintf(stderr, | |
495 | "gennorm2 error: " | |
496 | "U+%04lX's round-trip mapping's non-starter " | |
497 | "U+%04lX decomposes, " | |
498 | "not possible in Unicode normalization\n", | |
499 | (long)start, (long)c); | |
500 | exit(U_INVALID_FORMAT_ERROR); | |
501 | } | |
502 | } | |
503 | if(decomposed==NULL) { | |
504 | decomposed=new UnicodeString(m, 0, prev); | |
505 | } | |
506 | decomposed->append(*cNorm.mapping); | |
507 | } else if(Hangul::isHangul(c)) { | |
508 | UChar buffer[3]; | |
509 | int32_t hangulLength=Hangul::decompose(c, buffer); | |
4388f060 | 510 | if(norm.mappingType==Norm::ROUND_TRIP && prev!=0) { |
729e4ab9 A |
511 | fprintf(stderr, |
512 | "gennorm2 error: " | |
513 | "U+%04lX's round-trip mapping's non-starter " | |
514 | "U+%04lX decomposes, " | |
515 | "not possible in Unicode normalization\n", | |
516 | (long)start, (long)c); | |
517 | exit(U_INVALID_FORMAT_ERROR); | |
518 | } | |
519 | if(decomposed==NULL) { | |
520 | decomposed=new UnicodeString(m, 0, prev); | |
521 | } | |
522 | decomposed->append(buffer, hangulLength); | |
523 | } else if(decomposed!=NULL) { | |
524 | decomposed->append(m, prev, i-prev); | |
525 | } | |
526 | } | |
527 | if(decomposed!=NULL) { | |
4388f060 A |
528 | if(norm.rawMapping==NULL) { |
529 | // Remember the original mapping when decomposing recursively. | |
530 | norm.rawMapping=norm.mapping; | |
531 | } else { | |
532 | delete norm.mapping; | |
533 | } | |
534 | norm.mapping=decomposed; | |
535 | // Not norm.setMappingCP(); because the original mapping | |
729e4ab9 A |
536 | // is most likely to be encodable as a delta. |
537 | return TRUE; | |
538 | } | |
539 | } | |
540 | return FALSE; | |
541 | } | |
542 | ||
543 | class BuilderReorderingBuffer { | |
544 | public: | |
545 | BuilderReorderingBuffer() : fLength(0), fLastStarterIndex(-1), fDidReorder(FALSE) {} | |
546 | void reset() { | |
547 | fLength=0; | |
548 | fLastStarterIndex=-1; | |
549 | fDidReorder=FALSE; | |
550 | } | |
551 | int32_t length() const { return fLength; } | |
552 | UBool isEmpty() const { return fLength==0; } | |
553 | int32_t lastStarterIndex() const { return fLastStarterIndex; } | |
554 | UChar32 charAt(int32_t i) const { return fArray[i]>>8; } | |
555 | uint8_t ccAt(int32_t i) const { return (uint8_t)fArray[i]; } | |
556 | UBool didReorder() const { return fDidReorder; } | |
557 | void append(UChar32 c, uint8_t cc) { | |
558 | if(cc==0 || fLength==0 || ccAt(fLength-1)<=cc) { | |
559 | if(cc==0) { | |
560 | fLastStarterIndex=fLength; | |
561 | } | |
562 | fArray[fLength++]=(c<<8)|cc; | |
563 | return; | |
564 | } | |
565 | // Let this character bubble back to its canonical order. | |
566 | int32_t i=fLength-1; | |
567 | while(i>fLastStarterIndex && ccAt(i)>cc) { | |
568 | --i; | |
569 | } | |
570 | ++i; // after the last starter or prevCC<=cc | |
571 | // Move this and the following characters forward one to make space. | |
572 | for(int32_t j=fLength; i<j; --j) { | |
573 | fArray[j]=fArray[j-1]; | |
574 | } | |
575 | fArray[i]=(c<<8)|cc; | |
576 | ++fLength; | |
577 | fDidReorder=TRUE; | |
578 | } | |
579 | void toString(UnicodeString &dest) { | |
580 | dest.remove(); | |
581 | for(int32_t i=0; i<fLength; ++i) { | |
582 | dest.append(charAt(i)); | |
583 | } | |
584 | } | |
585 | void setComposite(UChar32 composite, int32_t combMarkIndex) { | |
586 | fArray[fLastStarterIndex]=composite<<8; | |
587 | // Remove the combining mark that contributed to the composite. | |
588 | --fLength; | |
589 | while(combMarkIndex<fLength) { | |
590 | fArray[combMarkIndex]=fArray[combMarkIndex+1]; | |
591 | ++combMarkIndex; | |
592 | } | |
593 | } | |
594 | private: | |
595 | int32_t fArray[Normalizer2Impl::MAPPING_LENGTH_MASK]; | |
596 | int32_t fLength; | |
597 | int32_t fLastStarterIndex; | |
598 | UBool fDidReorder; | |
599 | }; | |
600 | ||
601 | void | |
602 | Normalizer2DataBuilder::reorder(Norm *p, BuilderReorderingBuffer &buffer) { | |
603 | UnicodeString &m=*p->mapping; | |
604 | int32_t length=m.length(); | |
605 | if(length>Normalizer2Impl::MAPPING_LENGTH_MASK) { | |
606 | return; // writeMapping() will complain about it and print the code point. | |
607 | } | |
608 | const UChar *s=m.getBuffer(); | |
609 | int32_t i=0; | |
610 | UChar32 c; | |
611 | while(i<length) { | |
612 | U16_NEXT(s, i, length, c); | |
613 | buffer.append(c, getCC(c)); | |
614 | } | |
615 | if(buffer.didReorder()) { | |
616 | buffer.toString(m); | |
617 | } | |
618 | } | |
619 | ||
4388f060 A |
620 | /* |
621 | * Computes the flag for the last code branch in Normalizer2Impl::hasCompBoundaryAfter(). | |
622 | * A starter character with a mapping does not have a composition boundary after it | |
623 | * if the character itself combines-forward (which is tested by the caller of this function), | |
624 | * or it is deleted (mapped to the empty string), | |
625 | * or its mapping contains no starter, | |
626 | * or the last starter combines-forward. | |
627 | */ | |
729e4ab9 A |
628 | UBool Normalizer2DataBuilder::hasNoCompBoundaryAfter(BuilderReorderingBuffer &buffer) { |
629 | if(buffer.isEmpty()) { | |
4388f060 | 630 | return TRUE; // maps-to-empty-string is no boundary of any kind |
729e4ab9 A |
631 | } |
632 | int32_t lastStarterIndex=buffer.lastStarterIndex(); | |
633 | if(lastStarterIndex<0) { | |
634 | return TRUE; // no starter | |
635 | } | |
636 | UChar32 starter=buffer.charAt(lastStarterIndex); | |
637 | if( Hangul::isJamoL(starter) || | |
638 | (Hangul::isJamoV(starter) && | |
639 | 0<lastStarterIndex && Hangul::isJamoL(buffer.charAt(lastStarterIndex-1))) | |
640 | ) { | |
641 | // A Jamo leading consonant or an LV pair combines-forward if it is at the end, | |
642 | // otherwise it is blocked. | |
643 | return lastStarterIndex==buffer.length()-1; | |
644 | } | |
4388f060 | 645 | // Note: There can be no Hangul syllable in the fully decomposed mapping. |
729e4ab9 A |
646 | const Norm *starterNorm=&getNormRef(starter); |
647 | if(starterNorm->compositions==NULL) { | |
648 | return FALSE; // the last starter does not combine forward | |
649 | } | |
650 | // Compose as far as possible, and see if further compositions are possible. | |
651 | uint8_t prevCC=0; | |
652 | for(int32_t combMarkIndex=lastStarterIndex+1; combMarkIndex<buffer.length();) { | |
653 | uint8_t cc=buffer.ccAt(combMarkIndex); // !=0 because after last starter | |
654 | if(combinesWithCCBetween(*starterNorm, prevCC, cc)) { | |
655 | return TRUE; | |
656 | } | |
657 | if( prevCC<cc && | |
658 | (starter=combine(*starterNorm, buffer.charAt(combMarkIndex)))>=0 | |
659 | ) { | |
660 | buffer.setComposite(starter, combMarkIndex); | |
661 | starterNorm=&getNormRef(starter); | |
662 | if(starterNorm->compositions==NULL) { | |
663 | return FALSE; // the composite does not combine further | |
664 | } | |
665 | } else { | |
666 | prevCC=cc; | |
667 | ++combMarkIndex; | |
668 | } | |
669 | } | |
670 | // TRUE if the final, forward-combining starter is at the end. | |
671 | return prevCC==0; | |
672 | } | |
673 | ||
674 | // Requires p->hasMapping(). | |
4388f060 A |
675 | // Returns the offset of the "first unit" from the beginning of the extraData for c. |
676 | // That is the same as the length of the optional data for the raw mapping and the ccc/lccc word. | |
677 | int32_t Normalizer2DataBuilder::writeMapping(UChar32 c, const Norm *p, UnicodeString &dataString) { | |
729e4ab9 A |
678 | UnicodeString &m=*p->mapping; |
679 | int32_t length=m.length(); | |
680 | if(length>Normalizer2Impl::MAPPING_LENGTH_MASK) { | |
681 | fprintf(stderr, | |
682 | "gennorm2 error: " | |
683 | "mapping for U+%04lX longer than maximum of %d\n", | |
684 | (long)c, Normalizer2Impl::MAPPING_LENGTH_MASK); | |
685 | exit(U_INVALID_FORMAT_ERROR); | |
686 | } | |
687 | int32_t leadCC, trailCC; | |
688 | if(length==0) { | |
689 | leadCC=trailCC=0; | |
690 | } else { | |
691 | leadCC=getCC(m.char32At(0)); | |
692 | trailCC=getCC(m.char32At(length-1)); | |
693 | } | |
694 | if(c<Normalizer2Impl::MIN_CCC_LCCC_CP && (p->cc!=0 || leadCC!=0)) { | |
695 | fprintf(stderr, | |
696 | "gennorm2 error: " | |
697 | "U+%04lX below U+0300 has ccc!=0 or lccc!=0, not supported by ICU\n", | |
698 | (long)c); | |
699 | exit(U_INVALID_FORMAT_ERROR); | |
700 | } | |
4388f060 A |
701 | // Write small-FCD data. |
702 | if((leadCC|trailCC)!=0) { | |
703 | UChar32 lead= c<=0xffff ? c : U16_LEAD(c); | |
704 | smallFCD[lead>>8]|=(uint8_t)1<<((lead>>5)&7); | |
705 | } | |
706 | // Write the mapping & raw mapping extraData. | |
729e4ab9 | 707 | int32_t firstUnit=length|(trailCC<<8); |
4388f060 A |
708 | int32_t preMappingLength=0; |
709 | if(p->rawMapping!=NULL) { | |
710 | UnicodeString &rm=*p->rawMapping; | |
711 | int32_t rmLength=rm.length(); | |
712 | if(rmLength>Normalizer2Impl::MAPPING_LENGTH_MASK) { | |
713 | fprintf(stderr, | |
714 | "gennorm2 error: " | |
715 | "raw mapping for U+%04lX longer than maximum of %d\n", | |
716 | (long)c, Normalizer2Impl::MAPPING_LENGTH_MASK); | |
717 | exit(U_INVALID_FORMAT_ERROR); | |
718 | } | |
719 | UChar rm0=rm.charAt(0); | |
720 | if( rmLength==length-1 && | |
721 | // 99: overlong substring lengths get pinned to remainder lengths anyway | |
722 | 0==rm.compare(1, 99, m, 2, 99) && | |
723 | rm0>Normalizer2Impl::MAPPING_LENGTH_MASK | |
724 | ) { | |
725 | // Compression: | |
726 | // rawMapping=rm0+mapping.substring(2) -> store only rm0 | |
727 | // | |
728 | // The raw mapping is the same as the final mapping after replacing | |
729 | // the final mapping's first two code units with the raw mapping's first one. | |
730 | // In this case, we store only that first unit, rm0. | |
731 | // This helps with a few hundred mappings. | |
732 | dataString.append(rm0); | |
733 | preMappingLength=1; | |
734 | } else { | |
735 | // Store the raw mapping with its length. | |
736 | dataString.append(rm); | |
737 | dataString.append((UChar)rmLength); | |
738 | preMappingLength=rmLength+1; | |
739 | } | |
740 | firstUnit|=Normalizer2Impl::MAPPING_HAS_RAW_MAPPING; | |
729e4ab9 | 741 | } |
4388f060 A |
742 | int32_t cccLccc=p->cc|(leadCC<<8); |
743 | if(cccLccc!=0) { | |
744 | dataString.append((UChar)cccLccc); | |
745 | ++preMappingLength; | |
746 | firstUnit|=Normalizer2Impl::MAPPING_HAS_CCC_LCCC_WORD; | |
729e4ab9 A |
747 | } |
748 | if(p->hasNoCompBoundaryAfter) { | |
749 | firstUnit|=Normalizer2Impl::MAPPING_NO_COMP_BOUNDARY_AFTER; | |
750 | } | |
751 | dataString.append((UChar)firstUnit); | |
729e4ab9 | 752 | dataString.append(m); |
4388f060 | 753 | return preMappingLength; |
729e4ab9 A |
754 | } |
755 | ||
756 | // Requires p->compositions!=NULL. | |
757 | void Normalizer2DataBuilder::writeCompositions(UChar32 c, const Norm *p, UnicodeString &dataString) { | |
758 | if(p->cc!=0) { | |
759 | fprintf(stderr, | |
760 | "gennorm2 error: " | |
761 | "U+%04lX combines-forward and has ccc!=0, not possible in Unicode normalization\n", | |
762 | (long)c); | |
763 | exit(U_INVALID_FORMAT_ERROR); | |
764 | } | |
765 | int32_t length; | |
766 | const CompositionPair *pairs=p->getCompositionPairs(length); | |
767 | for(int32_t i=0; i<length; ++i) { | |
768 | const CompositionPair &pair=pairs[i]; | |
769 | // 22 bits for the composite character and whether it combines forward. | |
770 | UChar32 compositeAndFwd=pair.composite<<1; | |
771 | if(getNormRef(pair.composite).compositions!=NULL) { | |
772 | compositeAndFwd|=1; // The composite character also combines-forward. | |
773 | } | |
774 | // Encode most pairs in two units and some in three. | |
775 | int32_t firstUnit, secondUnit, thirdUnit; | |
776 | if(pair.trail<Normalizer2Impl::COMP_1_TRAIL_LIMIT) { | |
777 | if(compositeAndFwd<=0xffff) { | |
778 | firstUnit=pair.trail<<1; | |
779 | secondUnit=compositeAndFwd; | |
780 | thirdUnit=-1; | |
781 | } else { | |
782 | firstUnit=(pair.trail<<1)|Normalizer2Impl::COMP_1_TRIPLE; | |
783 | secondUnit=compositeAndFwd>>16; | |
784 | thirdUnit=compositeAndFwd; | |
785 | } | |
786 | } else { | |
787 | firstUnit=(Normalizer2Impl::COMP_1_TRAIL_LIMIT+ | |
788 | (pair.trail>>Normalizer2Impl::COMP_1_TRAIL_SHIFT))| | |
789 | Normalizer2Impl::COMP_1_TRIPLE; | |
790 | secondUnit=(pair.trail<<Normalizer2Impl::COMP_2_TRAIL_SHIFT)| | |
791 | (compositeAndFwd>>16); | |
792 | thirdUnit=compositeAndFwd; | |
793 | } | |
794 | // Set the high bit of the first unit if this is the last composition pair. | |
795 | if(i==(length-1)) { | |
796 | firstUnit|=Normalizer2Impl::COMP_1_LAST_TUPLE; | |
797 | } | |
798 | dataString.append((UChar)firstUnit).append((UChar)secondUnit); | |
799 | if(thirdUnit>=0) { | |
800 | dataString.append((UChar)thirdUnit); | |
801 | } | |
802 | } | |
803 | } | |
804 | ||
805 | class ExtraDataWriter : public Normalizer2DBEnumerator { | |
806 | public: | |
807 | ExtraDataWriter(Normalizer2DataBuilder &b) : | |
808 | Normalizer2DBEnumerator(b), | |
809 | yesYesCompositions(1000, (UChar32)0xffff, 2), // 0=inert, 1=Jamo L, 2=start of compositions | |
4388f060 | 810 | yesNoMappingsAndCompositions(1000, (UChar32)0, 1) {} // 0=Hangul, 1=start of normal data |
729e4ab9 A |
811 | virtual UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { |
812 | if(value!=0) { | |
813 | if(start!=end) { | |
814 | fprintf(stderr, | |
815 | "gennorm2 error: unexpected shared data for " | |
816 | "multiple code points U+%04lX..U+%04lX\n", | |
817 | (long)start, (long)end); | |
818 | exit(U_INTERNAL_PROGRAM_ERROR); | |
819 | } | |
820 | builder.writeExtraData(start, value, *this); | |
821 | } | |
822 | return TRUE; | |
823 | } | |
824 | UnicodeString maybeYesCompositions; | |
825 | UnicodeString yesYesCompositions; | |
4388f060 A |
826 | UnicodeString yesNoMappingsAndCompositions; |
827 | UnicodeString yesNoMappingsOnly; | |
729e4ab9 A |
828 | UnicodeString noNoMappings; |
829 | Hashtable previousNoNoMappings; // If constructed in runtime code, pass in UErrorCode. | |
830 | }; | |
831 | ||
832 | void Normalizer2DataBuilder::writeExtraData(UChar32 c, uint32_t value, ExtraDataWriter &writer) { | |
833 | Norm *p=norms+value; | |
4388f060 A |
834 | if(!p->hasMapping()) { |
835 | // Write small-FCD data. | |
836 | // There is similar code in writeMapping() for characters that do have a mapping. | |
837 | if(c<Normalizer2Impl::MIN_CCC_LCCC_CP && p->cc!=0) { | |
838 | fprintf(stderr, | |
839 | "gennorm2 error: " | |
840 | "U+%04lX below U+0300 has ccc!=0, not supported by ICU\n", | |
841 | (long)c); | |
842 | exit(U_INVALID_FORMAT_ERROR); | |
843 | } | |
844 | if(p->cc!=0) { | |
845 | UChar32 lead= c<=0xffff ? c : U16_LEAD(c); | |
846 | smallFCD[lead>>8]|=(uint8_t)1<<((lead>>5)&7); | |
847 | } | |
848 | } | |
729e4ab9 A |
849 | if(p->combinesBack) { |
850 | if(p->hasMapping()) { | |
851 | fprintf(stderr, | |
852 | "gennorm2 error: " | |
853 | "U+%04lX combines-back and decomposes, not possible in Unicode normalization\n", | |
854 | (long)c); | |
855 | exit(U_INVALID_FORMAT_ERROR); | |
856 | } | |
857 | if(p->compositions!=NULL) { | |
858 | p->offset= | |
859 | (writer.maybeYesCompositions.length()<<Norm::OFFSET_SHIFT)| | |
860 | Norm::OFFSET_MAYBE_YES; | |
861 | writeCompositions(c, p, writer.maybeYesCompositions); | |
862 | } | |
863 | } else if(!p->hasMapping()) { | |
864 | if(p->compositions!=NULL) { | |
865 | p->offset= | |
866 | (writer.yesYesCompositions.length()<<Norm::OFFSET_SHIFT)| | |
867 | Norm::OFFSET_YES_YES; | |
868 | writeCompositions(c, p, writer.yesYesCompositions); | |
869 | } | |
870 | } else if(p->mappingType==Norm::ROUND_TRIP) { | |
729e4ab9 | 871 | if(p->compositions!=NULL) { |
4388f060 A |
872 | int32_t offset=writer.yesNoMappingsAndCompositions.length()+ |
873 | writeMapping(c, p, writer.yesNoMappingsAndCompositions); | |
874 | p->offset=(offset<<Norm::OFFSET_SHIFT)|Norm::OFFSET_YES_NO_MAPPING_AND_COMPOSITION; | |
875 | writeCompositions(c, p, writer.yesNoMappingsAndCompositions); | |
876 | } else { | |
877 | int32_t offset=writer.yesNoMappingsOnly.length()+ | |
878 | writeMapping(c, p, writer.yesNoMappingsOnly); | |
879 | p->offset=(offset<<Norm::OFFSET_SHIFT)|Norm::OFFSET_YES_NO_MAPPING_ONLY; | |
729e4ab9 A |
880 | } |
881 | } else /* one-way */ { | |
882 | if(p->compositions!=NULL) { | |
883 | fprintf(stderr, | |
884 | "gennorm2 error: " | |
885 | "U+%04lX combines-forward and has a one-way mapping, " | |
886 | "not possible in Unicode normalization\n", | |
887 | (long)c); | |
888 | exit(U_INVALID_FORMAT_ERROR); | |
889 | } | |
890 | if(p->cc==0 && optimization!=OPTIMIZE_FAST) { | |
891 | // Try a compact, algorithmic encoding. | |
4388f060 A |
892 | // Only for ccc=0, because we can't store additional information |
893 | // and we do not recursively follow an algorithmic encoding for access to the ccc. | |
894 | // | |
895 | // Also, if hasNoCompBoundaryAfter is set, we can only use the algorithmic encoding | |
896 | // if the mappingCP decomposes further, to ensure that there is a place to store it. | |
897 | // We want to see that the final mapping does not have exactly 1 code point, | |
898 | // or else we would have to recursively ensure that the final mapping is stored | |
899 | // in normal extraData. | |
900 | if(p->mappingCP>=0 && (!p->hasNoCompBoundaryAfter || 1!=p->mapping->countChar32())) { | |
729e4ab9 A |
901 | int32_t delta=p->mappingCP-c; |
902 | if(-Normalizer2Impl::MAX_DELTA<=delta && delta<=Normalizer2Impl::MAX_DELTA) { | |
903 | p->offset=(delta<<Norm::OFFSET_SHIFT)|Norm::OFFSET_DELTA; | |
904 | } | |
905 | } | |
906 | } | |
907 | if(p->offset==0) { | |
908 | int32_t oldNoNoLength=writer.noNoMappings.length(); | |
4388f060 | 909 | int32_t offset=oldNoNoLength+writeMapping(c, p, writer.noNoMappings); |
729e4ab9 A |
910 | UnicodeString newMapping=writer.noNoMappings.tempSubString(oldNoNoLength); |
911 | int32_t previousOffset=writer.previousNoNoMappings.geti(newMapping); | |
912 | if(previousOffset!=0) { | |
913 | // Duplicate, remove the new units and point to the old ones. | |
914 | writer.noNoMappings.truncate(oldNoNoLength); | |
4388f060 | 915 | p->offset=((previousOffset-1)<<Norm::OFFSET_SHIFT)|Norm::OFFSET_NO_NO; |
729e4ab9 A |
916 | } else { |
917 | // Enter this new mapping into the hashtable, avoiding value 0 which is "not found". | |
918 | IcuToolErrorCode errorCode("gennorm2/writeExtraData()/Hashtable.puti()"); | |
4388f060 A |
919 | writer.previousNoNoMappings.puti(newMapping, offset+1, errorCode); |
920 | p->offset=(offset<<Norm::OFFSET_SHIFT)|Norm::OFFSET_NO_NO; | |
729e4ab9 A |
921 | } |
922 | } | |
923 | } | |
924 | } | |
925 | ||
926 | class Norm16Writer : public Normalizer2DBEnumerator { | |
927 | public: | |
928 | Norm16Writer(Normalizer2DataBuilder &b) : Normalizer2DBEnumerator(b) {} | |
929 | virtual UBool rangeHandler(UChar32 start, UChar32 end, uint32_t value) { | |
930 | builder.writeNorm16(start, end, value); | |
931 | return TRUE; | |
932 | } | |
933 | }; | |
934 | ||
935 | void Normalizer2DataBuilder::writeNorm16(UChar32 start, UChar32 end, uint32_t value) { | |
936 | if(value!=0) { | |
937 | const Norm *p=norms+value; | |
938 | int32_t offset=p->offset>>Norm::OFFSET_SHIFT; | |
939 | int32_t norm16=0; | |
940 | UBool isDecompNo=FALSE; | |
941 | UBool isCompNoMaybe=FALSE; | |
942 | switch(p->offset&Norm::OFFSET_MASK) { | |
943 | case Norm::OFFSET_NONE: | |
944 | // No mapping, no compositions list. | |
945 | if(p->combinesBack) { | |
946 | norm16=Normalizer2Impl::MIN_NORMAL_MAYBE_YES+p->cc; | |
947 | isDecompNo=(UBool)(p->cc!=0); | |
948 | isCompNoMaybe=TRUE; | |
949 | } else if(p->cc!=0) { | |
950 | norm16=Normalizer2Impl::MIN_YES_YES_WITH_CC-1+p->cc; | |
951 | isDecompNo=isCompNoMaybe=TRUE; | |
952 | } | |
953 | break; | |
954 | case Norm::OFFSET_MAYBE_YES: | |
955 | norm16=indexes[Normalizer2Impl::IX_MIN_MAYBE_YES]+offset; | |
956 | isCompNoMaybe=TRUE; | |
957 | break; | |
958 | case Norm::OFFSET_YES_YES: | |
959 | norm16=offset; | |
960 | break; | |
4388f060 | 961 | case Norm::OFFSET_YES_NO_MAPPING_AND_COMPOSITION: |
729e4ab9 A |
962 | norm16=indexes[Normalizer2Impl::IX_MIN_YES_NO]+offset; |
963 | isDecompNo=TRUE; | |
964 | break; | |
4388f060 A |
965 | case Norm::OFFSET_YES_NO_MAPPING_ONLY: |
966 | norm16=indexes[Normalizer2Impl::IX_MIN_YES_NO_MAPPINGS_ONLY]+offset; | |
967 | isDecompNo=TRUE; | |
968 | break; | |
729e4ab9 A |
969 | case Norm::OFFSET_NO_NO: |
970 | norm16=indexes[Normalizer2Impl::IX_MIN_NO_NO]+offset; | |
971 | isDecompNo=isCompNoMaybe=TRUE; | |
972 | break; | |
973 | case Norm::OFFSET_DELTA: | |
974 | norm16=getCenterNoNoDelta()+offset; | |
975 | isDecompNo=isCompNoMaybe=TRUE; | |
976 | break; | |
977 | default: // Should not occur. | |
978 | exit(U_INTERNAL_PROGRAM_ERROR); | |
979 | } | |
980 | IcuToolErrorCode errorCode("gennorm2/writeNorm16()"); | |
981 | utrie2_setRange32(norm16Trie, start, end, (uint32_t)norm16, TRUE, errorCode); | |
982 | if(isDecompNo && start<indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]) { | |
983 | indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]=start; | |
984 | } | |
985 | if(isCompNoMaybe && start<indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]) { | |
986 | indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]=start; | |
987 | } | |
988 | } | |
989 | } | |
990 | ||
991 | void Normalizer2DataBuilder::setHangulData() { | |
992 | HangulIterator hi; | |
993 | const HangulIterator::Range *range; | |
994 | // Check that none of the Hangul/Jamo code points have data. | |
995 | while((range=hi.nextRange())!=NULL) { | |
996 | for(UChar32 c=range->start; c<range->limit; ++c) { | |
997 | if(utrie2_get32(norm16Trie, c)!=0) { | |
998 | fprintf(stderr, | |
999 | "gennorm2 error: " | |
1000 | "illegal mapping/composition/ccc data for Hangul or Jamo U+%04lX\n", | |
1001 | (long)c); | |
1002 | exit(U_INVALID_FORMAT_ERROR); | |
1003 | } | |
1004 | } | |
1005 | } | |
1006 | // Set data for algorithmic runtime handling. | |
1007 | IcuToolErrorCode errorCode("gennorm2/setHangulData()"); | |
1008 | hi.reset(); | |
1009 | while((range=hi.nextRange())!=NULL) { | |
1010 | uint16_t norm16=range->norm16; | |
1011 | if(norm16==0) { | |
1012 | norm16=(uint16_t)indexes[Normalizer2Impl::IX_MIN_YES_NO]; // Hangul LV/LVT encoded as minYesNo | |
1013 | if(range->start<indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]) { | |
1014 | indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]=range->start; | |
1015 | } | |
1016 | } else { | |
1017 | if(range->start<indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]) { // Jamo V/T are maybeYes | |
1018 | indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]=range->start; | |
1019 | } | |
1020 | } | |
1021 | utrie2_setRange32(norm16Trie, range->start, range->limit-1, norm16, TRUE, errorCode); | |
1022 | errorCode.assertSuccess(); | |
1023 | } | |
1024 | } | |
1025 | ||
1026 | U_CDECL_BEGIN | |
1027 | ||
1028 | static UBool U_CALLCONV | |
1029 | enumRangeMaxValue(const void *context, UChar32 /*start*/, UChar32 /*end*/, uint32_t value) { | |
1030 | uint32_t *pMaxValue=(uint32_t *)context; | |
1031 | if(value>*pMaxValue) { | |
1032 | *pMaxValue=value; | |
1033 | } | |
1034 | return TRUE; | |
1035 | } | |
1036 | ||
1037 | U_CDECL_END | |
1038 | ||
1039 | void Normalizer2DataBuilder::processData() { | |
1040 | IcuToolErrorCode errorCode("gennorm2/processData()"); | |
1041 | norm16Trie=utrie2_open(0, 0, errorCode); | |
1042 | errorCode.assertSuccess(); | |
1043 | ||
1044 | utrie2_enum(normTrie, NULL, enumRangeHandler, CompositionBuilder(*this).ptr()); | |
1045 | ||
1046 | Decomposer decomposer(*this); | |
1047 | do { | |
1048 | decomposer.didDecompose=FALSE; | |
1049 | utrie2_enum(normTrie, NULL, enumRangeHandler, &decomposer); | |
1050 | } while(decomposer.didDecompose); | |
1051 | ||
1052 | BuilderReorderingBuffer buffer; | |
1053 | int32_t normsLength=utm_countItems(normMem); | |
1054 | for(int32_t i=1; i<normsLength; ++i) { | |
4388f060 A |
1055 | // Set the hasNoCompBoundaryAfter flag for use by the last code branch |
1056 | // in Normalizer2Impl::hasCompBoundaryAfter(). | |
1057 | // For details see the comments on hasNoCompBoundaryAfter(buffer). | |
1058 | const Norm &norm=norms[i]; | |
1059 | if(norm.hasMapping()) { | |
1060 | if(norm.compositions!=NULL) { | |
1061 | norms[i].hasNoCompBoundaryAfter=TRUE; | |
1062 | } else { | |
1063 | buffer.reset(); | |
1064 | reorder(norms+i, buffer); | |
1065 | norms[i].hasNoCompBoundaryAfter=hasNoCompBoundaryAfter(buffer); | |
1066 | } | |
729e4ab9 A |
1067 | } |
1068 | } | |
1069 | ||
1070 | indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]=0x110000; | |
1071 | indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]=0x110000; | |
1072 | ||
1073 | ExtraDataWriter extraDataWriter(*this); | |
1074 | utrie2_enum(normTrie, NULL, enumRangeHandler, &extraDataWriter); | |
1075 | ||
1076 | extraData=extraDataWriter.maybeYesCompositions; | |
1077 | extraData.append(extraDataWriter.yesYesCompositions). | |
4388f060 A |
1078 | append(extraDataWriter.yesNoMappingsAndCompositions). |
1079 | append(extraDataWriter.yesNoMappingsOnly). | |
729e4ab9 A |
1080 | append(extraDataWriter.noNoMappings); |
1081 | // Pad to even length for 4-byte alignment of following data. | |
1082 | if(extraData.length()&1) { | |
1083 | extraData.append((UChar)0); | |
1084 | } | |
1085 | ||
1086 | indexes[Normalizer2Impl::IX_MIN_YES_NO]= | |
1087 | extraDataWriter.yesYesCompositions.length(); | |
4388f060 | 1088 | indexes[Normalizer2Impl::IX_MIN_YES_NO_MAPPINGS_ONLY]= |
729e4ab9 | 1089 | indexes[Normalizer2Impl::IX_MIN_YES_NO]+ |
4388f060 A |
1090 | extraDataWriter.yesNoMappingsAndCompositions.length(); |
1091 | indexes[Normalizer2Impl::IX_MIN_NO_NO]= | |
1092 | indexes[Normalizer2Impl::IX_MIN_YES_NO_MAPPINGS_ONLY]+ | |
1093 | extraDataWriter.yesNoMappingsOnly.length(); | |
729e4ab9 A |
1094 | indexes[Normalizer2Impl::IX_LIMIT_NO_NO]= |
1095 | indexes[Normalizer2Impl::IX_MIN_NO_NO]+ | |
1096 | extraDataWriter.noNoMappings.length(); | |
1097 | indexes[Normalizer2Impl::IX_MIN_MAYBE_YES]= | |
1098 | Normalizer2Impl::MIN_NORMAL_MAYBE_YES- | |
1099 | extraDataWriter.maybeYesCompositions.length(); | |
1100 | ||
1101 | int32_t minNoNoDelta=getCenterNoNoDelta()-Normalizer2Impl::MAX_DELTA; | |
1102 | if(indexes[Normalizer2Impl::IX_LIMIT_NO_NO]>minNoNoDelta) { | |
1103 | fprintf(stderr, | |
1104 | "gennorm2 error: " | |
1105 | "data structure overflow, too much mapping composition data\n"); | |
1106 | exit(U_BUFFER_OVERFLOW_ERROR); | |
1107 | } | |
1108 | ||
1109 | utrie2_enum(normTrie, NULL, enumRangeHandler, Norm16Writer(*this).ptr()); | |
1110 | ||
1111 | setHangulData(); | |
1112 | ||
1113 | // Look for the "worst" norm16 value of any supplementary code point | |
1114 | // corresponding to a lead surrogate, and set it as that surrogate's value. | |
1115 | // Enables quick check inner loops to look at only code units. | |
1116 | // | |
1117 | // We could be more sophisticated: | |
1118 | // We could collect a bit set for whether there are values in the different | |
1119 | // norm16 ranges (yesNo, maybeYes, yesYesWithCC etc.) | |
1120 | // and select the best value that only breaks the composition and/or decomposition | |
1121 | // inner loops if necessary. | |
1122 | // However, that seems like overkill for an optimization for supplementary characters. | |
1123 | for(UChar lead=0xd800; lead<0xdc00; ++lead) { | |
1124 | uint32_t maxValue=utrie2_get32(norm16Trie, lead); | |
1125 | utrie2_enumForLeadSurrogate(norm16Trie, lead, NULL, enumRangeMaxValue, &maxValue); | |
1126 | if( maxValue>=(uint32_t)indexes[Normalizer2Impl::IX_LIMIT_NO_NO] && | |
1127 | maxValue>(uint32_t)indexes[Normalizer2Impl::IX_MIN_NO_NO] | |
1128 | ) { | |
1129 | // Set noNo ("worst" value) if it got into "less-bad" maybeYes or ccc!=0. | |
1130 | // Otherwise it might end up at something like JAMO_VT which stays in | |
1131 | // the inner decomposition quick check loop. | |
1132 | maxValue=(uint32_t)indexes[Normalizer2Impl::IX_LIMIT_NO_NO]-1; | |
1133 | } | |
1134 | utrie2_set32ForLeadSurrogateCodeUnit(norm16Trie, lead, maxValue, errorCode); | |
1135 | } | |
1136 | ||
1137 | // Adjust supplementary minimum code points to break quick check loops at their lead surrogates. | |
1138 | // For an empty data file, minCP=0x110000 turns into 0xdc00 (first trail surrogate) | |
1139 | // which is harmless. | |
1140 | // As a result, the minimum code points are always BMP code points. | |
1141 | int32_t minCP=indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]; | |
1142 | if(minCP>=0x10000) { | |
1143 | indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]=U16_LEAD(minCP); | |
1144 | } | |
1145 | minCP=indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]; | |
1146 | if(minCP>=0x10000) { | |
1147 | indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]=U16_LEAD(minCP); | |
1148 | } | |
729e4ab9 | 1149 | |
729e4ab9 | 1150 | utrie2_freeze(norm16Trie, UTRIE2_16_VALUE_BITS, errorCode); |
b331163b | 1151 | norm16TrieLength=utrie2_serialize(norm16Trie, NULL, 0, errorCode); |
729e4ab9 A |
1152 | if(errorCode.get()!=U_BUFFER_OVERFLOW_ERROR) { |
1153 | fprintf(stderr, "gennorm2 error: unable to freeze/serialize the normalization trie - %s\n", | |
1154 | errorCode.errorName()); | |
1155 | exit(errorCode.reset()); | |
1156 | } | |
1157 | errorCode.reset(); | |
729e4ab9 A |
1158 | |
1159 | int32_t offset=(int32_t)sizeof(indexes); | |
1160 | indexes[Normalizer2Impl::IX_NORM_TRIE_OFFSET]=offset; | |
1161 | offset+=norm16TrieLength; | |
1162 | indexes[Normalizer2Impl::IX_EXTRA_DATA_OFFSET]=offset; | |
4388f060 A |
1163 | offset+=extraData.length()*2; |
1164 | indexes[Normalizer2Impl::IX_SMALL_FCD_OFFSET]=offset; | |
1165 | offset+=sizeof(smallFCD); | |
1166 | int32_t totalSize=offset; | |
1167 | for(int32_t i=Normalizer2Impl::IX_RESERVED3_OFFSET; i<=Normalizer2Impl::IX_TOTAL_SIZE; ++i) { | |
729e4ab9 A |
1168 | indexes[i]=totalSize; |
1169 | } | |
1170 | ||
1171 | if(beVerbose) { | |
1172 | printf("size of normalization trie: %5ld bytes\n", (long)norm16TrieLength); | |
1173 | printf("size of 16-bit extra data: %5ld uint16_t\n", (long)extraData.length()); | |
4388f060 | 1174 | printf("size of small-FCD data: %5ld bytes\n", (long)sizeof(smallFCD)); |
729e4ab9 A |
1175 | printf("size of binary data file contents: %5ld bytes\n", (long)totalSize); |
1176 | printf("minDecompNoCodePoint: U+%04lX\n", (long)indexes[Normalizer2Impl::IX_MIN_DECOMP_NO_CP]); | |
1177 | printf("minCompNoMaybeCodePoint: U+%04lX\n", (long)indexes[Normalizer2Impl::IX_MIN_COMP_NO_MAYBE_CP]); | |
1178 | printf("minYesNo: 0x%04x\n", (int)indexes[Normalizer2Impl::IX_MIN_YES_NO]); | |
4388f060 | 1179 | printf("minYesNoMappingsOnly: 0x%04x\n", (int)indexes[Normalizer2Impl::IX_MIN_YES_NO_MAPPINGS_ONLY]); |
729e4ab9 A |
1180 | printf("minNoNo: 0x%04x\n", (int)indexes[Normalizer2Impl::IX_MIN_NO_NO]); |
1181 | printf("limitNoNo: 0x%04x\n", (int)indexes[Normalizer2Impl::IX_LIMIT_NO_NO]); | |
1182 | printf("minMaybeYes: 0x%04x\n", (int)indexes[Normalizer2Impl::IX_MIN_MAYBE_YES]); | |
1183 | } | |
1184 | ||
4388f060 A |
1185 | UVersionInfo nullVersion={ 0, 0, 0, 0 }; |
1186 | if(0==memcmp(nullVersion, unicodeVersion, 4)) { | |
1187 | u_versionFromString(unicodeVersion, U_UNICODE_VERSION); | |
1188 | } | |
729e4ab9 | 1189 | memcpy(dataInfo.dataVersion, unicodeVersion, 4); |
b331163b A |
1190 | } |
1191 | ||
1192 | void Normalizer2DataBuilder::writeBinaryFile(const char *filename) { | |
1193 | processData(); | |
1194 | ||
1195 | IcuToolErrorCode errorCode("gennorm2/writeBinaryFile()"); | |
1196 | LocalArray<uint8_t> norm16TrieBytes(new uint8_t[norm16TrieLength]); | |
1197 | utrie2_serialize(norm16Trie, norm16TrieBytes.getAlias(), norm16TrieLength, errorCode); | |
1198 | errorCode.assertSuccess(); | |
1199 | ||
729e4ab9 A |
1200 | UNewDataMemory *pData= |
1201 | udata_create(NULL, NULL, filename, &dataInfo, | |
1202 | haveCopyright ? U_COPYRIGHT_STRING : NULL, errorCode); | |
1203 | if(errorCode.isFailure()) { | |
1204 | fprintf(stderr, "gennorm2 error: unable to create the output file %s - %s\n", | |
1205 | filename, errorCode.errorName()); | |
1206 | exit(errorCode.reset()); | |
1207 | } | |
1208 | udata_writeBlock(pData, indexes, sizeof(indexes)); | |
1209 | udata_writeBlock(pData, norm16TrieBytes.getAlias(), norm16TrieLength); | |
1210 | udata_writeUString(pData, extraData.getBuffer(), extraData.length()); | |
4388f060 | 1211 | udata_writeBlock(pData, smallFCD, sizeof(smallFCD)); |
729e4ab9 A |
1212 | int32_t writtenSize=udata_finish(pData, errorCode); |
1213 | if(errorCode.isFailure()) { | |
1214 | fprintf(stderr, "gennorm2: error %s writing the output file\n", errorCode.errorName()); | |
1215 | exit(errorCode.reset()); | |
1216 | } | |
b331163b | 1217 | int32_t totalSize=indexes[Normalizer2Impl::IX_TOTAL_SIZE]; |
729e4ab9 A |
1218 | if(writtenSize!=totalSize) { |
1219 | fprintf(stderr, "gennorm2 error: written size %ld != calculated size %ld\n", | |
1220 | (long)writtenSize, (long)totalSize); | |
1221 | exit(U_INTERNAL_PROGRAM_ERROR); | |
1222 | } | |
1223 | } | |
1224 | ||
b331163b A |
1225 | void |
1226 | Normalizer2DataBuilder::writeCSourceFile(const char *filename) { | |
1227 | processData(); | |
1228 | ||
1229 | IcuToolErrorCode errorCode("gennorm2/writeCSourceFile()"); | |
1230 | const char *basename=findBasename(filename); | |
1231 | CharString path(filename, (int32_t)(basename-filename), errorCode); | |
1232 | CharString dataName(basename, errorCode); | |
1233 | const char *extension=strrchr(basename, '.'); | |
1234 | if(extension!=NULL) { | |
1235 | dataName.truncate((int32_t)(extension-basename)); | |
1236 | } | |
1237 | errorCode.assertSuccess(); | |
1238 | ||
1239 | LocalArray<uint8_t> norm16TrieBytes(new uint8_t[norm16TrieLength]); | |
1240 | utrie2_serialize(norm16Trie, norm16TrieBytes.getAlias(), norm16TrieLength, errorCode); | |
1241 | errorCode.assertSuccess(); | |
1242 | ||
1243 | FILE *f=usrc_create(path.data(), basename, "icu/source/tools/gennorm2/n2builder.cpp"); | |
1244 | if(f==NULL) { | |
1245 | fprintf(stderr, "gennorm2/writeCSourceFile() error: unable to create the output file %s\n", | |
1246 | filename); | |
1247 | exit(U_FILE_ACCESS_ERROR); | |
1248 | return; | |
1249 | } | |
1250 | char line[100]; | |
1251 | sprintf(line, "static const UVersionInfo %s_formatVersion={", dataName.data()); | |
1252 | usrc_writeArray(f, line, dataInfo.formatVersion, 8, 4, "};\n"); | |
1253 | sprintf(line, "static const UVersionInfo %s_dataVersion={", dataName.data()); | |
1254 | usrc_writeArray(f, line, dataInfo.dataVersion, 8, 4, "};\n\n"); | |
1255 | sprintf(line, "static const int32_t %s_indexes[Normalizer2Impl::IX_COUNT]={\n", | |
1256 | dataName.data()); | |
1257 | usrc_writeArray(f, | |
1258 | line, | |
1259 | indexes, 32, Normalizer2Impl::IX_COUNT, | |
1260 | "\n};\n\n"); | |
1261 | sprintf(line, "static const uint16_t %s_trieIndex[%%ld]={\n", dataName.data()); | |
1262 | usrc_writeUTrie2Arrays(f, | |
1263 | line, NULL, | |
1264 | norm16Trie, | |
1265 | "\n};\n\n"); | |
1266 | sprintf(line, "static const uint16_t %s_extraData[%%ld]={\n", dataName.data()); | |
1267 | usrc_writeArray(f, | |
1268 | line, | |
1269 | extraData.getBuffer(), 16, extraData.length(), | |
1270 | "\n};\n\n"); | |
1271 | sprintf(line, "static const uint8_t %s_smallFCD[%%ld]={\n", dataName.data()); | |
1272 | usrc_writeArray(f, | |
1273 | line, | |
1274 | smallFCD, 8, sizeof(smallFCD), | |
1275 | "\n};\n\n"); | |
1276 | /*fputs( // TODO | |
1277 | "static const UCaseProps %s_singleton={\n" | |
1278 | " NULL,\n" | |
1279 | " %s_indexes,\n" | |
1280 | " %s_extraData,\n" | |
1281 | " %s_smallFCD,\n", | |
1282 | f);*/ | |
1283 | sprintf(line, "static const UTrie2 %s_trie={\n", dataName.data()); | |
1284 | char line2[100]; | |
1285 | sprintf(line2, "%s_trieIndex", dataName.data()); | |
1286 | usrc_writeUTrie2Struct(f, | |
1287 | line, | |
1288 | norm16Trie, line2, NULL, | |
1289 | "};\n"); | |
1290 | fclose(f); | |
1291 | } | |
1292 | ||
729e4ab9 A |
1293 | U_NAMESPACE_END |
1294 | ||
1295 | #endif /* #if !UCONFIG_NO_NORMALIZATION */ | |
1296 | ||
1297 | /* | |
1298 | * Hey, Emacs, please set the following: | |
1299 | * | |
1300 | * Local Variables: | |
1301 | * indent-tabs-mode: nil | |
1302 | * End: | |
1303 | */ |