1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ******************************************************************************* |
5 | * Copyright (C) 2010-2014, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ******************************************************************************* |
8 | * utf16collationiterator.cpp |
9 | * |
10 | * created on: 2010oct27 |
11 | * created by: Markus W. Scherer |
12 | */ |
13 | |
14 | #include "unicode/utypes.h" |
15 | |
16 | #if !UCONFIG_NO_COLLATION |
17 | |
18 | #include "charstr.h" |
19 | #include "cmemory.h" |
20 | #include "collation.h" |
21 | #include "collationdata.h" |
22 | #include "collationfcd.h" |
23 | #include "collationiterator.h" |
24 | #include "normalizer2impl.h" |
25 | #include "uassert.h" |
26 | #include "utf16collationiterator.h" |
27 | |
28 | U_NAMESPACE_BEGIN |
29 | |
30 | UTF16CollationIterator::UTF16CollationIterator(const UTF16CollationIterator &other, |
31 | const UChar *newText) |
32 | : CollationIterator(other), |
33 | start(newText), |
34 | pos(newText + (other.pos - other.start)), |
35 | limit(other.limit == NULL ? NULL : newText + (other.limit - other.start)) { |
36 | } |
37 | |
38 | UTF16CollationIterator::~UTF16CollationIterator() {} |
39 | |
40 | UBool |
41 | UTF16CollationIterator::operator==(const CollationIterator &other) const { |
42 | if(!CollationIterator::operator==(other)) { return FALSE; } |
43 | const UTF16CollationIterator &o = static_cast<const UTF16CollationIterator &>(other); |
44 | // Compare the iterator state but not the text: Assume that the caller does that. |
45 | return (pos - start) == (o.pos - o.start); |
46 | } |
47 | |
48 | void |
49 | UTF16CollationIterator::resetToOffset(int32_t newOffset) { |
50 | reset(); |
51 | pos = start + newOffset; |
52 | } |
53 | |
54 | int32_t |
55 | UTF16CollationIterator::getOffset() const { |
56 | return (int32_t)(pos - start); |
57 | } |
58 | |
59 | uint32_t |
60 | UTF16CollationIterator::handleNextCE32(UChar32 &c, UErrorCode & /*errorCode*/) { |
61 | if(pos == limit) { |
62 | c = U_SENTINEL; |
63 | return Collation::FALLBACK_CE32; |
64 | } |
65 | c = *pos++; |
66 | return UTRIE2_GET32_FROM_U16_SINGLE_LEAD(trie, c); |
67 | } |
68 | |
69 | UChar |
70 | UTF16CollationIterator::handleGetTrailSurrogate() { |
71 | if(pos == limit) { return 0; } |
72 | UChar trail; |
73 | if(U16_IS_TRAIL(trail = *pos)) { ++pos; } |
74 | return trail; |
75 | } |
76 | |
77 | UBool |
78 | UTF16CollationIterator::foundNULTerminator() { |
79 | if(limit == NULL) { |
80 | limit = --pos; |
81 | return TRUE; |
82 | } else { |
83 | return FALSE; |
84 | } |
85 | } |
86 | |
87 | UChar32 |
88 | UTF16CollationIterator::nextCodePoint(UErrorCode & /*errorCode*/) { |
89 | if(pos == limit) { |
90 | return U_SENTINEL; |
91 | } |
92 | UChar32 c = *pos; |
93 | if(c == 0 && limit == NULL) { |
94 | limit = pos; |
95 | return U_SENTINEL; |
96 | } |
97 | ++pos; |
98 | UChar trail; |
99 | if(U16_IS_LEAD(c) && pos != limit && U16_IS_TRAIL(trail = *pos)) { |
100 | ++pos; |
101 | return U16_GET_SUPPLEMENTARY(c, trail); |
102 | } else { |
103 | return c; |
104 | } |
105 | } |
106 | |
107 | UChar32 |
108 | UTF16CollationIterator::previousCodePoint(UErrorCode & /*errorCode*/) { |
109 | if(pos == start) { |
110 | return U_SENTINEL; |
111 | } |
112 | UChar32 c = *--pos; |
113 | UChar lead; |
114 | if(U16_IS_TRAIL(c) && pos != start && U16_IS_LEAD(lead = *(pos - 1))) { |
115 | --pos; |
116 | return U16_GET_SUPPLEMENTARY(lead, c); |
117 | } else { |
118 | return c; |
119 | } |
120 | } |
121 | |
122 | void |
123 | UTF16CollationIterator::forwardNumCodePoints(int32_t num, UErrorCode & /*errorCode*/) { |
124 | while(num > 0 && pos != limit) { |
125 | UChar32 c = *pos; |
126 | if(c == 0 && limit == NULL) { |
127 | limit = pos; |
128 | break; |
129 | } |
130 | ++pos; |
131 | --num; |
132 | if(U16_IS_LEAD(c) && pos != limit && U16_IS_TRAIL(*pos)) { |
133 | ++pos; |
134 | } |
135 | } |
136 | } |
137 | |
138 | void |
139 | UTF16CollationIterator::backwardNumCodePoints(int32_t num, UErrorCode & /*errorCode*/) { |
140 | while(num > 0 && pos != start) { |
141 | UChar32 c = *--pos; |
142 | --num; |
143 | if(U16_IS_TRAIL(c) && pos != start && U16_IS_LEAD(*(pos-1))) { |
144 | --pos; |
145 | } |
146 | } |
147 | } |
148 | |
149 | // FCDUTF16CollationIterator ----------------------------------------------- *** |
150 | |
151 | FCDUTF16CollationIterator::FCDUTF16CollationIterator(const FCDUTF16CollationIterator &other, |
152 | const UChar *newText) |
153 | : UTF16CollationIterator(other), |
154 | rawStart(newText), |
155 | segmentStart(newText + (other.segmentStart - other.rawStart)), |
156 | segmentLimit(other.segmentLimit == NULL ? NULL : newText + (other.segmentLimit - other.rawStart)), |
157 | rawLimit(other.rawLimit == NULL ? NULL : newText + (other.rawLimit - other.rawStart)), |
158 | nfcImpl(other.nfcImpl), |
159 | normalized(other.normalized), |
160 | checkDir(other.checkDir) { |
161 | if(checkDir != 0 || other.start == other.segmentStart) { |
162 | start = newText + (other.start - other.rawStart); |
163 | pos = newText + (other.pos - other.rawStart); |
164 | limit = other.limit == NULL ? NULL : newText + (other.limit - other.rawStart); |
165 | } else { |
166 | start = normalized.getBuffer(); |
167 | pos = start + (other.pos - other.start); |
168 | limit = start + normalized.length(); |
169 | } |
170 | } |
171 | |
172 | FCDUTF16CollationIterator::~FCDUTF16CollationIterator() {} |
173 | |
174 | UBool |
175 | FCDUTF16CollationIterator::operator==(const CollationIterator &other) const { |
176 | // Skip the UTF16CollationIterator and call its parent. |
177 | if(!CollationIterator::operator==(other)) { return FALSE; } |
178 | const FCDUTF16CollationIterator &o = static_cast<const FCDUTF16CollationIterator &>(other); |
179 | // Compare the iterator state but not the text: Assume that the caller does that. |
180 | if(checkDir != o.checkDir) { return FALSE; } |
181 | if(checkDir == 0 && (start == segmentStart) != (o.start == o.segmentStart)) { return FALSE; } |
182 | if(checkDir != 0 || start == segmentStart) { |
183 | return (pos - rawStart) == (o.pos - o.rawStart); |
184 | } else { |
185 | return (segmentStart - rawStart) == (o.segmentStart - o.rawStart) && |
186 | (pos - start) == (o.pos - o.start); |
187 | } |
188 | } |
189 | |
190 | void |
191 | FCDUTF16CollationIterator::resetToOffset(int32_t newOffset) { |
192 | reset(); |
193 | start = segmentStart = pos = rawStart + newOffset; |
194 | limit = rawLimit; |
195 | checkDir = 1; |
196 | } |
197 | |
198 | int32_t |
199 | FCDUTF16CollationIterator::getOffset() const { |
200 | if(checkDir != 0 || start == segmentStart) { |
201 | return (int32_t)(pos - rawStart); |
202 | } else if(pos == start) { |
203 | return (int32_t)(segmentStart - rawStart); |
204 | } else { |
205 | return (int32_t)(segmentLimit - rawStart); |
206 | } |
207 | } |
208 | |
209 | uint32_t |
210 | FCDUTF16CollationIterator::handleNextCE32(UChar32 &c, UErrorCode &errorCode) { |
211 | for(;;) { |
212 | if(checkDir > 0) { |
213 | if(pos == limit) { |
214 | c = U_SENTINEL; |
215 | return Collation::FALLBACK_CE32; |
216 | } |
217 | c = *pos++; |
218 | if(CollationFCD::hasTccc(c)) { |
219 | if(CollationFCD::maybeTibetanCompositeVowel(c) || |
220 | (pos != limit && CollationFCD::hasLccc(*pos))) { |
221 | --pos; |
222 | if(!nextSegment(errorCode)) { |
223 | c = U_SENTINEL; |
224 | return Collation::FALLBACK_CE32; |
225 | } |
226 | c = *pos++; |
227 | } |
228 | } |
229 | break; |
230 | } else if(checkDir == 0 && pos != limit) { |
231 | c = *pos++; |
232 | break; |
233 | } else { |
234 | switchToForward(); |
235 | } |
236 | } |
237 | return UTRIE2_GET32_FROM_U16_SINGLE_LEAD(trie, c); |
238 | } |
239 | |
240 | UBool |
241 | FCDUTF16CollationIterator::foundNULTerminator() { |
242 | if(limit == NULL) { |
243 | limit = rawLimit = --pos; |
244 | return TRUE; |
245 | } else { |
246 | return FALSE; |
247 | } |
248 | } |
249 | |
250 | UChar32 |
251 | FCDUTF16CollationIterator::nextCodePoint(UErrorCode &errorCode) { |
252 | UChar32 c; |
253 | for(;;) { |
254 | if(checkDir > 0) { |
255 | if(pos == limit) { |
256 | return U_SENTINEL; |
257 | } |
258 | c = *pos++; |
259 | if(CollationFCD::hasTccc(c)) { |
260 | if(CollationFCD::maybeTibetanCompositeVowel(c) || |
261 | (pos != limit && CollationFCD::hasLccc(*pos))) { |
262 | --pos; |
263 | if(!nextSegment(errorCode)) { |
264 | return U_SENTINEL; |
265 | } |
266 | c = *pos++; |
267 | } |
268 | } else if(c == 0 && limit == NULL) { |
269 | limit = rawLimit = --pos; |
270 | return U_SENTINEL; |
271 | } |
272 | break; |
273 | } else if(checkDir == 0 && pos != limit) { |
274 | c = *pos++; |
275 | break; |
276 | } else { |
277 | switchToForward(); |
278 | } |
279 | } |
280 | UChar trail; |
281 | if(U16_IS_LEAD(c) && pos != limit && U16_IS_TRAIL(trail = *pos)) { |
282 | ++pos; |
283 | return U16_GET_SUPPLEMENTARY(c, trail); |
284 | } else { |
285 | return c; |
286 | } |
287 | } |
288 | |
289 | UChar32 |
290 | FCDUTF16CollationIterator::previousCodePoint(UErrorCode &errorCode) { |
291 | UChar32 c; |
292 | for(;;) { |
293 | if(checkDir < 0) { |
294 | if(pos == start) { |
295 | return U_SENTINEL; |
296 | } |
297 | c = *--pos; |
298 | if(CollationFCD::hasLccc(c)) { |
299 | if(CollationFCD::maybeTibetanCompositeVowel(c) || |
300 | (pos != start && CollationFCD::hasTccc(*(pos - 1)))) { |
301 | ++pos; |
302 | if(!previousSegment(errorCode)) { |
303 | return U_SENTINEL; |
304 | } |
305 | c = *--pos; |
306 | } |
307 | } |
308 | break; |
309 | } else if(checkDir == 0 && pos != start) { |
310 | c = *--pos; |
311 | break; |
312 | } else { |
313 | switchToBackward(); |
314 | } |
315 | } |
316 | UChar lead; |
317 | if(U16_IS_TRAIL(c) && pos != start && U16_IS_LEAD(lead = *(pos - 1))) { |
318 | --pos; |
319 | return U16_GET_SUPPLEMENTARY(lead, c); |
320 | } else { |
321 | return c; |
322 | } |
323 | } |
324 | |
325 | void |
326 | FCDUTF16CollationIterator::forwardNumCodePoints(int32_t num, UErrorCode &errorCode) { |
327 | // Specify the class to avoid a virtual-function indirection. |
328 | // In Java, we would declare this class final. |
329 | while(num > 0 && FCDUTF16CollationIterator::nextCodePoint(errorCode) >= 0) { |
330 | --num; |
331 | } |
332 | } |
333 | |
334 | void |
335 | FCDUTF16CollationIterator::backwardNumCodePoints(int32_t num, UErrorCode &errorCode) { |
336 | // Specify the class to avoid a virtual-function indirection. |
337 | // In Java, we would declare this class final. |
338 | while(num > 0 && FCDUTF16CollationIterator::previousCodePoint(errorCode) >= 0) { |
339 | --num; |
340 | } |
341 | } |
342 | |
343 | void |
344 | FCDUTF16CollationIterator::switchToForward() { |
345 | U_ASSERT(checkDir < 0 || (checkDir == 0 && pos == limit)); |
346 | if(checkDir < 0) { |
347 | // Turn around from backward checking. |
348 | start = segmentStart = pos; |
349 | if(pos == segmentLimit) { |
350 | limit = rawLimit; |
351 | checkDir = 1; // Check forward. |
352 | } else { // pos < segmentLimit |
353 | checkDir = 0; // Stay in FCD segment. |
354 | } |
355 | } else { |
356 | // Reached the end of the FCD segment. |
357 | if(start == segmentStart) { |
358 | // The input text segment is FCD, extend it forward. |
359 | } else { |
360 | // The input text segment needed to be normalized. |
361 | // Switch to checking forward from it. |
362 | pos = start = segmentStart = segmentLimit; |
363 | // Note: If this segment is at the end of the input text, |
364 | // then it might help to return FALSE to indicate that, so that |
365 | // we do not have to re-check and normalize when we turn around and go backwards. |
366 | // However, that would complicate the call sites for an optimization of an unusual case. |
367 | } |
368 | limit = rawLimit; |
369 | checkDir = 1; |
370 | } |
371 | } |
372 | |
373 | UBool |
374 | FCDUTF16CollationIterator::nextSegment(UErrorCode &errorCode) { |
375 | if(U_FAILURE(errorCode)) { return FALSE; } |
376 | U_ASSERT(checkDir > 0 && pos != limit); |
377 | // The input text [segmentStart..pos[ passes the FCD check. |
378 | const UChar *p = pos; |
379 | uint8_t prevCC = 0; |
380 | for(;;) { |
381 | // Fetch the next character's fcd16 value. |
382 | const UChar *q = p; |
383 | uint16_t fcd16 = nfcImpl.nextFCD16(p, rawLimit); |
384 | uint8_t leadCC = (uint8_t)(fcd16 >> 8); |
385 | if(leadCC == 0 && q != pos) { |
386 | // FCD boundary before the [q, p[ character. |
387 | limit = segmentLimit = q; |
388 | break; |
389 | } |
390 | if(leadCC != 0 && (prevCC > leadCC || CollationFCD::isFCD16OfTibetanCompositeVowel(fcd16))) { |
391 | // Fails FCD check. Find the next FCD boundary and normalize. |
392 | do { |
393 | q = p; |
394 | } while(p != rawLimit && nfcImpl.nextFCD16(p, rawLimit) > 0xff); |
395 | if(!normalize(pos, q, errorCode)) { return FALSE; } |
396 | pos = start; |
397 | break; |
398 | } |
399 | prevCC = (uint8_t)fcd16; |
400 | if(p == rawLimit || prevCC == 0) { |
401 | // FCD boundary after the last character. |
402 | limit = segmentLimit = p; |
403 | break; |
404 | } |
405 | } |
406 | U_ASSERT(pos != limit); |
407 | checkDir = 0; |
408 | return TRUE; |
409 | } |
410 | |
411 | void |
412 | FCDUTF16CollationIterator::switchToBackward() { |
413 | U_ASSERT(checkDir > 0 || (checkDir == 0 && pos == start)); |
414 | if(checkDir > 0) { |
415 | // Turn around from forward checking. |
416 | limit = segmentLimit = pos; |
417 | if(pos == segmentStart) { |
418 | start = rawStart; |
419 | checkDir = -1; // Check backward. |
420 | } else { // pos > segmentStart |
421 | checkDir = 0; // Stay in FCD segment. |
422 | } |
423 | } else { |
424 | // Reached the start of the FCD segment. |
425 | if(start == segmentStart) { |
426 | // The input text segment is FCD, extend it backward. |
427 | } else { |
428 | // The input text segment needed to be normalized. |
429 | // Switch to checking backward from it. |
430 | pos = limit = segmentLimit = segmentStart; |
431 | } |
432 | start = rawStart; |
433 | checkDir = -1; |
434 | } |
435 | } |
436 | |
437 | UBool |
438 | FCDUTF16CollationIterator::previousSegment(UErrorCode &errorCode) { |
439 | if(U_FAILURE(errorCode)) { return FALSE; } |
440 | U_ASSERT(checkDir < 0 && pos != start); |
441 | // The input text [pos..segmentLimit[ passes the FCD check. |
442 | const UChar *p = pos; |
443 | uint8_t nextCC = 0; |
444 | for(;;) { |
445 | // Fetch the previous character's fcd16 value. |
446 | const UChar *q = p; |
447 | uint16_t fcd16 = nfcImpl.previousFCD16(rawStart, p); |
448 | uint8_t trailCC = (uint8_t)fcd16; |
449 | if(trailCC == 0 && q != pos) { |
450 | // FCD boundary after the [p, q[ character. |
451 | start = segmentStart = q; |
452 | break; |
453 | } |
454 | if(trailCC != 0 && ((nextCC != 0 && trailCC > nextCC) || |
455 | CollationFCD::isFCD16OfTibetanCompositeVowel(fcd16))) { |
456 | // Fails FCD check. Find the previous FCD boundary and normalize. |
457 | do { |
458 | q = p; |
459 | } while(fcd16 > 0xff && p != rawStart && |
460 | (fcd16 = nfcImpl.previousFCD16(rawStart, p)) != 0); |
461 | if(!normalize(q, pos, errorCode)) { return FALSE; } |
462 | pos = limit; |
463 | break; |
464 | } |
465 | nextCC = (uint8_t)(fcd16 >> 8); |
466 | if(p == rawStart || nextCC == 0) { |
467 | // FCD boundary before the following character. |
468 | start = segmentStart = p; |
469 | break; |
470 | } |
471 | } |
472 | U_ASSERT(pos != start); |
473 | checkDir = 0; |
474 | return TRUE; |
475 | } |
476 | |
477 | UBool |
478 | FCDUTF16CollationIterator::normalize(const UChar *from, const UChar *to, UErrorCode &errorCode) { |
479 | // NFD without argument checking. |
480 | U_ASSERT(U_SUCCESS(errorCode)); |
481 | nfcImpl.decompose(from, to, normalized, (int32_t)(to - from), errorCode); |
482 | if(U_FAILURE(errorCode)) { return FALSE; } |
483 | // Switch collation processing into the FCD buffer |
484 | // with the result of normalizing [segmentStart, segmentLimit[. |
485 | segmentStart = from; |
486 | segmentLimit = to; |
487 | start = normalized.getBuffer(); |
488 | limit = start + normalized.length(); |
489 | return TRUE; |
490 | } |
491 | |
492 | U_NAMESPACE_END |
493 | |
494 | #endif // !UCONFIG_NO_COLLATION |
495 | |