• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2 **********************************************************************
3 *   Copyright (C) 2001-2011 IBM and others. All rights reserved.
4 **********************************************************************
5 *   Date        Name        Description
6 *  07/02/2001   synwee      Creation.
7 **********************************************************************
8 */
9 
10 #include "unicode/utypes.h"
11 
12 #if !UCONFIG_NO_COLLATION && !UCONFIG_NO_BREAK_ITERATION
13 
14 #include "unicode/usearch.h"
15 #include "unicode/ustring.h"
16 #include "unicode/uchar.h"
17 #include "normalizer2impl.h"
18 #include "ucol_imp.h"
19 #include "usrchimp.h"
20 #include "cmemory.h"
21 #include "ucln_in.h"
22 #include "uassert.h"
23 #include "ustr_imp.h"
24 
25 U_NAMESPACE_USE
26 
27 // don't use Boyer-Moore
28 // (and if we decide to turn this on again there are several new TODOs that will need to be addressed)
29 #define BOYER_MOORE 0
30 
31 #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
32 
33 // internal definition ---------------------------------------------------
34 
35 #define LAST_BYTE_MASK_          0xFF
36 #define SECOND_LAST_BYTE_SHIFT_  8
37 #define SUPPLEMENTARY_MIN_VALUE_ 0x10000
38 
39 static const uint16_t *fcdTrieIndex = NULL;
40 static UChar32 fcdHighStart = 0;
41 
42 // internal methods -------------------------------------------------
43 
44 /**
45 * Fast collation element iterator setOffset.
46 * This function does not check for bounds.
47 * @param coleiter collation element iterator
48 * @param offset to set
49 */
50 static
setColEIterOffset(UCollationElements * elems,int32_t offset)51 inline void setColEIterOffset(UCollationElements *elems,
52                       int32_t             offset)
53 {
54     collIterate *ci = &(elems->iteratordata_);
55     ci->pos         = ci->string + offset;
56     ci->CEpos       = ci->toReturn = ci->extendCEs ? ci->extendCEs : ci->CEs;
57     if (ci->flags & UCOL_ITER_INNORMBUF) {
58         ci->flags = ci->origFlags;
59     }
60     ci->fcdPosition = NULL;
61 
62     ci->offsetReturn = NULL;
63     ci->offsetStore  = ci->offsetBuffer;
64     ci->offsetRepeatCount = ci->offsetRepeatValue = 0;
65 }
66 
67 /**
68 * Getting the mask for collation strength
69 * @param strength collation strength
70 * @return collation element mask
71 */
72 static
getMask(UCollationStrength strength)73 inline uint32_t getMask(UCollationStrength strength)
74 {
75     switch (strength)
76     {
77     case UCOL_PRIMARY:
78         return UCOL_PRIMARYORDERMASK;
79     case UCOL_SECONDARY:
80         return UCOL_SECONDARYORDERMASK | UCOL_PRIMARYORDERMASK;
81     default:
82         return UCOL_TERTIARYORDERMASK | UCOL_SECONDARYORDERMASK |
83                UCOL_PRIMARYORDERMASK;
84     }
85 }
86 
87 /**
88 * This is to squeeze the 21bit ces into a 256 table
89 * @param ce collation element
90 * @return collapsed version of the collation element
91 */
92 static
hash(uint32_t ce)93 inline int hash(uint32_t ce)
94 {
95     // the old value UCOL_PRIMARYORDER(ce) % MAX_TABLE_SIZE_ does not work
96     // well with the new collation where most of the latin 1 characters
97     // are of the value xx000xxx. their hashes will most of the time be 0
98     // to be discussed on the hash algo.
99     return UCOL_PRIMARYORDER(ce) % MAX_TABLE_SIZE_;
100 }
101 
102 U_CDECL_BEGIN
103 static UBool U_CALLCONV
usearch_cleanup(void)104 usearch_cleanup(void) {
105     fcdTrieIndex = NULL;
106     return TRUE;
107 }
108 U_CDECL_END
109 
110 /**
111 * Initializing the fcd tables.
112 * Internal method, status assumed to be a success.
113 * @param status output error if any, caller to check status before calling
114 *               method, status assumed to be success when passed in.
115 */
116 static
initializeFCD(UErrorCode * status)117 inline void initializeFCD(UErrorCode *status)
118 {
119     if (fcdTrieIndex == NULL) {
120         fcdTrieIndex = unorm_getFCDTrieIndex(fcdHighStart, status);
121         ucln_i18n_registerCleanup(UCLN_I18N_USEARCH, usearch_cleanup);
122     }
123 }
124 
125 /**
126 * Gets the fcd value for a character at the argument index.
127 * This method takes into accounts of the supplementary characters.
128 * @param str UTF16 string where character for fcd retrieval resides
129 * @param offset position of the character whose fcd is to be retrieved, to be
130 *               overwritten with the next character position, taking
131 *               surrogate characters into consideration.
132 * @param strlength length of the argument string
133 * @return fcd value
134 */
135 static
getFCD(const UChar * str,int32_t * offset,int32_t strlength)136 uint16_t getFCD(const UChar   *str, int32_t *offset,
137                              int32_t  strlength)
138 {
139     const UChar *temp = str + *offset;
140     uint16_t    result = unorm_nextFCD16(fcdTrieIndex, fcdHighStart, temp, str + strlength);
141     *offset = (int32_t)(temp - str);
142     return result;
143 }
144 
145 /**
146 * Getting the modified collation elements taking into account the collation
147 * attributes
148 * @param strsrch string search data
149 * @param sourcece
150 * @return the modified collation element
151 */
152 static
getCE(const UStringSearch * strsrch,uint32_t sourcece)153 inline int32_t getCE(const UStringSearch *strsrch, uint32_t sourcece)
154 {
155     // note for tertiary we can't use the collator->tertiaryMask, that
156     // is a preprocessed mask that takes into account case options. since
157     // we are only concerned with exact matches, we don't need that.
158     sourcece &= strsrch->ceMask;
159 
160     if (strsrch->toShift) {
161         // alternate handling here, since only the 16 most significant digits
162         // is only used, we can safely do a compare without masking
163         // if the ce is a variable, we mask and get only the primary values
164         // no shifting to quartenary is required since all primary values
165         // less than variabletop will need to be masked off anyway.
166         if (strsrch->variableTop > sourcece) {
167             if (strsrch->strength >= UCOL_QUATERNARY) {
168                 sourcece &= UCOL_PRIMARYORDERMASK;
169             }
170             else {
171                 sourcece = UCOL_IGNORABLE;
172             }
173         }
174     } else if (strsrch->strength >= UCOL_QUATERNARY && sourcece == UCOL_IGNORABLE) {
175         sourcece = 0xFFFF;
176     }
177 
178     return sourcece;
179 }
180 
181 /**
182 * Allocate a memory and returns NULL if it failed.
183 * Internal method, status assumed to be a success.
184 * @param size to allocate
185 * @param status output error if any, caller to check status before calling
186 *               method, status assumed to be success when passed in.
187 * @return newly allocated array, NULL otherwise
188 */
189 static
allocateMemory(uint32_t size,UErrorCode * status)190 inline void * allocateMemory(uint32_t size, UErrorCode *status)
191 {
192     uint32_t *result = (uint32_t *)uprv_malloc(size);
193     if (result == NULL) {
194         *status = U_MEMORY_ALLOCATION_ERROR;
195     }
196     return result;
197 }
198 
199 /**
200 * Adds a uint32_t value to a destination array.
201 * Creates a new array if we run out of space. The caller will have to
202 * manually deallocate the newly allocated array.
203 * Internal method, status assumed to be success, caller has to check status
204 * before calling this method. destination not to be NULL and has at least
205 * size destinationlength.
206 * @param destination target array
207 * @param offset destination offset to add value
208 * @param destinationlength target array size, return value for the new size
209 * @param value to be added
210 * @param increments incremental size expected
211 * @param status output error if any, caller to check status before calling
212 *               method, status assumed to be success when passed in.
213 * @return new destination array, destination if there was no new allocation
214 */
215 static
addTouint32_tArray(int32_t * destination,uint32_t offset,uint32_t * destinationlength,uint32_t value,uint32_t increments,UErrorCode * status)216 inline int32_t * addTouint32_tArray(int32_t    *destination,
217                                     uint32_t    offset,
218                                     uint32_t   *destinationlength,
219                                     uint32_t    value,
220                                     uint32_t    increments,
221                                     UErrorCode *status)
222 {
223     uint32_t newlength = *destinationlength;
224     if (offset + 1 == newlength) {
225         newlength += increments;
226         int32_t *temp = (int32_t *)allocateMemory(
227                                          sizeof(int32_t) * newlength, status);
228         if (U_FAILURE(*status)) {
229             return NULL;
230         }
231         uprv_memcpy(temp, destination, sizeof(int32_t) * offset);
232         *destinationlength = newlength;
233         destination        = temp;
234     }
235     destination[offset] = value;
236     return destination;
237 }
238 
239 /**
240 * Adds a uint64_t value to a destination array.
241 * Creates a new array if we run out of space. The caller will have to
242 * manually deallocate the newly allocated array.
243 * Internal method, status assumed to be success, caller has to check status
244 * before calling this method. destination not to be NULL and has at least
245 * size destinationlength.
246 * @param destination target array
247 * @param offset destination offset to add value
248 * @param destinationlength target array size, return value for the new size
249 * @param value to be added
250 * @param increments incremental size expected
251 * @param status output error if any, caller to check status before calling
252 *               method, status assumed to be success when passed in.
253 * @return new destination array, destination if there was no new allocation
254 */
255 static
addTouint64_tArray(int64_t * destination,uint32_t offset,uint32_t * destinationlength,uint64_t value,uint32_t increments,UErrorCode * status)256 inline int64_t * addTouint64_tArray(int64_t    *destination,
257                                     uint32_t    offset,
258                                     uint32_t   *destinationlength,
259                                     uint64_t    value,
260                                     uint32_t    increments,
261                                     UErrorCode *status)
262 {
263     uint32_t newlength = *destinationlength;
264     if (offset + 1 == newlength) {
265         newlength += increments;
266         int64_t *temp = (int64_t *)allocateMemory(
267                                          sizeof(int64_t) * newlength, status);
268 
269         if (U_FAILURE(*status)) {
270             return NULL;
271         }
272 
273         uprv_memcpy(temp, destination, sizeof(int64_t) * offset);
274         *destinationlength = newlength;
275         destination        = temp;
276     }
277 
278     destination[offset] = value;
279 
280     return destination;
281 }
282 
283 /**
284 * Initializing the ce table for a pattern.
285 * Stores non-ignorable collation keys.
286 * Table size will be estimated by the size of the pattern text. Table
287 * expansion will be perform as we go along. Adding 1 to ensure that the table
288 * size definitely increases.
289 * Internal method, status assumed to be a success.
290 * @param strsrch string search data
291 * @param status output error if any, caller to check status before calling
292 *               method, status assumed to be success when passed in.
293 * @return total number of expansions
294 */
295 static
initializePatternCETable(UStringSearch * strsrch,UErrorCode * status)296 inline uint16_t initializePatternCETable(UStringSearch *strsrch,
297                                          UErrorCode    *status)
298 {
299     UPattern *pattern            = &(strsrch->pattern);
300     uint32_t  cetablesize        = INITIAL_ARRAY_SIZE_;
301     int32_t  *cetable            = pattern->CEBuffer;
302     uint32_t  patternlength      = pattern->textLength;
303     UCollationElements *coleiter = strsrch->utilIter;
304 
305     if (coleiter == NULL) {
306         coleiter = ucol_openElements(strsrch->collator, pattern->text,
307                                      patternlength, status);
308         // status will be checked in ucol_next(..) later and if it is an
309         // error UCOL_NULLORDER the result of ucol_next(..) and 0 will be
310         // returned.
311         strsrch->utilIter = coleiter;
312     }
313     else {
314         uprv_init_collIterate(strsrch->collator, pattern->text,
315                          pattern->textLength,
316                          &coleiter->iteratordata_,
317                          status);
318     }
319     if(U_FAILURE(*status)) {
320         return 0;
321     }
322 
323     if (pattern->CE != cetable && pattern->CE) {
324         uprv_free(pattern->CE);
325     }
326 
327     uint16_t  offset      = 0;
328     uint16_t  result      = 0;
329     int32_t   ce;
330 
331     while ((ce = ucol_next(coleiter, status)) != UCOL_NULLORDER &&
332            U_SUCCESS(*status)) {
333         uint32_t newce = getCE(strsrch, ce);
334         if (newce) {
335             int32_t *temp = addTouint32_tArray(cetable, offset, &cetablesize,
336                                   newce,
337                                   patternlength - ucol_getOffset(coleiter) + 1,
338                                   status);
339             if (U_FAILURE(*status)) {
340                 return 0;
341             }
342             offset ++;
343             if (cetable != temp && cetable != pattern->CEBuffer) {
344                 uprv_free(cetable);
345             }
346             cetable = temp;
347         }
348         result += (uint16_t)(ucol_getMaxExpansion(coleiter, ce) - 1);
349     }
350 
351     cetable[offset]   = 0;
352     pattern->CE       = cetable;
353     pattern->CELength = offset;
354 
355     return result;
356 }
357 
358 /**
359 * Initializing the pce table for a pattern.
360 * Stores non-ignorable collation keys.
361 * Table size will be estimated by the size of the pattern text. Table
362 * expansion will be perform as we go along. Adding 1 to ensure that the table
363 * size definitely increases.
364 * Internal method, status assumed to be a success.
365 * @param strsrch string search data
366 * @param status output error if any, caller to check status before calling
367 *               method, status assumed to be success when passed in.
368 * @return total number of expansions
369 */
370 static
initializePatternPCETable(UStringSearch * strsrch,UErrorCode * status)371 inline uint16_t initializePatternPCETable(UStringSearch *strsrch,
372                                           UErrorCode    *status)
373 {
374     UPattern *pattern            = &(strsrch->pattern);
375     uint32_t  pcetablesize       = INITIAL_ARRAY_SIZE_;
376     int64_t  *pcetable           = pattern->PCEBuffer;
377     uint32_t  patternlength      = pattern->textLength;
378     UCollationElements *coleiter = strsrch->utilIter;
379 
380     if (coleiter == NULL) {
381         coleiter = ucol_openElements(strsrch->collator, pattern->text,
382                                      patternlength, status);
383         // status will be checked in ucol_next(..) later and if it is an
384         // error UCOL_NULLORDER the result of ucol_next(..) and 0 will be
385         // returned.
386         strsrch->utilIter = coleiter;
387     } else {
388         uprv_init_collIterate(strsrch->collator, pattern->text,
389                               pattern->textLength,
390                               &coleiter->iteratordata_,
391                               status);
392     }
393     if(U_FAILURE(*status)) {
394         return 0;
395     }
396 
397     if (pattern->PCE != pcetable && pattern->PCE != NULL) {
398         uprv_free(pattern->PCE);
399     }
400 
401     uint16_t  offset = 0;
402     uint16_t  result = 0;
403     int64_t   pce;
404 
405     uprv_init_pce(coleiter);
406 
407     // ** Should processed CEs be signed or unsigned?
408     // ** (the rest of the code in this file seems to play fast-and-loose with
409     // **  whether a CE is signed or unsigned. For example, look at routine above this one.)
410     while ((pce = ucol_nextProcessed(coleiter, NULL, NULL, status)) != UCOL_PROCESSED_NULLORDER &&
411            U_SUCCESS(*status)) {
412         int64_t *temp = addTouint64_tArray(pcetable, offset, &pcetablesize,
413                               pce,
414                               patternlength - ucol_getOffset(coleiter) + 1,
415                               status);
416 
417         if (U_FAILURE(*status)) {
418             return 0;
419         }
420 
421         offset += 1;
422 
423         if (pcetable != temp && pcetable != pattern->PCEBuffer) {
424             uprv_free(pcetable);
425         }
426 
427         pcetable = temp;
428         //result += (uint16_t)(ucol_getMaxExpansion(coleiter, ce) - 1);
429     }
430 
431     pcetable[offset]   = 0;
432     pattern->PCE       = pcetable;
433     pattern->PCELength = offset;
434 
435     return result;
436 }
437 
438 /**
439 * Initializes the pattern struct.
440 * Internal method, status assumed to be success.
441 * @param strsrch UStringSearch data storage
442 * @param status output error if any, caller to check status before calling
443 *               method, status assumed to be success when passed in.
444 * @return expansionsize the total expansion size of the pattern
445 */
446 static
initializePattern(UStringSearch * strsrch,UErrorCode * status)447 inline int16_t initializePattern(UStringSearch *strsrch, UErrorCode *status)
448 {
449           UPattern   *pattern     = &(strsrch->pattern);
450     const UChar      *patterntext = pattern->text;
451           int32_t     length      = pattern->textLength;
452           int32_t index       = 0;
453 
454     // Since the strength is primary, accents are ignored in the pattern.
455     if (strsrch->strength == UCOL_PRIMARY) {
456         pattern->hasPrefixAccents = 0;
457         pattern->hasSuffixAccents = 0;
458     } else {
459         pattern->hasPrefixAccents = getFCD(patterntext, &index, length) >>
460                                                          SECOND_LAST_BYTE_SHIFT_;
461         index = length;
462         UTF_BACK_1(patterntext, 0, index);
463         pattern->hasSuffixAccents = getFCD(patterntext, &index, length) &
464                                                                  LAST_BYTE_MASK_;
465     }
466 
467     // ** HACK **
468     if (strsrch->pattern.PCE != NULL) {
469         if (strsrch->pattern.PCE != strsrch->pattern.PCEBuffer) {
470             uprv_free(strsrch->pattern.PCE);
471         }
472 
473         strsrch->pattern.PCE = NULL;
474     }
475 
476     // since intializePattern is an internal method status is a success.
477     return initializePatternCETable(strsrch, status);
478 }
479 
480 /**
481 * Initializing shift tables, with the default values.
482 * If a corresponding default value is 0, the shift table is not set.
483 * @param shift table for forwards shift
484 * @param backshift table for backwards shift
485 * @param cetable table containing pattern ce
486 * @param cesize size of the pattern ces
487 * @param expansionsize total size of the expansions
488 * @param defaultforward the default forward value
489 * @param defaultbackward the default backward value
490 */
491 static
setShiftTable(int16_t shift[],int16_t backshift[],int32_t * cetable,int32_t cesize,int16_t expansionsize,int16_t defaultforward,int16_t defaultbackward)492 inline void setShiftTable(int16_t   shift[], int16_t backshift[],
493                           int32_t  *cetable, int32_t cesize,
494                           int16_t   expansionsize,
495                           int16_t   defaultforward,
496                           int16_t   defaultbackward)
497 {
498     // estimate the value to shift. to do that we estimate the smallest
499     // number of characters to give the relevant ces, ie approximately
500     // the number of ces minus their expansion, since expansions can come
501     // from a character.
502     int32_t count;
503     for (count = 0; count < MAX_TABLE_SIZE_; count ++) {
504         shift[count] = defaultforward;
505     }
506     cesize --; // down to the last index
507     for (count = 0; count < cesize; count ++) {
508         // number of ces from right of array to the count
509         int temp = defaultforward - count - 1;
510         shift[hash(cetable[count])] = temp > 1 ? temp : 1;
511     }
512     shift[hash(cetable[cesize])] = 1;
513     // for ignorables we just shift by one. see test examples.
514     shift[hash(0)] = 1;
515 
516     for (count = 0; count < MAX_TABLE_SIZE_; count ++) {
517         backshift[count] = defaultbackward;
518     }
519     for (count = cesize; count > 0; count --) {
520         // the original value count does not seem to work
521         backshift[hash(cetable[count])] = count > expansionsize ?
522                                           (int16_t)(count - expansionsize) : 1;
523     }
524     backshift[hash(cetable[0])] = 1;
525     backshift[hash(0)] = 1;
526 }
527 
528 /**
529 * Building of the pattern collation element list and the boyer moore strsrch
530 * table.
531 * The canonical match will only be performed after the default match fails.
532 * For both cases we need to remember the size of the composed and decomposed
533 * versions of the string. Since the Boyer-Moore shift calculations shifts by
534 * a number of characters in the text and tries to match the pattern from that
535 * offset, the shift value can not be too large in case we miss some
536 * characters. To choose a right shift size, we estimate the NFC form of the
537 * and use its size as a shift guide. The NFC form should be the small
538 * possible representation of the pattern. Anyways, we'll err on the smaller
539 * shift size. Hence the calculation for minlength.
540 * Canonical match will be performed slightly differently. We'll split the
541 * pattern into 3 parts, the prefix accents (PA), the middle string bounded by
542 * the first and last base character (MS), the ending accents (EA). Matches
543 * will be done on MS first, and only when we match MS then some processing
544 * will be required for the prefix and end accents in order to determine if
545 * they match PA and EA. Hence the default shift values
546 * for the canonical match will take the size of either end's accent into
547 * consideration. Forwards search will take the end accents into consideration
548 * for the default shift values and the backwards search will take the prefix
549 * accents into consideration.
550 * If pattern has no non-ignorable ce, we return a illegal argument error.
551 * Internal method, status assumed to be success.
552 * @param strsrch UStringSearch data storage
553 * @param status  for output errors if it occurs, status is assumed to be a
554 *                success when it is passed in.
555 */
556 static
initialize(UStringSearch * strsrch,UErrorCode * status)557 inline void initialize(UStringSearch *strsrch, UErrorCode *status)
558 {
559     int16_t expandlength  = initializePattern(strsrch, status);
560     if (U_SUCCESS(*status) && strsrch->pattern.CELength > 0) {
561         UPattern *pattern = &strsrch->pattern;
562         int32_t   cesize  = pattern->CELength;
563 
564         int16_t minlength = cesize > expandlength
565                             ? (int16_t)cesize - expandlength : 1;
566         pattern->defaultShiftSize    = minlength;
567         setShiftTable(pattern->shift, pattern->backShift, pattern->CE,
568                       cesize, expandlength, minlength, minlength);
569         return;
570     }
571     strsrch->pattern.defaultShiftSize = 0;
572 }
573 
574 #if BOYER_MOORE
575 /**
576 * Check to make sure that the match length is at the end of the character by
577 * using the breakiterator.
578 * @param strsrch string search data
579 * @param start target text start offset
580 * @param end target text end offset
581 */
582 static
checkBreakBoundary(const UStringSearch * strsrch,int32_t *,int32_t * end)583 void checkBreakBoundary(const UStringSearch *strsrch, int32_t * /*start*/,
584                                int32_t *end)
585 {
586 #if !UCONFIG_NO_BREAK_ITERATION
587     UBreakIterator *breakiterator = strsrch->search->internalBreakIter;
588     if (breakiterator) {
589         int32_t matchend = *end;
590         //int32_t matchstart = *start;
591 
592         if (!ubrk_isBoundary(breakiterator, matchend)) {
593             *end = ubrk_following(breakiterator, matchend);
594         }
595 
596         /* Check the start of the matched text to make sure it doesn't have any accents
597          * before it.  This code may not be necessary and so it is commented out */
598         /*if (!ubrk_isBoundary(breakiterator, matchstart) && !ubrk_isBoundary(breakiterator, matchstart-1)) {
599             *start = ubrk_preceding(breakiterator, matchstart);
600         }*/
601     }
602 #endif
603 }
604 
605 /**
606 * Determine whether the target text in UStringSearch bounded by the offset
607 * start and end is one or more whole units of text as
608 * determined by the breakiterator in UStringSearch.
609 * @param strsrch string search data
610 * @param start target text start offset
611 * @param end target text end offset
612 */
613 static
isBreakUnit(const UStringSearch * strsrch,int32_t start,int32_t end)614 UBool isBreakUnit(const UStringSearch *strsrch, int32_t start,
615                                int32_t    end)
616 {
617 #if !UCONFIG_NO_BREAK_ITERATION
618     UBreakIterator *breakiterator = strsrch->search->breakIter;
619     //TODO: Add here.
620     if (breakiterator) {
621         int32_t startindex = ubrk_first(breakiterator);
622         int32_t endindex   = ubrk_last(breakiterator);
623 
624         // out-of-range indexes are never boundary positions
625         if (start < startindex || start > endindex ||
626             end < startindex || end > endindex) {
627             return FALSE;
628         }
629         // otherwise, we can use following() on the position before the
630         // specified one and return true of the position we get back is the
631         // one the user specified
632         UBool result = (start == startindex ||
633                 ubrk_following(breakiterator, start - 1) == start) &&
634                (end == endindex ||
635                 ubrk_following(breakiterator, end - 1) == end);
636         if (result) {
637             // iterates the individual ces
638                   UCollationElements *coleiter  = strsrch->utilIter;
639             const UChar              *text      = strsrch->search->text +
640                                                                       start;
641                   UErrorCode          status    = U_ZERO_ERROR;
642             ucol_setText(coleiter, text, end - start, &status);
643             for (int32_t count = 0; count < strsrch->pattern.CELength;
644                  count ++) {
645                 int32_t ce = getCE(strsrch, ucol_next(coleiter, &status));
646                 if (ce == UCOL_IGNORABLE) {
647                     count --;
648                     continue;
649                 }
650                 if (U_FAILURE(status) || ce != strsrch->pattern.CE[count]) {
651                     return FALSE;
652                 }
653             }
654             int32_t nextce = ucol_next(coleiter, &status);
655             while (ucol_getOffset(coleiter) == (end - start)
656                    && getCE(strsrch, nextce) == UCOL_IGNORABLE) {
657                 nextce = ucol_next(coleiter, &status);
658             }
659             if (ucol_getOffset(coleiter) == (end - start)
660                 && nextce != UCOL_NULLORDER) {
661                 // extra collation elements at the end of the match
662                 return FALSE;
663             }
664         }
665         return result;
666     }
667 #endif
668     return TRUE;
669 }
670 
671 /**
672 * Getting the next base character offset if current offset is an accent,
673 * or the current offset if the current character contains a base character.
674 * accents the following base character will be returned
675 * @param text string
676 * @param textoffset current offset
677 * @param textlength length of text string
678 * @return the next base character or the current offset
679 *         if the current character is contains a base character.
680 */
681 static
getNextBaseOffset(const UChar * text,int32_t textoffset,int32_t textlength)682 inline int32_t getNextBaseOffset(const UChar       *text,
683                                            int32_t  textoffset,
684                                            int32_t      textlength)
685 {
686     if (textoffset < textlength) {
687         int32_t temp = textoffset;
688         if (getFCD(text, &temp, textlength) >> SECOND_LAST_BYTE_SHIFT_) {
689             while (temp < textlength) {
690                 int32_t result = temp;
691                 if ((getFCD(text, &temp, textlength) >>
692                      SECOND_LAST_BYTE_SHIFT_) == 0) {
693                     return result;
694                 }
695             }
696             return textlength;
697         }
698     }
699     return textoffset;
700 }
701 
702 /**
703 * Gets the next base character offset depending on the string search pattern
704 * data
705 * @param strsrch string search data
706 * @param textoffset current offset, one offset away from the last character
707 *                   to search for.
708 * @return start index of the next base character or the current offset
709 *         if the current character is contains a base character.
710 */
711 static
getNextUStringSearchBaseOffset(UStringSearch * strsrch,int32_t textoffset)712 inline int32_t getNextUStringSearchBaseOffset(UStringSearch *strsrch,
713                                                   int32_t    textoffset)
714 {
715     int32_t textlength = strsrch->search->textLength;
716     if (strsrch->pattern.hasSuffixAccents &&
717         textoffset < textlength) {
718               int32_t  temp       = textoffset;
719         const UChar       *text       = strsrch->search->text;
720         UTF_BACK_1(text, 0, temp);
721         if (getFCD(text, &temp, textlength) & LAST_BYTE_MASK_) {
722             return getNextBaseOffset(text, textoffset, textlength);
723         }
724     }
725     return textoffset;
726 }
727 
728 /**
729 * Shifting the collation element iterator position forward to prepare for
730 * a following match. If the last character is a unsafe character, we'll only
731 * shift by 1 to capture contractions, normalization etc.
732 * Internal method, status assumed to be success.
733 * @param text strsrch string search data
734 * @param textoffset start text position to do search
735 * @param ce the text ce which failed the match.
736 * @param patternceindex index of the ce within the pattern ce buffer which
737 *        failed the match
738 * @return final offset
739 */
740 static
shiftForward(UStringSearch * strsrch,int32_t textoffset,int32_t ce,int32_t patternceindex)741 inline int32_t shiftForward(UStringSearch *strsrch,
742                                 int32_t    textoffset,
743                                 int32_t       ce,
744                                 int32_t        patternceindex)
745 {
746     UPattern *pattern = &(strsrch->pattern);
747     if (ce != UCOL_NULLORDER) {
748         int32_t shift = pattern->shift[hash(ce)];
749         // this is to adjust for characters in the middle of the
750         // substring for matching that failed.
751         int32_t adjust = pattern->CELength - patternceindex;
752         if (adjust > 1 && shift >= adjust) {
753             shift -= adjust - 1;
754         }
755         textoffset += shift;
756     }
757     else {
758         textoffset += pattern->defaultShiftSize;
759     }
760 
761     textoffset = getNextUStringSearchBaseOffset(strsrch, textoffset);
762     // check for unsafe characters
763     // * if it is the start or middle of a contraction: to be done after
764     //   a initial match is found
765     // * thai or lao base consonant character: similar to contraction
766     // * high surrogate character: similar to contraction
767     // * next character is a accent: shift to the next base character
768     return textoffset;
769 }
770 #endif // #if BOYER_MOORE
771 
772 /**
773 * sets match not found
774 * @param strsrch string search data
775 */
776 static
setMatchNotFound(UStringSearch * strsrch)777 inline void setMatchNotFound(UStringSearch *strsrch)
778 {
779     // this method resets the match result regardless of the error status.
780     strsrch->search->matchedIndex = USEARCH_DONE;
781     strsrch->search->matchedLength = 0;
782     if (strsrch->search->isForwardSearching) {
783         setColEIterOffset(strsrch->textIter, strsrch->search->textLength);
784     }
785     else {
786         setColEIterOffset(strsrch->textIter, 0);
787     }
788 }
789 
790 #if BOYER_MOORE
791 /**
792 * Gets the offset to the next safe point in text.
793 * ie. not the middle of a contraction, swappable characters or supplementary
794 * characters.
795 * @param collator collation sata
796 * @param text string to work with
797 * @param textoffset offset in string
798 * @param textlength length of text string
799 * @return offset to the next safe character
800 */
801 static
getNextSafeOffset(const UCollator * collator,const UChar * text,int32_t textoffset,int32_t textlength)802 inline int32_t getNextSafeOffset(const UCollator   *collator,
803                                      const UChar       *text,
804                                            int32_t  textoffset,
805                                            int32_t      textlength)
806 {
807     int32_t result = textoffset; // first contraction character
808     while (result != textlength && ucol_unsafeCP(text[result], collator)) {
809         result ++;
810     }
811     return result;
812 }
813 
814 /**
815 * This checks for accents in the potential match started with a .
816 * composite character.
817 * This is really painful... we have to check that composite character do not
818 * have any extra accents. We have to normalize the potential match and find
819 * the immediate decomposed character before the match.
820 * The first composite character would have been taken care of by the fcd
821 * checks in checkForwardExactMatch.
822 * This is the slow path after the fcd of the first character and
823 * the last character has been checked by checkForwardExactMatch and we
824 * determine that the potential match has extra non-ignorable preceding
825 * ces.
826 * E.g. looking for \u0301 acute in \u01FA A ring above and acute,
827 * checkExtraMatchAccent should fail since there is a middle ring in \u01FA
828 * Note here that accents checking are slow and cautioned in the API docs.
829 * Internal method, status assumed to be a success, caller should check status
830 * before calling this method
831 * @param strsrch string search data
832 * @param start index of the potential unfriendly composite character
833 * @param end index of the potential unfriendly composite character
834 * @param status output error status if any.
835 * @return TRUE if there is non-ignorable accents before at the beginning
836 *              of the match, FALSE otherwise.
837 */
838 
839 static
checkExtraMatchAccents(const UStringSearch * strsrch,int32_t start,int32_t end,UErrorCode * status)840 UBool checkExtraMatchAccents(const UStringSearch *strsrch, int32_t start,
841                                    int32_t    end,
842                                    UErrorCode    *status)
843 {
844     UBool result = FALSE;
845     if (strsrch->pattern.hasPrefixAccents) {
846               int32_t  length = end - start;
847               int32_t  offset = 0;
848         const UChar       *text   = strsrch->search->text + start;
849 
850         UTF_FWD_1(text, offset, length);
851         // we are only concerned with the first composite character
852         if (unorm_quickCheck(text, offset, UNORM_NFD, status) == UNORM_NO) {
853             int32_t safeoffset = getNextSafeOffset(strsrch->collator,
854                                                        text, 0, length);
855             if (safeoffset != length) {
856                 safeoffset ++;
857             }
858             UChar   *norm = NULL;
859             UChar    buffer[INITIAL_ARRAY_SIZE_];
860             int32_t  size = unorm_normalize(text, safeoffset, UNORM_NFD, 0,
861                                             buffer, INITIAL_ARRAY_SIZE_,
862                                             status);
863             if (U_FAILURE(*status)) {
864                 return FALSE;
865             }
866             if (size >= INITIAL_ARRAY_SIZE_) {
867                 norm = (UChar *)allocateMemory((size + 1) * sizeof(UChar),
868                                                status);
869                 // if allocation failed, status will be set to
870                 // U_MEMORY_ALLOCATION_ERROR and unorm_normalize internally
871                 // checks for it.
872                 size = unorm_normalize(text, safeoffset, UNORM_NFD, 0, norm,
873                                        size, status);
874                 if (U_FAILURE(*status) && norm != NULL) {
875                     uprv_free(norm);
876                     return FALSE;
877                 }
878             }
879             else {
880                 norm = buffer;
881             }
882 
883             UCollationElements *coleiter  = strsrch->utilIter;
884             ucol_setText(coleiter, norm, size, status);
885             uint32_t            firstce   = strsrch->pattern.CE[0];
886             UBool               ignorable = TRUE;
887             uint32_t            ce        = UCOL_IGNORABLE;
888             while (U_SUCCESS(*status) && ce != firstce && ce != (uint32_t)UCOL_NULLORDER) {
889                 offset = ucol_getOffset(coleiter);
890                 if (ce != firstce && ce != UCOL_IGNORABLE) {
891                     ignorable = FALSE;
892                 }
893                 ce = ucol_next(coleiter, status);
894             }
895             UChar32 codepoint;
896             UTF_PREV_CHAR(norm, 0, offset, codepoint);
897             result = !ignorable && (u_getCombiningClass(codepoint) != 0);
898 
899             if (norm != buffer) {
900                 uprv_free(norm);
901             }
902         }
903     }
904 
905     return result;
906 }
907 
908 /**
909 * Used by exact matches, checks if there are accents before the match.
910 * This is really painful... we have to check that composite characters at
911 * the start of the matches have to not have any extra accents.
912 * We check the FCD of the character first, if it starts with an accent and
913 * the first pattern ce does not match the first ce of the character, we bail.
914 * Otherwise we try normalizing the first composite
915 * character and find the immediate decomposed character before the match to
916 * see if it is an non-ignorable accent.
917 * Now normalizing the first composite character is enough because we ensure
918 * that when the match is passed in here with extra beginning ces, the
919 * first or last ce that match has to occur within the first character.
920 * E.g. looking for \u0301 acute in \u01FA A ring above and acute,
921 * checkExtraMatchAccent should fail since there is a middle ring in \u01FA
922 * Note here that accents checking are slow and cautioned in the API docs.
923 * @param strsrch string search data
924 * @param start offset
925 * @param end offset
926 * @return TRUE if there are accents on either side of the match,
927 *         FALSE otherwise
928 */
929 static
hasAccentsBeforeMatch(const UStringSearch * strsrch,int32_t start,int32_t end)930 UBool hasAccentsBeforeMatch(const UStringSearch *strsrch, int32_t start,
931                                   int32_t    end)
932 {
933     if (strsrch->pattern.hasPrefixAccents) {
934         UCollationElements *coleiter  = strsrch->textIter;
935         UErrorCode          status    = U_ZERO_ERROR;
936         // we have been iterating forwards previously
937         uint32_t            ignorable = TRUE;
938         int32_t             firstce   = strsrch->pattern.CE[0];
939 
940         setColEIterOffset(coleiter, start);
941         int32_t ce  = getCE(strsrch, ucol_next(coleiter, &status));
942         if (U_FAILURE(status)) {
943             return TRUE;
944         }
945         while (ce != firstce) {
946             if (ce != UCOL_IGNORABLE) {
947                 ignorable = FALSE;
948             }
949             ce = getCE(strsrch, ucol_next(coleiter, &status));
950             if (U_FAILURE(status) || ce == UCOL_NULLORDER) {
951                 return TRUE;
952             }
953         }
954         if (!ignorable && inNormBuf(coleiter)) {
955             // within normalization buffer, discontiguous handled here
956             return TRUE;
957         }
958 
959         // within text
960         int32_t temp = start;
961         // original code
962         // accent = (getFCD(strsrch->search->text, &temp,
963         //                  strsrch->search->textLength)
964         //            >> SECOND_LAST_BYTE_SHIFT_);
965         // however this code does not work well with VC7 .net in release mode.
966         // maybe the inlines for getFCD combined with shifting has bugs in
967         // VC7. anyways this is a work around.
968         UBool accent = getFCD(strsrch->search->text, &temp,
969                               strsrch->search->textLength) > 0xFF;
970         if (!accent) {
971             return checkExtraMatchAccents(strsrch, start, end, &status);
972         }
973         if (!ignorable) {
974             return TRUE;
975         }
976         if (start > 0) {
977             temp = start;
978             UTF_BACK_1(strsrch->search->text, 0, temp);
979             if (getFCD(strsrch->search->text, &temp,
980                        strsrch->search->textLength) & LAST_BYTE_MASK_) {
981                 setColEIterOffset(coleiter, start);
982                 ce = ucol_previous(coleiter, &status);
983                 if (U_FAILURE(status) ||
984                     (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE)) {
985                     return TRUE;
986                 }
987             }
988         }
989     }
990 
991     return FALSE;
992 }
993 
994 /**
995 * Used by exact matches, checks if there are accents bounding the match.
996 * Note this is the initial boundary check. If the potential match
997 * starts or ends with composite characters, the accents in those
998 * characters will be determined later.
999 * Not doing backwards iteration here, since discontiguos contraction for
1000 * backwards collation element iterator, use up too many characters.
1001 * E.g. looking for \u030A ring in \u01FA A ring above and acute,
1002 * should fail since there is a acute at the end of \u01FA
1003 * Note here that accents checking are slow and cautioned in the API docs.
1004 * @param strsrch string search data
1005 * @param start offset of match
1006 * @param end end offset of the match
1007 * @return TRUE if there are accents on either side of the match,
1008 *         FALSE otherwise
1009 */
1010 static
hasAccentsAfterMatch(const UStringSearch * strsrch,int32_t start,int32_t end)1011 UBool hasAccentsAfterMatch(const UStringSearch *strsrch, int32_t start,
1012                                  int32_t    end)
1013 {
1014     if (strsrch->pattern.hasSuffixAccents) {
1015         const UChar       *text       = strsrch->search->text;
1016               int32_t  temp       = end;
1017               int32_t      textlength = strsrch->search->textLength;
1018         UTF_BACK_1(text, 0, temp);
1019         if (getFCD(text, &temp, textlength) & LAST_BYTE_MASK_) {
1020             int32_t             firstce  = strsrch->pattern.CE[0];
1021             UCollationElements *coleiter = strsrch->textIter;
1022             UErrorCode          status   = U_ZERO_ERROR;
1023             int32_t ce;
1024             setColEIterOffset(coleiter, start);
1025             while ((ce = getCE(strsrch, ucol_next(coleiter, &status))) != firstce) {
1026                 if (U_FAILURE(status) || ce == UCOL_NULLORDER) {
1027                     return TRUE;
1028                 }
1029             }
1030             int32_t count = 1;
1031             while (count < strsrch->pattern.CELength) {
1032                 if (getCE(strsrch, ucol_next(coleiter, &status))
1033                     == UCOL_IGNORABLE) {
1034                     // Thai can give an ignorable here.
1035                     count --;
1036                 }
1037                 if (U_FAILURE(status)) {
1038                     return TRUE;
1039                 }
1040                 count ++;
1041             }
1042 
1043             ce = ucol_next(coleiter, &status);
1044             if (U_FAILURE(status)) {
1045                 return TRUE;
1046             }
1047             if (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE) {
1048                 ce = getCE(strsrch, ce);
1049             }
1050             if (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE) {
1051                 if (ucol_getOffset(coleiter) <= end) {
1052                     return TRUE;
1053                 }
1054                 if (getFCD(text, &end, textlength) >> SECOND_LAST_BYTE_SHIFT_) {
1055                     return TRUE;
1056                 }
1057             }
1058         }
1059     }
1060     return FALSE;
1061 }
1062 #endif // #if BOYER_MOORE
1063 
1064 /**
1065 * Checks if the offset runs out of the text string
1066 * @param offset
1067 * @param textlength of the text string
1068 * @return TRUE if offset is out of bounds, FALSE otherwise
1069 */
1070 static
isOutOfBounds(int32_t textlength,int32_t offset)1071 inline UBool isOutOfBounds(int32_t textlength, int32_t offset)
1072 {
1073     return offset < 0 || offset > textlength;
1074 }
1075 
1076 /**
1077 * Checks for identical match
1078 * @param strsrch string search data
1079 * @param start offset of possible match
1080 * @param end offset of possible match
1081 * @return TRUE if identical match is found
1082 */
1083 static
checkIdentical(const UStringSearch * strsrch,int32_t start,int32_t end)1084 inline UBool checkIdentical(const UStringSearch *strsrch, int32_t start,
1085                                   int32_t    end)
1086 {
1087     if (strsrch->strength != UCOL_IDENTICAL) {
1088         return TRUE;
1089     }
1090 
1091     // Note: We could use Normalizer::compare() or similar, but for short strings
1092     // which may not be in FCD it might be faster to just NFD them.
1093     UErrorCode status = U_ZERO_ERROR;
1094     UnicodeString t2, p2;
1095     strsrch->nfd->normalize(
1096         UnicodeString(FALSE, strsrch->search->text + start, end - start), t2, status);
1097     strsrch->nfd->normalize(
1098         UnicodeString(FALSE, strsrch->pattern.text, strsrch->pattern.textLength), p2, status);
1099     // return FALSE if NFD failed
1100     return U_SUCCESS(status) && t2 == p2;
1101 }
1102 
1103 #if BOYER_MOORE
1104 /**
1105 * Checks to see if the match is repeated
1106 * @param strsrch string search data
1107 * @param start new match start index
1108 * @param end new match end index
1109 * @return TRUE if the the match is repeated, FALSE otherwise
1110 */
1111 static
checkRepeatedMatch(UStringSearch * strsrch,int32_t start,int32_t end)1112 inline UBool checkRepeatedMatch(UStringSearch *strsrch,
1113                                 int32_t    start,
1114                                 int32_t    end)
1115 {
1116     int32_t lastmatchindex = strsrch->search->matchedIndex;
1117     UBool       result;
1118     if (lastmatchindex == USEARCH_DONE) {
1119         return FALSE;
1120     }
1121     if (strsrch->search->isForwardSearching) {
1122         result = start <= lastmatchindex;
1123     }
1124     else {
1125         result = start >= lastmatchindex;
1126     }
1127     if (!result && !strsrch->search->isOverlap) {
1128         if (strsrch->search->isForwardSearching) {
1129             result = start < lastmatchindex + strsrch->search->matchedLength;
1130         }
1131         else {
1132             result = end > lastmatchindex;
1133         }
1134     }
1135     return result;
1136 }
1137 
1138 /**
1139 * Gets the collation element iterator's current offset.
1140 * @param coleiter collation element iterator
1141 * @param forwards flag TRUE if we are moving in th forwards direction
1142 * @return current offset
1143 */
1144 static
getColElemIterOffset(const UCollationElements * coleiter,UBool forwards)1145 inline int32_t getColElemIterOffset(const UCollationElements *coleiter,
1146                                               UBool               forwards)
1147 {
1148     int32_t result = ucol_getOffset(coleiter);
1149     // intricacies of the the backwards collation element iterator
1150     if (FALSE && !forwards && inNormBuf(coleiter) && !isFCDPointerNull(coleiter)) {
1151         result ++;
1152     }
1153     return result;
1154 }
1155 
1156 /**
1157 * Checks match for contraction.
1158 * If the match ends with a partial contraction we fail.
1159 * If the match starts too far off (because of backwards iteration) we try to
1160 * chip off the extra characters depending on whether a breakiterator has
1161 * been used.
1162 * Internal method, error assumed to be success, caller has to check status
1163 * before calling this method.
1164 * @param strsrch string search data
1165 * @param start offset of potential match, to be modified if necessary
1166 * @param end offset of potential match, to be modified if necessary
1167 * @param status output error status if any
1168 * @return TRUE if match passes the contraction test, FALSE otherwise
1169 */
1170 
1171 static
checkNextExactContractionMatch(UStringSearch * strsrch,int32_t * start,int32_t * end,UErrorCode * status)1172 UBool checkNextExactContractionMatch(UStringSearch *strsrch,
1173                                      int32_t   *start,
1174                                      int32_t   *end, UErrorCode  *status)
1175 {
1176           UCollationElements *coleiter   = strsrch->textIter;
1177           int32_t             textlength = strsrch->search->textLength;
1178           int32_t             temp       = *start;
1179     const UCollator          *collator   = strsrch->collator;
1180     const UChar              *text       = strsrch->search->text;
1181     // This part checks if either ends of the match contains potential
1182     // contraction. If so we'll have to iterate through them
1183     // The start contraction needs to be checked since ucol_previous dumps
1184     // all characters till the first safe character into the buffer.
1185     // *start + 1 is used to test for the unsafe characters instead of *start
1186     // because ucol_prev takes all unsafe characters till the first safe
1187     // character ie *start. so by testing *start + 1, we can estimate if
1188     // excess prefix characters has been included in the potential search
1189     // results.
1190     if ((*end < textlength && ucol_unsafeCP(text[*end], collator)) ||
1191         (*start + 1 < textlength
1192          && ucol_unsafeCP(text[*start + 1], collator))) {
1193         int32_t expansion  = getExpansionPrefix(coleiter);
1194         UBool   expandflag = expansion > 0;
1195         setColEIterOffset(coleiter, *start);
1196         while (expansion > 0) {
1197             // getting rid of the redundant ce, caused by setOffset.
1198             // since backward contraction/expansion may have extra ces if we
1199             // are in the normalization buffer, hasAccentsBeforeMatch would
1200             // have taken care of it.
1201             // E.g. the character \u01FA will have an expansion of 3, but if
1202             // we are only looking for acute and ring \u030A and \u0301, we'll
1203             // have to skip the first ce in the expansion buffer.
1204             ucol_next(coleiter, status);
1205             if (U_FAILURE(*status)) {
1206                 return FALSE;
1207             }
1208             if (ucol_getOffset(coleiter) != temp) {
1209                 *start = temp;
1210                 temp  = ucol_getOffset(coleiter);
1211             }
1212             expansion --;
1213         }
1214 
1215         int32_t  *patternce       = strsrch->pattern.CE;
1216         int32_t   patterncelength = strsrch->pattern.CELength;
1217         int32_t   count           = 0;
1218         while (count < patterncelength) {
1219             int32_t ce = getCE(strsrch, ucol_next(coleiter, status));
1220             if (ce == UCOL_IGNORABLE) {
1221                 continue;
1222             }
1223             if (expandflag && count == 0 && ucol_getOffset(coleiter) != temp) {
1224                 *start = temp;
1225                 temp   = ucol_getOffset(coleiter);
1226             }
1227             if (U_FAILURE(*status) || ce != patternce[count]) {
1228                 (*end) ++;
1229                 *end = getNextUStringSearchBaseOffset(strsrch, *end);
1230                 return FALSE;
1231             }
1232             count ++;
1233         }
1234     }
1235     return TRUE;
1236 }
1237 
1238 /**
1239 * Checks and sets the match information if found.
1240 * Checks
1241 * <ul>
1242 * <li> the potential match does not repeat the previous match
1243 * <li> boundaries are correct
1244 * <li> exact matches has no extra accents
1245 * <li> identical matchesb
1246 * <li> potential match does not end in the middle of a contraction
1247 * <\ul>
1248 * Otherwise the offset will be shifted to the next character.
1249 * Internal method, status assumed to be success, caller has to check status
1250 * before calling this method.
1251 * @param strsrch string search data
1252 * @param textoffset offset in the collation element text. the returned value
1253 *        will be the truncated end offset of the match or the new start
1254 *        search offset.
1255 * @param status output error status if any
1256 * @return TRUE if the match is valid, FALSE otherwise
1257 */
1258 static
checkNextExactMatch(UStringSearch * strsrch,int32_t * textoffset,UErrorCode * status)1259 inline UBool checkNextExactMatch(UStringSearch *strsrch,
1260                                  int32_t   *textoffset, UErrorCode *status)
1261 {
1262     UCollationElements *coleiter = strsrch->textIter;
1263     int32_t         start    = getColElemIterOffset(coleiter, FALSE);
1264 
1265     if (!checkNextExactContractionMatch(strsrch, &start, textoffset, status)) {
1266         return FALSE;
1267     }
1268 
1269     // this totally matches, however we need to check if it is repeating
1270     if (!isBreakUnit(strsrch, start, *textoffset) ||
1271         checkRepeatedMatch(strsrch, start, *textoffset) ||
1272         hasAccentsBeforeMatch(strsrch, start, *textoffset) ||
1273         !checkIdentical(strsrch, start, *textoffset) ||
1274         hasAccentsAfterMatch(strsrch, start, *textoffset)) {
1275 
1276         (*textoffset) ++;
1277         *textoffset = getNextUStringSearchBaseOffset(strsrch, *textoffset);
1278         return FALSE;
1279     }
1280 
1281     //Add breakiterator boundary check for primary strength search.
1282     if (!strsrch->search->breakIter && strsrch->strength == UCOL_PRIMARY) {
1283         checkBreakBoundary(strsrch, &start, textoffset);
1284     }
1285 
1286     // totally match, we will get rid of the ending ignorables.
1287     strsrch->search->matchedIndex  = start;
1288     strsrch->search->matchedLength = *textoffset - start;
1289     return TRUE;
1290 }
1291 
1292 /**
1293 * Getting the previous base character offset, or the current offset if the
1294 * current character is a base character
1295 * @param text string
1296 * @param textoffset one offset after the current character
1297 * @return the offset of the next character after the base character or the first
1298 *         composed character with accents
1299 */
1300 static
getPreviousBaseOffset(const UChar * text,int32_t textoffset)1301 inline int32_t getPreviousBaseOffset(const UChar       *text,
1302                                                int32_t  textoffset)
1303 {
1304     if (textoffset > 0) {
1305         for (;;) {
1306             int32_t result = textoffset;
1307             UTF_BACK_1(text, 0, textoffset);
1308             int32_t temp = textoffset;
1309             uint16_t fcd = getFCD(text, &temp, result);
1310             if ((fcd >> SECOND_LAST_BYTE_SHIFT_) == 0) {
1311                 if (fcd & LAST_BYTE_MASK_) {
1312                     return textoffset;
1313                 }
1314                 return result;
1315             }
1316             if (textoffset == 0) {
1317                 return 0;
1318             }
1319         }
1320     }
1321     return textoffset;
1322 }
1323 
1324 /**
1325 * Getting the indexes of the accents that are not blocked in the argument
1326 * accent array
1327 * @param accents array of accents in nfd terminated by a 0.
1328 * @param accentsindex array of indexes of the accents that are not blocked
1329 */
1330 static
getUnblockedAccentIndex(UChar * accents,int32_t * accentsindex)1331 inline int getUnblockedAccentIndex(UChar *accents, int32_t *accentsindex)
1332 {
1333     int32_t index     = 0;
1334     int32_t     length    = u_strlen(accents);
1335     UChar32     codepoint = 0;
1336     int         cclass    = 0;
1337     int         result    = 0;
1338     int32_t temp;
1339     while (index < length) {
1340         temp = index;
1341         UTF_NEXT_CHAR(accents, index, length, codepoint);
1342         if (u_getCombiningClass(codepoint) != cclass) {
1343             cclass        = u_getCombiningClass(codepoint);
1344             accentsindex[result] = temp;
1345             result ++;
1346         }
1347     }
1348     accentsindex[result] = length;
1349     return result;
1350 }
1351 
1352 /**
1353 * Appends 3 UChar arrays to a destination array.
1354 * Creates a new array if we run out of space. The caller will have to
1355 * manually deallocate the newly allocated array.
1356 * Internal method, status assumed to be success, caller has to check status
1357 * before calling this method. destination not to be NULL and has at least
1358 * size destinationlength.
1359 * @param destination target array
1360 * @param destinationlength target array size, returning the appended length
1361 * @param source1 null-terminated first array
1362 * @param source2 second array
1363 * @param source2length length of seond array
1364 * @param source3 null-terminated third array
1365 * @param status error status if any
1366 * @return new destination array, destination if there was no new allocation
1367 */
1368 static
addToUCharArray(UChar * destination,int32_t * destinationlength,const UChar * source1,const UChar * source2,int32_t source2length,const UChar * source3,UErrorCode * status)1369 inline UChar * addToUCharArray(      UChar      *destination,
1370                                      int32_t    *destinationlength,
1371                                const UChar      *source1,
1372                                const UChar      *source2,
1373                                      int32_t     source2length,
1374                                const UChar      *source3,
1375                                      UErrorCode *status)
1376 {
1377     int32_t source1length = source1 ? u_strlen(source1) : 0;
1378     int32_t source3length = source3 ? u_strlen(source3) : 0;
1379     if (*destinationlength < source1length + source2length + source3length +
1380                                                                            1)
1381     {
1382         destination = (UChar *)allocateMemory(
1383           (source1length + source2length + source3length + 1) * sizeof(UChar),
1384           status);
1385         // if error allocating memory, status will be
1386         // U_MEMORY_ALLOCATION_ERROR
1387         if (U_FAILURE(*status)) {
1388             *destinationlength = 0;
1389             return NULL;
1390         }
1391     }
1392     if (source1length != 0) {
1393         uprv_memcpy(destination, source1, sizeof(UChar) * source1length);
1394     }
1395     if (source2length != 0) {
1396         uprv_memcpy(destination + source1length, source2,
1397                     sizeof(UChar) * source2length);
1398     }
1399     if (source3length != 0) {
1400         uprv_memcpy(destination + source1length + source2length, source3,
1401                     sizeof(UChar) * source3length);
1402     }
1403     *destinationlength = source1length + source2length + source3length;
1404     return destination;
1405 }
1406 
1407 /**
1408 * Running through a collation element iterator to see if the contents matches
1409 * pattern in string search data
1410 * @param strsrch string search data
1411 * @param coleiter collation element iterator
1412 * @return TRUE if a match if found, FALSE otherwise
1413 */
1414 static
checkCollationMatch(const UStringSearch * strsrch,UCollationElements * coleiter)1415 inline UBool checkCollationMatch(const UStringSearch      *strsrch,
1416                                        UCollationElements *coleiter)
1417 {
1418     int         patternceindex = strsrch->pattern.CELength;
1419     int32_t    *patternce      = strsrch->pattern.CE;
1420     UErrorCode  status = U_ZERO_ERROR;
1421     while (patternceindex > 0) {
1422         int32_t ce = getCE(strsrch, ucol_next(coleiter, &status));
1423         if (ce == UCOL_IGNORABLE) {
1424             continue;
1425         }
1426         if (U_FAILURE(status) || ce != *patternce) {
1427             return FALSE;
1428         }
1429         patternce ++;
1430         patternceindex --;
1431     }
1432     return TRUE;
1433 }
1434 
1435 /**
1436 * Rearranges the front accents to try matching.
1437 * Prefix accents in the text will be grouped according to their combining
1438 * class and the groups will be mixed and matched to try find the perfect
1439 * match with the pattern.
1440 * So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1441 * step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1442 *         "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1443 *         "\u0301\u0325".
1444 * step 2: check if any of the generated substrings matches the pattern.
1445 * Internal method, status is assumed to be success, caller has to check status
1446 * before calling this method.
1447 * @param strsrch string search match
1448 * @param start first offset of the accents to start searching
1449 * @param end start of the last accent set
1450 * @param status output error status if any
1451 * @return USEARCH_DONE if a match is not found, otherwise return the starting
1452 *         offset of the match. Note this start includes all preceding accents.
1453 */
1454 static
doNextCanonicalPrefixMatch(UStringSearch * strsrch,int32_t start,int32_t end,UErrorCode * status)1455 int32_t doNextCanonicalPrefixMatch(UStringSearch *strsrch,
1456                                        int32_t    start,
1457                                        int32_t    end,
1458                                        UErrorCode    *status)
1459 {
1460     const UChar       *text       = strsrch->search->text;
1461           int32_t      textlength = strsrch->search->textLength;
1462           int32_t  tempstart  = start;
1463 
1464     if ((getFCD(text, &tempstart, textlength) & LAST_BYTE_MASK_) == 0) {
1465         // die... failed at a base character
1466         return USEARCH_DONE;
1467     }
1468 
1469     int32_t offset = getNextBaseOffset(text, tempstart, textlength);
1470     start = getPreviousBaseOffset(text, tempstart);
1471 
1472     UChar       accents[INITIAL_ARRAY_SIZE_];
1473     // normalizing the offensive string
1474     unorm_normalize(text + start, offset - start, UNORM_NFD, 0, accents,
1475                     INITIAL_ARRAY_SIZE_, status);
1476     if (U_FAILURE(*status)) {
1477         return USEARCH_DONE;
1478     }
1479 
1480     int32_t         accentsindex[INITIAL_ARRAY_SIZE_];
1481     int32_t         accentsize = getUnblockedAccentIndex(accents,
1482                                                                  accentsindex);
1483     int32_t         count      = (2 << (accentsize - 1)) - 1;
1484     UChar               buffer[INITIAL_ARRAY_SIZE_];
1485     UCollationElements *coleiter   = strsrch->utilIter;
1486     while (U_SUCCESS(*status) && count > 0) {
1487         UChar *rearrange = strsrch->canonicalPrefixAccents;
1488         // copy the base characters
1489         for (int k = 0; k < accentsindex[0]; k ++) {
1490             *rearrange ++ = accents[k];
1491         }
1492         // forming all possible canonical rearrangement by dropping
1493         // sets of accents
1494         for (int i = 0; i <= accentsize - 1; i ++) {
1495             int32_t mask = 1 << (accentsize - i - 1);
1496             if (count & mask) {
1497                 for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1498                     *rearrange ++ = accents[j];
1499                 }
1500             }
1501         }
1502         *rearrange = 0;
1503         int32_t  matchsize = INITIAL_ARRAY_SIZE_;
1504         UChar   *match     = addToUCharArray(buffer, &matchsize,
1505                                            strsrch->canonicalPrefixAccents,
1506                                            strsrch->search->text + offset,
1507                                            end - offset,
1508                                            strsrch->canonicalSuffixAccents,
1509                                            status);
1510 
1511         // if status is a failure, ucol_setText does nothing.
1512         // run the collator iterator through this match
1513         ucol_setText(coleiter, match, matchsize, status);
1514         if (U_SUCCESS(*status)) {
1515             if (checkCollationMatch(strsrch, coleiter)) {
1516                 if (match != buffer) {
1517                     uprv_free(match);
1518                 }
1519                 return start;
1520             }
1521         }
1522         count --;
1523     }
1524     return USEARCH_DONE;
1525 }
1526 
1527 /**
1528 * Gets the offset to the safe point in text before textoffset.
1529 * ie. not the middle of a contraction, swappable characters or supplementary
1530 * characters.
1531 * @param collator collation sata
1532 * @param text string to work with
1533 * @param textoffset offset in string
1534 * @param textlength length of text string
1535 * @return offset to the previous safe character
1536 */
1537 static
getPreviousSafeOffset(const UCollator * collator,const UChar * text,int32_t textoffset)1538 inline uint32_t getPreviousSafeOffset(const UCollator   *collator,
1539                                       const UChar       *text,
1540                                             int32_t  textoffset)
1541 {
1542     int32_t result = textoffset; // first contraction character
1543     while (result != 0 && ucol_unsafeCP(text[result - 1], collator)) {
1544         result --;
1545     }
1546     if (result != 0) {
1547         // the first contraction character is consider unsafe here
1548         result --;
1549     }
1550     return result;
1551 }
1552 
1553 /**
1554 * Cleaning up after we passed the safe zone
1555 * @param strsrch string search data
1556 * @param safetext safe text array
1557 * @param safebuffer safe text buffer
1558 * @param coleiter collation element iterator for safe text
1559 */
1560 static
cleanUpSafeText(const UStringSearch * strsrch,UChar * safetext,UChar * safebuffer)1561 inline void cleanUpSafeText(const UStringSearch *strsrch, UChar *safetext,
1562                                   UChar         *safebuffer)
1563 {
1564     if (safetext != safebuffer && safetext != strsrch->canonicalSuffixAccents)
1565     {
1566        uprv_free(safetext);
1567     }
1568 }
1569 
1570 /**
1571 * Take the rearranged end accents and tries matching. If match failed at
1572 * a seperate preceding set of accents (seperated from the rearranged on by
1573 * at least a base character) then we rearrange the preceding accents and
1574 * tries matching again.
1575 * We allow skipping of the ends of the accent set if the ces do not match.
1576 * However if the failure is found before the accent set, it fails.
1577 * Internal method, status assumed to be success, caller has to check status
1578 * before calling this method.
1579 * @param strsrch string search data
1580 * @param textoffset of the start of the rearranged accent
1581 * @param status output error status if any
1582 * @return USEARCH_DONE if a match is not found, otherwise return the starting
1583 *         offset of the match. Note this start includes all preceding accents.
1584 */
1585 static
doNextCanonicalSuffixMatch(UStringSearch * strsrch,int32_t textoffset,UErrorCode * status)1586 int32_t doNextCanonicalSuffixMatch(UStringSearch *strsrch,
1587                                        int32_t    textoffset,
1588                                        UErrorCode    *status)
1589 {
1590     const UChar              *text           = strsrch->search->text;
1591     const UCollator          *collator       = strsrch->collator;
1592           int32_t             safelength     = 0;
1593           UChar              *safetext;
1594           int32_t             safetextlength;
1595           UChar               safebuffer[INITIAL_ARRAY_SIZE_];
1596           UCollationElements *coleiter       = strsrch->utilIter;
1597           int32_t         safeoffset     = textoffset;
1598 
1599     if (textoffset != 0 && ucol_unsafeCP(strsrch->canonicalSuffixAccents[0],
1600                                          collator)) {
1601         safeoffset     = getPreviousSafeOffset(collator, text, textoffset);
1602         safelength     = textoffset - safeoffset;
1603         safetextlength = INITIAL_ARRAY_SIZE_;
1604         safetext       = addToUCharArray(safebuffer, &safetextlength, NULL,
1605                                          text + safeoffset, safelength,
1606                                          strsrch->canonicalSuffixAccents,
1607                                          status);
1608     }
1609     else {
1610         safetextlength = u_strlen(strsrch->canonicalSuffixAccents);
1611         safetext       = strsrch->canonicalSuffixAccents;
1612     }
1613 
1614     // if status is a failure, ucol_setText does nothing
1615     ucol_setText(coleiter, safetext, safetextlength, status);
1616     // status checked in loop below
1617 
1618     int32_t  *ce        = strsrch->pattern.CE;
1619     int32_t   celength  = strsrch->pattern.CELength;
1620     int       ceindex   = celength - 1;
1621     UBool     isSafe    = TRUE; // indication flag for position in safe zone
1622 
1623     while (ceindex >= 0) {
1624         int32_t textce = ucol_previous(coleiter, status);
1625         if (U_FAILURE(*status)) {
1626             if (isSafe) {
1627                 cleanUpSafeText(strsrch, safetext, safebuffer);
1628             }
1629             return USEARCH_DONE;
1630         }
1631         if (textce == UCOL_NULLORDER) {
1632             // check if we have passed the safe buffer
1633             if (coleiter == strsrch->textIter) {
1634                 cleanUpSafeText(strsrch, safetext, safebuffer);
1635                 return USEARCH_DONE;
1636             }
1637             cleanUpSafeText(strsrch, safetext, safebuffer);
1638             safetext = safebuffer;
1639             coleiter = strsrch->textIter;
1640             setColEIterOffset(coleiter, safeoffset);
1641             // status checked at the start of the loop
1642             isSafe = FALSE;
1643             continue;
1644         }
1645         textce = getCE(strsrch, textce);
1646         if (textce != UCOL_IGNORABLE && textce != ce[ceindex]) {
1647             // do the beginning stuff
1648             int32_t failedoffset = getColElemIterOffset(coleiter, FALSE);
1649             if (isSafe && failedoffset >= safelength) {
1650                 // alas... no hope. failed at rearranged accent set
1651                 cleanUpSafeText(strsrch, safetext, safebuffer);
1652                 return USEARCH_DONE;
1653             }
1654             else {
1655                 if (isSafe) {
1656                     failedoffset += safeoffset;
1657                     cleanUpSafeText(strsrch, safetext, safebuffer);
1658                 }
1659 
1660                 // try rearranging the front accents
1661                 int32_t result = doNextCanonicalPrefixMatch(strsrch,
1662                                         failedoffset, textoffset, status);
1663                 if (result != USEARCH_DONE) {
1664                     // if status is a failure, ucol_setOffset does nothing
1665                     setColEIterOffset(strsrch->textIter, result);
1666                 }
1667                 if (U_FAILURE(*status)) {
1668                     return USEARCH_DONE;
1669                 }
1670                 return result;
1671             }
1672         }
1673         if (textce == ce[ceindex]) {
1674             ceindex --;
1675         }
1676     }
1677     // set offset here
1678     if (isSafe) {
1679         int32_t result     = getColElemIterOffset(coleiter, FALSE);
1680         // sets the text iterator here with the correct expansion and offset
1681         int32_t    leftoverces = getExpansionPrefix(coleiter);
1682         cleanUpSafeText(strsrch, safetext, safebuffer);
1683         if (result >= safelength) {
1684             result = textoffset;
1685         }
1686         else {
1687             result += safeoffset;
1688         }
1689         setColEIterOffset(strsrch->textIter, result);
1690         strsrch->textIter->iteratordata_.toReturn =
1691                        setExpansionPrefix(strsrch->textIter, leftoverces);
1692         return result;
1693     }
1694 
1695     return ucol_getOffset(coleiter);
1696 }
1697 
1698 /**
1699 * Trying out the substring and sees if it can be a canonical match.
1700 * This will try normalizing the end accents and arranging them into canonical
1701 * equivalents and check their corresponding ces with the pattern ce.
1702 * Suffix accents in the text will be grouped according to their combining
1703 * class and the groups will be mixed and matched to try find the perfect
1704 * match with the pattern.
1705 * So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1706 * step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1707 *         "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1708 *         "\u0301\u0325".
1709 * step 2: check if any of the generated substrings matches the pattern.
1710 * Internal method, status assumed to be success, caller has to check status
1711 * before calling this method.
1712 * @param strsrch string search data
1713 * @param textoffset end offset in the collation element text that ends with
1714 *                   the accents to be rearranged
1715 * @param status error status if any
1716 * @return TRUE if the match is valid, FALSE otherwise
1717 */
1718 static
doNextCanonicalMatch(UStringSearch * strsrch,int32_t textoffset,UErrorCode * status)1719 UBool doNextCanonicalMatch(UStringSearch *strsrch,
1720                            int32_t    textoffset,
1721                            UErrorCode    *status)
1722 {
1723     const UChar       *text = strsrch->search->text;
1724           int32_t  temp = textoffset;
1725     UTF_BACK_1(text, 0, temp);
1726     if ((getFCD(text, &temp, textoffset) & LAST_BYTE_MASK_) == 0) {
1727         UCollationElements *coleiter = strsrch->textIter;
1728         int32_t         offset   = getColElemIterOffset(coleiter, FALSE);
1729         if (strsrch->pattern.hasPrefixAccents) {
1730             offset = doNextCanonicalPrefixMatch(strsrch, offset, textoffset,
1731                                                 status);
1732             if (U_SUCCESS(*status) && offset != USEARCH_DONE) {
1733                 setColEIterOffset(coleiter, offset);
1734                 return TRUE;
1735             }
1736         }
1737         return FALSE;
1738     }
1739 
1740     if (!strsrch->pattern.hasSuffixAccents) {
1741         return FALSE;
1742     }
1743 
1744     UChar       accents[INITIAL_ARRAY_SIZE_];
1745     // offset to the last base character in substring to search
1746     int32_t baseoffset = getPreviousBaseOffset(text, textoffset);
1747     // normalizing the offensive string
1748     unorm_normalize(text + baseoffset, textoffset - baseoffset, UNORM_NFD,
1749                                0, accents, INITIAL_ARRAY_SIZE_, status);
1750     // status checked in loop below
1751 
1752     int32_t accentsindex[INITIAL_ARRAY_SIZE_];
1753     int32_t size = getUnblockedAccentIndex(accents, accentsindex);
1754 
1755     // 2 power n - 1 plus the full set of accents
1756     int32_t  count = (2 << (size - 1)) - 1;
1757     while (U_SUCCESS(*status) && count > 0) {
1758         UChar *rearrange = strsrch->canonicalSuffixAccents;
1759         // copy the base characters
1760         for (int k = 0; k < accentsindex[0]; k ++) {
1761             *rearrange ++ = accents[k];
1762         }
1763         // forming all possible canonical rearrangement by dropping
1764         // sets of accents
1765         for (int i = 0; i <= size - 1; i ++) {
1766             int32_t mask = 1 << (size - i - 1);
1767             if (count & mask) {
1768                 for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1769                     *rearrange ++ = accents[j];
1770                 }
1771             }
1772         }
1773         *rearrange = 0;
1774         int32_t offset = doNextCanonicalSuffixMatch(strsrch, baseoffset,
1775                                                         status);
1776         if (offset != USEARCH_DONE) {
1777             return TRUE; // match found
1778         }
1779         count --;
1780     }
1781     return FALSE;
1782 }
1783 
1784 /**
1785 * Gets the previous base character offset depending on the string search
1786 * pattern data
1787 * @param strsrch string search data
1788 * @param textoffset current offset, current character
1789 * @return the offset of the next character after this base character or itself
1790 *         if it is a composed character with accents
1791 */
1792 static
getPreviousUStringSearchBaseOffset(UStringSearch * strsrch,int32_t textoffset)1793 inline int32_t getPreviousUStringSearchBaseOffset(UStringSearch *strsrch,
1794                                                       int32_t textoffset)
1795 {
1796     if (strsrch->pattern.hasPrefixAccents && textoffset > 0) {
1797         const UChar       *text = strsrch->search->text;
1798               int32_t  offset = textoffset;
1799         if (getFCD(text, &offset, strsrch->search->textLength) >>
1800                                                    SECOND_LAST_BYTE_SHIFT_) {
1801             return getPreviousBaseOffset(text, textoffset);
1802         }
1803     }
1804     return textoffset;
1805 }
1806 
1807 /**
1808 * Checks match for contraction.
1809 * If the match ends with a partial contraction we fail.
1810 * If the match starts too far off (because of backwards iteration) we try to
1811 * chip off the extra characters
1812 * Internal method, status assumed to be success, caller has to check status
1813 * before calling this method.
1814 * @param strsrch string search data
1815 * @param start offset of potential match, to be modified if necessary
1816 * @param end offset of potential match, to be modified if necessary
1817 * @param status output error status if any
1818 * @return TRUE if match passes the contraction test, FALSE otherwise
1819 */
1820 static
checkNextCanonicalContractionMatch(UStringSearch * strsrch,int32_t * start,int32_t * end,UErrorCode * status)1821 UBool checkNextCanonicalContractionMatch(UStringSearch *strsrch,
1822                                          int32_t   *start,
1823                                          int32_t   *end,
1824                                          UErrorCode    *status)
1825 {
1826           UCollationElements *coleiter   = strsrch->textIter;
1827           int32_t             textlength = strsrch->search->textLength;
1828           int32_t         temp       = *start;
1829     const UCollator          *collator   = strsrch->collator;
1830     const UChar              *text       = strsrch->search->text;
1831     // This part checks if either ends of the match contains potential
1832     // contraction. If so we'll have to iterate through them
1833     if ((*end < textlength && ucol_unsafeCP(text[*end], collator)) ||
1834         (*start + 1 < textlength
1835          && ucol_unsafeCP(text[*start + 1], collator))) {
1836         int32_t expansion  = getExpansionPrefix(coleiter);
1837         UBool   expandflag = expansion > 0;
1838         setColEIterOffset(coleiter, *start);
1839         while (expansion > 0) {
1840             // getting rid of the redundant ce, caused by setOffset.
1841             // since backward contraction/expansion may have extra ces if we
1842             // are in the normalization buffer, hasAccentsBeforeMatch would
1843             // have taken care of it.
1844             // E.g. the character \u01FA will have an expansion of 3, but if
1845             // we are only looking for acute and ring \u030A and \u0301, we'll
1846             // have to skip the first ce in the expansion buffer.
1847             ucol_next(coleiter, status);
1848             if (U_FAILURE(*status)) {
1849                 return FALSE;
1850             }
1851             if (ucol_getOffset(coleiter) != temp) {
1852                 *start = temp;
1853                 temp  = ucol_getOffset(coleiter);
1854             }
1855             expansion --;
1856         }
1857 
1858         int32_t  *patternce       = strsrch->pattern.CE;
1859         int32_t   patterncelength = strsrch->pattern.CELength;
1860         int32_t   count           = 0;
1861         int32_t   textlength      = strsrch->search->textLength;
1862         while (count < patterncelength) {
1863             int32_t ce = getCE(strsrch, ucol_next(coleiter, status));
1864             // status checked below, note that if status is a failure
1865             // ucol_next returns UCOL_NULLORDER
1866             if (ce == UCOL_IGNORABLE) {
1867                 continue;
1868             }
1869             if (expandflag && count == 0 && ucol_getOffset(coleiter) != temp) {
1870                 *start = temp;
1871                 temp   = ucol_getOffset(coleiter);
1872             }
1873 
1874             if (count == 0 && ce != patternce[0]) {
1875                 // accents may have extra starting ces, this occurs when a
1876                 // pure accent pattern is matched without rearrangement
1877                 // text \u0325\u0300 and looking for \u0300
1878                 int32_t expected = patternce[0];
1879                 if (getFCD(text, start, textlength) & LAST_BYTE_MASK_) {
1880                     ce = getCE(strsrch, ucol_next(coleiter, status));
1881                     while (U_SUCCESS(*status) && ce != expected &&
1882                            ce != UCOL_NULLORDER &&
1883                            ucol_getOffset(coleiter) <= *end) {
1884                         ce = getCE(strsrch, ucol_next(coleiter, status));
1885                     }
1886                 }
1887             }
1888             if (U_FAILURE(*status) || ce != patternce[count]) {
1889                 (*end) ++;
1890                 *end = getNextUStringSearchBaseOffset(strsrch, *end);
1891                 return FALSE;
1892             }
1893             count ++;
1894         }
1895     }
1896     return TRUE;
1897 }
1898 
1899 /**
1900 * Checks and sets the match information if found.
1901 * Checks
1902 * <ul>
1903 * <li> the potential match does not repeat the previous match
1904 * <li> boundaries are correct
1905 * <li> potential match does not end in the middle of a contraction
1906 * <li> identical matches
1907 * <\ul>
1908 * Otherwise the offset will be shifted to the next character.
1909 * Internal method, status assumed to be success, caller has to check the
1910 * status before calling this method.
1911 * @param strsrch string search data
1912 * @param textoffset offset in the collation element text. the returned value
1913 *        will be the truncated end offset of the match or the new start
1914 *        search offset.
1915 * @param status output error status if any
1916 * @return TRUE if the match is valid, FALSE otherwise
1917 */
1918 static
checkNextCanonicalMatch(UStringSearch * strsrch,int32_t * textoffset,UErrorCode * status)1919 inline UBool checkNextCanonicalMatch(UStringSearch *strsrch,
1920                                      int32_t   *textoffset,
1921                                      UErrorCode    *status)
1922 {
1923     // to ensure that the start and ends are not composite characters
1924     UCollationElements *coleiter = strsrch->textIter;
1925     // if we have a canonical accent match
1926     if ((strsrch->pattern.hasSuffixAccents &&
1927         strsrch->canonicalSuffixAccents[0]) ||
1928         (strsrch->pattern.hasPrefixAccents &&
1929         strsrch->canonicalPrefixAccents[0])) {
1930         strsrch->search->matchedIndex  = getPreviousUStringSearchBaseOffset(
1931                                                     strsrch,
1932                                                     ucol_getOffset(coleiter));
1933         strsrch->search->matchedLength = *textoffset -
1934                                                 strsrch->search->matchedIndex;
1935         return TRUE;
1936     }
1937 
1938     int32_t start = getColElemIterOffset(coleiter, FALSE);
1939     if (!checkNextCanonicalContractionMatch(strsrch, &start, textoffset,
1940                                             status) || U_FAILURE(*status)) {
1941         return FALSE;
1942     }
1943 
1944     start = getPreviousUStringSearchBaseOffset(strsrch, start);
1945     // this totally matches, however we need to check if it is repeating
1946     if (checkRepeatedMatch(strsrch, start, *textoffset) ||
1947         !isBreakUnit(strsrch, start, *textoffset) ||
1948         !checkIdentical(strsrch, start, *textoffset)) {
1949         (*textoffset) ++;
1950         *textoffset = getNextBaseOffset(strsrch->search->text, *textoffset,
1951                                         strsrch->search->textLength);
1952         return FALSE;
1953     }
1954 
1955     strsrch->search->matchedIndex  = start;
1956     strsrch->search->matchedLength = *textoffset - start;
1957     return TRUE;
1958 }
1959 
1960 /**
1961 * Shifting the collation element iterator position forward to prepare for
1962 * a preceding match. If the first character is a unsafe character, we'll only
1963 * shift by 1 to capture contractions, normalization etc.
1964 * Internal method, status assumed to be success, caller has to check status
1965 * before calling this method.
1966 * @param text strsrch string search data
1967 * @param textoffset start text position to do search
1968 * @param ce the text ce which failed the match.
1969 * @param patternceindex index of the ce within the pattern ce buffer which
1970 *        failed the match
1971 * @return final offset
1972 */
1973 static
reverseShift(UStringSearch * strsrch,int32_t textoffset,int32_t ce,int32_t patternceindex)1974 inline int32_t reverseShift(UStringSearch *strsrch,
1975                                 int32_t    textoffset,
1976                                 int32_t       ce,
1977                                 int32_t        patternceindex)
1978 {
1979     if (strsrch->search->isOverlap) {
1980         if (textoffset != strsrch->search->textLength) {
1981             textoffset --;
1982         }
1983         else {
1984             textoffset -= strsrch->pattern.defaultShiftSize;
1985         }
1986     }
1987     else {
1988         if (ce != UCOL_NULLORDER) {
1989             int32_t shift = strsrch->pattern.backShift[hash(ce)];
1990 
1991             // this is to adjust for characters in the middle of the substring
1992             // for matching that failed.
1993             int32_t adjust = patternceindex;
1994             if (adjust > 1 && shift > adjust) {
1995                 shift -= adjust - 1;
1996             }
1997             textoffset -= shift;
1998         }
1999         else {
2000             textoffset -= strsrch->pattern.defaultShiftSize;
2001         }
2002     }
2003     textoffset = getPreviousUStringSearchBaseOffset(strsrch, textoffset);
2004     return textoffset;
2005 }
2006 
2007 /**
2008 * Checks match for contraction.
2009 * If the match starts with a partial contraction we fail.
2010 * Internal method, status assumed to be success, caller has to check status
2011 * before calling this method.
2012 * @param strsrch string search data
2013 * @param start offset of potential match, to be modified if necessary
2014 * @param end offset of potential match, to be modified if necessary
2015 * @param status output error status if any
2016 * @return TRUE if match passes the contraction test, FALSE otherwise
2017 */
2018 static
checkPreviousExactContractionMatch(UStringSearch * strsrch,int32_t * start,int32_t * end,UErrorCode * status)2019 UBool checkPreviousExactContractionMatch(UStringSearch *strsrch,
2020                                      int32_t   *start,
2021                                      int32_t   *end, UErrorCode  *status)
2022 {
2023           UCollationElements *coleiter   = strsrch->textIter;
2024           int32_t             textlength = strsrch->search->textLength;
2025           int32_t             temp       = *end;
2026     const UCollator          *collator   = strsrch->collator;
2027     const UChar              *text       = strsrch->search->text;
2028     // This part checks if either if the start of the match contains potential
2029     // contraction. If so we'll have to iterate through them
2030     // Since we used ucol_next while previously looking for the potential
2031     // match, this guarantees that our end will not be a partial contraction,
2032     // or a partial supplementary character.
2033     if (*start < textlength && ucol_unsafeCP(text[*start], collator)) {
2034         int32_t expansion  = getExpansionSuffix(coleiter);
2035         UBool   expandflag = expansion > 0;
2036         setColEIterOffset(coleiter, *end);
2037         while (U_SUCCESS(*status) && expansion > 0) {
2038             // getting rid of the redundant ce
2039             // since forward contraction/expansion may have extra ces
2040             // if we are in the normalization buffer, hasAccentsBeforeMatch
2041             // would have taken care of it.
2042             // E.g. the character \u01FA will have an expansion of 3, but if
2043             // we are only looking for A ring A\u030A, we'll have to skip the
2044             // last ce in the expansion buffer
2045             ucol_previous(coleiter, status);
2046             if (U_FAILURE(*status)) {
2047                 return FALSE;
2048             }
2049             if (ucol_getOffset(coleiter) != temp) {
2050                 *end = temp;
2051                 temp  = ucol_getOffset(coleiter);
2052             }
2053             expansion --;
2054         }
2055 
2056         int32_t  *patternce       = strsrch->pattern.CE;
2057         int32_t   patterncelength = strsrch->pattern.CELength;
2058         int32_t   count           = patterncelength;
2059         while (count > 0) {
2060             int32_t ce = getCE(strsrch, ucol_previous(coleiter, status));
2061             // status checked below, note that if status is a failure
2062             // ucol_previous returns UCOL_NULLORDER
2063             if (ce == UCOL_IGNORABLE) {
2064                 continue;
2065             }
2066             if (expandflag && count == 0 &&
2067                 getColElemIterOffset(coleiter, FALSE) != temp) {
2068                 *end = temp;
2069                 temp  = ucol_getOffset(coleiter);
2070             }
2071             if (U_FAILURE(*status) || ce != patternce[count - 1]) {
2072                 (*start) --;
2073                 *start = getPreviousBaseOffset(text, *start);
2074                 return FALSE;
2075             }
2076             count --;
2077         }
2078     }
2079     return TRUE;
2080 }
2081 
2082 /**
2083 * Checks and sets the match information if found.
2084 * Checks
2085 * <ul>
2086 * <li> the current match does not repeat the last match
2087 * <li> boundaries are correct
2088 * <li> exact matches has no extra accents
2089 * <li> identical matches
2090 * <\ul>
2091 * Otherwise the offset will be shifted to the preceding character.
2092 * Internal method, status assumed to be success, caller has to check status
2093 * before calling this method.
2094 * @param strsrch string search data
2095 * @param collator
2096 * @param coleiter collation element iterator
2097 * @param text string
2098 * @param textoffset offset in the collation element text. the returned value
2099 *        will be the truncated start offset of the match or the new start
2100 *        search offset.
2101 * @param status output error status if any
2102 * @return TRUE if the match is valid, FALSE otherwise
2103 */
2104 static
checkPreviousExactMatch(UStringSearch * strsrch,int32_t * textoffset,UErrorCode * status)2105 inline UBool checkPreviousExactMatch(UStringSearch *strsrch,
2106                                      int32_t   *textoffset,
2107                                      UErrorCode    *status)
2108 {
2109     // to ensure that the start and ends are not composite characters
2110     int32_t end = ucol_getOffset(strsrch->textIter);
2111     if (!checkPreviousExactContractionMatch(strsrch, textoffset, &end, status)
2112         || U_FAILURE(*status)) {
2113             return FALSE;
2114     }
2115 
2116     // this totally matches, however we need to check if it is repeating
2117     // the old match
2118     if (checkRepeatedMatch(strsrch, *textoffset, end) ||
2119         !isBreakUnit(strsrch, *textoffset, end) ||
2120         hasAccentsBeforeMatch(strsrch, *textoffset, end) ||
2121         !checkIdentical(strsrch, *textoffset, end) ||
2122         hasAccentsAfterMatch(strsrch, *textoffset, end)) {
2123         (*textoffset) --;
2124         *textoffset = getPreviousBaseOffset(strsrch->search->text,
2125                                             *textoffset);
2126         return FALSE;
2127     }
2128 
2129     //Add breakiterator boundary check for primary strength search.
2130     if (!strsrch->search->breakIter && strsrch->strength == UCOL_PRIMARY) {
2131         checkBreakBoundary(strsrch, textoffset, &end);
2132     }
2133 
2134     strsrch->search->matchedIndex = *textoffset;
2135     strsrch->search->matchedLength = end - *textoffset;
2136     return TRUE;
2137 }
2138 
2139 /**
2140 * Rearranges the end accents to try matching.
2141 * Suffix accents in the text will be grouped according to their combining
2142 * class and the groups will be mixed and matched to try find the perfect
2143 * match with the pattern.
2144 * So for instance looking for "\u0301" in "\u030A\u0301\u0325"
2145 * step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
2146 *         "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
2147 *         "\u0301\u0325".
2148 * step 2: check if any of the generated substrings matches the pattern.
2149 * Internal method, status assumed to be success, user has to check status
2150 * before calling this method.
2151 * @param strsrch string search match
2152 * @param start offset of the first base character
2153 * @param end start of the last accent set
2154 * @param status only error status if any
2155 * @return USEARCH_DONE if a match is not found, otherwise return the ending
2156 *         offset of the match. Note this start includes all following accents.
2157 */
2158 static
doPreviousCanonicalSuffixMatch(UStringSearch * strsrch,int32_t start,int32_t end,UErrorCode * status)2159 int32_t doPreviousCanonicalSuffixMatch(UStringSearch *strsrch,
2160                                            int32_t    start,
2161                                            int32_t    end,
2162                                            UErrorCode    *status)
2163 {
2164     const UChar       *text       = strsrch->search->text;
2165           int32_t  tempend    = end;
2166 
2167     UTF_BACK_1(text, 0, tempend);
2168     if (!(getFCD(text, &tempend, strsrch->search->textLength) &
2169                                                            LAST_BYTE_MASK_)) {
2170         // die... failed at a base character
2171         return USEARCH_DONE;
2172     }
2173     end = getNextBaseOffset(text, end, strsrch->search->textLength);
2174 
2175     if (U_SUCCESS(*status)) {
2176         UChar       accents[INITIAL_ARRAY_SIZE_];
2177         int32_t offset = getPreviousBaseOffset(text, end);
2178         // normalizing the offensive string
2179         unorm_normalize(text + offset, end - offset, UNORM_NFD, 0, accents,
2180                         INITIAL_ARRAY_SIZE_, status);
2181 
2182         int32_t         accentsindex[INITIAL_ARRAY_SIZE_];
2183         int32_t         accentsize = getUnblockedAccentIndex(accents,
2184                                                          accentsindex);
2185         int32_t         count      = (2 << (accentsize - 1)) - 1;
2186         UChar               buffer[INITIAL_ARRAY_SIZE_];
2187         UCollationElements *coleiter = strsrch->utilIter;
2188         while (U_SUCCESS(*status) && count > 0) {
2189             UChar *rearrange = strsrch->canonicalSuffixAccents;
2190             // copy the base characters
2191             for (int k = 0; k < accentsindex[0]; k ++) {
2192                 *rearrange ++ = accents[k];
2193             }
2194             // forming all possible canonical rearrangement by dropping
2195             // sets of accents
2196             for (int i = 0; i <= accentsize - 1; i ++) {
2197                 int32_t mask = 1 << (accentsize - i - 1);
2198                 if (count & mask) {
2199                     for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
2200                         *rearrange ++ = accents[j];
2201                     }
2202                 }
2203             }
2204             *rearrange = 0;
2205             int32_t  matchsize = INITIAL_ARRAY_SIZE_;
2206             UChar   *match     = addToUCharArray(buffer, &matchsize,
2207                                            strsrch->canonicalPrefixAccents,
2208                                            strsrch->search->text + start,
2209                                            offset - start,
2210                                            strsrch->canonicalSuffixAccents,
2211                                            status);
2212 
2213             // run the collator iterator through this match
2214             // if status is a failure ucol_setText does nothing
2215             ucol_setText(coleiter, match, matchsize, status);
2216             if (U_SUCCESS(*status)) {
2217                 if (checkCollationMatch(strsrch, coleiter)) {
2218                     if (match != buffer) {
2219                         uprv_free(match);
2220                     }
2221                     return end;
2222                 }
2223             }
2224             count --;
2225         }
2226     }
2227     return USEARCH_DONE;
2228 }
2229 
2230 /**
2231 * Take the rearranged start accents and tries matching. If match failed at
2232 * a seperate following set of accents (seperated from the rearranged on by
2233 * at least a base character) then we rearrange the preceding accents and
2234 * tries matching again.
2235 * We allow skipping of the ends of the accent set if the ces do not match.
2236 * However if the failure is found before the accent set, it fails.
2237 * Internal method, status assumed to be success, caller has to check status
2238 * before calling this method.
2239 * @param strsrch string search data
2240 * @param textoffset of the ends of the rearranged accent
2241 * @param status output error status if any
2242 * @return USEARCH_DONE if a match is not found, otherwise return the ending
2243 *         offset of the match. Note this start includes all following accents.
2244 */
2245 static
doPreviousCanonicalPrefixMatch(UStringSearch * strsrch,int32_t textoffset,UErrorCode * status)2246 int32_t doPreviousCanonicalPrefixMatch(UStringSearch *strsrch,
2247                                            int32_t    textoffset,
2248                                            UErrorCode    *status)
2249 {
2250     const UChar       *text       = strsrch->search->text;
2251     const UCollator   *collator   = strsrch->collator;
2252           int32_t      safelength = 0;
2253           UChar       *safetext;
2254           int32_t      safetextlength;
2255           UChar        safebuffer[INITIAL_ARRAY_SIZE_];
2256           int32_t  safeoffset = textoffset;
2257 
2258     if (textoffset &&
2259         ucol_unsafeCP(strsrch->canonicalPrefixAccents[
2260                                  u_strlen(strsrch->canonicalPrefixAccents) - 1
2261                                          ], collator)) {
2262         safeoffset     = getNextSafeOffset(collator, text, textoffset,
2263                                            strsrch->search->textLength);
2264         safelength     = safeoffset - textoffset;
2265         safetextlength = INITIAL_ARRAY_SIZE_;
2266         safetext       = addToUCharArray(safebuffer, &safetextlength,
2267                                          strsrch->canonicalPrefixAccents,
2268                                          text + textoffset, safelength,
2269                                          NULL, status);
2270     }
2271     else {
2272         safetextlength = u_strlen(strsrch->canonicalPrefixAccents);
2273         safetext       = strsrch->canonicalPrefixAccents;
2274     }
2275 
2276     UCollationElements *coleiter = strsrch->utilIter;
2277      // if status is a failure, ucol_setText does nothing
2278     ucol_setText(coleiter, safetext, safetextlength, status);
2279     // status checked in loop below
2280 
2281     int32_t  *ce           = strsrch->pattern.CE;
2282     int32_t   celength     = strsrch->pattern.CELength;
2283     int       ceindex      = 0;
2284     UBool     isSafe       = TRUE; // safe zone indication flag for position
2285     int32_t   prefixlength = u_strlen(strsrch->canonicalPrefixAccents);
2286 
2287     while (ceindex < celength) {
2288         int32_t textce = ucol_next(coleiter, status);
2289         if (U_FAILURE(*status)) {
2290             if (isSafe) {
2291                 cleanUpSafeText(strsrch, safetext, safebuffer);
2292             }
2293             return USEARCH_DONE;
2294         }
2295         if (textce == UCOL_NULLORDER) {
2296             // check if we have passed the safe buffer
2297             if (coleiter == strsrch->textIter) {
2298                 cleanUpSafeText(strsrch, safetext, safebuffer);
2299                 return USEARCH_DONE;
2300             }
2301             cleanUpSafeText(strsrch, safetext, safebuffer);
2302             safetext = safebuffer;
2303             coleiter = strsrch->textIter;
2304             setColEIterOffset(coleiter, safeoffset);
2305             // status checked at the start of the loop
2306             isSafe = FALSE;
2307             continue;
2308         }
2309         textce = getCE(strsrch, textce);
2310         if (textce != UCOL_IGNORABLE && textce != ce[ceindex]) {
2311             // do the beginning stuff
2312             int32_t failedoffset = ucol_getOffset(coleiter);
2313             if (isSafe && failedoffset <= prefixlength) {
2314                 // alas... no hope. failed at rearranged accent set
2315                 cleanUpSafeText(strsrch, safetext, safebuffer);
2316                 return USEARCH_DONE;
2317             }
2318             else {
2319                 if (isSafe) {
2320                     failedoffset = safeoffset - failedoffset;
2321                     cleanUpSafeText(strsrch, safetext, safebuffer);
2322                 }
2323 
2324                 // try rearranging the end accents
2325                 int32_t result = doPreviousCanonicalSuffixMatch(strsrch,
2326                                         textoffset, failedoffset, status);
2327                 if (result != USEARCH_DONE) {
2328                     // if status is a failure, ucol_setOffset does nothing
2329                     setColEIterOffset(strsrch->textIter, result);
2330                 }
2331                 if (U_FAILURE(*status)) {
2332                     return USEARCH_DONE;
2333                 }
2334                 return result;
2335             }
2336         }
2337         if (textce == ce[ceindex]) {
2338             ceindex ++;
2339         }
2340     }
2341     // set offset here
2342     if (isSafe) {
2343         int32_t result      = ucol_getOffset(coleiter);
2344         // sets the text iterator here with the correct expansion and offset
2345         int32_t     leftoverces = getExpansionSuffix(coleiter);
2346         cleanUpSafeText(strsrch, safetext, safebuffer);
2347         if (result <= prefixlength) {
2348             result = textoffset;
2349         }
2350         else {
2351             result = textoffset + (safeoffset - result);
2352         }
2353         setColEIterOffset(strsrch->textIter, result);
2354         setExpansionSuffix(strsrch->textIter, leftoverces);
2355         return result;
2356     }
2357 
2358     return ucol_getOffset(coleiter);
2359 }
2360 
2361 /**
2362 * Trying out the substring and sees if it can be a canonical match.
2363 * This will try normalizing the starting accents and arranging them into
2364 * canonical equivalents and check their corresponding ces with the pattern ce.
2365 * Prefix accents in the text will be grouped according to their combining
2366 * class and the groups will be mixed and matched to try find the perfect
2367 * match with the pattern.
2368 * So for instance looking for "\u0301" in "\u030A\u0301\u0325"
2369 * step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
2370 *         "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
2371 *         "\u0301\u0325".
2372 * step 2: check if any of the generated substrings matches the pattern.
2373 * Internal method, status assumed to be success, caller has to check status
2374 * before calling this method.
2375 * @param strsrch string search data
2376 * @param textoffset start offset in the collation element text that starts
2377 *                   with the accents to be rearranged
2378 * @param status output error status if any
2379 * @return TRUE if the match is valid, FALSE otherwise
2380 */
2381 static
doPreviousCanonicalMatch(UStringSearch * strsrch,int32_t textoffset,UErrorCode * status)2382 UBool doPreviousCanonicalMatch(UStringSearch *strsrch,
2383                                int32_t    textoffset,
2384                                UErrorCode    *status)
2385 {
2386     const UChar       *text       = strsrch->search->text;
2387           int32_t  temp       = textoffset;
2388           int32_t      textlength = strsrch->search->textLength;
2389     if ((getFCD(text, &temp, textlength) >> SECOND_LAST_BYTE_SHIFT_) == 0) {
2390         UCollationElements *coleiter = strsrch->textIter;
2391         int32_t         offset   = ucol_getOffset(coleiter);
2392         if (strsrch->pattern.hasSuffixAccents) {
2393             offset = doPreviousCanonicalSuffixMatch(strsrch, textoffset,
2394                                                     offset, status);
2395             if (U_SUCCESS(*status) && offset != USEARCH_DONE) {
2396                 setColEIterOffset(coleiter, offset);
2397                 return TRUE;
2398             }
2399         }
2400         return FALSE;
2401     }
2402 
2403     if (!strsrch->pattern.hasPrefixAccents) {
2404         return FALSE;
2405     }
2406 
2407     UChar       accents[INITIAL_ARRAY_SIZE_];
2408     // offset to the last base character in substring to search
2409     int32_t baseoffset = getNextBaseOffset(text, textoffset, textlength);
2410     // normalizing the offensive string
2411     unorm_normalize(text + textoffset, baseoffset - textoffset, UNORM_NFD,
2412                                0, accents, INITIAL_ARRAY_SIZE_, status);
2413     // status checked in loop
2414 
2415     int32_t accentsindex[INITIAL_ARRAY_SIZE_];
2416     int32_t size = getUnblockedAccentIndex(accents, accentsindex);
2417 
2418     // 2 power n - 1 plus the full set of accents
2419     int32_t  count = (2 << (size - 1)) - 1;
2420     while (U_SUCCESS(*status) && count > 0) {
2421         UChar *rearrange = strsrch->canonicalPrefixAccents;
2422         // copy the base characters
2423         for (int k = 0; k < accentsindex[0]; k ++) {
2424             *rearrange ++ = accents[k];
2425         }
2426         // forming all possible canonical rearrangement by dropping
2427         // sets of accents
2428         for (int i = 0; i <= size - 1; i ++) {
2429             int32_t mask = 1 << (size - i - 1);
2430             if (count & mask) {
2431                 for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
2432                     *rearrange ++ = accents[j];
2433                 }
2434             }
2435         }
2436         *rearrange = 0;
2437         int32_t offset = doPreviousCanonicalPrefixMatch(strsrch,
2438                                                           baseoffset, status);
2439         if (offset != USEARCH_DONE) {
2440             return TRUE; // match found
2441         }
2442         count --;
2443     }
2444     return FALSE;
2445 }
2446 
2447 /**
2448 * Checks match for contraction.
2449 * If the match starts with a partial contraction we fail.
2450 * Internal method, status assumed to be success, caller has to check status
2451 * before calling this method.
2452 * @param strsrch string search data
2453 * @param start offset of potential match, to be modified if necessary
2454 * @param end offset of potential match, to be modified if necessary
2455 * @param status only error status if any
2456 * @return TRUE if match passes the contraction test, FALSE otherwise
2457 */
2458 static
checkPreviousCanonicalContractionMatch(UStringSearch * strsrch,int32_t * start,int32_t * end,UErrorCode * status)2459 UBool checkPreviousCanonicalContractionMatch(UStringSearch *strsrch,
2460                                      int32_t   *start,
2461                                      int32_t   *end, UErrorCode  *status)
2462 {
2463           UCollationElements *coleiter   = strsrch->textIter;
2464           int32_t             textlength = strsrch->search->textLength;
2465           int32_t         temp       = *end;
2466     const UCollator          *collator   = strsrch->collator;
2467     const UChar              *text       = strsrch->search->text;
2468     // This part checks if either if the start of the match contains potential
2469     // contraction. If so we'll have to iterate through them
2470     // Since we used ucol_next while previously looking for the potential
2471     // match, this guarantees that our end will not be a partial contraction,
2472     // or a partial supplementary character.
2473     if (*start < textlength && ucol_unsafeCP(text[*start], collator)) {
2474         int32_t expansion  = getExpansionSuffix(coleiter);
2475         UBool   expandflag = expansion > 0;
2476         setColEIterOffset(coleiter, *end);
2477         while (expansion > 0) {
2478             // getting rid of the redundant ce
2479             // since forward contraction/expansion may have extra ces
2480             // if we are in the normalization buffer, hasAccentsBeforeMatch
2481             // would have taken care of it.
2482             // E.g. the character \u01FA will have an expansion of 3, but if
2483             // we are only looking for A ring A\u030A, we'll have to skip the
2484             // last ce in the expansion buffer
2485             ucol_previous(coleiter, status);
2486             if (U_FAILURE(*status)) {
2487                 return FALSE;
2488             }
2489             if (ucol_getOffset(coleiter) != temp) {
2490                 *end = temp;
2491                 temp  = ucol_getOffset(coleiter);
2492             }
2493             expansion --;
2494         }
2495 
2496         int32_t  *patternce       = strsrch->pattern.CE;
2497         int32_t   patterncelength = strsrch->pattern.CELength;
2498         int32_t   count           = patterncelength;
2499         while (count > 0) {
2500             int32_t ce = getCE(strsrch, ucol_previous(coleiter, status));
2501             // status checked below, note that if status is a failure
2502             // ucol_previous returns UCOL_NULLORDER
2503             if (ce == UCOL_IGNORABLE) {
2504                 continue;
2505             }
2506             if (expandflag && count == 0 &&
2507                 getColElemIterOffset(coleiter, FALSE) != temp) {
2508                 *end = temp;
2509                 temp  = ucol_getOffset(coleiter);
2510             }
2511             if (count == patterncelength &&
2512                 ce != patternce[patterncelength - 1]) {
2513                 // accents may have extra starting ces, this occurs when a
2514                 // pure accent pattern is matched without rearrangement
2515                 int32_t    expected = patternce[patterncelength - 1];
2516                 UTF_BACK_1(text, 0, *end);
2517                 if (getFCD(text, end, textlength) & LAST_BYTE_MASK_) {
2518                     ce = getCE(strsrch, ucol_previous(coleiter, status));
2519                     while (U_SUCCESS(*status) && ce != expected &&
2520                            ce != UCOL_NULLORDER &&
2521                            ucol_getOffset(coleiter) <= *start) {
2522                         ce = getCE(strsrch, ucol_previous(coleiter, status));
2523                     }
2524                 }
2525             }
2526             if (U_FAILURE(*status) || ce != patternce[count - 1]) {
2527                 (*start) --;
2528                 *start = getPreviousBaseOffset(text, *start);
2529                 return FALSE;
2530             }
2531             count --;
2532         }
2533     }
2534     return TRUE;
2535 }
2536 
2537 /**
2538 * Checks and sets the match information if found.
2539 * Checks
2540 * <ul>
2541 * <li> the potential match does not repeat the previous match
2542 * <li> boundaries are correct
2543 * <li> potential match does not end in the middle of a contraction
2544 * <li> identical matches
2545 * <\ul>
2546 * Otherwise the offset will be shifted to the next character.
2547 * Internal method, status assumed to be success, caller has to check status
2548 * before calling this method.
2549 * @param strsrch string search data
2550 * @param textoffset offset in the collation element text. the returned value
2551 *        will be the truncated start offset of the match or the new start
2552 *        search offset.
2553 * @param status only error status if any
2554 * @return TRUE if the match is valid, FALSE otherwise
2555 */
2556 static
checkPreviousCanonicalMatch(UStringSearch * strsrch,int32_t * textoffset,UErrorCode * status)2557 inline UBool checkPreviousCanonicalMatch(UStringSearch *strsrch,
2558                                          int32_t   *textoffset,
2559                                          UErrorCode    *status)
2560 {
2561     // to ensure that the start and ends are not composite characters
2562     UCollationElements *coleiter = strsrch->textIter;
2563     // if we have a canonical accent match
2564     if ((strsrch->pattern.hasSuffixAccents &&
2565         strsrch->canonicalSuffixAccents[0]) ||
2566         (strsrch->pattern.hasPrefixAccents &&
2567         strsrch->canonicalPrefixAccents[0])) {
2568         strsrch->search->matchedIndex  = *textoffset;
2569         strsrch->search->matchedLength =
2570             getNextUStringSearchBaseOffset(strsrch,
2571                                       getColElemIterOffset(coleiter, FALSE))
2572             - *textoffset;
2573         return TRUE;
2574     }
2575 
2576     int32_t end = ucol_getOffset(coleiter);
2577     if (!checkPreviousCanonicalContractionMatch(strsrch, textoffset, &end,
2578                                                 status) ||
2579          U_FAILURE(*status)) {
2580         return FALSE;
2581     }
2582 
2583     end = getNextUStringSearchBaseOffset(strsrch, end);
2584     // this totally matches, however we need to check if it is repeating
2585     if (checkRepeatedMatch(strsrch, *textoffset, end) ||
2586         !isBreakUnit(strsrch, *textoffset, end) ||
2587         !checkIdentical(strsrch, *textoffset, end)) {
2588         (*textoffset) --;
2589         *textoffset = getPreviousBaseOffset(strsrch->search->text,
2590                                             *textoffset);
2591         return FALSE;
2592     }
2593 
2594     strsrch->search->matchedIndex  = *textoffset;
2595     strsrch->search->matchedLength = end - *textoffset;
2596     return TRUE;
2597 }
2598 #endif // #if BOYER_MOORE
2599 
2600 // constructors and destructor -------------------------------------------
2601 
usearch_open(const UChar * pattern,int32_t patternlength,const UChar * text,int32_t textlength,const char * locale,UBreakIterator * breakiter,UErrorCode * status)2602 U_CAPI UStringSearch * U_EXPORT2 usearch_open(const UChar *pattern,
2603                                           int32_t         patternlength,
2604                                     const UChar          *text,
2605                                           int32_t         textlength,
2606                                     const char           *locale,
2607                                           UBreakIterator *breakiter,
2608                                           UErrorCode     *status)
2609 {
2610     if (U_FAILURE(*status)) {
2611         return NULL;
2612     }
2613 #if UCONFIG_NO_BREAK_ITERATION
2614     if (breakiter != NULL) {
2615         *status = U_UNSUPPORTED_ERROR;
2616         return NULL;
2617     }
2618 #endif
2619     if (locale) {
2620         // ucol_open internally checks for status
2621         UCollator     *collator = ucol_open(locale, status);
2622         // pattern, text checks are done in usearch_openFromCollator
2623         UStringSearch *result   = usearch_openFromCollator(pattern,
2624                                               patternlength, text, textlength,
2625                                               collator, breakiter, status);
2626 
2627         if (result == NULL || U_FAILURE(*status)) {
2628             if (collator) {
2629                 ucol_close(collator);
2630             }
2631             return NULL;
2632         }
2633         else {
2634             result->ownCollator = TRUE;
2635         }
2636         return result;
2637     }
2638     *status = U_ILLEGAL_ARGUMENT_ERROR;
2639     return NULL;
2640 }
2641 
usearch_openFromCollator(const UChar * pattern,int32_t patternlength,const UChar * text,int32_t textlength,const UCollator * collator,UBreakIterator * breakiter,UErrorCode * status)2642 U_CAPI UStringSearch * U_EXPORT2 usearch_openFromCollator(
2643                                   const UChar          *pattern,
2644                                         int32_t         patternlength,
2645                                   const UChar          *text,
2646                                         int32_t         textlength,
2647                                   const UCollator      *collator,
2648                                         UBreakIterator *breakiter,
2649                                         UErrorCode     *status)
2650 {
2651     if (U_FAILURE(*status)) {
2652         return NULL;
2653     }
2654 #if UCONFIG_NO_BREAK_ITERATION
2655     if (breakiter != NULL) {
2656         *status = U_UNSUPPORTED_ERROR;
2657         return NULL;
2658     }
2659 #endif
2660     if (pattern == NULL || text == NULL || collator == NULL) {
2661         *status = U_ILLEGAL_ARGUMENT_ERROR;
2662         return NULL;
2663     }
2664 
2665     // string search does not really work when numeric collation is turned on
2666     if(ucol_getAttribute(collator, UCOL_NUMERIC_COLLATION, status) == UCOL_ON) {
2667         *status = U_UNSUPPORTED_ERROR;
2668         return NULL;
2669     }
2670 
2671     if (U_SUCCESS(*status)) {
2672         initializeFCD(status);
2673         if (U_FAILURE(*status)) {
2674             return NULL;
2675         }
2676 
2677         UStringSearch *result;
2678         if (textlength == -1) {
2679             textlength = u_strlen(text);
2680         }
2681         if (patternlength == -1) {
2682             patternlength = u_strlen(pattern);
2683         }
2684         if (textlength <= 0 || patternlength <= 0) {
2685             *status = U_ILLEGAL_ARGUMENT_ERROR;
2686             return NULL;
2687         }
2688 
2689         result = (UStringSearch *)uprv_malloc(sizeof(UStringSearch));
2690         if (result == NULL) {
2691             *status = U_MEMORY_ALLOCATION_ERROR;
2692             return NULL;
2693         }
2694 
2695         result->collator    = collator;
2696         result->strength    = ucol_getStrength(collator);
2697         result->ceMask      = getMask(result->strength);
2698         result->toShift     =
2699              ucol_getAttribute(collator, UCOL_ALTERNATE_HANDLING, status) ==
2700                                                             UCOL_SHIFTED;
2701         result->variableTop = ucol_getVariableTop(collator, status);
2702 
2703         result->nfd         = Normalizer2Factory::getNFDInstance(*status);
2704 
2705         if (U_FAILURE(*status)) {
2706             uprv_free(result);
2707             return NULL;
2708         }
2709 
2710         result->search             = (USearch *)uprv_malloc(sizeof(USearch));
2711         if (result->search == NULL) {
2712             *status = U_MEMORY_ALLOCATION_ERROR;
2713             uprv_free(result);
2714             return NULL;
2715         }
2716 
2717         result->search->text       = text;
2718         result->search->textLength = textlength;
2719 
2720         result->pattern.text       = pattern;
2721         result->pattern.textLength = patternlength;
2722         result->pattern.CE         = NULL;
2723         result->pattern.PCE        = NULL;
2724 
2725         result->search->breakIter  = breakiter;
2726 #if !UCONFIG_NO_BREAK_ITERATION
2727         result->search->internalBreakIter = ubrk_open(UBRK_CHARACTER, ucol_getLocaleByType(result->collator, ULOC_VALID_LOCALE, status), text, textlength, status);
2728         if (breakiter) {
2729             ubrk_setText(breakiter, text, textlength, status);
2730         }
2731 #endif
2732 
2733         result->ownCollator           = FALSE;
2734         result->search->matchedLength = 0;
2735         result->search->matchedIndex  = USEARCH_DONE;
2736         result->utilIter              = NULL;
2737         result->textIter              = ucol_openElements(collator, text,
2738                                                           textlength, status);
2739         if (U_FAILURE(*status)) {
2740             usearch_close(result);
2741             return NULL;
2742         }
2743 
2744         result->search->isOverlap          = FALSE;
2745         result->search->isCanonicalMatch   = FALSE;
2746         result->search->elementComparisonType = 0;
2747         result->search->isForwardSearching = TRUE;
2748         result->search->reset              = TRUE;
2749 
2750         initialize(result, status);
2751 
2752         if (U_FAILURE(*status)) {
2753             usearch_close(result);
2754             return NULL;
2755         }
2756 
2757         return result;
2758     }
2759     return NULL;
2760 }
2761 
usearch_close(UStringSearch * strsrch)2762 U_CAPI void U_EXPORT2 usearch_close(UStringSearch *strsrch)
2763 {
2764     if (strsrch) {
2765         if (strsrch->pattern.CE != strsrch->pattern.CEBuffer &&
2766             strsrch->pattern.CE) {
2767             uprv_free(strsrch->pattern.CE);
2768         }
2769 
2770         if (strsrch->pattern.PCE != NULL &&
2771             strsrch->pattern.PCE != strsrch->pattern.PCEBuffer) {
2772             uprv_free(strsrch->pattern.PCE);
2773         }
2774 
2775         ucol_closeElements(strsrch->textIter);
2776         ucol_closeElements(strsrch->utilIter);
2777 
2778         if (strsrch->ownCollator && strsrch->collator) {
2779             ucol_close((UCollator *)strsrch->collator);
2780         }
2781 
2782 #if !UCONFIG_NO_BREAK_ITERATION
2783         if (strsrch->search->internalBreakIter) {
2784             ubrk_close(strsrch->search->internalBreakIter);
2785         }
2786 #endif
2787 
2788         uprv_free(strsrch->search);
2789         uprv_free(strsrch);
2790     }
2791 }
2792 
2793 // set and get methods --------------------------------------------------
2794 
usearch_setOffset(UStringSearch * strsrch,int32_t position,UErrorCode * status)2795 U_CAPI void U_EXPORT2 usearch_setOffset(UStringSearch *strsrch,
2796                                         int32_t    position,
2797                                         UErrorCode    *status)
2798 {
2799     if (U_SUCCESS(*status) && strsrch) {
2800         if (isOutOfBounds(strsrch->search->textLength, position)) {
2801             *status = U_INDEX_OUTOFBOUNDS_ERROR;
2802         }
2803         else {
2804             setColEIterOffset(strsrch->textIter, position);
2805         }
2806         strsrch->search->matchedIndex  = USEARCH_DONE;
2807         strsrch->search->matchedLength = 0;
2808         strsrch->search->reset         = FALSE;
2809     }
2810 }
2811 
usearch_getOffset(const UStringSearch * strsrch)2812 U_CAPI int32_t U_EXPORT2 usearch_getOffset(const UStringSearch *strsrch)
2813 {
2814     if (strsrch) {
2815         int32_t result = ucol_getOffset(strsrch->textIter);
2816         if (isOutOfBounds(strsrch->search->textLength, result)) {
2817             return USEARCH_DONE;
2818         }
2819         return result;
2820     }
2821     return USEARCH_DONE;
2822 }
2823 
usearch_setAttribute(UStringSearch * strsrch,USearchAttribute attribute,USearchAttributeValue value,UErrorCode * status)2824 U_CAPI void U_EXPORT2 usearch_setAttribute(UStringSearch *strsrch,
2825                                  USearchAttribute attribute,
2826                                  USearchAttributeValue value,
2827                                  UErrorCode *status)
2828 {
2829     if (U_SUCCESS(*status) && strsrch) {
2830         switch (attribute)
2831         {
2832         case USEARCH_OVERLAP :
2833             strsrch->search->isOverlap = (value == USEARCH_ON ? TRUE : FALSE);
2834             break;
2835         case USEARCH_CANONICAL_MATCH :
2836             strsrch->search->isCanonicalMatch = (value == USEARCH_ON ? TRUE :
2837                                                                       FALSE);
2838             break;
2839         case USEARCH_ELEMENT_COMPARISON :
2840             if (value == USEARCH_PATTERN_BASE_WEIGHT_IS_WILDCARD || value == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD) {
2841                 strsrch->search->elementComparisonType = (int16_t)value;
2842             } else {
2843                 strsrch->search->elementComparisonType = 0;
2844             }
2845             break;
2846         case USEARCH_ATTRIBUTE_COUNT :
2847         default:
2848             *status = U_ILLEGAL_ARGUMENT_ERROR;
2849         }
2850     }
2851     if (value == USEARCH_ATTRIBUTE_VALUE_COUNT) {
2852         *status = U_ILLEGAL_ARGUMENT_ERROR;
2853     }
2854 }
2855 
usearch_getAttribute(const UStringSearch * strsrch,USearchAttribute attribute)2856 U_CAPI USearchAttributeValue U_EXPORT2 usearch_getAttribute(
2857                                                 const UStringSearch *strsrch,
2858                                                 USearchAttribute attribute)
2859 {
2860     if (strsrch) {
2861         switch (attribute) {
2862         case USEARCH_OVERLAP :
2863             return (strsrch->search->isOverlap == TRUE ? USEARCH_ON :
2864                                                         USEARCH_OFF);
2865         case USEARCH_CANONICAL_MATCH :
2866             return (strsrch->search->isCanonicalMatch == TRUE ? USEARCH_ON :
2867                                                                USEARCH_OFF);
2868         case USEARCH_ELEMENT_COMPARISON :
2869             {
2870                 int16_t value = strsrch->search->elementComparisonType;
2871                 if (value == USEARCH_PATTERN_BASE_WEIGHT_IS_WILDCARD || value == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD) {
2872                     return (USearchAttributeValue)value;
2873                 } else {
2874                     return USEARCH_STANDARD_ELEMENT_COMPARISON;
2875                 }
2876             }
2877         case USEARCH_ATTRIBUTE_COUNT :
2878             return USEARCH_DEFAULT;
2879         }
2880     }
2881     return USEARCH_DEFAULT;
2882 }
2883 
usearch_getMatchedStart(const UStringSearch * strsrch)2884 U_CAPI int32_t U_EXPORT2 usearch_getMatchedStart(
2885                                                 const UStringSearch *strsrch)
2886 {
2887     if (strsrch == NULL) {
2888         return USEARCH_DONE;
2889     }
2890     return strsrch->search->matchedIndex;
2891 }
2892 
2893 
usearch_getMatchedText(const UStringSearch * strsrch,UChar * result,int32_t resultCapacity,UErrorCode * status)2894 U_CAPI int32_t U_EXPORT2 usearch_getMatchedText(const UStringSearch *strsrch,
2895                                             UChar         *result,
2896                                             int32_t        resultCapacity,
2897                                             UErrorCode    *status)
2898 {
2899     if (U_FAILURE(*status)) {
2900         return USEARCH_DONE;
2901     }
2902     if (strsrch == NULL || resultCapacity < 0 || (resultCapacity > 0 &&
2903         result == NULL)) {
2904         *status = U_ILLEGAL_ARGUMENT_ERROR;
2905         return USEARCH_DONE;
2906     }
2907 
2908     int32_t     copylength = strsrch->search->matchedLength;
2909     int32_t copyindex  = strsrch->search->matchedIndex;
2910     if (copyindex == USEARCH_DONE) {
2911         u_terminateUChars(result, resultCapacity, 0, status);
2912         return USEARCH_DONE;
2913     }
2914 
2915     if (resultCapacity < copylength) {
2916         copylength = resultCapacity;
2917     }
2918     if (copylength > 0) {
2919         uprv_memcpy(result, strsrch->search->text + copyindex,
2920                     copylength * sizeof(UChar));
2921     }
2922     return u_terminateUChars(result, resultCapacity,
2923                              strsrch->search->matchedLength, status);
2924 }
2925 
usearch_getMatchedLength(const UStringSearch * strsrch)2926 U_CAPI int32_t U_EXPORT2 usearch_getMatchedLength(
2927                                               const UStringSearch *strsrch)
2928 {
2929     if (strsrch) {
2930         return strsrch->search->matchedLength;
2931     }
2932     return USEARCH_DONE;
2933 }
2934 
2935 #if !UCONFIG_NO_BREAK_ITERATION
2936 
usearch_setBreakIterator(UStringSearch * strsrch,UBreakIterator * breakiter,UErrorCode * status)2937 U_CAPI void U_EXPORT2 usearch_setBreakIterator(UStringSearch  *strsrch,
2938                                                UBreakIterator *breakiter,
2939                                                UErrorCode     *status)
2940 {
2941     if (U_SUCCESS(*status) && strsrch) {
2942         strsrch->search->breakIter = breakiter;
2943         if (breakiter) {
2944             ubrk_setText(breakiter, strsrch->search->text,
2945                          strsrch->search->textLength, status);
2946         }
2947     }
2948 }
2949 
2950 U_CAPI const UBreakIterator* U_EXPORT2
usearch_getBreakIterator(const UStringSearch * strsrch)2951 usearch_getBreakIterator(const UStringSearch *strsrch)
2952 {
2953     if (strsrch) {
2954         return strsrch->search->breakIter;
2955     }
2956     return NULL;
2957 }
2958 
2959 #endif
2960 
usearch_setText(UStringSearch * strsrch,const UChar * text,int32_t textlength,UErrorCode * status)2961 U_CAPI void U_EXPORT2 usearch_setText(      UStringSearch *strsrch,
2962                                       const UChar         *text,
2963                                             int32_t        textlength,
2964                                             UErrorCode    *status)
2965 {
2966     if (U_SUCCESS(*status)) {
2967         if (strsrch == NULL || text == NULL || textlength < -1 ||
2968             textlength == 0) {
2969             *status = U_ILLEGAL_ARGUMENT_ERROR;
2970         }
2971         else {
2972             if (textlength == -1) {
2973                 textlength = u_strlen(text);
2974             }
2975             strsrch->search->text       = text;
2976             strsrch->search->textLength = textlength;
2977             ucol_setText(strsrch->textIter, text, textlength, status);
2978             strsrch->search->matchedIndex  = USEARCH_DONE;
2979             strsrch->search->matchedLength = 0;
2980             strsrch->search->reset         = TRUE;
2981 #if !UCONFIG_NO_BREAK_ITERATION
2982             if (strsrch->search->breakIter != NULL) {
2983                 ubrk_setText(strsrch->search->breakIter, text,
2984                              textlength, status);
2985             }
2986             ubrk_setText(strsrch->search->internalBreakIter, text, textlength, status);
2987 #endif
2988         }
2989     }
2990 }
2991 
usearch_getText(const UStringSearch * strsrch,int32_t * length)2992 U_CAPI const UChar * U_EXPORT2 usearch_getText(const UStringSearch *strsrch,
2993                                                      int32_t       *length)
2994 {
2995     if (strsrch) {
2996         *length = strsrch->search->textLength;
2997         return strsrch->search->text;
2998     }
2999     return NULL;
3000 }
3001 
usearch_setCollator(UStringSearch * strsrch,const UCollator * collator,UErrorCode * status)3002 U_CAPI void U_EXPORT2 usearch_setCollator(      UStringSearch *strsrch,
3003                                           const UCollator     *collator,
3004                                                 UErrorCode    *status)
3005 {
3006     if (U_SUCCESS(*status)) {
3007         if (collator == NULL) {
3008             *status = U_ILLEGAL_ARGUMENT_ERROR;
3009             return;
3010         }
3011 
3012         if (strsrch) {
3013             if (strsrch->ownCollator && (strsrch->collator != collator)) {
3014                 ucol_close((UCollator *)strsrch->collator);
3015                 strsrch->ownCollator = FALSE;
3016             }
3017             strsrch->collator    = collator;
3018             strsrch->strength    = ucol_getStrength(collator);
3019             strsrch->ceMask      = getMask(strsrch->strength);
3020 #if !UCONFIG_NO_BREAK_ITERATION
3021             ubrk_close(strsrch->search->internalBreakIter);
3022             strsrch->search->internalBreakIter = ubrk_open(UBRK_CHARACTER, ucol_getLocaleByType(collator, ULOC_VALID_LOCALE, status),
3023                                                      strsrch->search->text, strsrch->search->textLength, status);
3024 #endif
3025             // if status is a failure, ucol_getAttribute returns UCOL_DEFAULT
3026             strsrch->toShift     =
3027                ucol_getAttribute(collator, UCOL_ALTERNATE_HANDLING, status) ==
3028                                                                 UCOL_SHIFTED;
3029             // if status is a failure, ucol_getVariableTop returns 0
3030             strsrch->variableTop = ucol_getVariableTop(collator, status);
3031             if (U_SUCCESS(*status)) {
3032                 initialize(strsrch, status);
3033                 if (U_SUCCESS(*status)) {
3034                     /* free offset buffer to avoid memory leak before initializing. */
3035                     ucol_freeOffsetBuffer(&(strsrch->textIter->iteratordata_));
3036                     uprv_init_collIterate(collator, strsrch->search->text,
3037                                           strsrch->search->textLength,
3038                                           &(strsrch->textIter->iteratordata_),
3039                                           status);
3040                     strsrch->utilIter->iteratordata_.coll = collator;
3041                 }
3042             }
3043         }
3044 
3045         // **** are these calls needed?
3046         // **** we call uprv_init_pce in initializePatternPCETable
3047         // **** and the CEBuffer constructor...
3048 #if 0
3049         uprv_init_pce(strsrch->textIter);
3050         uprv_init_pce(strsrch->utilIter);
3051 #endif
3052     }
3053 }
3054 
usearch_getCollator(const UStringSearch * strsrch)3055 U_CAPI UCollator * U_EXPORT2 usearch_getCollator(const UStringSearch *strsrch)
3056 {
3057     if (strsrch) {
3058         return (UCollator *)strsrch->collator;
3059     }
3060     return NULL;
3061 }
3062 
usearch_setPattern(UStringSearch * strsrch,const UChar * pattern,int32_t patternlength,UErrorCode * status)3063 U_CAPI void U_EXPORT2 usearch_setPattern(      UStringSearch *strsrch,
3064                                          const UChar         *pattern,
3065                                                int32_t        patternlength,
3066                                                UErrorCode    *status)
3067 {
3068     if (U_SUCCESS(*status)) {
3069         if (strsrch == NULL || pattern == NULL) {
3070             *status = U_ILLEGAL_ARGUMENT_ERROR;
3071         }
3072         else {
3073             if (patternlength == -1) {
3074                 patternlength = u_strlen(pattern);
3075             }
3076             if (patternlength == 0) {
3077                 *status = U_ILLEGAL_ARGUMENT_ERROR;
3078                 return;
3079             }
3080             strsrch->pattern.text       = pattern;
3081             strsrch->pattern.textLength = patternlength;
3082             initialize(strsrch, status);
3083         }
3084     }
3085 }
3086 
3087 U_CAPI const UChar* U_EXPORT2
usearch_getPattern(const UStringSearch * strsrch,int32_t * length)3088 usearch_getPattern(const UStringSearch *strsrch,
3089                    int32_t       *length)
3090 {
3091     if (strsrch) {
3092         *length = strsrch->pattern.textLength;
3093         return strsrch->pattern.text;
3094     }
3095     return NULL;
3096 }
3097 
3098 // miscellanous methods --------------------------------------------------
3099 
usearch_first(UStringSearch * strsrch,UErrorCode * status)3100 U_CAPI int32_t U_EXPORT2 usearch_first(UStringSearch *strsrch,
3101                                            UErrorCode    *status)
3102 {
3103     if (strsrch && U_SUCCESS(*status)) {
3104         strsrch->search->isForwardSearching = TRUE;
3105         usearch_setOffset(strsrch, 0, status);
3106         if (U_SUCCESS(*status)) {
3107             return usearch_next(strsrch, status);
3108         }
3109     }
3110     return USEARCH_DONE;
3111 }
3112 
usearch_following(UStringSearch * strsrch,int32_t position,UErrorCode * status)3113 U_CAPI int32_t U_EXPORT2 usearch_following(UStringSearch *strsrch,
3114                                                int32_t    position,
3115                                                UErrorCode    *status)
3116 {
3117     if (strsrch && U_SUCCESS(*status)) {
3118         strsrch->search->isForwardSearching = TRUE;
3119         // position checked in usearch_setOffset
3120         usearch_setOffset(strsrch, position, status);
3121         if (U_SUCCESS(*status)) {
3122             return usearch_next(strsrch, status);
3123         }
3124     }
3125     return USEARCH_DONE;
3126 }
3127 
usearch_last(UStringSearch * strsrch,UErrorCode * status)3128 U_CAPI int32_t U_EXPORT2 usearch_last(UStringSearch *strsrch,
3129                                           UErrorCode    *status)
3130 {
3131     if (strsrch && U_SUCCESS(*status)) {
3132         strsrch->search->isForwardSearching = FALSE;
3133         usearch_setOffset(strsrch, strsrch->search->textLength, status);
3134         if (U_SUCCESS(*status)) {
3135             return usearch_previous(strsrch, status);
3136         }
3137     }
3138     return USEARCH_DONE;
3139 }
3140 
usearch_preceding(UStringSearch * strsrch,int32_t position,UErrorCode * status)3141 U_CAPI int32_t U_EXPORT2 usearch_preceding(UStringSearch *strsrch,
3142                                                int32_t    position,
3143                                                UErrorCode    *status)
3144 {
3145     if (strsrch && U_SUCCESS(*status)) {
3146         strsrch->search->isForwardSearching = FALSE;
3147         // position checked in usearch_setOffset
3148         usearch_setOffset(strsrch, position, status);
3149         if (U_SUCCESS(*status)) {
3150             return usearch_previous(strsrch, status);
3151         }
3152     }
3153     return USEARCH_DONE;
3154 }
3155 
3156 /**
3157 * If a direction switch is required, we'll count the number of ces till the
3158 * beginning of the collation element iterator and iterate forwards that
3159 * number of times. This is so that we get to the correct point within the
3160 * string to continue the search in. Imagine when we are in the middle of the
3161 * normalization buffer when the change in direction is request. arrrgghh....
3162 * After searching the offset within the collation element iterator will be
3163 * shifted to the start of the match. If a match is not found, the offset would
3164 * have been set to the end of the text string in the collation element
3165 * iterator.
3166 * Okay, here's my take on normalization buffer. The only time when there can
3167 * be 2 matches within the same normalization is when the pattern is consists
3168 * of all accents. But since the offset returned is from the text string, we
3169 * should not confuse the caller by returning the second match within the
3170 * same normalization buffer. If we do, the 2 results will have the same match
3171 * offsets, and that'll be confusing. I'll return the next match that doesn't
3172 * fall within the same normalization buffer. Note this does not affect the
3173 * results of matches spanning the text and the normalization buffer.
3174 * The position to start searching is taken from the collation element
3175 * iterator. Callers of this API would have to set the offset in the collation
3176 * element iterator before using this method.
3177 */
usearch_next(UStringSearch * strsrch,UErrorCode * status)3178 U_CAPI int32_t U_EXPORT2 usearch_next(UStringSearch *strsrch,
3179                                           UErrorCode    *status)
3180 {
3181     if (U_SUCCESS(*status) && strsrch) {
3182         // note offset is either equivalent to the start of the previous match
3183         // or is set by the user
3184         int32_t      offset       = usearch_getOffset(strsrch);
3185         USearch     *search       = strsrch->search;
3186         search->reset             = FALSE;
3187         int32_t      textlength   = search->textLength;
3188         if (search->isForwardSearching) {
3189 #if BOYER_MOORE
3190             if (offset == textlength
3191                 || (!search->isOverlap &&
3192                     (offset + strsrch->pattern.defaultShiftSize > textlength ||
3193                     (search->matchedIndex != USEARCH_DONE &&
3194                      offset + search->matchedLength >= textlength)))) {
3195                 // not enough characters to match
3196                 setMatchNotFound(strsrch);
3197                 return USEARCH_DONE;
3198             }
3199 #else
3200             if (offset == textlength ||
3201                 (! search->isOverlap &&
3202                 (search->matchedIndex != USEARCH_DONE &&
3203                 offset + search->matchedLength > textlength))) {
3204                     // not enough characters to match
3205                     setMatchNotFound(strsrch);
3206                     return USEARCH_DONE;
3207             }
3208 #endif
3209         }
3210         else {
3211             // switching direction.
3212             // if matchedIndex == USEARCH_DONE, it means that either a
3213             // setOffset has been called or that previous ran off the text
3214             // string. the iterator would have been set to offset 0 if a
3215             // match is not found.
3216             search->isForwardSearching = TRUE;
3217             if (search->matchedIndex != USEARCH_DONE) {
3218                 // there's no need to set the collation element iterator
3219                 // the next call to next will set the offset.
3220                 return search->matchedIndex;
3221             }
3222         }
3223 
3224         if (U_SUCCESS(*status)) {
3225             if (strsrch->pattern.CELength == 0) {
3226                 if (search->matchedIndex == USEARCH_DONE) {
3227                     search->matchedIndex = offset;
3228                 }
3229                 else { // moves by codepoints
3230                     UTF_FWD_1(search->text, search->matchedIndex, textlength);
3231                 }
3232 
3233                 search->matchedLength = 0;
3234                 setColEIterOffset(strsrch->textIter, search->matchedIndex);
3235                 // status checked below
3236                 if (search->matchedIndex == textlength) {
3237                     search->matchedIndex = USEARCH_DONE;
3238                 }
3239             }
3240             else {
3241                 if (search->matchedLength > 0) {
3242                     // if matchlength is 0 we are at the start of the iteration
3243                     if (search->isOverlap) {
3244                         ucol_setOffset(strsrch->textIter, offset + 1, status);
3245                     }
3246                     else {
3247                         ucol_setOffset(strsrch->textIter,
3248                                        offset + search->matchedLength, status);
3249                     }
3250                 }
3251                 else {
3252                     // for boundary check purposes. this will ensure that the
3253                     // next match will not preceed the current offset
3254                     // note search->matchedIndex will always be set to something
3255                     // in the code
3256                     search->matchedIndex = offset - 1;
3257                 }
3258 
3259                 if (search->isCanonicalMatch) {
3260                     // can't use exact here since extra accents are allowed.
3261                     usearch_handleNextCanonical(strsrch, status);
3262                 }
3263                 else {
3264                     usearch_handleNextExact(strsrch, status);
3265                 }
3266             }
3267 
3268             if (U_FAILURE(*status)) {
3269                 return USEARCH_DONE;
3270             }
3271 
3272 #if !BOYER_MOORE
3273             if (search->matchedIndex == USEARCH_DONE) {
3274                 ucol_setOffset(strsrch->textIter, search->textLength, status);
3275             } else {
3276                 ucol_setOffset(strsrch->textIter, search->matchedIndex, status);
3277             }
3278 #endif
3279 
3280             return search->matchedIndex;
3281         }
3282     }
3283     return USEARCH_DONE;
3284 }
3285 
usearch_previous(UStringSearch * strsrch,UErrorCode * status)3286 U_CAPI int32_t U_EXPORT2 usearch_previous(UStringSearch *strsrch,
3287                                               UErrorCode *status)
3288 {
3289     if (U_SUCCESS(*status) && strsrch) {
3290         int32_t offset;
3291         USearch *search = strsrch->search;
3292         if (search->reset) {
3293             offset                     = search->textLength;
3294             search->isForwardSearching = FALSE;
3295             search->reset              = FALSE;
3296             setColEIterOffset(strsrch->textIter, offset);
3297         }
3298         else {
3299             offset = usearch_getOffset(strsrch);
3300         }
3301 
3302         int32_t matchedindex = search->matchedIndex;
3303         if (search->isForwardSearching == TRUE) {
3304             // switching direction.
3305             // if matchedIndex == USEARCH_DONE, it means that either a
3306             // setOffset has been called or that next ran off the text
3307             // string. the iterator would have been set to offset textLength if
3308             // a match is not found.
3309             search->isForwardSearching = FALSE;
3310             if (matchedindex != USEARCH_DONE) {
3311                 return matchedindex;
3312             }
3313         }
3314         else {
3315 #if BOYER_MOORE
3316             if (offset == 0 || matchedindex == 0 ||
3317                 (!search->isOverlap &&
3318                     (offset < strsrch->pattern.defaultShiftSize ||
3319                     (matchedindex != USEARCH_DONE &&
3320                     matchedindex < strsrch->pattern.defaultShiftSize)))) {
3321                 // not enough characters to match
3322                 setMatchNotFound(strsrch);
3323                 return USEARCH_DONE;
3324             }
3325 #else
3326             // Could check pattern length, but the
3327             // linear search will do the right thing
3328             if (offset == 0 || matchedindex == 0) {
3329                 setMatchNotFound(strsrch);
3330                 return USEARCH_DONE;
3331             }
3332 #endif
3333         }
3334 
3335         if (U_SUCCESS(*status)) {
3336             if (strsrch->pattern.CELength == 0) {
3337                 search->matchedIndex =
3338                       (matchedindex == USEARCH_DONE ? offset : matchedindex);
3339                 if (search->matchedIndex == 0) {
3340                     setMatchNotFound(strsrch);
3341                     // status checked below
3342                 }
3343                 else { // move by codepoints
3344                     UTF_BACK_1(search->text, 0, search->matchedIndex);
3345                     setColEIterOffset(strsrch->textIter, search->matchedIndex);
3346                     // status checked below
3347                     search->matchedLength = 0;
3348                 }
3349             }
3350             else {
3351                 if (strsrch->search->isCanonicalMatch) {
3352                     // can't use exact here since extra accents are allowed.
3353                     usearch_handlePreviousCanonical(strsrch, status);
3354                     // status checked below
3355                 }
3356                 else {
3357                     usearch_handlePreviousExact(strsrch, status);
3358                     // status checked below
3359                 }
3360             }
3361 
3362             if (U_FAILURE(*status)) {
3363                 return USEARCH_DONE;
3364             }
3365 
3366             return search->matchedIndex;
3367         }
3368     }
3369     return USEARCH_DONE;
3370 }
3371 
3372 
3373 
usearch_reset(UStringSearch * strsrch)3374 U_CAPI void U_EXPORT2 usearch_reset(UStringSearch *strsrch)
3375 {
3376     /*
3377     reset is setting the attributes that are already in
3378     string search, hence all attributes in the collator should
3379     be retrieved without any problems
3380     */
3381     if (strsrch) {
3382         UErrorCode status            = U_ZERO_ERROR;
3383         UBool      sameCollAttribute = TRUE;
3384         uint32_t   ceMask;
3385         UBool      shift;
3386         uint32_t   varTop;
3387 
3388         // **** hack to deal w/ how processed CEs encode quaternary ****
3389         UCollationStrength newStrength = ucol_getStrength(strsrch->collator);
3390         if ((strsrch->strength < UCOL_QUATERNARY && newStrength >= UCOL_QUATERNARY) ||
3391             (strsrch->strength >= UCOL_QUATERNARY && newStrength < UCOL_QUATERNARY)) {
3392                 sameCollAttribute = FALSE;
3393         }
3394 
3395         strsrch->strength    = ucol_getStrength(strsrch->collator);
3396         ceMask = getMask(strsrch->strength);
3397         if (strsrch->ceMask != ceMask) {
3398             strsrch->ceMask = ceMask;
3399             sameCollAttribute = FALSE;
3400         }
3401 
3402         // if status is a failure, ucol_getAttribute returns UCOL_DEFAULT
3403         shift = ucol_getAttribute(strsrch->collator, UCOL_ALTERNATE_HANDLING,
3404                                   &status) == UCOL_SHIFTED;
3405         if (strsrch->toShift != shift) {
3406             strsrch->toShift  = shift;
3407             sameCollAttribute = FALSE;
3408         }
3409 
3410         // if status is a failure, ucol_getVariableTop returns 0
3411         varTop = ucol_getVariableTop(strsrch->collator, &status);
3412         if (strsrch->variableTop != varTop) {
3413             strsrch->variableTop = varTop;
3414             sameCollAttribute    = FALSE;
3415         }
3416         if (!sameCollAttribute) {
3417             initialize(strsrch, &status);
3418         }
3419         /* free offset buffer to avoid memory leak before initializing. */
3420         ucol_freeOffsetBuffer(&(strsrch->textIter->iteratordata_));
3421         uprv_init_collIterate(strsrch->collator, strsrch->search->text,
3422                               strsrch->search->textLength,
3423                               &(strsrch->textIter->iteratordata_),
3424                               &status);
3425         strsrch->search->matchedLength      = 0;
3426         strsrch->search->matchedIndex       = USEARCH_DONE;
3427         strsrch->search->isOverlap          = FALSE;
3428         strsrch->search->isCanonicalMatch   = FALSE;
3429         strsrch->search->elementComparisonType = 0;
3430         strsrch->search->isForwardSearching = TRUE;
3431         strsrch->search->reset              = TRUE;
3432     }
3433 }
3434 
3435 //
3436 //  CEI  Collation Element + source text index.
3437 //       These structs are kept in the circular buffer.
3438 //
3439 struct  CEI {
3440     int64_t ce;
3441     int32_t lowIndex;
3442     int32_t highIndex;
3443 };
3444 
3445 U_NAMESPACE_BEGIN
3446 
3447 
3448 //
3449 //  CEBuffer   A circular buffer of CEs from the text being searched.
3450 //
3451 #define   DEFAULT_CEBUFFER_SIZE 96
3452 #define   CEBUFFER_EXTRA 32
3453 // Some typical max values to make buffer size more reasonable for asymmetric search.
3454 // #8694 is for a better long-term solution to allocation of this buffer.
3455 #define   MAX_TARGET_IGNORABLES_PER_PAT_JAMO_L 8
3456 #define   MAX_TARGET_IGNORABLES_PER_PAT_OTHER 3
3457 #define   MIGHT_BE_JAMO_L(c) ((c >= 0x1100 && c <= 0x115E) || (c >= 0x3131 && c <= 0x314E) || (c >= 0x3165 && c <= 0x3186))
3458 struct CEBuffer {
3459     CEI                  defBuf[DEFAULT_CEBUFFER_SIZE];
3460     CEI                 *buf;
3461     int32_t              bufSize;
3462     int32_t              firstIx;
3463     int32_t              limitIx;
3464     UCollationElements  *ceIter;
3465     UStringSearch       *strSearch;
3466 
3467 
3468 
3469                CEBuffer(UStringSearch *ss, UErrorCode *status);
3470                ~CEBuffer();
3471    const CEI   *get(int32_t index);
3472    const CEI   *getPrevious(int32_t index);
3473 };
3474 
3475 
CEBuffer(UStringSearch * ss,UErrorCode * status)3476 CEBuffer::CEBuffer(UStringSearch *ss, UErrorCode *status) {
3477     buf = defBuf;
3478     strSearch = ss;
3479     bufSize = ss->pattern.PCELength + CEBUFFER_EXTRA;
3480     if (ss->search->elementComparisonType != 0) {
3481         const UChar * patText = ss->pattern.text;
3482         if (patText) {
3483             const UChar * patTextLimit = patText + ss->pattern.textLength;
3484             while ( patText < patTextLimit ) {
3485                 UChar c = *patText++;
3486                 if (MIGHT_BE_JAMO_L(c)) {
3487                     bufSize += MAX_TARGET_IGNORABLES_PER_PAT_JAMO_L;
3488                 } else {
3489                     // No check for surrogates, we might allocate slightly more buffer than necessary.
3490                     bufSize += MAX_TARGET_IGNORABLES_PER_PAT_OTHER;
3491                 }
3492             }
3493         }
3494     }
3495     ceIter    = ss->textIter;
3496     firstIx = 0;
3497     limitIx = 0;
3498 
3499     uprv_init_pce(ceIter);
3500 
3501     if (bufSize>DEFAULT_CEBUFFER_SIZE) {
3502         buf = (CEI *)uprv_malloc(bufSize * sizeof(CEI));
3503         if (buf == NULL) {
3504             *status = U_MEMORY_ALLOCATION_ERROR;
3505         }
3506     }
3507 }
3508 
3509 // TODO: add a reset or init function so that allocated
3510 //       buffers can be retained & reused.
3511 
~CEBuffer()3512 CEBuffer::~CEBuffer() {
3513     if (buf != defBuf) {
3514         uprv_free(buf);
3515     }
3516 }
3517 
3518 
3519 // Get the CE with the specified index.
3520 //   Index must be in the range
3521 //          n-history_size < index < n+1
3522 //   where n is the largest index to have been fetched by some previous call to this function.
3523 //   The CE value will be UCOL__PROCESSED_NULLORDER at end of input.
3524 //
get(int32_t index)3525 const CEI *CEBuffer::get(int32_t index) {
3526     int i = index % bufSize;
3527 
3528     if (index>=firstIx && index<limitIx) {
3529         // The request was for an entry already in our buffer.
3530         //  Just return it.
3531         return &buf[i];
3532     }
3533 
3534     // Caller is requesting a new, never accessed before, CE.
3535     //   Verify that it is the next one in sequence, which is all
3536     //   that is allowed.
3537     if (index != limitIx) {
3538         U_ASSERT(FALSE);
3539 
3540         return NULL;
3541     }
3542 
3543     // Manage the circular CE buffer indexing
3544     limitIx++;
3545 
3546     if (limitIx - firstIx >= bufSize) {
3547         // The buffer is full, knock out the lowest-indexed entry.
3548         firstIx++;
3549     }
3550 
3551     UErrorCode status = U_ZERO_ERROR;
3552 
3553     buf[i].ce = ucol_nextProcessed(ceIter, &buf[i].lowIndex, &buf[i].highIndex, &status);
3554 
3555     return &buf[i];
3556 }
3557 
3558 // Get the CE with the specified index.
3559 //   Index must be in the range
3560 //          n-history_size < index < n+1
3561 //   where n is the largest index to have been fetched by some previous call to this function.
3562 //   The CE value will be UCOL__PROCESSED_NULLORDER at end of input.
3563 //
getPrevious(int32_t index)3564 const CEI *CEBuffer::getPrevious(int32_t index) {
3565     int i = index % bufSize;
3566 
3567     if (index>=firstIx && index<limitIx) {
3568         // The request was for an entry already in our buffer.
3569         //  Just return it.
3570         return &buf[i];
3571     }
3572 
3573     // Caller is requesting a new, never accessed before, CE.
3574     //   Verify that it is the next one in sequence, which is all
3575     //   that is allowed.
3576     if (index != limitIx) {
3577         U_ASSERT(FALSE);
3578 
3579         return NULL;
3580     }
3581 
3582     // Manage the circular CE buffer indexing
3583     limitIx++;
3584 
3585     if (limitIx - firstIx >= bufSize) {
3586         // The buffer is full, knock out the lowest-indexed entry.
3587         firstIx++;
3588     }
3589 
3590     UErrorCode status = U_ZERO_ERROR;
3591 
3592     buf[i].ce = ucol_previousProcessed(ceIter, &buf[i].lowIndex, &buf[i].highIndex, &status);
3593 
3594     return &buf[i];
3595 }
3596 
3597 U_NAMESPACE_END
3598 
3599 
3600 // #define USEARCH_DEBUG
3601 
3602 #ifdef USEARCH_DEBUG
3603 #include <stdio.h>
3604 #include <stdlib.h>
3605 #endif
3606 
3607 /*
3608  * Find the next break boundary after startIndex. If the UStringSearch object
3609  * has an external break iterator, use that. Otherwise use the internal character
3610  * break iterator.
3611  */
nextBoundaryAfter(UStringSearch * strsrch,int32_t startIndex)3612 static int32_t nextBoundaryAfter(UStringSearch *strsrch, int32_t startIndex) {
3613 #if 0
3614     const UChar *text = strsrch->search->text;
3615     int32_t textLen   = strsrch->search->textLength;
3616 
3617     U_ASSERT(startIndex>=0);
3618     U_ASSERT(startIndex<=textLen);
3619 
3620     if (startIndex >= textLen) {
3621         return startIndex;
3622     }
3623 
3624     UChar32  c;
3625     int32_t  i = startIndex;
3626     U16_NEXT(text, i, textLen, c);
3627 
3628     // If we are on a control character, stop without looking for combining marks.
3629     //    Control characters do not combine.
3630     int32_t gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
3631     if (gcProperty==U_GCB_CONTROL || gcProperty==U_GCB_LF || gcProperty==U_GCB_CR) {
3632         return i;
3633     }
3634 
3635     // The initial character was not a control, and can thus accept trailing
3636     //   combining characters.  Advance over however many of them there are.
3637     int32_t  indexOfLastCharChecked;
3638     for (;;) {
3639         indexOfLastCharChecked = i;
3640         if (i>=textLen) {
3641             break;
3642         }
3643         U16_NEXT(text, i, textLen, c);
3644         gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
3645         if (gcProperty != U_GCB_EXTEND && gcProperty != U_GCB_SPACING_MARK) {
3646             break;
3647         }
3648     }
3649     return indexOfLastCharChecked;
3650 #elif !UCONFIG_NO_BREAK_ITERATION
3651     UBreakIterator *breakiterator = strsrch->search->breakIter;
3652 
3653     if (breakiterator == NULL) {
3654         breakiterator = strsrch->search->internalBreakIter;
3655     }
3656 
3657     if (breakiterator != NULL) {
3658         return ubrk_following(breakiterator, startIndex);
3659     }
3660 
3661     return startIndex;
3662 #else
3663     // **** or should we use the original code? ****
3664     return startIndex;
3665 #endif
3666 
3667 }
3668 
3669 /*
3670  * Returns TRUE if index is on a break boundary. If the UStringSearch
3671  * has an external break iterator, test using that, otherwise test
3672  * using the internal character break iterator.
3673  */
isBreakBoundary(UStringSearch * strsrch,int32_t index)3674 static UBool isBreakBoundary(UStringSearch *strsrch, int32_t index) {
3675 #if 0
3676     const UChar *text = strsrch->search->text;
3677     int32_t textLen   = strsrch->search->textLength;
3678 
3679     U_ASSERT(index>=0);
3680     U_ASSERT(index<=textLen);
3681 
3682     if (index>=textLen || index<=0) {
3683         return TRUE;
3684     }
3685 
3686     // If the character at the current index is not a GRAPHEME_EXTEND
3687     //    then we can not be within a combining sequence.
3688     UChar32  c;
3689     U16_GET(text, 0, index, textLen, c);
3690     int32_t gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
3691     if (gcProperty != U_GCB_EXTEND && gcProperty != U_GCB_SPACING_MARK) {
3692         return TRUE;
3693     }
3694 
3695     // We are at a combining mark.  If the preceding character is anything
3696     //   except a CONTROL, CR or LF, we are in a combining sequence.
3697     U16_PREV(text, 0, index, c);
3698     gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
3699     UBool combining =  !(gcProperty==U_GCB_CONTROL || gcProperty==U_GCB_LF || gcProperty==U_GCB_CR);
3700     return !combining;
3701 #elif !UCONFIG_NO_BREAK_ITERATION
3702     UBreakIterator *breakiterator = strsrch->search->breakIter;
3703 
3704     if (breakiterator == NULL) {
3705         breakiterator = strsrch->search->internalBreakIter;
3706     }
3707 
3708     return (breakiterator != NULL && ubrk_isBoundary(breakiterator, index));
3709 #else
3710     // **** or use the original code? ****
3711     return TRUE;
3712 #endif
3713 }
3714 
3715 #if 0
3716 static UBool onBreakBoundaries(const UStringSearch *strsrch, int32_t start, int32_t end)
3717 {
3718 #if !UCONFIG_NO_BREAK_ITERATION
3719     UBreakIterator *breakiterator = strsrch->search->breakIter;
3720 
3721     if (breakiterator != NULL) {
3722         int32_t startindex = ubrk_first(breakiterator);
3723         int32_t endindex   = ubrk_last(breakiterator);
3724 
3725         // out-of-range indexes are never boundary positions
3726         if (start < startindex || start > endindex ||
3727             end < startindex || end > endindex) {
3728             return FALSE;
3729         }
3730 
3731         return ubrk_isBoundary(breakiterator, start) &&
3732                ubrk_isBoundary(breakiterator, end);
3733     }
3734 #endif
3735 
3736     return TRUE;
3737 }
3738 #endif
3739 
3740 typedef enum {
3741     U_CE_MATCH = -1,
3742     U_CE_NO_MATCH = 0,
3743     U_CE_SKIP_TARG,
3744     U_CE_SKIP_PATN
3745 } UCompareCEsResult;
3746 #define U_CE_LEVEL2_BASE 0x00000005
3747 #define U_CE_LEVEL3_BASE 0x00050000
3748 
compareCE64s(int64_t targCE,int64_t patCE,int16_t compareType)3749 static UCompareCEsResult compareCE64s(int64_t targCE, int64_t patCE, int16_t compareType) {
3750     if (targCE == patCE) {
3751         return U_CE_MATCH;
3752     }
3753     if (compareType == 0) {
3754         return U_CE_NO_MATCH;
3755     }
3756 
3757     int64_t targCEshifted = targCE >> 32;
3758     int64_t patCEshifted = patCE >> 32;
3759     int64_t mask;
3760 
3761     mask = 0xFFFF0000;
3762     int32_t targLev1 = (int32_t)(targCEshifted & mask);
3763     int32_t patLev1 = (int32_t)(patCEshifted & mask);
3764     if ( targLev1 != patLev1 ) {
3765         if ( targLev1 == 0 ) {
3766             return U_CE_SKIP_TARG;
3767         }
3768         if ( patLev1 == 0 && compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD ) {
3769             return U_CE_SKIP_PATN;
3770         }
3771         return U_CE_NO_MATCH;
3772     }
3773 
3774     mask = 0x0000FFFF;
3775     int32_t targLev2 = (int32_t)(targCEshifted & mask);
3776     int32_t patLev2 = (int32_t)(patCEshifted & mask);
3777     if ( targLev2 != patLev2 ) {
3778         if ( targLev2 == 0 ) {
3779             return U_CE_SKIP_TARG;
3780         }
3781         if ( patLev2 == 0 && compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD ) {
3782             return U_CE_SKIP_PATN;
3783         }
3784         return (patLev2 == U_CE_LEVEL2_BASE || (compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD && targLev2 == U_CE_LEVEL2_BASE) )?
3785             U_CE_MATCH: U_CE_NO_MATCH;
3786     }
3787 
3788     mask = 0xFFFF0000;
3789     int32_t targLev3 = (int32_t)(targCE & mask);
3790     int32_t patLev3 = (int32_t)(patCE & mask);
3791     if ( targLev3 != patLev3 ) {
3792         return (patLev3 == U_CE_LEVEL3_BASE || (compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD && targLev3 == U_CE_LEVEL3_BASE) )?
3793             U_CE_MATCH: U_CE_NO_MATCH;
3794    }
3795 
3796     return U_CE_MATCH;
3797 }
3798 
3799 #if BOYER_MOORE
3800 // TODO: #if BOYER_MOORE, need 32-bit version of compareCE64s
3801 #endif
3802 
usearch_search(UStringSearch * strsrch,int32_t startIdx,int32_t * matchStart,int32_t * matchLimit,UErrorCode * status)3803 U_CAPI UBool U_EXPORT2 usearch_search(UStringSearch  *strsrch,
3804                                        int32_t        startIdx,
3805                                        int32_t        *matchStart,
3806                                        int32_t        *matchLimit,
3807                                        UErrorCode     *status)
3808 {
3809     if (U_FAILURE(*status)) {
3810         return FALSE;
3811     }
3812 
3813     // TODO:  reject search patterns beginning with a combining char.
3814 
3815 #ifdef USEARCH_DEBUG
3816     if (getenv("USEARCH_DEBUG") != NULL) {
3817         printf("Pattern CEs\n");
3818         for (int ii=0; ii<strsrch->pattern.CELength; ii++) {
3819             printf(" %8x", strsrch->pattern.CE[ii]);
3820         }
3821         printf("\n");
3822     }
3823 
3824 #endif
3825     // Input parameter sanity check.
3826     //  TODO:  should input indicies clip to the text length
3827     //         in the same way that UText does.
3828     if(strsrch->pattern.CELength == 0         ||
3829        startIdx < 0                           ||
3830        startIdx > strsrch->search->textLength ||
3831        strsrch->pattern.CE == NULL) {
3832            *status = U_ILLEGAL_ARGUMENT_ERROR;
3833            return FALSE;
3834     }
3835 
3836     if (strsrch->pattern.PCE == NULL) {
3837         initializePatternPCETable(strsrch, status);
3838     }
3839 
3840     ucol_setOffset(strsrch->textIter, startIdx, status);
3841     CEBuffer ceb(strsrch, status);
3842 
3843 
3844     int32_t    targetIx = 0;
3845     const CEI *targetCEI = NULL;
3846     int32_t    patIx;
3847     UBool      found;
3848 
3849     int32_t  mStart = -1;
3850     int32_t  mLimit = -1;
3851     int32_t  minLimit;
3852     int32_t  maxLimit;
3853 
3854 
3855 
3856     // Outer loop moves over match starting positions in the
3857     //      target CE space.
3858     // Here we see the target as a sequence of collation elements, resulting from the following:
3859     // 1. Target characters were decomposed, and (if appropriate) other compressions and expansions are applied
3860     //    (for example, digraphs such as IJ may be broken into two characters).
3861     // 2. An int64_t CE weight is determined for each resulting unit (high 16 bits are primary strength, next
3862     //    16 bits are secondary, next 16 (the high 16 bits of the low 32-bit half) are tertiary. Any of these
3863     //    fields that are for strengths below that of the collator are set to 0. If this makes the int64_t
3864     //    CE weight 0 (as for a combining diacritic with secondary weight when the collator strentgh is primary),
3865     //    then the CE is deleted, so the following code sees only CEs that are relevant.
3866     // For each CE, the lowIndex and highIndex correspond to where this CE begins and ends in the original text.
3867     // If lowIndex==highIndex, either the CE resulted from an expansion/decomposition of one of the original text
3868     // characters, or the CE marks the limit of the target text (in which case the CE weight is UCOL_PROCESSED_NULLORDER).
3869     //
3870     for(targetIx=0; ; targetIx++)
3871     {
3872         found = TRUE;
3873         //  Inner loop checks for a match beginning at each
3874         //  position from the outer loop.
3875         int32_t targetIxOffset = 0;
3876         int64_t patCE = 0;
3877         // For targetIx > 0, this ceb.get gets a CE that is as far back in the ring buffer
3878         // (compared to the last CE fetched for the previous targetIx value) as we need to go
3879         // for this targetIx value, so if it is non-NULL then other ceb.get calls should be OK.
3880         const CEI *firstCEI = ceb.get(targetIx);
3881         if (firstCEI == NULL) {
3882             *status = U_INTERNAL_PROGRAM_ERROR;
3883             found = FALSE;
3884             break;
3885         }
3886 
3887         for (patIx=0; patIx<strsrch->pattern.PCELength; patIx++) {
3888             patCE = strsrch->pattern.PCE[patIx];
3889             targetCEI = ceb.get(targetIx+patIx+targetIxOffset);
3890             //  Compare CE from target string with CE from the pattern.
3891             //    Note that the target CE will be UCOL_PROCESSED_NULLORDER if we reach the end of input,
3892             //    which will fail the compare, below.
3893             UCompareCEsResult ceMatch = compareCE64s(targetCEI->ce, patCE, strsrch->search->elementComparisonType);
3894             if ( ceMatch == U_CE_NO_MATCH ) {
3895                 found = FALSE;
3896                 break;
3897             } else if ( ceMatch > U_CE_NO_MATCH ) {
3898                 if ( ceMatch == U_CE_SKIP_TARG ) {
3899                     // redo with same patCE, next targCE
3900                     patIx--;
3901                     targetIxOffset++;
3902                 } else { // ceMatch == U_CE_SKIP_PATN
3903                     // redo with same targCE, next patCE
3904                     targetIxOffset--;
3905                 }
3906             }
3907         }
3908         targetIxOffset += strsrch->pattern.PCELength; // this is now the offset in target CE space to end of the match so far
3909 
3910         if (!found && ((targetCEI == NULL) || (targetCEI->ce != UCOL_PROCESSED_NULLORDER))) {
3911             // No match at this targetIx.  Try again at the next.
3912             continue;
3913         }
3914 
3915         if (!found) {
3916             // No match at all, we have run off the end of the target text.
3917             break;
3918         }
3919 
3920 
3921         // We have found a match in CE space.
3922         // Now determine the bounds in string index space.
3923         //  There still is a chance of match failure if the CE range not correspond to
3924         //     an acceptable character range.
3925         //
3926         const CEI *lastCEI  = ceb.get(targetIx + targetIxOffset - 1);
3927 
3928         mStart   = firstCEI->lowIndex;
3929         minLimit = lastCEI->lowIndex;
3930 
3931         // Look at the CE following the match.  If it is UCOL_NULLORDER the match
3932         //   extended to the end of input, and the match is good.
3933 
3934         // Look at the high and low indices of the CE following the match. If
3935         // they are the same it means one of two things:
3936         //    1. The match extended to the last CE from the target text, which is OK, or
3937         //    2. The last CE that was part of the match is in an expansion that extends
3938         //       to the first CE after the match. In this case, we reject the match.
3939         const CEI *nextCEI = 0;
3940         if (strsrch->search->elementComparisonType == 0) {
3941             nextCEI  = ceb.get(targetIx + targetIxOffset);
3942             maxLimit = nextCEI->lowIndex;
3943             if (nextCEI->lowIndex == nextCEI->highIndex && nextCEI->ce != UCOL_PROCESSED_NULLORDER) {
3944                 found = FALSE;
3945             }
3946         } else {
3947             for ( ; ; ++targetIxOffset ) {
3948                 nextCEI = ceb.get(targetIx + targetIxOffset);
3949                 maxLimit = nextCEI->lowIndex;
3950                 // If we are at the end of the target too, match succeeds
3951                 if (  nextCEI->ce == UCOL_PROCESSED_NULLORDER ) {
3952                     break;
3953                 }
3954                 // As long as the next CE has primary weight of 0,
3955                 // it is part of the last target element matched by the pattern;
3956                 // make sure it can be part of a match with the last patCE
3957                 if ( (((nextCEI->ce) >> 32) & 0xFFFF0000UL) == 0 ) {
3958                     UCompareCEsResult ceMatch = compareCE64s(nextCEI->ce, patCE, strsrch->search->elementComparisonType);
3959                     if ( ceMatch == U_CE_NO_MATCH || ceMatch == U_CE_SKIP_PATN ) {
3960                         found = FALSE;
3961                         break;
3962                     }
3963                 // If lowIndex == highIndex, this target CE is part of an expansion of the last matched
3964                 // target element, but it has non-zero primary weight => match fails
3965                 } else if ( nextCEI->lowIndex == nextCEI->highIndex ) {
3966                     found = false;
3967                     break;
3968                 // Else the target CE is not part of an expansion of the last matched element, match succeeds
3969                 } else {
3970                     break;
3971                 }
3972             }
3973         }
3974 
3975 
3976         // Check for the start of the match being within a combining sequence.
3977         //   This can happen if the pattern itself begins with a combining char, and
3978         //   the match found combining marks in the target text that were attached
3979         //    to something else.
3980         //   This type of match should be rejected for not completely consuming a
3981         //   combining sequence.
3982         if (!isBreakBoundary(strsrch, mStart)) {
3983             found = FALSE;
3984         }
3985 
3986         // Check for the start of the match being within an Collation Element Expansion,
3987         //   meaning that the first char of the match is only partially matched.
3988         //   With exapnsions, the first CE will report the index of the source
3989         //   character, and all subsequent (expansions) CEs will report the source index of the
3990         //    _following_ character.
3991         int32_t secondIx = firstCEI->highIndex;
3992         if (mStart == secondIx) {
3993             found = FALSE;
3994         }
3995 
3996         //  Advance the match end position to the first acceptable match boundary.
3997         //    This advances the index over any combining charcters.
3998         mLimit = maxLimit;
3999         if (minLimit < maxLimit) {
4000             // When the last CE's low index is same with its high index, the CE is likely
4001             // a part of expansion. In this case, the index is located just after the
4002             // character corresponding to the CEs compared above. If the index is right
4003             // at the break boundary, move the position to the next boundary will result
4004             // incorrect match length when there are ignorable characters exist between
4005             // the position and the next character produces CE(s). See ticket#8482.
4006             if (minLimit == lastCEI->highIndex && isBreakBoundary(strsrch, minLimit)) {
4007                 mLimit = minLimit;
4008             } else {
4009                 int32_t nba = nextBoundaryAfter(strsrch, minLimit);
4010                 if (nba >= lastCEI->highIndex) {
4011                     mLimit = nba;
4012                 }
4013             }
4014         }
4015 
4016     #ifdef USEARCH_DEBUG
4017         if (getenv("USEARCH_DEBUG") != NULL) {
4018             printf("minLimit, maxLimit, mLimit = %d, %d, %d\n", minLimit, maxLimit, mLimit);
4019         }
4020     #endif
4021 
4022         // If advancing to the end of a combining sequence in character indexing space
4023         //   advanced us beyond the end of the match in CE space, reject this match.
4024         if (mLimit > maxLimit) {
4025             found = FALSE;
4026         }
4027 
4028         if (!isBreakBoundary(strsrch, mLimit)) {
4029             found = FALSE;
4030         }
4031 
4032         if (! checkIdentical(strsrch, mStart, mLimit)) {
4033             found = FALSE;
4034         }
4035 
4036         if (found) {
4037             break;
4038         }
4039     }
4040 
4041     #ifdef USEARCH_DEBUG
4042     if (getenv("USEARCH_DEBUG") != NULL) {
4043         printf("Target CEs [%d .. %d]\n", ceb.firstIx, ceb.limitIx);
4044         int32_t  lastToPrint = ceb.limitIx+2;
4045         for (int ii=ceb.firstIx; ii<lastToPrint; ii++) {
4046             printf("%8x@%d ", ceb.get(ii)->ce, ceb.get(ii)->srcIndex);
4047         }
4048         printf("\n%s\n", found? "match found" : "no match");
4049     }
4050     #endif
4051 
4052     // All Done.  Store back the match bounds to the caller.
4053     //
4054     if (found==FALSE) {
4055         mLimit = -1;
4056         mStart = -1;
4057     }
4058 
4059     if (matchStart != NULL) {
4060         *matchStart= mStart;
4061     }
4062 
4063     if (matchLimit != NULL) {
4064         *matchLimit = mLimit;
4065     }
4066 
4067     return found;
4068 }
4069 
usearch_searchBackwards(UStringSearch * strsrch,int32_t startIdx,int32_t * matchStart,int32_t * matchLimit,UErrorCode * status)4070 U_CAPI UBool U_EXPORT2 usearch_searchBackwards(UStringSearch  *strsrch,
4071                                                 int32_t        startIdx,
4072                                                 int32_t        *matchStart,
4073                                                 int32_t        *matchLimit,
4074                                                 UErrorCode     *status)
4075 {
4076     if (U_FAILURE(*status)) {
4077         return FALSE;
4078     }
4079 
4080     // TODO:  reject search patterns beginning with a combining char.
4081 
4082 #ifdef USEARCH_DEBUG
4083     if (getenv("USEARCH_DEBUG") != NULL) {
4084         printf("Pattern CEs\n");
4085         for (int ii=0; ii<strsrch->pattern.CELength; ii++) {
4086             printf(" %8x", strsrch->pattern.CE[ii]);
4087         }
4088         printf("\n");
4089     }
4090 
4091 #endif
4092     // Input parameter sanity check.
4093     //  TODO:  should input indicies clip to the text length
4094     //         in the same way that UText does.
4095     if(strsrch->pattern.CELength == 0         ||
4096        startIdx < 0                           ||
4097        startIdx > strsrch->search->textLength ||
4098        strsrch->pattern.CE == NULL) {
4099            *status = U_ILLEGAL_ARGUMENT_ERROR;
4100            return FALSE;
4101     }
4102 
4103     if (strsrch->pattern.PCE == NULL) {
4104         initializePatternPCETable(strsrch, status);
4105     }
4106 
4107     CEBuffer ceb(strsrch, status);
4108     int32_t    targetIx = 0;
4109 
4110     /*
4111      * Pre-load the buffer with the CE's for the grapheme
4112      * after our starting position so that we're sure that
4113      * we can look at the CE following the match when we
4114      * check the match boundaries.
4115      *
4116      * This will also pre-fetch the first CE that we'll
4117      * consider for the match.
4118      */
4119     if (startIdx < strsrch->search->textLength) {
4120         UBreakIterator *bi = strsrch->search->internalBreakIter;
4121         int32_t next = ubrk_following(bi, startIdx);
4122 
4123         ucol_setOffset(strsrch->textIter, next, status);
4124 
4125         for (targetIx = 0; ; targetIx += 1) {
4126             if (ceb.getPrevious(targetIx)->lowIndex < startIdx) {
4127                 break;
4128             }
4129         }
4130     } else {
4131         ucol_setOffset(strsrch->textIter, startIdx, status);
4132     }
4133 
4134 
4135     const CEI *targetCEI = NULL;
4136     int32_t    patIx;
4137     UBool      found;
4138 
4139     int32_t  limitIx = targetIx;
4140     int32_t  mStart = -1;
4141     int32_t  mLimit = -1;
4142     int32_t  minLimit;
4143     int32_t  maxLimit;
4144 
4145 
4146 
4147     // Outer loop moves over match starting positions in the
4148     //      target CE space.
4149     // Here, targetIx values increase toward the beginning of the base text (i.e. we get the text CEs in reverse order).
4150     // But  patIx is 0 at the beginning of the pattern and increases toward the end.
4151     // So this loop performs a comparison starting with the end of pattern, and prcessd toward the beginning of the pattern
4152     // and the beginning of the base text.
4153     for(targetIx = limitIx; ; targetIx += 1)
4154     {
4155         found = TRUE;
4156         // For targetIx > limitIx, this ceb.getPrevious gets a CE that is as far back in the ring buffer
4157         // (compared to the last CE fetched for the previous targetIx value) as we need to go
4158         // for this targetIx value, so if it is non-NULL then other ceb.getPrevious calls should be OK.
4159         const CEI *lastCEI  = ceb.getPrevious(targetIx);
4160         if (lastCEI == NULL) {
4161             *status = U_INTERNAL_PROGRAM_ERROR;
4162             found = FALSE;
4163              break;
4164         }
4165         //  Inner loop checks for a match beginning at each
4166         //  position from the outer loop.
4167         int32_t targetIxOffset = 0;
4168         for (patIx = strsrch->pattern.PCELength - 1; patIx >= 0; patIx -= 1) {
4169             int64_t patCE = strsrch->pattern.PCE[patIx];
4170 
4171             targetCEI = ceb.getPrevious(targetIx + strsrch->pattern.PCELength - 1 - patIx + targetIxOffset);
4172             //  Compare CE from target string with CE from the pattern.
4173             //    Note that the target CE will be UCOL_NULLORDER if we reach the end of input,
4174             //    which will fail the compare, below.
4175             UCompareCEsResult ceMatch = compareCE64s(targetCEI->ce, patCE, strsrch->search->elementComparisonType);
4176             if ( ceMatch == U_CE_NO_MATCH ) {
4177                 found = FALSE;
4178                 break;
4179             } else if ( ceMatch > U_CE_NO_MATCH ) {
4180                 if ( ceMatch == U_CE_SKIP_TARG ) {
4181                     // redo with same patCE, next targCE
4182                     patIx++;
4183                     targetIxOffset++;
4184                 } else { // ceMatch == U_CE_SKIP_PATN
4185                     // redo with same targCE, next patCE
4186                     targetIxOffset--;
4187                 }
4188             }
4189         }
4190 
4191         if (!found && ((targetCEI == NULL) || (targetCEI->ce != UCOL_PROCESSED_NULLORDER))) {
4192             // No match at this targetIx.  Try again at the next.
4193             continue;
4194         }
4195 
4196         if (!found) {
4197             // No match at all, we have run off the end of the target text.
4198             break;
4199         }
4200 
4201 
4202         // We have found a match in CE space.
4203         // Now determine the bounds in string index space.
4204         //  There still is a chance of match failure if the CE range not correspond to
4205         //     an acceptable character range.
4206         //
4207         const CEI *firstCEI = ceb.getPrevious(targetIx + strsrch->pattern.PCELength - 1 + targetIxOffset);
4208         mStart   = firstCEI->lowIndex;
4209 
4210         // Check for the start of the match being within a combining sequence.
4211         //   This can happen if the pattern itself begins with a combining char, and
4212         //   the match found combining marks in the target text that were attached
4213         //    to something else.
4214         //   This type of match should be rejected for not completely consuming a
4215         //   combining sequence.
4216         if (!isBreakBoundary(strsrch, mStart)) {
4217             found = FALSE;
4218         }
4219 
4220         // Look at the high index of the first CE in the match. If it's the same as the
4221         // low index, the first CE in the match is in the middle of an expansion.
4222         if (mStart == firstCEI->highIndex) {
4223             found = FALSE;
4224         }
4225 
4226 
4227         minLimit = lastCEI->lowIndex;
4228 
4229         if (targetIx > 0) {
4230             // Look at the CE following the match.  If it is UCOL_NULLORDER the match
4231             //   extended to the end of input, and the match is good.
4232 
4233             // Look at the high and low indices of the CE following the match. If
4234             // they are the same it means one of two things:
4235             //    1. The match extended to the last CE from the target text, which is OK, or
4236             //    2. The last CE that was part of the match is in an expansion that extends
4237             //       to the first CE after the match. In this case, we reject the match.
4238             const CEI *nextCEI  = ceb.getPrevious(targetIx - 1);
4239 
4240             if (nextCEI->lowIndex == nextCEI->highIndex && nextCEI->ce != UCOL_PROCESSED_NULLORDER) {
4241                 found = FALSE;
4242             }
4243 
4244             mLimit = maxLimit = nextCEI->lowIndex;
4245 
4246             //  Advance the match end position to the first acceptable match boundary.
4247             //    This advances the index over any combining charcters.
4248             if (minLimit < maxLimit) {
4249                 int32_t nba = nextBoundaryAfter(strsrch, minLimit);
4250 
4251                 if (nba >= lastCEI->highIndex) {
4252                     mLimit = nba;
4253                 }
4254             }
4255 
4256             // If advancing to the end of a combining sequence in character indexing space
4257             //   advanced us beyond the end of the match in CE space, reject this match.
4258             if (mLimit > maxLimit) {
4259                 found = FALSE;
4260             }
4261 
4262             // Make sure the end of the match is on a break boundary
4263             if (!isBreakBoundary(strsrch, mLimit)) {
4264                 found = FALSE;
4265             }
4266 
4267         } else {
4268             // No non-ignorable CEs after this point.
4269             // The maximum position is detected by boundary after
4270             // the last non-ignorable CE. Combining sequence
4271             // across the start index will be truncated.
4272             int32_t nba = nextBoundaryAfter(strsrch, minLimit);
4273             mLimit = maxLimit = (nba > 0) && (startIdx > nba) ? nba : startIdx;
4274         }
4275 
4276     #ifdef USEARCH_DEBUG
4277         if (getenv("USEARCH_DEBUG") != NULL) {
4278             printf("minLimit, maxLimit, mLimit = %d, %d, %d\n", minLimit, maxLimit, mLimit);
4279         }
4280     #endif
4281 
4282 
4283         if (! checkIdentical(strsrch, mStart, mLimit)) {
4284             found = FALSE;
4285         }
4286 
4287         if (found) {
4288             break;
4289         }
4290     }
4291 
4292     #ifdef USEARCH_DEBUG
4293     if (getenv("USEARCH_DEBUG") != NULL) {
4294         printf("Target CEs [%d .. %d]\n", ceb.firstIx, ceb.limitIx);
4295         int32_t  lastToPrint = ceb.limitIx+2;
4296         for (int ii=ceb.firstIx; ii<lastToPrint; ii++) {
4297             printf("%8x@%d ", ceb.get(ii)->ce, ceb.get(ii)->srcIndex);
4298         }
4299         printf("\n%s\n", found? "match found" : "no match");
4300     }
4301     #endif
4302 
4303     // All Done.  Store back the match bounds to the caller.
4304     //
4305     if (found==FALSE) {
4306         mLimit = -1;
4307         mStart = -1;
4308     }
4309 
4310     if (matchStart != NULL) {
4311         *matchStart= mStart;
4312     }
4313 
4314     if (matchLimit != NULL) {
4315         *matchLimit = mLimit;
4316     }
4317 
4318     return found;
4319 }
4320 
4321 // internal use methods declared in usrchimp.h -----------------------------
4322 
usearch_handleNextExact(UStringSearch * strsrch,UErrorCode * status)4323 UBool usearch_handleNextExact(UStringSearch *strsrch, UErrorCode *status)
4324 {
4325     if (U_FAILURE(*status)) {
4326         setMatchNotFound(strsrch);
4327         return FALSE;
4328     }
4329 
4330 #if BOYER_MOORE
4331     UCollationElements *coleiter        = strsrch->textIter;
4332     int32_t             textlength      = strsrch->search->textLength;
4333     int32_t            *patternce       = strsrch->pattern.CE;
4334     int32_t             patterncelength = strsrch->pattern.CELength;
4335     int32_t             textoffset      = ucol_getOffset(coleiter);
4336 
4337     // status used in setting coleiter offset, since offset is checked in
4338     // shiftForward before setting the coleiter offset, status never
4339     // a failure
4340     textoffset = shiftForward(strsrch, textoffset, UCOL_NULLORDER,
4341                               patterncelength);
4342     while (textoffset <= textlength)
4343     {
4344         uint32_t    patternceindex = patterncelength - 1;
4345         int32_t     targetce;
4346         UBool       found          = FALSE;
4347         int32_t    lastce          = UCOL_NULLORDER;
4348 
4349         setColEIterOffset(coleiter, textoffset);
4350 
4351         for (;;) {
4352             // finding the last pattern ce match, imagine composite characters
4353             // for example: search for pattern A in text \u00C0
4354             // we'll have to skip \u0300 the grave first before we get to A
4355             targetce = ucol_previous(coleiter, status);
4356             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4357                 found = FALSE;
4358                 break;
4359             }
4360             targetce = getCE(strsrch, targetce);
4361             if (targetce == UCOL_IGNORABLE && inNormBuf(coleiter)) {
4362                 // this is for the text \u0315\u0300 that requires
4363                 // normalization and pattern \u0300, where \u0315 is ignorable
4364                 continue;
4365             }
4366             if (lastce == UCOL_NULLORDER || lastce == UCOL_IGNORABLE) {
4367                 lastce = targetce;
4368             }
4369             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4370             if (targetce == patternce[patternceindex]) {
4371                 // the first ce can be a contraction
4372                 found = TRUE;
4373                 break;
4374             }
4375             if (!hasExpansion(coleiter)) {
4376                 found = FALSE;
4377                 break;
4378             }
4379         }
4380 
4381         //targetce = lastce;
4382 
4383         while (found && patternceindex > 0) {
4384             lastce = targetce;
4385             targetce    = ucol_previous(coleiter, status);
4386             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4387                 found = FALSE;
4388                 break;
4389             }
4390             targetce    = getCE(strsrch, targetce);
4391             if (targetce == UCOL_IGNORABLE) {
4392                 continue;
4393             }
4394 
4395             patternceindex --;
4396             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4397             found = found && targetce == patternce[patternceindex];
4398         }
4399 
4400         targetce = lastce;
4401 
4402         if (!found) {
4403             if (U_FAILURE(*status)) {
4404                 break;
4405             }
4406             textoffset = shiftForward(strsrch, textoffset, lastce,
4407                                       patternceindex);
4408             // status checked at loop.
4409             patternceindex = patterncelength;
4410             continue;
4411         }
4412 
4413         if (checkNextExactMatch(strsrch, &textoffset, status)) {
4414             // status checked in ucol_setOffset
4415             setColEIterOffset(coleiter, strsrch->search->matchedIndex);
4416             return TRUE;
4417         }
4418     }
4419     setMatchNotFound(strsrch);
4420     return FALSE;
4421 #else
4422     int32_t textOffset = ucol_getOffset(strsrch->textIter);
4423     int32_t start = -1;
4424     int32_t end = -1;
4425 
4426     if (usearch_search(strsrch, textOffset, &start, &end, status)) {
4427         strsrch->search->matchedIndex  = start;
4428         strsrch->search->matchedLength = end - start;
4429         return TRUE;
4430     } else {
4431         setMatchNotFound(strsrch);
4432         return FALSE;
4433     }
4434 #endif
4435 }
4436 
usearch_handleNextCanonical(UStringSearch * strsrch,UErrorCode * status)4437 UBool usearch_handleNextCanonical(UStringSearch *strsrch, UErrorCode *status)
4438 {
4439     if (U_FAILURE(*status)) {
4440         setMatchNotFound(strsrch);
4441         return FALSE;
4442     }
4443 
4444 #if BOYER_MOORE
4445     UCollationElements *coleiter        = strsrch->textIter;
4446     int32_t             textlength      = strsrch->search->textLength;
4447     int32_t            *patternce       = strsrch->pattern.CE;
4448     int32_t             patterncelength = strsrch->pattern.CELength;
4449     int32_t             textoffset      = ucol_getOffset(coleiter);
4450     UBool               hasPatternAccents =
4451        strsrch->pattern.hasSuffixAccents || strsrch->pattern.hasPrefixAccents;
4452 
4453     textoffset = shiftForward(strsrch, textoffset, UCOL_NULLORDER,
4454                               patterncelength);
4455     strsrch->canonicalPrefixAccents[0] = 0;
4456     strsrch->canonicalSuffixAccents[0] = 0;
4457 
4458     while (textoffset <= textlength)
4459     {
4460         int32_t     patternceindex = patterncelength - 1;
4461         int32_t     targetce;
4462         UBool       found          = FALSE;
4463         int32_t     lastce         = UCOL_NULLORDER;
4464 
4465         setColEIterOffset(coleiter, textoffset);
4466 
4467         for (;;) {
4468             // finding the last pattern ce match, imagine composite characters
4469             // for example: search for pattern A in text \u00C0
4470             // we'll have to skip \u0300 the grave first before we get to A
4471             targetce = ucol_previous(coleiter, status);
4472             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4473                 found = FALSE;
4474                 break;
4475             }
4476             targetce = getCE(strsrch, targetce);
4477             if (lastce == UCOL_NULLORDER || lastce == UCOL_IGNORABLE) {
4478                 lastce = targetce;
4479             }
4480             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4481             if (targetce == patternce[patternceindex]) {
4482                 // the first ce can be a contraction
4483                 found = TRUE;
4484                 break;
4485             }
4486             if (!hasExpansion(coleiter)) {
4487                 found = FALSE;
4488                 break;
4489             }
4490         }
4491 
4492         while (found && patternceindex > 0) {
4493             targetce    = ucol_previous(coleiter, status);
4494             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4495                 found = FALSE;
4496                 break;
4497             }
4498             targetce    = getCE(strsrch, targetce);
4499             if (targetce == UCOL_IGNORABLE) {
4500                 continue;
4501             }
4502 
4503             patternceindex --;
4504             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4505             found = found && targetce == patternce[patternceindex];
4506         }
4507 
4508         // initializing the rearranged accent array
4509         if (hasPatternAccents && !found) {
4510             strsrch->canonicalPrefixAccents[0] = 0;
4511             strsrch->canonicalSuffixAccents[0] = 0;
4512             if (U_FAILURE(*status)) {
4513                 break;
4514             }
4515             found = doNextCanonicalMatch(strsrch, textoffset, status);
4516         }
4517 
4518         if (!found) {
4519             if (U_FAILURE(*status)) {
4520                 break;
4521             }
4522             textoffset = shiftForward(strsrch, textoffset, lastce,
4523                                       patternceindex);
4524             // status checked at loop
4525             patternceindex = patterncelength;
4526             continue;
4527         }
4528 
4529         if (checkNextCanonicalMatch(strsrch, &textoffset, status)) {
4530             setColEIterOffset(coleiter, strsrch->search->matchedIndex);
4531             return TRUE;
4532         }
4533     }
4534     setMatchNotFound(strsrch);
4535     return FALSE;
4536 #else
4537     int32_t textOffset = ucol_getOffset(strsrch->textIter);
4538     int32_t start = -1;
4539     int32_t end = -1;
4540 
4541     if (usearch_search(strsrch, textOffset, &start, &end, status)) {
4542         strsrch->search->matchedIndex  = start;
4543         strsrch->search->matchedLength = end - start;
4544         return TRUE;
4545     } else {
4546         setMatchNotFound(strsrch);
4547         return FALSE;
4548     }
4549 #endif
4550 }
4551 
usearch_handlePreviousExact(UStringSearch * strsrch,UErrorCode * status)4552 UBool usearch_handlePreviousExact(UStringSearch *strsrch, UErrorCode *status)
4553 {
4554     if (U_FAILURE(*status)) {
4555         setMatchNotFound(strsrch);
4556         return FALSE;
4557     }
4558 
4559 #if BOYER_MOORE
4560     UCollationElements *coleiter        = strsrch->textIter;
4561     int32_t            *patternce       = strsrch->pattern.CE;
4562     int32_t             patterncelength = strsrch->pattern.CELength;
4563     int32_t             textoffset      = ucol_getOffset(coleiter);
4564 
4565     // shifting it check for setting offset
4566     // if setOffset is called previously or there was no previous match, we
4567     // leave the offset as it is.
4568     if (strsrch->search->matchedIndex != USEARCH_DONE) {
4569         textoffset = strsrch->search->matchedIndex;
4570     }
4571 
4572     textoffset = reverseShift(strsrch, textoffset, UCOL_NULLORDER,
4573                               patterncelength);
4574 
4575     while (textoffset >= 0)
4576     {
4577         int32_t     patternceindex = 1;
4578         int32_t     targetce;
4579         UBool       found          = FALSE;
4580         int32_t     firstce        = UCOL_NULLORDER;
4581 
4582         // if status is a failure, ucol_setOffset does nothing
4583         setColEIterOffset(coleiter, textoffset);
4584 
4585         for (;;) {
4586             // finding the first pattern ce match, imagine composite
4587             // characters. for example: search for pattern \u0300 in text
4588             // \u00C0, we'll have to skip A first before we get to
4589             // \u0300 the grave accent
4590             targetce = ucol_next(coleiter, status);
4591             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4592                 found = FALSE;
4593                 break;
4594             }
4595             targetce = getCE(strsrch, targetce);
4596             if (firstce == UCOL_NULLORDER || firstce == UCOL_IGNORABLE) {
4597                 firstce = targetce;
4598             }
4599             if (targetce == UCOL_IGNORABLE && strsrch->strength != UCOL_PRIMARY) {
4600                 continue;
4601             }
4602             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4603             if (targetce == patternce[0]) {
4604                 found = TRUE;
4605                 break;
4606             }
4607             if (!hasExpansion(coleiter)) {
4608                 // checking for accents in composite character
4609                 found = FALSE;
4610                 break;
4611             }
4612         }
4613 
4614         //targetce = firstce;
4615 
4616         while (found && (patternceindex < patterncelength)) {
4617             firstce = targetce;
4618             targetce    = ucol_next(coleiter, status);
4619             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4620                 found = FALSE;
4621                 break;
4622             }
4623             targetce    = getCE(strsrch, targetce);
4624             if (targetce == UCOL_IGNORABLE) {
4625                 continue;
4626             }
4627 
4628             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4629             found = found && targetce == patternce[patternceindex];
4630             patternceindex ++;
4631         }
4632 
4633         targetce = firstce;
4634 
4635         if (!found) {
4636             if (U_FAILURE(*status)) {
4637                 break;
4638             }
4639 
4640             textoffset = reverseShift(strsrch, textoffset, targetce,
4641                                       patternceindex);
4642             patternceindex = 0;
4643             continue;
4644         }
4645 
4646         if (checkPreviousExactMatch(strsrch, &textoffset, status)) {
4647             setColEIterOffset(coleiter, textoffset);
4648             return TRUE;
4649         }
4650     }
4651     setMatchNotFound(strsrch);
4652     return FALSE;
4653 #else
4654     int32_t textOffset;
4655 
4656     if (strsrch->search->isOverlap) {
4657         if (strsrch->search->matchedIndex != USEARCH_DONE) {
4658             textOffset = strsrch->search->matchedIndex + strsrch->search->matchedLength - 1;
4659         } else {
4660             // move the start position at the end of possible match
4661             initializePatternPCETable(strsrch, status);
4662             for (int32_t nPCEs = 0; nPCEs < strsrch->pattern.PCELength - 1; nPCEs++) {
4663                 int64_t pce = ucol_nextProcessed(strsrch->textIter, NULL, NULL, status);
4664                 if (pce == UCOL_PROCESSED_NULLORDER) {
4665                     // at the end of the text
4666                     break;
4667                 }
4668             }
4669             if (U_FAILURE(*status)) {
4670                 setMatchNotFound(strsrch);
4671                 return FALSE;
4672             }
4673             textOffset = ucol_getOffset(strsrch->textIter);
4674         }
4675     } else {
4676         textOffset = ucol_getOffset(strsrch->textIter);
4677     }
4678 
4679     int32_t start = -1;
4680     int32_t end = -1;
4681 
4682     if (usearch_searchBackwards(strsrch, textOffset, &start, &end, status)) {
4683         strsrch->search->matchedIndex = start;
4684         strsrch->search->matchedLength = end - start;
4685         return TRUE;
4686     } else {
4687         setMatchNotFound(strsrch);
4688         return FALSE;
4689     }
4690 #endif
4691 }
4692 
usearch_handlePreviousCanonical(UStringSearch * strsrch,UErrorCode * status)4693 UBool usearch_handlePreviousCanonical(UStringSearch *strsrch,
4694                                       UErrorCode    *status)
4695 {
4696     if (U_FAILURE(*status)) {
4697         setMatchNotFound(strsrch);
4698         return FALSE;
4699     }
4700 
4701 #if BOYER_MOORE
4702     UCollationElements *coleiter        = strsrch->textIter;
4703     int32_t            *patternce       = strsrch->pattern.CE;
4704     int32_t             patterncelength = strsrch->pattern.CELength;
4705     int32_t             textoffset      = ucol_getOffset(coleiter);
4706     UBool               hasPatternAccents =
4707        strsrch->pattern.hasSuffixAccents || strsrch->pattern.hasPrefixAccents;
4708 
4709     // shifting it check for setting offset
4710     // if setOffset is called previously or there was no previous match, we
4711     // leave the offset as it is.
4712     if (strsrch->search->matchedIndex != USEARCH_DONE) {
4713         textoffset = strsrch->search->matchedIndex;
4714     }
4715 
4716     textoffset = reverseShift(strsrch, textoffset, UCOL_NULLORDER,
4717                               patterncelength);
4718     strsrch->canonicalPrefixAccents[0] = 0;
4719     strsrch->canonicalSuffixAccents[0] = 0;
4720 
4721     while (textoffset >= 0)
4722     {
4723         int32_t     patternceindex = 1;
4724         int32_t     targetce;
4725         UBool       found          = FALSE;
4726         int32_t     firstce        = UCOL_NULLORDER;
4727 
4728         setColEIterOffset(coleiter, textoffset);
4729         for (;;) {
4730             // finding the first pattern ce match, imagine composite
4731             // characters. for example: search for pattern \u0300 in text
4732             // \u00C0, we'll have to skip A first before we get to
4733             // \u0300 the grave accent
4734             targetce = ucol_next(coleiter, status);
4735             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4736                 found = FALSE;
4737                 break;
4738             }
4739             targetce = getCE(strsrch, targetce);
4740             if (firstce == UCOL_NULLORDER || firstce == UCOL_IGNORABLE) {
4741                 firstce = targetce;
4742             }
4743 
4744             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4745             if (targetce == patternce[0]) {
4746                 // the first ce can be a contraction
4747                 found = TRUE;
4748                 break;
4749             }
4750             if (!hasExpansion(coleiter)) {
4751                 // checking for accents in composite character
4752                 found = FALSE;
4753                 break;
4754             }
4755         }
4756 
4757         targetce = firstce;
4758 
4759         while (found && patternceindex < patterncelength) {
4760             targetce    = ucol_next(coleiter, status);
4761             if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
4762                 found = FALSE;
4763                 break;
4764             }
4765             targetce = getCE(strsrch, targetce);
4766             if (targetce == UCOL_IGNORABLE) {
4767                 continue;
4768             }
4769 
4770             // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
4771             found = found && targetce == patternce[patternceindex];
4772             patternceindex ++;
4773         }
4774 
4775         // initializing the rearranged accent array
4776         if (hasPatternAccents && !found) {
4777             strsrch->canonicalPrefixAccents[0] = 0;
4778             strsrch->canonicalSuffixAccents[0] = 0;
4779             if (U_FAILURE(*status)) {
4780                 break;
4781             }
4782             found = doPreviousCanonicalMatch(strsrch, textoffset, status);
4783         }
4784 
4785         if (!found) {
4786             if (U_FAILURE(*status)) {
4787                 break;
4788             }
4789             textoffset = reverseShift(strsrch, textoffset, targetce,
4790                                       patternceindex);
4791             patternceindex = 0;
4792             continue;
4793         }
4794 
4795         if (checkPreviousCanonicalMatch(strsrch, &textoffset, status)) {
4796             setColEIterOffset(coleiter, textoffset);
4797             return TRUE;
4798         }
4799     }
4800     setMatchNotFound(strsrch);
4801     return FALSE;
4802 #else
4803     int32_t textOffset;
4804 
4805     if (strsrch->search->isOverlap) {
4806         if (strsrch->search->matchedIndex != USEARCH_DONE) {
4807             textOffset = strsrch->search->matchedIndex + strsrch->search->matchedLength - 1;
4808         } else {
4809             // move the start position at the end of possible match
4810             initializePatternPCETable(strsrch, status);
4811             for (int32_t nPCEs = 0; nPCEs < strsrch->pattern.PCELength - 1; nPCEs++) {
4812                 int64_t pce = ucol_nextProcessed(strsrch->textIter, NULL, NULL, status);
4813                 if (pce == UCOL_PROCESSED_NULLORDER) {
4814                     // at the end of the text
4815                     break;
4816                 }
4817             }
4818             if (U_FAILURE(*status)) {
4819                 setMatchNotFound(strsrch);
4820                 return FALSE;
4821             }
4822             textOffset = ucol_getOffset(strsrch->textIter);
4823         }
4824     } else {
4825         textOffset = ucol_getOffset(strsrch->textIter);
4826     }
4827 
4828     int32_t start = -1;
4829     int32_t end = -1;
4830 
4831     if (usearch_searchBackwards(strsrch, textOffset, &start, &end, status)) {
4832         strsrch->search->matchedIndex = start;
4833         strsrch->search->matchedLength = end - start;
4834         return TRUE;
4835     } else {
4836         setMatchNotFound(strsrch);
4837         return FALSE;
4838     }
4839 #endif
4840 }
4841 
4842 #endif /* #if !UCONFIG_NO_COLLATION */
4843