1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/i18n/usearch.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,4842 @@
1.4 +/*
1.5 +**********************************************************************
1.6 +* Copyright (C) 2001-2011 IBM and others. All rights reserved.
1.7 +**********************************************************************
1.8 +* Date Name Description
1.9 +* 07/02/2001 synwee Creation.
1.10 +**********************************************************************
1.11 +*/
1.12 +
1.13 +#include "unicode/utypes.h"
1.14 +
1.15 +#if !UCONFIG_NO_COLLATION && !UCONFIG_NO_BREAK_ITERATION
1.16 +
1.17 +#include "unicode/usearch.h"
1.18 +#include "unicode/ustring.h"
1.19 +#include "unicode/uchar.h"
1.20 +#include "unicode/utf16.h"
1.21 +#include "normalizer2impl.h"
1.22 +#include "ucol_imp.h"
1.23 +#include "usrchimp.h"
1.24 +#include "cmemory.h"
1.25 +#include "ucln_in.h"
1.26 +#include "uassert.h"
1.27 +#include "ustr_imp.h"
1.28 +
1.29 +U_NAMESPACE_USE
1.30 +
1.31 +// don't use Boyer-Moore
1.32 +// (and if we decide to turn this on again there are several new TODOs that will need to be addressed)
1.33 +#define BOYER_MOORE 0
1.34 +
1.35 +#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
1.36 +
1.37 +// internal definition ---------------------------------------------------
1.38 +
1.39 +#define LAST_BYTE_MASK_ 0xFF
1.40 +#define SECOND_LAST_BYTE_SHIFT_ 8
1.41 +#define SUPPLEMENTARY_MIN_VALUE_ 0x10000
1.42 +
1.43 +static const Normalizer2Impl *g_nfcImpl = NULL;
1.44 +
1.45 +// internal methods -------------------------------------------------
1.46 +
1.47 +/**
1.48 +* Fast collation element iterator setOffset.
1.49 +* This function does not check for bounds.
1.50 +* @param coleiter collation element iterator
1.51 +* @param offset to set
1.52 +*/
1.53 +static
1.54 +inline void setColEIterOffset(UCollationElements *elems,
1.55 + int32_t offset)
1.56 +{
1.57 + collIterate *ci = &(elems->iteratordata_);
1.58 + ci->pos = ci->string + offset;
1.59 + ci->CEpos = ci->toReturn = ci->extendCEs ? ci->extendCEs : ci->CEs;
1.60 + if (ci->flags & UCOL_ITER_INNORMBUF) {
1.61 + ci->flags = ci->origFlags;
1.62 + }
1.63 + ci->fcdPosition = NULL;
1.64 +
1.65 + ci->offsetReturn = NULL;
1.66 + ci->offsetStore = ci->offsetBuffer;
1.67 + ci->offsetRepeatCount = ci->offsetRepeatValue = 0;
1.68 +}
1.69 +
1.70 +/**
1.71 +* Getting the mask for collation strength
1.72 +* @param strength collation strength
1.73 +* @return collation element mask
1.74 +*/
1.75 +static
1.76 +inline uint32_t getMask(UCollationStrength strength)
1.77 +{
1.78 + switch (strength)
1.79 + {
1.80 + case UCOL_PRIMARY:
1.81 + return UCOL_PRIMARYORDERMASK;
1.82 + case UCOL_SECONDARY:
1.83 + return UCOL_SECONDARYORDERMASK | UCOL_PRIMARYORDERMASK;
1.84 + default:
1.85 + return UCOL_TERTIARYORDERMASK | UCOL_SECONDARYORDERMASK |
1.86 + UCOL_PRIMARYORDERMASK;
1.87 + }
1.88 +}
1.89 +
1.90 +/**
1.91 +* This is to squeeze the 21bit ces into a 256 table
1.92 +* @param ce collation element
1.93 +* @return collapsed version of the collation element
1.94 +*/
1.95 +static
1.96 +inline int hash(uint32_t ce)
1.97 +{
1.98 + // the old value UCOL_PRIMARYORDER(ce) % MAX_TABLE_SIZE_ does not work
1.99 + // well with the new collation where most of the latin 1 characters
1.100 + // are of the value xx000xxx. their hashes will most of the time be 0
1.101 + // to be discussed on the hash algo.
1.102 + return UCOL_PRIMARYORDER(ce) % MAX_TABLE_SIZE_;
1.103 +}
1.104 +
1.105 +U_CDECL_BEGIN
1.106 +static UBool U_CALLCONV
1.107 +usearch_cleanup(void) {
1.108 + g_nfcImpl = NULL;
1.109 + return TRUE;
1.110 +}
1.111 +U_CDECL_END
1.112 +
1.113 +/**
1.114 +* Initializing the fcd tables.
1.115 +* Internal method, status assumed to be a success.
1.116 +* @param status output error if any, caller to check status before calling
1.117 +* method, status assumed to be success when passed in.
1.118 +*/
1.119 +static
1.120 +inline void initializeFCD(UErrorCode *status)
1.121 +{
1.122 + if (g_nfcImpl == NULL) {
1.123 + g_nfcImpl = Normalizer2Factory::getNFCImpl(*status);
1.124 + ucln_i18n_registerCleanup(UCLN_I18N_USEARCH, usearch_cleanup);
1.125 + }
1.126 +}
1.127 +
1.128 +/**
1.129 +* Gets the fcd value for a character at the argument index.
1.130 +* This method takes into accounts of the supplementary characters.
1.131 +* @param str UTF16 string where character for fcd retrieval resides
1.132 +* @param offset position of the character whose fcd is to be retrieved, to be
1.133 +* overwritten with the next character position, taking
1.134 +* surrogate characters into consideration.
1.135 +* @param strlength length of the argument string
1.136 +* @return fcd value
1.137 +*/
1.138 +static
1.139 +uint16_t getFCD(const UChar *str, int32_t *offset,
1.140 + int32_t strlength)
1.141 +{
1.142 + const UChar *temp = str + *offset;
1.143 + uint16_t result = g_nfcImpl->nextFCD16(temp, str + strlength);
1.144 + *offset = (int32_t)(temp - str);
1.145 + return result;
1.146 +}
1.147 +
1.148 +/**
1.149 +* Getting the modified collation elements taking into account the collation
1.150 +* attributes
1.151 +* @param strsrch string search data
1.152 +* @param sourcece
1.153 +* @return the modified collation element
1.154 +*/
1.155 +static
1.156 +inline int32_t getCE(const UStringSearch *strsrch, uint32_t sourcece)
1.157 +{
1.158 + // note for tertiary we can't use the collator->tertiaryMask, that
1.159 + // is a preprocessed mask that takes into account case options. since
1.160 + // we are only concerned with exact matches, we don't need that.
1.161 + sourcece &= strsrch->ceMask;
1.162 +
1.163 + if (strsrch->toShift) {
1.164 + // alternate handling here, since only the 16 most significant digits
1.165 + // is only used, we can safely do a compare without masking
1.166 + // if the ce is a variable, we mask and get only the primary values
1.167 + // no shifting to quartenary is required since all primary values
1.168 + // less than variabletop will need to be masked off anyway.
1.169 + if (strsrch->variableTop > sourcece) {
1.170 + if (strsrch->strength >= UCOL_QUATERNARY) {
1.171 + sourcece &= UCOL_PRIMARYORDERMASK;
1.172 + }
1.173 + else {
1.174 + sourcece = UCOL_IGNORABLE;
1.175 + }
1.176 + }
1.177 + } else if (strsrch->strength >= UCOL_QUATERNARY && sourcece == UCOL_IGNORABLE) {
1.178 + sourcece = 0xFFFF;
1.179 + }
1.180 +
1.181 + return sourcece;
1.182 +}
1.183 +
1.184 +/**
1.185 +* Allocate a memory and returns NULL if it failed.
1.186 +* Internal method, status assumed to be a success.
1.187 +* @param size to allocate
1.188 +* @param status output error if any, caller to check status before calling
1.189 +* method, status assumed to be success when passed in.
1.190 +* @return newly allocated array, NULL otherwise
1.191 +*/
1.192 +static
1.193 +inline void * allocateMemory(uint32_t size, UErrorCode *status)
1.194 +{
1.195 + uint32_t *result = (uint32_t *)uprv_malloc(size);
1.196 + if (result == NULL) {
1.197 + *status = U_MEMORY_ALLOCATION_ERROR;
1.198 + }
1.199 + return result;
1.200 +}
1.201 +
1.202 +/**
1.203 +* Adds a uint32_t value to a destination array.
1.204 +* Creates a new array if we run out of space. The caller will have to
1.205 +* manually deallocate the newly allocated array.
1.206 +* Internal method, status assumed to be success, caller has to check status
1.207 +* before calling this method. destination not to be NULL and has at least
1.208 +* size destinationlength.
1.209 +* @param destination target array
1.210 +* @param offset destination offset to add value
1.211 +* @param destinationlength target array size, return value for the new size
1.212 +* @param value to be added
1.213 +* @param increments incremental size expected
1.214 +* @param status output error if any, caller to check status before calling
1.215 +* method, status assumed to be success when passed in.
1.216 +* @return new destination array, destination if there was no new allocation
1.217 +*/
1.218 +static
1.219 +inline int32_t * addTouint32_tArray(int32_t *destination,
1.220 + uint32_t offset,
1.221 + uint32_t *destinationlength,
1.222 + uint32_t value,
1.223 + uint32_t increments,
1.224 + UErrorCode *status)
1.225 +{
1.226 + uint32_t newlength = *destinationlength;
1.227 + if (offset + 1 == newlength) {
1.228 + newlength += increments;
1.229 + int32_t *temp = (int32_t *)allocateMemory(
1.230 + sizeof(int32_t) * newlength, status);
1.231 + if (U_FAILURE(*status)) {
1.232 + return NULL;
1.233 + }
1.234 + uprv_memcpy(temp, destination, sizeof(int32_t) * offset);
1.235 + *destinationlength = newlength;
1.236 + destination = temp;
1.237 + }
1.238 + destination[offset] = value;
1.239 + return destination;
1.240 +}
1.241 +
1.242 +/**
1.243 +* Adds a uint64_t value to a destination array.
1.244 +* Creates a new array if we run out of space. The caller will have to
1.245 +* manually deallocate the newly allocated array.
1.246 +* Internal method, status assumed to be success, caller has to check status
1.247 +* before calling this method. destination not to be NULL and has at least
1.248 +* size destinationlength.
1.249 +* @param destination target array
1.250 +* @param offset destination offset to add value
1.251 +* @param destinationlength target array size, return value for the new size
1.252 +* @param value to be added
1.253 +* @param increments incremental size expected
1.254 +* @param status output error if any, caller to check status before calling
1.255 +* method, status assumed to be success when passed in.
1.256 +* @return new destination array, destination if there was no new allocation
1.257 +*/
1.258 +static
1.259 +inline int64_t * addTouint64_tArray(int64_t *destination,
1.260 + uint32_t offset,
1.261 + uint32_t *destinationlength,
1.262 + uint64_t value,
1.263 + uint32_t increments,
1.264 + UErrorCode *status)
1.265 +{
1.266 + uint32_t newlength = *destinationlength;
1.267 + if (offset + 1 == newlength) {
1.268 + newlength += increments;
1.269 + int64_t *temp = (int64_t *)allocateMemory(
1.270 + sizeof(int64_t) * newlength, status);
1.271 +
1.272 + if (U_FAILURE(*status)) {
1.273 + return NULL;
1.274 + }
1.275 +
1.276 + uprv_memcpy(temp, destination, sizeof(int64_t) * offset);
1.277 + *destinationlength = newlength;
1.278 + destination = temp;
1.279 + }
1.280 +
1.281 + destination[offset] = value;
1.282 +
1.283 + return destination;
1.284 +}
1.285 +
1.286 +/**
1.287 +* Initializing the ce table for a pattern.
1.288 +* Stores non-ignorable collation keys.
1.289 +* Table size will be estimated by the size of the pattern text. Table
1.290 +* expansion will be perform as we go along. Adding 1 to ensure that the table
1.291 +* size definitely increases.
1.292 +* Internal method, status assumed to be a success.
1.293 +* @param strsrch string search data
1.294 +* @param status output error if any, caller to check status before calling
1.295 +* method, status assumed to be success when passed in.
1.296 +* @return total number of expansions
1.297 +*/
1.298 +static
1.299 +inline uint16_t initializePatternCETable(UStringSearch *strsrch,
1.300 + UErrorCode *status)
1.301 +{
1.302 + UPattern *pattern = &(strsrch->pattern);
1.303 + uint32_t cetablesize = INITIAL_ARRAY_SIZE_;
1.304 + int32_t *cetable = pattern->CEBuffer;
1.305 + uint32_t patternlength = pattern->textLength;
1.306 + UCollationElements *coleiter = strsrch->utilIter;
1.307 +
1.308 + if (coleiter == NULL) {
1.309 + coleiter = ucol_openElements(strsrch->collator, pattern->text,
1.310 + patternlength, status);
1.311 + // status will be checked in ucol_next(..) later and if it is an
1.312 + // error UCOL_NULLORDER the result of ucol_next(..) and 0 will be
1.313 + // returned.
1.314 + strsrch->utilIter = coleiter;
1.315 + }
1.316 + else {
1.317 + uprv_init_collIterate(strsrch->collator, pattern->text,
1.318 + pattern->textLength,
1.319 + &coleiter->iteratordata_,
1.320 + status);
1.321 + }
1.322 + if(U_FAILURE(*status)) {
1.323 + return 0;
1.324 + }
1.325 +
1.326 + if (pattern->CE != cetable && pattern->CE) {
1.327 + uprv_free(pattern->CE);
1.328 + }
1.329 +
1.330 + uint16_t offset = 0;
1.331 + uint16_t result = 0;
1.332 + int32_t ce;
1.333 +
1.334 + while ((ce = ucol_next(coleiter, status)) != UCOL_NULLORDER &&
1.335 + U_SUCCESS(*status)) {
1.336 + uint32_t newce = getCE(strsrch, ce);
1.337 + if (newce) {
1.338 + int32_t *temp = addTouint32_tArray(cetable, offset, &cetablesize,
1.339 + newce,
1.340 + patternlength - ucol_getOffset(coleiter) + 1,
1.341 + status);
1.342 + if (U_FAILURE(*status)) {
1.343 + return 0;
1.344 + }
1.345 + offset ++;
1.346 + if (cetable != temp && cetable != pattern->CEBuffer) {
1.347 + uprv_free(cetable);
1.348 + }
1.349 + cetable = temp;
1.350 + }
1.351 + result += (uint16_t)(ucol_getMaxExpansion(coleiter, ce) - 1);
1.352 + }
1.353 +
1.354 + cetable[offset] = 0;
1.355 + pattern->CE = cetable;
1.356 + pattern->CELength = offset;
1.357 +
1.358 + return result;
1.359 +}
1.360 +
1.361 +/**
1.362 +* Initializing the pce table for a pattern.
1.363 +* Stores non-ignorable collation keys.
1.364 +* Table size will be estimated by the size of the pattern text. Table
1.365 +* expansion will be perform as we go along. Adding 1 to ensure that the table
1.366 +* size definitely increases.
1.367 +* Internal method, status assumed to be a success.
1.368 +* @param strsrch string search data
1.369 +* @param status output error if any, caller to check status before calling
1.370 +* method, status assumed to be success when passed in.
1.371 +* @return total number of expansions
1.372 +*/
1.373 +static
1.374 +inline uint16_t initializePatternPCETable(UStringSearch *strsrch,
1.375 + UErrorCode *status)
1.376 +{
1.377 + UPattern *pattern = &(strsrch->pattern);
1.378 + uint32_t pcetablesize = INITIAL_ARRAY_SIZE_;
1.379 + int64_t *pcetable = pattern->PCEBuffer;
1.380 + uint32_t patternlength = pattern->textLength;
1.381 + UCollationElements *coleiter = strsrch->utilIter;
1.382 +
1.383 + if (coleiter == NULL) {
1.384 + coleiter = ucol_openElements(strsrch->collator, pattern->text,
1.385 + patternlength, status);
1.386 + // status will be checked in ucol_next(..) later and if it is an
1.387 + // error UCOL_NULLORDER the result of ucol_next(..) and 0 will be
1.388 + // returned.
1.389 + strsrch->utilIter = coleiter;
1.390 + } else {
1.391 + uprv_init_collIterate(strsrch->collator, pattern->text,
1.392 + pattern->textLength,
1.393 + &coleiter->iteratordata_,
1.394 + status);
1.395 + }
1.396 + if(U_FAILURE(*status)) {
1.397 + return 0;
1.398 + }
1.399 +
1.400 + if (pattern->PCE != pcetable && pattern->PCE != NULL) {
1.401 + uprv_free(pattern->PCE);
1.402 + }
1.403 +
1.404 + uint16_t offset = 0;
1.405 + uint16_t result = 0;
1.406 + int64_t pce;
1.407 +
1.408 + uprv_init_pce(coleiter);
1.409 +
1.410 + // ** Should processed CEs be signed or unsigned?
1.411 + // ** (the rest of the code in this file seems to play fast-and-loose with
1.412 + // ** whether a CE is signed or unsigned. For example, look at routine above this one.)
1.413 + while ((pce = ucol_nextProcessed(coleiter, NULL, NULL, status)) != UCOL_PROCESSED_NULLORDER &&
1.414 + U_SUCCESS(*status)) {
1.415 + int64_t *temp = addTouint64_tArray(pcetable, offset, &pcetablesize,
1.416 + pce,
1.417 + patternlength - ucol_getOffset(coleiter) + 1,
1.418 + status);
1.419 +
1.420 + if (U_FAILURE(*status)) {
1.421 + return 0;
1.422 + }
1.423 +
1.424 + offset += 1;
1.425 +
1.426 + if (pcetable != temp && pcetable != pattern->PCEBuffer) {
1.427 + uprv_free(pcetable);
1.428 + }
1.429 +
1.430 + pcetable = temp;
1.431 + //result += (uint16_t)(ucol_getMaxExpansion(coleiter, ce) - 1);
1.432 + }
1.433 +
1.434 + pcetable[offset] = 0;
1.435 + pattern->PCE = pcetable;
1.436 + pattern->PCELength = offset;
1.437 +
1.438 + return result;
1.439 +}
1.440 +
1.441 +/**
1.442 +* Initializes the pattern struct.
1.443 +* Internal method, status assumed to be success.
1.444 +* @param strsrch UStringSearch data storage
1.445 +* @param status output error if any, caller to check status before calling
1.446 +* method, status assumed to be success when passed in.
1.447 +* @return expansionsize the total expansion size of the pattern
1.448 +*/
1.449 +static
1.450 +inline int16_t initializePattern(UStringSearch *strsrch, UErrorCode *status)
1.451 +{
1.452 + UPattern *pattern = &(strsrch->pattern);
1.453 + const UChar *patterntext = pattern->text;
1.454 + int32_t length = pattern->textLength;
1.455 + int32_t index = 0;
1.456 +
1.457 + // Since the strength is primary, accents are ignored in the pattern.
1.458 + if (strsrch->strength == UCOL_PRIMARY) {
1.459 + pattern->hasPrefixAccents = 0;
1.460 + pattern->hasSuffixAccents = 0;
1.461 + } else {
1.462 + pattern->hasPrefixAccents = getFCD(patterntext, &index, length) >>
1.463 + SECOND_LAST_BYTE_SHIFT_;
1.464 + index = length;
1.465 + U16_BACK_1(patterntext, 0, index);
1.466 + pattern->hasSuffixAccents = getFCD(patterntext, &index, length) &
1.467 + LAST_BYTE_MASK_;
1.468 + }
1.469 +
1.470 + // ** HACK **
1.471 + if (strsrch->pattern.PCE != NULL) {
1.472 + if (strsrch->pattern.PCE != strsrch->pattern.PCEBuffer) {
1.473 + uprv_free(strsrch->pattern.PCE);
1.474 + }
1.475 +
1.476 + strsrch->pattern.PCE = NULL;
1.477 + }
1.478 +
1.479 + // since intializePattern is an internal method status is a success.
1.480 + return initializePatternCETable(strsrch, status);
1.481 +}
1.482 +
1.483 +/**
1.484 +* Initializing shift tables, with the default values.
1.485 +* If a corresponding default value is 0, the shift table is not set.
1.486 +* @param shift table for forwards shift
1.487 +* @param backshift table for backwards shift
1.488 +* @param cetable table containing pattern ce
1.489 +* @param cesize size of the pattern ces
1.490 +* @param expansionsize total size of the expansions
1.491 +* @param defaultforward the default forward value
1.492 +* @param defaultbackward the default backward value
1.493 +*/
1.494 +static
1.495 +inline void setShiftTable(int16_t shift[], int16_t backshift[],
1.496 + int32_t *cetable, int32_t cesize,
1.497 + int16_t expansionsize,
1.498 + int16_t defaultforward,
1.499 + int16_t defaultbackward)
1.500 +{
1.501 + // estimate the value to shift. to do that we estimate the smallest
1.502 + // number of characters to give the relevant ces, ie approximately
1.503 + // the number of ces minus their expansion, since expansions can come
1.504 + // from a character.
1.505 + int32_t count;
1.506 + for (count = 0; count < MAX_TABLE_SIZE_; count ++) {
1.507 + shift[count] = defaultforward;
1.508 + }
1.509 + cesize --; // down to the last index
1.510 + for (count = 0; count < cesize; count ++) {
1.511 + // number of ces from right of array to the count
1.512 + int temp = defaultforward - count - 1;
1.513 + shift[hash(cetable[count])] = temp > 1 ? temp : 1;
1.514 + }
1.515 + shift[hash(cetable[cesize])] = 1;
1.516 + // for ignorables we just shift by one. see test examples.
1.517 + shift[hash(0)] = 1;
1.518 +
1.519 + for (count = 0; count < MAX_TABLE_SIZE_; count ++) {
1.520 + backshift[count] = defaultbackward;
1.521 + }
1.522 + for (count = cesize; count > 0; count --) {
1.523 + // the original value count does not seem to work
1.524 + backshift[hash(cetable[count])] = count > expansionsize ?
1.525 + (int16_t)(count - expansionsize) : 1;
1.526 + }
1.527 + backshift[hash(cetable[0])] = 1;
1.528 + backshift[hash(0)] = 1;
1.529 +}
1.530 +
1.531 +/**
1.532 +* Building of the pattern collation element list and the boyer moore strsrch
1.533 +* table.
1.534 +* The canonical match will only be performed after the default match fails.
1.535 +* For both cases we need to remember the size of the composed and decomposed
1.536 +* versions of the string. Since the Boyer-Moore shift calculations shifts by
1.537 +* a number of characters in the text and tries to match the pattern from that
1.538 +* offset, the shift value can not be too large in case we miss some
1.539 +* characters. To choose a right shift size, we estimate the NFC form of the
1.540 +* and use its size as a shift guide. The NFC form should be the small
1.541 +* possible representation of the pattern. Anyways, we'll err on the smaller
1.542 +* shift size. Hence the calculation for minlength.
1.543 +* Canonical match will be performed slightly differently. We'll split the
1.544 +* pattern into 3 parts, the prefix accents (PA), the middle string bounded by
1.545 +* the first and last base character (MS), the ending accents (EA). Matches
1.546 +* will be done on MS first, and only when we match MS then some processing
1.547 +* will be required for the prefix and end accents in order to determine if
1.548 +* they match PA and EA. Hence the default shift values
1.549 +* for the canonical match will take the size of either end's accent into
1.550 +* consideration. Forwards search will take the end accents into consideration
1.551 +* for the default shift values and the backwards search will take the prefix
1.552 +* accents into consideration.
1.553 +* If pattern has no non-ignorable ce, we return a illegal argument error.
1.554 +* Internal method, status assumed to be success.
1.555 +* @param strsrch UStringSearch data storage
1.556 +* @param status for output errors if it occurs, status is assumed to be a
1.557 +* success when it is passed in.
1.558 +*/
1.559 +static
1.560 +inline void initialize(UStringSearch *strsrch, UErrorCode *status)
1.561 +{
1.562 + int16_t expandlength = initializePattern(strsrch, status);
1.563 + if (U_SUCCESS(*status) && strsrch->pattern.CELength > 0) {
1.564 + UPattern *pattern = &strsrch->pattern;
1.565 + int32_t cesize = pattern->CELength;
1.566 +
1.567 + int16_t minlength = cesize > expandlength
1.568 + ? (int16_t)cesize - expandlength : 1;
1.569 + pattern->defaultShiftSize = minlength;
1.570 + setShiftTable(pattern->shift, pattern->backShift, pattern->CE,
1.571 + cesize, expandlength, minlength, minlength);
1.572 + return;
1.573 + }
1.574 + strsrch->pattern.defaultShiftSize = 0;
1.575 +}
1.576 +
1.577 +#if BOYER_MOORE
1.578 +/**
1.579 +* Check to make sure that the match length is at the end of the character by
1.580 +* using the breakiterator.
1.581 +* @param strsrch string search data
1.582 +* @param start target text start offset
1.583 +* @param end target text end offset
1.584 +*/
1.585 +static
1.586 +void checkBreakBoundary(const UStringSearch *strsrch, int32_t * /*start*/,
1.587 + int32_t *end)
1.588 +{
1.589 +#if !UCONFIG_NO_BREAK_ITERATION
1.590 + UBreakIterator *breakiterator = strsrch->search->internalBreakIter;
1.591 + if (breakiterator) {
1.592 + int32_t matchend = *end;
1.593 + //int32_t matchstart = *start;
1.594 +
1.595 + if (!ubrk_isBoundary(breakiterator, matchend)) {
1.596 + *end = ubrk_following(breakiterator, matchend);
1.597 + }
1.598 +
1.599 + /* Check the start of the matched text to make sure it doesn't have any accents
1.600 + * before it. This code may not be necessary and so it is commented out */
1.601 + /*if (!ubrk_isBoundary(breakiterator, matchstart) && !ubrk_isBoundary(breakiterator, matchstart-1)) {
1.602 + *start = ubrk_preceding(breakiterator, matchstart);
1.603 + }*/
1.604 + }
1.605 +#endif
1.606 +}
1.607 +
1.608 +/**
1.609 +* Determine whether the target text in UStringSearch bounded by the offset
1.610 +* start and end is one or more whole units of text as
1.611 +* determined by the breakiterator in UStringSearch.
1.612 +* @param strsrch string search data
1.613 +* @param start target text start offset
1.614 +* @param end target text end offset
1.615 +*/
1.616 +static
1.617 +UBool isBreakUnit(const UStringSearch *strsrch, int32_t start,
1.618 + int32_t end)
1.619 +{
1.620 +#if !UCONFIG_NO_BREAK_ITERATION
1.621 + UBreakIterator *breakiterator = strsrch->search->breakIter;
1.622 + //TODO: Add here.
1.623 + if (breakiterator) {
1.624 + int32_t startindex = ubrk_first(breakiterator);
1.625 + int32_t endindex = ubrk_last(breakiterator);
1.626 +
1.627 + // out-of-range indexes are never boundary positions
1.628 + if (start < startindex || start > endindex ||
1.629 + end < startindex || end > endindex) {
1.630 + return FALSE;
1.631 + }
1.632 + // otherwise, we can use following() on the position before the
1.633 + // specified one and return true of the position we get back is the
1.634 + // one the user specified
1.635 + UBool result = (start == startindex ||
1.636 + ubrk_following(breakiterator, start - 1) == start) &&
1.637 + (end == endindex ||
1.638 + ubrk_following(breakiterator, end - 1) == end);
1.639 + if (result) {
1.640 + // iterates the individual ces
1.641 + UCollationElements *coleiter = strsrch->utilIter;
1.642 + const UChar *text = strsrch->search->text +
1.643 + start;
1.644 + UErrorCode status = U_ZERO_ERROR;
1.645 + ucol_setText(coleiter, text, end - start, &status);
1.646 + for (int32_t count = 0; count < strsrch->pattern.CELength;
1.647 + count ++) {
1.648 + int32_t ce = getCE(strsrch, ucol_next(coleiter, &status));
1.649 + if (ce == UCOL_IGNORABLE) {
1.650 + count --;
1.651 + continue;
1.652 + }
1.653 + if (U_FAILURE(status) || ce != strsrch->pattern.CE[count]) {
1.654 + return FALSE;
1.655 + }
1.656 + }
1.657 + int32_t nextce = ucol_next(coleiter, &status);
1.658 + while (ucol_getOffset(coleiter) == (end - start)
1.659 + && getCE(strsrch, nextce) == UCOL_IGNORABLE) {
1.660 + nextce = ucol_next(coleiter, &status);
1.661 + }
1.662 + if (ucol_getOffset(coleiter) == (end - start)
1.663 + && nextce != UCOL_NULLORDER) {
1.664 + // extra collation elements at the end of the match
1.665 + return FALSE;
1.666 + }
1.667 + }
1.668 + return result;
1.669 + }
1.670 +#endif
1.671 + return TRUE;
1.672 +}
1.673 +
1.674 +/**
1.675 +* Getting the next base character offset if current offset is an accent,
1.676 +* or the current offset if the current character contains a base character.
1.677 +* accents the following base character will be returned
1.678 +* @param text string
1.679 +* @param textoffset current offset
1.680 +* @param textlength length of text string
1.681 +* @return the next base character or the current offset
1.682 +* if the current character is contains a base character.
1.683 +*/
1.684 +static
1.685 +inline int32_t getNextBaseOffset(const UChar *text,
1.686 + int32_t textoffset,
1.687 + int32_t textlength)
1.688 +{
1.689 + if (textoffset < textlength) {
1.690 + int32_t temp = textoffset;
1.691 + if (getFCD(text, &temp, textlength) >> SECOND_LAST_BYTE_SHIFT_) {
1.692 + while (temp < textlength) {
1.693 + int32_t result = temp;
1.694 + if ((getFCD(text, &temp, textlength) >>
1.695 + SECOND_LAST_BYTE_SHIFT_) == 0) {
1.696 + return result;
1.697 + }
1.698 + }
1.699 + return textlength;
1.700 + }
1.701 + }
1.702 + return textoffset;
1.703 +}
1.704 +
1.705 +/**
1.706 +* Gets the next base character offset depending on the string search pattern
1.707 +* data
1.708 +* @param strsrch string search data
1.709 +* @param textoffset current offset, one offset away from the last character
1.710 +* to search for.
1.711 +* @return start index of the next base character or the current offset
1.712 +* if the current character is contains a base character.
1.713 +*/
1.714 +static
1.715 +inline int32_t getNextUStringSearchBaseOffset(UStringSearch *strsrch,
1.716 + int32_t textoffset)
1.717 +{
1.718 + int32_t textlength = strsrch->search->textLength;
1.719 + if (strsrch->pattern.hasSuffixAccents &&
1.720 + textoffset < textlength) {
1.721 + int32_t temp = textoffset;
1.722 + const UChar *text = strsrch->search->text;
1.723 + U16_BACK_1(text, 0, temp);
1.724 + if (getFCD(text, &temp, textlength) & LAST_BYTE_MASK_) {
1.725 + return getNextBaseOffset(text, textoffset, textlength);
1.726 + }
1.727 + }
1.728 + return textoffset;
1.729 +}
1.730 +
1.731 +/**
1.732 +* Shifting the collation element iterator position forward to prepare for
1.733 +* a following match. If the last character is a unsafe character, we'll only
1.734 +* shift by 1 to capture contractions, normalization etc.
1.735 +* Internal method, status assumed to be success.
1.736 +* @param text strsrch string search data
1.737 +* @param textoffset start text position to do search
1.738 +* @param ce the text ce which failed the match.
1.739 +* @param patternceindex index of the ce within the pattern ce buffer which
1.740 +* failed the match
1.741 +* @return final offset
1.742 +*/
1.743 +static
1.744 +inline int32_t shiftForward(UStringSearch *strsrch,
1.745 + int32_t textoffset,
1.746 + int32_t ce,
1.747 + int32_t patternceindex)
1.748 +{
1.749 + UPattern *pattern = &(strsrch->pattern);
1.750 + if (ce != UCOL_NULLORDER) {
1.751 + int32_t shift = pattern->shift[hash(ce)];
1.752 + // this is to adjust for characters in the middle of the
1.753 + // substring for matching that failed.
1.754 + int32_t adjust = pattern->CELength - patternceindex;
1.755 + if (adjust > 1 && shift >= adjust) {
1.756 + shift -= adjust - 1;
1.757 + }
1.758 + textoffset += shift;
1.759 + }
1.760 + else {
1.761 + textoffset += pattern->defaultShiftSize;
1.762 + }
1.763 +
1.764 + textoffset = getNextUStringSearchBaseOffset(strsrch, textoffset);
1.765 + // check for unsafe characters
1.766 + // * if it is the start or middle of a contraction: to be done after
1.767 + // a initial match is found
1.768 + // * thai or lao base consonant character: similar to contraction
1.769 + // * high surrogate character: similar to contraction
1.770 + // * next character is a accent: shift to the next base character
1.771 + return textoffset;
1.772 +}
1.773 +#endif // #if BOYER_MOORE
1.774 +
1.775 +/**
1.776 +* sets match not found
1.777 +* @param strsrch string search data
1.778 +*/
1.779 +static
1.780 +inline void setMatchNotFound(UStringSearch *strsrch)
1.781 +{
1.782 + // this method resets the match result regardless of the error status.
1.783 + strsrch->search->matchedIndex = USEARCH_DONE;
1.784 + strsrch->search->matchedLength = 0;
1.785 + if (strsrch->search->isForwardSearching) {
1.786 + setColEIterOffset(strsrch->textIter, strsrch->search->textLength);
1.787 + }
1.788 + else {
1.789 + setColEIterOffset(strsrch->textIter, 0);
1.790 + }
1.791 +}
1.792 +
1.793 +#if BOYER_MOORE
1.794 +/**
1.795 +* Gets the offset to the next safe point in text.
1.796 +* ie. not the middle of a contraction, swappable characters or supplementary
1.797 +* characters.
1.798 +* @param collator collation sata
1.799 +* @param text string to work with
1.800 +* @param textoffset offset in string
1.801 +* @param textlength length of text string
1.802 +* @return offset to the next safe character
1.803 +*/
1.804 +static
1.805 +inline int32_t getNextSafeOffset(const UCollator *collator,
1.806 + const UChar *text,
1.807 + int32_t textoffset,
1.808 + int32_t textlength)
1.809 +{
1.810 + int32_t result = textoffset; // first contraction character
1.811 + while (result != textlength && ucol_unsafeCP(text[result], collator)) {
1.812 + result ++;
1.813 + }
1.814 + return result;
1.815 +}
1.816 +
1.817 +/**
1.818 +* This checks for accents in the potential match started with a .
1.819 +* composite character.
1.820 +* This is really painful... we have to check that composite character do not
1.821 +* have any extra accents. We have to normalize the potential match and find
1.822 +* the immediate decomposed character before the match.
1.823 +* The first composite character would have been taken care of by the fcd
1.824 +* checks in checkForwardExactMatch.
1.825 +* This is the slow path after the fcd of the first character and
1.826 +* the last character has been checked by checkForwardExactMatch and we
1.827 +* determine that the potential match has extra non-ignorable preceding
1.828 +* ces.
1.829 +* E.g. looking for \u0301 acute in \u01FA A ring above and acute,
1.830 +* checkExtraMatchAccent should fail since there is a middle ring in \u01FA
1.831 +* Note here that accents checking are slow and cautioned in the API docs.
1.832 +* Internal method, status assumed to be a success, caller should check status
1.833 +* before calling this method
1.834 +* @param strsrch string search data
1.835 +* @param start index of the potential unfriendly composite character
1.836 +* @param end index of the potential unfriendly composite character
1.837 +* @param status output error status if any.
1.838 +* @return TRUE if there is non-ignorable accents before at the beginning
1.839 +* of the match, FALSE otherwise.
1.840 +*/
1.841 +
1.842 +static
1.843 +UBool checkExtraMatchAccents(const UStringSearch *strsrch, int32_t start,
1.844 + int32_t end,
1.845 + UErrorCode *status)
1.846 +{
1.847 + UBool result = FALSE;
1.848 + if (strsrch->pattern.hasPrefixAccents) {
1.849 + int32_t length = end - start;
1.850 + int32_t offset = 0;
1.851 + const UChar *text = strsrch->search->text + start;
1.852 +
1.853 + U16_FWD_1(text, offset, length);
1.854 + // we are only concerned with the first composite character
1.855 + if (unorm_quickCheck(text, offset, UNORM_NFD, status) == UNORM_NO) {
1.856 + int32_t safeoffset = getNextSafeOffset(strsrch->collator,
1.857 + text, 0, length);
1.858 + if (safeoffset != length) {
1.859 + safeoffset ++;
1.860 + }
1.861 + UChar *norm = NULL;
1.862 + UChar buffer[INITIAL_ARRAY_SIZE_];
1.863 + int32_t size = unorm_normalize(text, safeoffset, UNORM_NFD, 0,
1.864 + buffer, INITIAL_ARRAY_SIZE_,
1.865 + status);
1.866 + if (U_FAILURE(*status)) {
1.867 + return FALSE;
1.868 + }
1.869 + if (size >= INITIAL_ARRAY_SIZE_) {
1.870 + norm = (UChar *)allocateMemory((size + 1) * sizeof(UChar),
1.871 + status);
1.872 + // if allocation failed, status will be set to
1.873 + // U_MEMORY_ALLOCATION_ERROR and unorm_normalize internally
1.874 + // checks for it.
1.875 + size = unorm_normalize(text, safeoffset, UNORM_NFD, 0, norm,
1.876 + size, status);
1.877 + if (U_FAILURE(*status) && norm != NULL) {
1.878 + uprv_free(norm);
1.879 + return FALSE;
1.880 + }
1.881 + }
1.882 + else {
1.883 + norm = buffer;
1.884 + }
1.885 +
1.886 + UCollationElements *coleiter = strsrch->utilIter;
1.887 + ucol_setText(coleiter, norm, size, status);
1.888 + uint32_t firstce = strsrch->pattern.CE[0];
1.889 + UBool ignorable = TRUE;
1.890 + uint32_t ce = UCOL_IGNORABLE;
1.891 + while (U_SUCCESS(*status) && ce != firstce && ce != (uint32_t)UCOL_NULLORDER) {
1.892 + offset = ucol_getOffset(coleiter);
1.893 + if (ce != firstce && ce != UCOL_IGNORABLE) {
1.894 + ignorable = FALSE;
1.895 + }
1.896 + ce = ucol_next(coleiter, status);
1.897 + }
1.898 + UChar32 codepoint;
1.899 + U16_PREV(norm, 0, offset, codepoint);
1.900 + result = !ignorable && (u_getCombiningClass(codepoint) != 0);
1.901 +
1.902 + if (norm != buffer) {
1.903 + uprv_free(norm);
1.904 + }
1.905 + }
1.906 + }
1.907 +
1.908 + return result;
1.909 +}
1.910 +
1.911 +/**
1.912 +* Used by exact matches, checks if there are accents before the match.
1.913 +* This is really painful... we have to check that composite characters at
1.914 +* the start of the matches have to not have any extra accents.
1.915 +* We check the FCD of the character first, if it starts with an accent and
1.916 +* the first pattern ce does not match the first ce of the character, we bail.
1.917 +* Otherwise we try normalizing the first composite
1.918 +* character and find the immediate decomposed character before the match to
1.919 +* see if it is an non-ignorable accent.
1.920 +* Now normalizing the first composite character is enough because we ensure
1.921 +* that when the match is passed in here with extra beginning ces, the
1.922 +* first or last ce that match has to occur within the first character.
1.923 +* E.g. looking for \u0301 acute in \u01FA A ring above and acute,
1.924 +* checkExtraMatchAccent should fail since there is a middle ring in \u01FA
1.925 +* Note here that accents checking are slow and cautioned in the API docs.
1.926 +* @param strsrch string search data
1.927 +* @param start offset
1.928 +* @param end offset
1.929 +* @return TRUE if there are accents on either side of the match,
1.930 +* FALSE otherwise
1.931 +*/
1.932 +static
1.933 +UBool hasAccentsBeforeMatch(const UStringSearch *strsrch, int32_t start,
1.934 + int32_t end)
1.935 +{
1.936 + if (strsrch->pattern.hasPrefixAccents) {
1.937 + UCollationElements *coleiter = strsrch->textIter;
1.938 + UErrorCode status = U_ZERO_ERROR;
1.939 + // we have been iterating forwards previously
1.940 + uint32_t ignorable = TRUE;
1.941 + int32_t firstce = strsrch->pattern.CE[0];
1.942 +
1.943 + setColEIterOffset(coleiter, start);
1.944 + int32_t ce = getCE(strsrch, ucol_next(coleiter, &status));
1.945 + if (U_FAILURE(status)) {
1.946 + return TRUE;
1.947 + }
1.948 + while (ce != firstce) {
1.949 + if (ce != UCOL_IGNORABLE) {
1.950 + ignorable = FALSE;
1.951 + }
1.952 + ce = getCE(strsrch, ucol_next(coleiter, &status));
1.953 + if (U_FAILURE(status) || ce == UCOL_NULLORDER) {
1.954 + return TRUE;
1.955 + }
1.956 + }
1.957 + if (!ignorable && inNormBuf(coleiter)) {
1.958 + // within normalization buffer, discontiguous handled here
1.959 + return TRUE;
1.960 + }
1.961 +
1.962 + // within text
1.963 + int32_t temp = start;
1.964 + // original code
1.965 + // accent = (getFCD(strsrch->search->text, &temp,
1.966 + // strsrch->search->textLength)
1.967 + // >> SECOND_LAST_BYTE_SHIFT_);
1.968 + // however this code does not work well with VC7 .net in release mode.
1.969 + // maybe the inlines for getFCD combined with shifting has bugs in
1.970 + // VC7. anyways this is a work around.
1.971 + UBool accent = getFCD(strsrch->search->text, &temp,
1.972 + strsrch->search->textLength) > 0xFF;
1.973 + if (!accent) {
1.974 + return checkExtraMatchAccents(strsrch, start, end, &status);
1.975 + }
1.976 + if (!ignorable) {
1.977 + return TRUE;
1.978 + }
1.979 + if (start > 0) {
1.980 + temp = start;
1.981 + U16_BACK_1(strsrch->search->text, 0, temp);
1.982 + if (getFCD(strsrch->search->text, &temp,
1.983 + strsrch->search->textLength) & LAST_BYTE_MASK_) {
1.984 + setColEIterOffset(coleiter, start);
1.985 + ce = ucol_previous(coleiter, &status);
1.986 + if (U_FAILURE(status) ||
1.987 + (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE)) {
1.988 + return TRUE;
1.989 + }
1.990 + }
1.991 + }
1.992 + }
1.993 +
1.994 + return FALSE;
1.995 +}
1.996 +
1.997 +/**
1.998 +* Used by exact matches, checks if there are accents bounding the match.
1.999 +* Note this is the initial boundary check. If the potential match
1.1000 +* starts or ends with composite characters, the accents in those
1.1001 +* characters will be determined later.
1.1002 +* Not doing backwards iteration here, since discontiguos contraction for
1.1003 +* backwards collation element iterator, use up too many characters.
1.1004 +* E.g. looking for \u030A ring in \u01FA A ring above and acute,
1.1005 +* should fail since there is a acute at the end of \u01FA
1.1006 +* Note here that accents checking are slow and cautioned in the API docs.
1.1007 +* @param strsrch string search data
1.1008 +* @param start offset of match
1.1009 +* @param end end offset of the match
1.1010 +* @return TRUE if there are accents on either side of the match,
1.1011 +* FALSE otherwise
1.1012 +*/
1.1013 +static
1.1014 +UBool hasAccentsAfterMatch(const UStringSearch *strsrch, int32_t start,
1.1015 + int32_t end)
1.1016 +{
1.1017 + if (strsrch->pattern.hasSuffixAccents) {
1.1018 + const UChar *text = strsrch->search->text;
1.1019 + int32_t temp = end;
1.1020 + int32_t textlength = strsrch->search->textLength;
1.1021 + U16_BACK_1(text, 0, temp);
1.1022 + if (getFCD(text, &temp, textlength) & LAST_BYTE_MASK_) {
1.1023 + int32_t firstce = strsrch->pattern.CE[0];
1.1024 + UCollationElements *coleiter = strsrch->textIter;
1.1025 + UErrorCode status = U_ZERO_ERROR;
1.1026 + int32_t ce;
1.1027 + setColEIterOffset(coleiter, start);
1.1028 + while ((ce = getCE(strsrch, ucol_next(coleiter, &status))) != firstce) {
1.1029 + if (U_FAILURE(status) || ce == UCOL_NULLORDER) {
1.1030 + return TRUE;
1.1031 + }
1.1032 + }
1.1033 + int32_t count = 1;
1.1034 + while (count < strsrch->pattern.CELength) {
1.1035 + if (getCE(strsrch, ucol_next(coleiter, &status))
1.1036 + == UCOL_IGNORABLE) {
1.1037 + // Thai can give an ignorable here.
1.1038 + count --;
1.1039 + }
1.1040 + if (U_FAILURE(status)) {
1.1041 + return TRUE;
1.1042 + }
1.1043 + count ++;
1.1044 + }
1.1045 +
1.1046 + ce = ucol_next(coleiter, &status);
1.1047 + if (U_FAILURE(status)) {
1.1048 + return TRUE;
1.1049 + }
1.1050 + if (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE) {
1.1051 + ce = getCE(strsrch, ce);
1.1052 + }
1.1053 + if (ce != UCOL_NULLORDER && ce != UCOL_IGNORABLE) {
1.1054 + if (ucol_getOffset(coleiter) <= end) {
1.1055 + return TRUE;
1.1056 + }
1.1057 + if (getFCD(text, &end, textlength) >> SECOND_LAST_BYTE_SHIFT_) {
1.1058 + return TRUE;
1.1059 + }
1.1060 + }
1.1061 + }
1.1062 + }
1.1063 + return FALSE;
1.1064 +}
1.1065 +#endif // #if BOYER_MOORE
1.1066 +
1.1067 +/**
1.1068 +* Checks if the offset runs out of the text string
1.1069 +* @param offset
1.1070 +* @param textlength of the text string
1.1071 +* @return TRUE if offset is out of bounds, FALSE otherwise
1.1072 +*/
1.1073 +static
1.1074 +inline UBool isOutOfBounds(int32_t textlength, int32_t offset)
1.1075 +{
1.1076 + return offset < 0 || offset > textlength;
1.1077 +}
1.1078 +
1.1079 +/**
1.1080 +* Checks for identical match
1.1081 +* @param strsrch string search data
1.1082 +* @param start offset of possible match
1.1083 +* @param end offset of possible match
1.1084 +* @return TRUE if identical match is found
1.1085 +*/
1.1086 +static
1.1087 +inline UBool checkIdentical(const UStringSearch *strsrch, int32_t start,
1.1088 + int32_t end)
1.1089 +{
1.1090 + if (strsrch->strength != UCOL_IDENTICAL) {
1.1091 + return TRUE;
1.1092 + }
1.1093 +
1.1094 + // Note: We could use Normalizer::compare() or similar, but for short strings
1.1095 + // which may not be in FCD it might be faster to just NFD them.
1.1096 + UErrorCode status = U_ZERO_ERROR;
1.1097 + UnicodeString t2, p2;
1.1098 + strsrch->nfd->normalize(
1.1099 + UnicodeString(FALSE, strsrch->search->text + start, end - start), t2, status);
1.1100 + strsrch->nfd->normalize(
1.1101 + UnicodeString(FALSE, strsrch->pattern.text, strsrch->pattern.textLength), p2, status);
1.1102 + // return FALSE if NFD failed
1.1103 + return U_SUCCESS(status) && t2 == p2;
1.1104 +}
1.1105 +
1.1106 +#if BOYER_MOORE
1.1107 +/**
1.1108 +* Checks to see if the match is repeated
1.1109 +* @param strsrch string search data
1.1110 +* @param start new match start index
1.1111 +* @param end new match end index
1.1112 +* @return TRUE if the the match is repeated, FALSE otherwise
1.1113 +*/
1.1114 +static
1.1115 +inline UBool checkRepeatedMatch(UStringSearch *strsrch,
1.1116 + int32_t start,
1.1117 + int32_t end)
1.1118 +{
1.1119 + int32_t lastmatchindex = strsrch->search->matchedIndex;
1.1120 + UBool result;
1.1121 + if (lastmatchindex == USEARCH_DONE) {
1.1122 + return FALSE;
1.1123 + }
1.1124 + if (strsrch->search->isForwardSearching) {
1.1125 + result = start <= lastmatchindex;
1.1126 + }
1.1127 + else {
1.1128 + result = start >= lastmatchindex;
1.1129 + }
1.1130 + if (!result && !strsrch->search->isOverlap) {
1.1131 + if (strsrch->search->isForwardSearching) {
1.1132 + result = start < lastmatchindex + strsrch->search->matchedLength;
1.1133 + }
1.1134 + else {
1.1135 + result = end > lastmatchindex;
1.1136 + }
1.1137 + }
1.1138 + return result;
1.1139 +}
1.1140 +
1.1141 +/**
1.1142 +* Gets the collation element iterator's current offset.
1.1143 +* @param coleiter collation element iterator
1.1144 +* @param forwards flag TRUE if we are moving in th forwards direction
1.1145 +* @return current offset
1.1146 +*/
1.1147 +static
1.1148 +inline int32_t getColElemIterOffset(const UCollationElements *coleiter,
1.1149 + UBool forwards)
1.1150 +{
1.1151 + int32_t result = ucol_getOffset(coleiter);
1.1152 + // intricacies of the the backwards collation element iterator
1.1153 + if (FALSE && !forwards && inNormBuf(coleiter) && !isFCDPointerNull(coleiter)) {
1.1154 + result ++;
1.1155 + }
1.1156 + return result;
1.1157 +}
1.1158 +
1.1159 +/**
1.1160 +* Checks match for contraction.
1.1161 +* If the match ends with a partial contraction we fail.
1.1162 +* If the match starts too far off (because of backwards iteration) we try to
1.1163 +* chip off the extra characters depending on whether a breakiterator has
1.1164 +* been used.
1.1165 +* Internal method, error assumed to be success, caller has to check status
1.1166 +* before calling this method.
1.1167 +* @param strsrch string search data
1.1168 +* @param start offset of potential match, to be modified if necessary
1.1169 +* @param end offset of potential match, to be modified if necessary
1.1170 +* @param status output error status if any
1.1171 +* @return TRUE if match passes the contraction test, FALSE otherwise
1.1172 +*/
1.1173 +
1.1174 +static
1.1175 +UBool checkNextExactContractionMatch(UStringSearch *strsrch,
1.1176 + int32_t *start,
1.1177 + int32_t *end, UErrorCode *status)
1.1178 +{
1.1179 + UCollationElements *coleiter = strsrch->textIter;
1.1180 + int32_t textlength = strsrch->search->textLength;
1.1181 + int32_t temp = *start;
1.1182 + const UCollator *collator = strsrch->collator;
1.1183 + const UChar *text = strsrch->search->text;
1.1184 + // This part checks if either ends of the match contains potential
1.1185 + // contraction. If so we'll have to iterate through them
1.1186 + // The start contraction needs to be checked since ucol_previous dumps
1.1187 + // all characters till the first safe character into the buffer.
1.1188 + // *start + 1 is used to test for the unsafe characters instead of *start
1.1189 + // because ucol_prev takes all unsafe characters till the first safe
1.1190 + // character ie *start. so by testing *start + 1, we can estimate if
1.1191 + // excess prefix characters has been included in the potential search
1.1192 + // results.
1.1193 + if ((*end < textlength && ucol_unsafeCP(text[*end], collator)) ||
1.1194 + (*start + 1 < textlength
1.1195 + && ucol_unsafeCP(text[*start + 1], collator))) {
1.1196 + int32_t expansion = getExpansionPrefix(coleiter);
1.1197 + UBool expandflag = expansion > 0;
1.1198 + setColEIterOffset(coleiter, *start);
1.1199 + while (expansion > 0) {
1.1200 + // getting rid of the redundant ce, caused by setOffset.
1.1201 + // since backward contraction/expansion may have extra ces if we
1.1202 + // are in the normalization buffer, hasAccentsBeforeMatch would
1.1203 + // have taken care of it.
1.1204 + // E.g. the character \u01FA will have an expansion of 3, but if
1.1205 + // we are only looking for acute and ring \u030A and \u0301, we'll
1.1206 + // have to skip the first ce in the expansion buffer.
1.1207 + ucol_next(coleiter, status);
1.1208 + if (U_FAILURE(*status)) {
1.1209 + return FALSE;
1.1210 + }
1.1211 + if (ucol_getOffset(coleiter) != temp) {
1.1212 + *start = temp;
1.1213 + temp = ucol_getOffset(coleiter);
1.1214 + }
1.1215 + expansion --;
1.1216 + }
1.1217 +
1.1218 + int32_t *patternce = strsrch->pattern.CE;
1.1219 + int32_t patterncelength = strsrch->pattern.CELength;
1.1220 + int32_t count = 0;
1.1221 + while (count < patterncelength) {
1.1222 + int32_t ce = getCE(strsrch, ucol_next(coleiter, status));
1.1223 + if (ce == UCOL_IGNORABLE) {
1.1224 + continue;
1.1225 + }
1.1226 + if (expandflag && count == 0 && ucol_getOffset(coleiter) != temp) {
1.1227 + *start = temp;
1.1228 + temp = ucol_getOffset(coleiter);
1.1229 + }
1.1230 + if (U_FAILURE(*status) || ce != patternce[count]) {
1.1231 + (*end) ++;
1.1232 + *end = getNextUStringSearchBaseOffset(strsrch, *end);
1.1233 + return FALSE;
1.1234 + }
1.1235 + count ++;
1.1236 + }
1.1237 + }
1.1238 + return TRUE;
1.1239 +}
1.1240 +
1.1241 +/**
1.1242 +* Checks and sets the match information if found.
1.1243 +* Checks
1.1244 +* <ul>
1.1245 +* <li> the potential match does not repeat the previous match
1.1246 +* <li> boundaries are correct
1.1247 +* <li> exact matches has no extra accents
1.1248 +* <li> identical matchesb
1.1249 +* <li> potential match does not end in the middle of a contraction
1.1250 +* <\ul>
1.1251 +* Otherwise the offset will be shifted to the next character.
1.1252 +* Internal method, status assumed to be success, caller has to check status
1.1253 +* before calling this method.
1.1254 +* @param strsrch string search data
1.1255 +* @param textoffset offset in the collation element text. the returned value
1.1256 +* will be the truncated end offset of the match or the new start
1.1257 +* search offset.
1.1258 +* @param status output error status if any
1.1259 +* @return TRUE if the match is valid, FALSE otherwise
1.1260 +*/
1.1261 +static
1.1262 +inline UBool checkNextExactMatch(UStringSearch *strsrch,
1.1263 + int32_t *textoffset, UErrorCode *status)
1.1264 +{
1.1265 + UCollationElements *coleiter = strsrch->textIter;
1.1266 + int32_t start = getColElemIterOffset(coleiter, FALSE);
1.1267 +
1.1268 + if (!checkNextExactContractionMatch(strsrch, &start, textoffset, status)) {
1.1269 + return FALSE;
1.1270 + }
1.1271 +
1.1272 + // this totally matches, however we need to check if it is repeating
1.1273 + if (!isBreakUnit(strsrch, start, *textoffset) ||
1.1274 + checkRepeatedMatch(strsrch, start, *textoffset) ||
1.1275 + hasAccentsBeforeMatch(strsrch, start, *textoffset) ||
1.1276 + !checkIdentical(strsrch, start, *textoffset) ||
1.1277 + hasAccentsAfterMatch(strsrch, start, *textoffset)) {
1.1278 +
1.1279 + (*textoffset) ++;
1.1280 + *textoffset = getNextUStringSearchBaseOffset(strsrch, *textoffset);
1.1281 + return FALSE;
1.1282 + }
1.1283 +
1.1284 + //Add breakiterator boundary check for primary strength search.
1.1285 + if (!strsrch->search->breakIter && strsrch->strength == UCOL_PRIMARY) {
1.1286 + checkBreakBoundary(strsrch, &start, textoffset);
1.1287 + }
1.1288 +
1.1289 + // totally match, we will get rid of the ending ignorables.
1.1290 + strsrch->search->matchedIndex = start;
1.1291 + strsrch->search->matchedLength = *textoffset - start;
1.1292 + return TRUE;
1.1293 +}
1.1294 +
1.1295 +/**
1.1296 +* Getting the previous base character offset, or the current offset if the
1.1297 +* current character is a base character
1.1298 +* @param text string
1.1299 +* @param textoffset one offset after the current character
1.1300 +* @return the offset of the next character after the base character or the first
1.1301 +* composed character with accents
1.1302 +*/
1.1303 +static
1.1304 +inline int32_t getPreviousBaseOffset(const UChar *text,
1.1305 + int32_t textoffset)
1.1306 +{
1.1307 + if (textoffset > 0) {
1.1308 + for (;;) {
1.1309 + int32_t result = textoffset;
1.1310 + U16_BACK_1(text, 0, textoffset);
1.1311 + int32_t temp = textoffset;
1.1312 + uint16_t fcd = getFCD(text, &temp, result);
1.1313 + if ((fcd >> SECOND_LAST_BYTE_SHIFT_) == 0) {
1.1314 + if (fcd & LAST_BYTE_MASK_) {
1.1315 + return textoffset;
1.1316 + }
1.1317 + return result;
1.1318 + }
1.1319 + if (textoffset == 0) {
1.1320 + return 0;
1.1321 + }
1.1322 + }
1.1323 + }
1.1324 + return textoffset;
1.1325 +}
1.1326 +
1.1327 +/**
1.1328 +* Getting the indexes of the accents that are not blocked in the argument
1.1329 +* accent array
1.1330 +* @param accents array of accents in nfd terminated by a 0.
1.1331 +* @param accentsindex array of indexes of the accents that are not blocked
1.1332 +*/
1.1333 +static
1.1334 +inline int getUnblockedAccentIndex(UChar *accents, int32_t *accentsindex)
1.1335 +{
1.1336 + int32_t index = 0;
1.1337 + int32_t length = u_strlen(accents);
1.1338 + UChar32 codepoint = 0;
1.1339 + int cclass = 0;
1.1340 + int result = 0;
1.1341 + int32_t temp;
1.1342 + while (index < length) {
1.1343 + temp = index;
1.1344 + U16_NEXT(accents, index, length, codepoint);
1.1345 + if (u_getCombiningClass(codepoint) != cclass) {
1.1346 + cclass = u_getCombiningClass(codepoint);
1.1347 + accentsindex[result] = temp;
1.1348 + result ++;
1.1349 + }
1.1350 + }
1.1351 + accentsindex[result] = length;
1.1352 + return result;
1.1353 +}
1.1354 +
1.1355 +/**
1.1356 +* Appends 3 UChar arrays to a destination array.
1.1357 +* Creates a new array if we run out of space. The caller will have to
1.1358 +* manually deallocate the newly allocated array.
1.1359 +* Internal method, status assumed to be success, caller has to check status
1.1360 +* before calling this method. destination not to be NULL and has at least
1.1361 +* size destinationlength.
1.1362 +* @param destination target array
1.1363 +* @param destinationlength target array size, returning the appended length
1.1364 +* @param source1 null-terminated first array
1.1365 +* @param source2 second array
1.1366 +* @param source2length length of seond array
1.1367 +* @param source3 null-terminated third array
1.1368 +* @param status error status if any
1.1369 +* @return new destination array, destination if there was no new allocation
1.1370 +*/
1.1371 +static
1.1372 +inline UChar * addToUCharArray( UChar *destination,
1.1373 + int32_t *destinationlength,
1.1374 + const UChar *source1,
1.1375 + const UChar *source2,
1.1376 + int32_t source2length,
1.1377 + const UChar *source3,
1.1378 + UErrorCode *status)
1.1379 +{
1.1380 + int32_t source1length = source1 ? u_strlen(source1) : 0;
1.1381 + int32_t source3length = source3 ? u_strlen(source3) : 0;
1.1382 + if (*destinationlength < source1length + source2length + source3length +
1.1383 + 1)
1.1384 + {
1.1385 + destination = (UChar *)allocateMemory(
1.1386 + (source1length + source2length + source3length + 1) * sizeof(UChar),
1.1387 + status);
1.1388 + // if error allocating memory, status will be
1.1389 + // U_MEMORY_ALLOCATION_ERROR
1.1390 + if (U_FAILURE(*status)) {
1.1391 + *destinationlength = 0;
1.1392 + return NULL;
1.1393 + }
1.1394 + }
1.1395 + if (source1length != 0) {
1.1396 + uprv_memcpy(destination, source1, sizeof(UChar) * source1length);
1.1397 + }
1.1398 + if (source2length != 0) {
1.1399 + uprv_memcpy(destination + source1length, source2,
1.1400 + sizeof(UChar) * source2length);
1.1401 + }
1.1402 + if (source3length != 0) {
1.1403 + uprv_memcpy(destination + source1length + source2length, source3,
1.1404 + sizeof(UChar) * source3length);
1.1405 + }
1.1406 + *destinationlength = source1length + source2length + source3length;
1.1407 + return destination;
1.1408 +}
1.1409 +
1.1410 +/**
1.1411 +* Running through a collation element iterator to see if the contents matches
1.1412 +* pattern in string search data
1.1413 +* @param strsrch string search data
1.1414 +* @param coleiter collation element iterator
1.1415 +* @return TRUE if a match if found, FALSE otherwise
1.1416 +*/
1.1417 +static
1.1418 +inline UBool checkCollationMatch(const UStringSearch *strsrch,
1.1419 + UCollationElements *coleiter)
1.1420 +{
1.1421 + int patternceindex = strsrch->pattern.CELength;
1.1422 + int32_t *patternce = strsrch->pattern.CE;
1.1423 + UErrorCode status = U_ZERO_ERROR;
1.1424 + while (patternceindex > 0) {
1.1425 + int32_t ce = getCE(strsrch, ucol_next(coleiter, &status));
1.1426 + if (ce == UCOL_IGNORABLE) {
1.1427 + continue;
1.1428 + }
1.1429 + if (U_FAILURE(status) || ce != *patternce) {
1.1430 + return FALSE;
1.1431 + }
1.1432 + patternce ++;
1.1433 + patternceindex --;
1.1434 + }
1.1435 + return TRUE;
1.1436 +}
1.1437 +
1.1438 +/**
1.1439 +* Rearranges the front accents to try matching.
1.1440 +* Prefix accents in the text will be grouped according to their combining
1.1441 +* class and the groups will be mixed and matched to try find the perfect
1.1442 +* match with the pattern.
1.1443 +* So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1.1444 +* step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1.1445 +* "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1.1446 +* "\u0301\u0325".
1.1447 +* step 2: check if any of the generated substrings matches the pattern.
1.1448 +* Internal method, status is assumed to be success, caller has to check status
1.1449 +* before calling this method.
1.1450 +* @param strsrch string search match
1.1451 +* @param start first offset of the accents to start searching
1.1452 +* @param end start of the last accent set
1.1453 +* @param status output error status if any
1.1454 +* @return USEARCH_DONE if a match is not found, otherwise return the starting
1.1455 +* offset of the match. Note this start includes all preceding accents.
1.1456 +*/
1.1457 +static
1.1458 +int32_t doNextCanonicalPrefixMatch(UStringSearch *strsrch,
1.1459 + int32_t start,
1.1460 + int32_t end,
1.1461 + UErrorCode *status)
1.1462 +{
1.1463 + const UChar *text = strsrch->search->text;
1.1464 + int32_t textlength = strsrch->search->textLength;
1.1465 + int32_t tempstart = start;
1.1466 +
1.1467 + if ((getFCD(text, &tempstart, textlength) & LAST_BYTE_MASK_) == 0) {
1.1468 + // die... failed at a base character
1.1469 + return USEARCH_DONE;
1.1470 + }
1.1471 +
1.1472 + int32_t offset = getNextBaseOffset(text, tempstart, textlength);
1.1473 + start = getPreviousBaseOffset(text, tempstart);
1.1474 +
1.1475 + UChar accents[INITIAL_ARRAY_SIZE_];
1.1476 + // normalizing the offensive string
1.1477 + unorm_normalize(text + start, offset - start, UNORM_NFD, 0, accents,
1.1478 + INITIAL_ARRAY_SIZE_, status);
1.1479 + if (U_FAILURE(*status)) {
1.1480 + return USEARCH_DONE;
1.1481 + }
1.1482 +
1.1483 + int32_t accentsindex[INITIAL_ARRAY_SIZE_];
1.1484 + int32_t accentsize = getUnblockedAccentIndex(accents,
1.1485 + accentsindex);
1.1486 + int32_t count = (2 << (accentsize - 1)) - 1;
1.1487 + UChar buffer[INITIAL_ARRAY_SIZE_];
1.1488 + UCollationElements *coleiter = strsrch->utilIter;
1.1489 + while (U_SUCCESS(*status) && count > 0) {
1.1490 + UChar *rearrange = strsrch->canonicalPrefixAccents;
1.1491 + // copy the base characters
1.1492 + for (int k = 0; k < accentsindex[0]; k ++) {
1.1493 + *rearrange ++ = accents[k];
1.1494 + }
1.1495 + // forming all possible canonical rearrangement by dropping
1.1496 + // sets of accents
1.1497 + for (int i = 0; i <= accentsize - 1; i ++) {
1.1498 + int32_t mask = 1 << (accentsize - i - 1);
1.1499 + if (count & mask) {
1.1500 + for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1.1501 + *rearrange ++ = accents[j];
1.1502 + }
1.1503 + }
1.1504 + }
1.1505 + *rearrange = 0;
1.1506 + int32_t matchsize = INITIAL_ARRAY_SIZE_;
1.1507 + UChar *match = addToUCharArray(buffer, &matchsize,
1.1508 + strsrch->canonicalPrefixAccents,
1.1509 + strsrch->search->text + offset,
1.1510 + end - offset,
1.1511 + strsrch->canonicalSuffixAccents,
1.1512 + status);
1.1513 +
1.1514 + // if status is a failure, ucol_setText does nothing.
1.1515 + // run the collator iterator through this match
1.1516 + ucol_setText(coleiter, match, matchsize, status);
1.1517 + if (U_SUCCESS(*status)) {
1.1518 + if (checkCollationMatch(strsrch, coleiter)) {
1.1519 + if (match != buffer) {
1.1520 + uprv_free(match);
1.1521 + }
1.1522 + return start;
1.1523 + }
1.1524 + }
1.1525 + count --;
1.1526 + }
1.1527 + return USEARCH_DONE;
1.1528 +}
1.1529 +
1.1530 +/**
1.1531 +* Gets the offset to the safe point in text before textoffset.
1.1532 +* ie. not the middle of a contraction, swappable characters or supplementary
1.1533 +* characters.
1.1534 +* @param collator collation sata
1.1535 +* @param text string to work with
1.1536 +* @param textoffset offset in string
1.1537 +* @param textlength length of text string
1.1538 +* @return offset to the previous safe character
1.1539 +*/
1.1540 +static
1.1541 +inline uint32_t getPreviousSafeOffset(const UCollator *collator,
1.1542 + const UChar *text,
1.1543 + int32_t textoffset)
1.1544 +{
1.1545 + int32_t result = textoffset; // first contraction character
1.1546 + while (result != 0 && ucol_unsafeCP(text[result - 1], collator)) {
1.1547 + result --;
1.1548 + }
1.1549 + if (result != 0) {
1.1550 + // the first contraction character is consider unsafe here
1.1551 + result --;
1.1552 + }
1.1553 + return result;
1.1554 +}
1.1555 +
1.1556 +/**
1.1557 +* Cleaning up after we passed the safe zone
1.1558 +* @param strsrch string search data
1.1559 +* @param safetext safe text array
1.1560 +* @param safebuffer safe text buffer
1.1561 +* @param coleiter collation element iterator for safe text
1.1562 +*/
1.1563 +static
1.1564 +inline void cleanUpSafeText(const UStringSearch *strsrch, UChar *safetext,
1.1565 + UChar *safebuffer)
1.1566 +{
1.1567 + if (safetext != safebuffer && safetext != strsrch->canonicalSuffixAccents)
1.1568 + {
1.1569 + uprv_free(safetext);
1.1570 + }
1.1571 +}
1.1572 +
1.1573 +/**
1.1574 +* Take the rearranged end accents and tries matching. If match failed at
1.1575 +* a seperate preceding set of accents (seperated from the rearranged on by
1.1576 +* at least a base character) then we rearrange the preceding accents and
1.1577 +* tries matching again.
1.1578 +* We allow skipping of the ends of the accent set if the ces do not match.
1.1579 +* However if the failure is found before the accent set, it fails.
1.1580 +* Internal method, status assumed to be success, caller has to check status
1.1581 +* before calling this method.
1.1582 +* @param strsrch string search data
1.1583 +* @param textoffset of the start of the rearranged accent
1.1584 +* @param status output error status if any
1.1585 +* @return USEARCH_DONE if a match is not found, otherwise return the starting
1.1586 +* offset of the match. Note this start includes all preceding accents.
1.1587 +*/
1.1588 +static
1.1589 +int32_t doNextCanonicalSuffixMatch(UStringSearch *strsrch,
1.1590 + int32_t textoffset,
1.1591 + UErrorCode *status)
1.1592 +{
1.1593 + const UChar *text = strsrch->search->text;
1.1594 + const UCollator *collator = strsrch->collator;
1.1595 + int32_t safelength = 0;
1.1596 + UChar *safetext;
1.1597 + int32_t safetextlength;
1.1598 + UChar safebuffer[INITIAL_ARRAY_SIZE_];
1.1599 + UCollationElements *coleiter = strsrch->utilIter;
1.1600 + int32_t safeoffset = textoffset;
1.1601 +
1.1602 + if (textoffset != 0 && ucol_unsafeCP(strsrch->canonicalSuffixAccents[0],
1.1603 + collator)) {
1.1604 + safeoffset = getPreviousSafeOffset(collator, text, textoffset);
1.1605 + safelength = textoffset - safeoffset;
1.1606 + safetextlength = INITIAL_ARRAY_SIZE_;
1.1607 + safetext = addToUCharArray(safebuffer, &safetextlength, NULL,
1.1608 + text + safeoffset, safelength,
1.1609 + strsrch->canonicalSuffixAccents,
1.1610 + status);
1.1611 + }
1.1612 + else {
1.1613 + safetextlength = u_strlen(strsrch->canonicalSuffixAccents);
1.1614 + safetext = strsrch->canonicalSuffixAccents;
1.1615 + }
1.1616 +
1.1617 + // if status is a failure, ucol_setText does nothing
1.1618 + ucol_setText(coleiter, safetext, safetextlength, status);
1.1619 + // status checked in loop below
1.1620 +
1.1621 + int32_t *ce = strsrch->pattern.CE;
1.1622 + int32_t celength = strsrch->pattern.CELength;
1.1623 + int ceindex = celength - 1;
1.1624 + UBool isSafe = TRUE; // indication flag for position in safe zone
1.1625 +
1.1626 + while (ceindex >= 0) {
1.1627 + int32_t textce = ucol_previous(coleiter, status);
1.1628 + if (U_FAILURE(*status)) {
1.1629 + if (isSafe) {
1.1630 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1631 + }
1.1632 + return USEARCH_DONE;
1.1633 + }
1.1634 + if (textce == UCOL_NULLORDER) {
1.1635 + // check if we have passed the safe buffer
1.1636 + if (coleiter == strsrch->textIter) {
1.1637 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1638 + return USEARCH_DONE;
1.1639 + }
1.1640 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1641 + safetext = safebuffer;
1.1642 + coleiter = strsrch->textIter;
1.1643 + setColEIterOffset(coleiter, safeoffset);
1.1644 + // status checked at the start of the loop
1.1645 + isSafe = FALSE;
1.1646 + continue;
1.1647 + }
1.1648 + textce = getCE(strsrch, textce);
1.1649 + if (textce != UCOL_IGNORABLE && textce != ce[ceindex]) {
1.1650 + // do the beginning stuff
1.1651 + int32_t failedoffset = getColElemIterOffset(coleiter, FALSE);
1.1652 + if (isSafe && failedoffset >= safelength) {
1.1653 + // alas... no hope. failed at rearranged accent set
1.1654 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1655 + return USEARCH_DONE;
1.1656 + }
1.1657 + else {
1.1658 + if (isSafe) {
1.1659 + failedoffset += safeoffset;
1.1660 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1661 + }
1.1662 +
1.1663 + // try rearranging the front accents
1.1664 + int32_t result = doNextCanonicalPrefixMatch(strsrch,
1.1665 + failedoffset, textoffset, status);
1.1666 + if (result != USEARCH_DONE) {
1.1667 + // if status is a failure, ucol_setOffset does nothing
1.1668 + setColEIterOffset(strsrch->textIter, result);
1.1669 + }
1.1670 + if (U_FAILURE(*status)) {
1.1671 + return USEARCH_DONE;
1.1672 + }
1.1673 + return result;
1.1674 + }
1.1675 + }
1.1676 + if (textce == ce[ceindex]) {
1.1677 + ceindex --;
1.1678 + }
1.1679 + }
1.1680 + // set offset here
1.1681 + if (isSafe) {
1.1682 + int32_t result = getColElemIterOffset(coleiter, FALSE);
1.1683 + // sets the text iterator here with the correct expansion and offset
1.1684 + int32_t leftoverces = getExpansionPrefix(coleiter);
1.1685 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.1686 + if (result >= safelength) {
1.1687 + result = textoffset;
1.1688 + }
1.1689 + else {
1.1690 + result += safeoffset;
1.1691 + }
1.1692 + setColEIterOffset(strsrch->textIter, result);
1.1693 + strsrch->textIter->iteratordata_.toReturn =
1.1694 + setExpansionPrefix(strsrch->textIter, leftoverces);
1.1695 + return result;
1.1696 + }
1.1697 +
1.1698 + return ucol_getOffset(coleiter);
1.1699 +}
1.1700 +
1.1701 +/**
1.1702 +* Trying out the substring and sees if it can be a canonical match.
1.1703 +* This will try normalizing the end accents and arranging them into canonical
1.1704 +* equivalents and check their corresponding ces with the pattern ce.
1.1705 +* Suffix accents in the text will be grouped according to their combining
1.1706 +* class and the groups will be mixed and matched to try find the perfect
1.1707 +* match with the pattern.
1.1708 +* So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1.1709 +* step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1.1710 +* "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1.1711 +* "\u0301\u0325".
1.1712 +* step 2: check if any of the generated substrings matches the pattern.
1.1713 +* Internal method, status assumed to be success, caller has to check status
1.1714 +* before calling this method.
1.1715 +* @param strsrch string search data
1.1716 +* @param textoffset end offset in the collation element text that ends with
1.1717 +* the accents to be rearranged
1.1718 +* @param status error status if any
1.1719 +* @return TRUE if the match is valid, FALSE otherwise
1.1720 +*/
1.1721 +static
1.1722 +UBool doNextCanonicalMatch(UStringSearch *strsrch,
1.1723 + int32_t textoffset,
1.1724 + UErrorCode *status)
1.1725 +{
1.1726 + const UChar *text = strsrch->search->text;
1.1727 + int32_t temp = textoffset;
1.1728 + U16_BACK_1(text, 0, temp);
1.1729 + if ((getFCD(text, &temp, textoffset) & LAST_BYTE_MASK_) == 0) {
1.1730 + UCollationElements *coleiter = strsrch->textIter;
1.1731 + int32_t offset = getColElemIterOffset(coleiter, FALSE);
1.1732 + if (strsrch->pattern.hasPrefixAccents) {
1.1733 + offset = doNextCanonicalPrefixMatch(strsrch, offset, textoffset,
1.1734 + status);
1.1735 + if (U_SUCCESS(*status) && offset != USEARCH_DONE) {
1.1736 + setColEIterOffset(coleiter, offset);
1.1737 + return TRUE;
1.1738 + }
1.1739 + }
1.1740 + return FALSE;
1.1741 + }
1.1742 +
1.1743 + if (!strsrch->pattern.hasSuffixAccents) {
1.1744 + return FALSE;
1.1745 + }
1.1746 +
1.1747 + UChar accents[INITIAL_ARRAY_SIZE_];
1.1748 + // offset to the last base character in substring to search
1.1749 + int32_t baseoffset = getPreviousBaseOffset(text, textoffset);
1.1750 + // normalizing the offensive string
1.1751 + unorm_normalize(text + baseoffset, textoffset - baseoffset, UNORM_NFD,
1.1752 + 0, accents, INITIAL_ARRAY_SIZE_, status);
1.1753 + // status checked in loop below
1.1754 +
1.1755 + int32_t accentsindex[INITIAL_ARRAY_SIZE_];
1.1756 + int32_t size = getUnblockedAccentIndex(accents, accentsindex);
1.1757 +
1.1758 + // 2 power n - 1 plus the full set of accents
1.1759 + int32_t count = (2 << (size - 1)) - 1;
1.1760 + while (U_SUCCESS(*status) && count > 0) {
1.1761 + UChar *rearrange = strsrch->canonicalSuffixAccents;
1.1762 + // copy the base characters
1.1763 + for (int k = 0; k < accentsindex[0]; k ++) {
1.1764 + *rearrange ++ = accents[k];
1.1765 + }
1.1766 + // forming all possible canonical rearrangement by dropping
1.1767 + // sets of accents
1.1768 + for (int i = 0; i <= size - 1; i ++) {
1.1769 + int32_t mask = 1 << (size - i - 1);
1.1770 + if (count & mask) {
1.1771 + for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1.1772 + *rearrange ++ = accents[j];
1.1773 + }
1.1774 + }
1.1775 + }
1.1776 + *rearrange = 0;
1.1777 + int32_t offset = doNextCanonicalSuffixMatch(strsrch, baseoffset,
1.1778 + status);
1.1779 + if (offset != USEARCH_DONE) {
1.1780 + return TRUE; // match found
1.1781 + }
1.1782 + count --;
1.1783 + }
1.1784 + return FALSE;
1.1785 +}
1.1786 +
1.1787 +/**
1.1788 +* Gets the previous base character offset depending on the string search
1.1789 +* pattern data
1.1790 +* @param strsrch string search data
1.1791 +* @param textoffset current offset, current character
1.1792 +* @return the offset of the next character after this base character or itself
1.1793 +* if it is a composed character with accents
1.1794 +*/
1.1795 +static
1.1796 +inline int32_t getPreviousUStringSearchBaseOffset(UStringSearch *strsrch,
1.1797 + int32_t textoffset)
1.1798 +{
1.1799 + if (strsrch->pattern.hasPrefixAccents && textoffset > 0) {
1.1800 + const UChar *text = strsrch->search->text;
1.1801 + int32_t offset = textoffset;
1.1802 + if (getFCD(text, &offset, strsrch->search->textLength) >>
1.1803 + SECOND_LAST_BYTE_SHIFT_) {
1.1804 + return getPreviousBaseOffset(text, textoffset);
1.1805 + }
1.1806 + }
1.1807 + return textoffset;
1.1808 +}
1.1809 +
1.1810 +/**
1.1811 +* Checks match for contraction.
1.1812 +* If the match ends with a partial contraction we fail.
1.1813 +* If the match starts too far off (because of backwards iteration) we try to
1.1814 +* chip off the extra characters
1.1815 +* Internal method, status assumed to be success, caller has to check status
1.1816 +* before calling this method.
1.1817 +* @param strsrch string search data
1.1818 +* @param start offset of potential match, to be modified if necessary
1.1819 +* @param end offset of potential match, to be modified if necessary
1.1820 +* @param status output error status if any
1.1821 +* @return TRUE if match passes the contraction test, FALSE otherwise
1.1822 +*/
1.1823 +static
1.1824 +UBool checkNextCanonicalContractionMatch(UStringSearch *strsrch,
1.1825 + int32_t *start,
1.1826 + int32_t *end,
1.1827 + UErrorCode *status)
1.1828 +{
1.1829 + UCollationElements *coleiter = strsrch->textIter;
1.1830 + int32_t textlength = strsrch->search->textLength;
1.1831 + int32_t temp = *start;
1.1832 + const UCollator *collator = strsrch->collator;
1.1833 + const UChar *text = strsrch->search->text;
1.1834 + // This part checks if either ends of the match contains potential
1.1835 + // contraction. If so we'll have to iterate through them
1.1836 + if ((*end < textlength && ucol_unsafeCP(text[*end], collator)) ||
1.1837 + (*start + 1 < textlength
1.1838 + && ucol_unsafeCP(text[*start + 1], collator))) {
1.1839 + int32_t expansion = getExpansionPrefix(coleiter);
1.1840 + UBool expandflag = expansion > 0;
1.1841 + setColEIterOffset(coleiter, *start);
1.1842 + while (expansion > 0) {
1.1843 + // getting rid of the redundant ce, caused by setOffset.
1.1844 + // since backward contraction/expansion may have extra ces if we
1.1845 + // are in the normalization buffer, hasAccentsBeforeMatch would
1.1846 + // have taken care of it.
1.1847 + // E.g. the character \u01FA will have an expansion of 3, but if
1.1848 + // we are only looking for acute and ring \u030A and \u0301, we'll
1.1849 + // have to skip the first ce in the expansion buffer.
1.1850 + ucol_next(coleiter, status);
1.1851 + if (U_FAILURE(*status)) {
1.1852 + return FALSE;
1.1853 + }
1.1854 + if (ucol_getOffset(coleiter) != temp) {
1.1855 + *start = temp;
1.1856 + temp = ucol_getOffset(coleiter);
1.1857 + }
1.1858 + expansion --;
1.1859 + }
1.1860 +
1.1861 + int32_t *patternce = strsrch->pattern.CE;
1.1862 + int32_t patterncelength = strsrch->pattern.CELength;
1.1863 + int32_t count = 0;
1.1864 + int32_t textlength = strsrch->search->textLength;
1.1865 + while (count < patterncelength) {
1.1866 + int32_t ce = getCE(strsrch, ucol_next(coleiter, status));
1.1867 + // status checked below, note that if status is a failure
1.1868 + // ucol_next returns UCOL_NULLORDER
1.1869 + if (ce == UCOL_IGNORABLE) {
1.1870 + continue;
1.1871 + }
1.1872 + if (expandflag && count == 0 && ucol_getOffset(coleiter) != temp) {
1.1873 + *start = temp;
1.1874 + temp = ucol_getOffset(coleiter);
1.1875 + }
1.1876 +
1.1877 + if (count == 0 && ce != patternce[0]) {
1.1878 + // accents may have extra starting ces, this occurs when a
1.1879 + // pure accent pattern is matched without rearrangement
1.1880 + // text \u0325\u0300 and looking for \u0300
1.1881 + int32_t expected = patternce[0];
1.1882 + if (getFCD(text, start, textlength) & LAST_BYTE_MASK_) {
1.1883 + ce = getCE(strsrch, ucol_next(coleiter, status));
1.1884 + while (U_SUCCESS(*status) && ce != expected &&
1.1885 + ce != UCOL_NULLORDER &&
1.1886 + ucol_getOffset(coleiter) <= *end) {
1.1887 + ce = getCE(strsrch, ucol_next(coleiter, status));
1.1888 + }
1.1889 + }
1.1890 + }
1.1891 + if (U_FAILURE(*status) || ce != patternce[count]) {
1.1892 + (*end) ++;
1.1893 + *end = getNextUStringSearchBaseOffset(strsrch, *end);
1.1894 + return FALSE;
1.1895 + }
1.1896 + count ++;
1.1897 + }
1.1898 + }
1.1899 + return TRUE;
1.1900 +}
1.1901 +
1.1902 +/**
1.1903 +* Checks and sets the match information if found.
1.1904 +* Checks
1.1905 +* <ul>
1.1906 +* <li> the potential match does not repeat the previous match
1.1907 +* <li> boundaries are correct
1.1908 +* <li> potential match does not end in the middle of a contraction
1.1909 +* <li> identical matches
1.1910 +* <\ul>
1.1911 +* Otherwise the offset will be shifted to the next character.
1.1912 +* Internal method, status assumed to be success, caller has to check the
1.1913 +* status before calling this method.
1.1914 +* @param strsrch string search data
1.1915 +* @param textoffset offset in the collation element text. the returned value
1.1916 +* will be the truncated end offset of the match or the new start
1.1917 +* search offset.
1.1918 +* @param status output error status if any
1.1919 +* @return TRUE if the match is valid, FALSE otherwise
1.1920 +*/
1.1921 +static
1.1922 +inline UBool checkNextCanonicalMatch(UStringSearch *strsrch,
1.1923 + int32_t *textoffset,
1.1924 + UErrorCode *status)
1.1925 +{
1.1926 + // to ensure that the start and ends are not composite characters
1.1927 + UCollationElements *coleiter = strsrch->textIter;
1.1928 + // if we have a canonical accent match
1.1929 + if ((strsrch->pattern.hasSuffixAccents &&
1.1930 + strsrch->canonicalSuffixAccents[0]) ||
1.1931 + (strsrch->pattern.hasPrefixAccents &&
1.1932 + strsrch->canonicalPrefixAccents[0])) {
1.1933 + strsrch->search->matchedIndex = getPreviousUStringSearchBaseOffset(
1.1934 + strsrch,
1.1935 + ucol_getOffset(coleiter));
1.1936 + strsrch->search->matchedLength = *textoffset -
1.1937 + strsrch->search->matchedIndex;
1.1938 + return TRUE;
1.1939 + }
1.1940 +
1.1941 + int32_t start = getColElemIterOffset(coleiter, FALSE);
1.1942 + if (!checkNextCanonicalContractionMatch(strsrch, &start, textoffset,
1.1943 + status) || U_FAILURE(*status)) {
1.1944 + return FALSE;
1.1945 + }
1.1946 +
1.1947 + start = getPreviousUStringSearchBaseOffset(strsrch, start);
1.1948 + // this totally matches, however we need to check if it is repeating
1.1949 + if (checkRepeatedMatch(strsrch, start, *textoffset) ||
1.1950 + !isBreakUnit(strsrch, start, *textoffset) ||
1.1951 + !checkIdentical(strsrch, start, *textoffset)) {
1.1952 + (*textoffset) ++;
1.1953 + *textoffset = getNextBaseOffset(strsrch->search->text, *textoffset,
1.1954 + strsrch->search->textLength);
1.1955 + return FALSE;
1.1956 + }
1.1957 +
1.1958 + strsrch->search->matchedIndex = start;
1.1959 + strsrch->search->matchedLength = *textoffset - start;
1.1960 + return TRUE;
1.1961 +}
1.1962 +
1.1963 +/**
1.1964 +* Shifting the collation element iterator position forward to prepare for
1.1965 +* a preceding match. If the first character is a unsafe character, we'll only
1.1966 +* shift by 1 to capture contractions, normalization etc.
1.1967 +* Internal method, status assumed to be success, caller has to check status
1.1968 +* before calling this method.
1.1969 +* @param text strsrch string search data
1.1970 +* @param textoffset start text position to do search
1.1971 +* @param ce the text ce which failed the match.
1.1972 +* @param patternceindex index of the ce within the pattern ce buffer which
1.1973 +* failed the match
1.1974 +* @return final offset
1.1975 +*/
1.1976 +static
1.1977 +inline int32_t reverseShift(UStringSearch *strsrch,
1.1978 + int32_t textoffset,
1.1979 + int32_t ce,
1.1980 + int32_t patternceindex)
1.1981 +{
1.1982 + if (strsrch->search->isOverlap) {
1.1983 + if (textoffset != strsrch->search->textLength) {
1.1984 + textoffset --;
1.1985 + }
1.1986 + else {
1.1987 + textoffset -= strsrch->pattern.defaultShiftSize;
1.1988 + }
1.1989 + }
1.1990 + else {
1.1991 + if (ce != UCOL_NULLORDER) {
1.1992 + int32_t shift = strsrch->pattern.backShift[hash(ce)];
1.1993 +
1.1994 + // this is to adjust for characters in the middle of the substring
1.1995 + // for matching that failed.
1.1996 + int32_t adjust = patternceindex;
1.1997 + if (adjust > 1 && shift > adjust) {
1.1998 + shift -= adjust - 1;
1.1999 + }
1.2000 + textoffset -= shift;
1.2001 + }
1.2002 + else {
1.2003 + textoffset -= strsrch->pattern.defaultShiftSize;
1.2004 + }
1.2005 + }
1.2006 + textoffset = getPreviousUStringSearchBaseOffset(strsrch, textoffset);
1.2007 + return textoffset;
1.2008 +}
1.2009 +
1.2010 +/**
1.2011 +* Checks match for contraction.
1.2012 +* If the match starts with a partial contraction we fail.
1.2013 +* Internal method, status assumed to be success, caller has to check status
1.2014 +* before calling this method.
1.2015 +* @param strsrch string search data
1.2016 +* @param start offset of potential match, to be modified if necessary
1.2017 +* @param end offset of potential match, to be modified if necessary
1.2018 +* @param status output error status if any
1.2019 +* @return TRUE if match passes the contraction test, FALSE otherwise
1.2020 +*/
1.2021 +static
1.2022 +UBool checkPreviousExactContractionMatch(UStringSearch *strsrch,
1.2023 + int32_t *start,
1.2024 + int32_t *end, UErrorCode *status)
1.2025 +{
1.2026 + UCollationElements *coleiter = strsrch->textIter;
1.2027 + int32_t textlength = strsrch->search->textLength;
1.2028 + int32_t temp = *end;
1.2029 + const UCollator *collator = strsrch->collator;
1.2030 + const UChar *text = strsrch->search->text;
1.2031 + // This part checks if either if the start of the match contains potential
1.2032 + // contraction. If so we'll have to iterate through them
1.2033 + // Since we used ucol_next while previously looking for the potential
1.2034 + // match, this guarantees that our end will not be a partial contraction,
1.2035 + // or a partial supplementary character.
1.2036 + if (*start < textlength && ucol_unsafeCP(text[*start], collator)) {
1.2037 + int32_t expansion = getExpansionSuffix(coleiter);
1.2038 + UBool expandflag = expansion > 0;
1.2039 + setColEIterOffset(coleiter, *end);
1.2040 + while (U_SUCCESS(*status) && expansion > 0) {
1.2041 + // getting rid of the redundant ce
1.2042 + // since forward contraction/expansion may have extra ces
1.2043 + // if we are in the normalization buffer, hasAccentsBeforeMatch
1.2044 + // would have taken care of it.
1.2045 + // E.g. the character \u01FA will have an expansion of 3, but if
1.2046 + // we are only looking for A ring A\u030A, we'll have to skip the
1.2047 + // last ce in the expansion buffer
1.2048 + ucol_previous(coleiter, status);
1.2049 + if (U_FAILURE(*status)) {
1.2050 + return FALSE;
1.2051 + }
1.2052 + if (ucol_getOffset(coleiter) != temp) {
1.2053 + *end = temp;
1.2054 + temp = ucol_getOffset(coleiter);
1.2055 + }
1.2056 + expansion --;
1.2057 + }
1.2058 +
1.2059 + int32_t *patternce = strsrch->pattern.CE;
1.2060 + int32_t patterncelength = strsrch->pattern.CELength;
1.2061 + int32_t count = patterncelength;
1.2062 + while (count > 0) {
1.2063 + int32_t ce = getCE(strsrch, ucol_previous(coleiter, status));
1.2064 + // status checked below, note that if status is a failure
1.2065 + // ucol_previous returns UCOL_NULLORDER
1.2066 + if (ce == UCOL_IGNORABLE) {
1.2067 + continue;
1.2068 + }
1.2069 + if (expandflag && count == 0 &&
1.2070 + getColElemIterOffset(coleiter, FALSE) != temp) {
1.2071 + *end = temp;
1.2072 + temp = ucol_getOffset(coleiter);
1.2073 + }
1.2074 + if (U_FAILURE(*status) || ce != patternce[count - 1]) {
1.2075 + (*start) --;
1.2076 + *start = getPreviousBaseOffset(text, *start);
1.2077 + return FALSE;
1.2078 + }
1.2079 + count --;
1.2080 + }
1.2081 + }
1.2082 + return TRUE;
1.2083 +}
1.2084 +
1.2085 +/**
1.2086 +* Checks and sets the match information if found.
1.2087 +* Checks
1.2088 +* <ul>
1.2089 +* <li> the current match does not repeat the last match
1.2090 +* <li> boundaries are correct
1.2091 +* <li> exact matches has no extra accents
1.2092 +* <li> identical matches
1.2093 +* <\ul>
1.2094 +* Otherwise the offset will be shifted to the preceding character.
1.2095 +* Internal method, status assumed to be success, caller has to check status
1.2096 +* before calling this method.
1.2097 +* @param strsrch string search data
1.2098 +* @param collator
1.2099 +* @param coleiter collation element iterator
1.2100 +* @param text string
1.2101 +* @param textoffset offset in the collation element text. the returned value
1.2102 +* will be the truncated start offset of the match or the new start
1.2103 +* search offset.
1.2104 +* @param status output error status if any
1.2105 +* @return TRUE if the match is valid, FALSE otherwise
1.2106 +*/
1.2107 +static
1.2108 +inline UBool checkPreviousExactMatch(UStringSearch *strsrch,
1.2109 + int32_t *textoffset,
1.2110 + UErrorCode *status)
1.2111 +{
1.2112 + // to ensure that the start and ends are not composite characters
1.2113 + int32_t end = ucol_getOffset(strsrch->textIter);
1.2114 + if (!checkPreviousExactContractionMatch(strsrch, textoffset, &end, status)
1.2115 + || U_FAILURE(*status)) {
1.2116 + return FALSE;
1.2117 + }
1.2118 +
1.2119 + // this totally matches, however we need to check if it is repeating
1.2120 + // the old match
1.2121 + if (checkRepeatedMatch(strsrch, *textoffset, end) ||
1.2122 + !isBreakUnit(strsrch, *textoffset, end) ||
1.2123 + hasAccentsBeforeMatch(strsrch, *textoffset, end) ||
1.2124 + !checkIdentical(strsrch, *textoffset, end) ||
1.2125 + hasAccentsAfterMatch(strsrch, *textoffset, end)) {
1.2126 + (*textoffset) --;
1.2127 + *textoffset = getPreviousBaseOffset(strsrch->search->text,
1.2128 + *textoffset);
1.2129 + return FALSE;
1.2130 + }
1.2131 +
1.2132 + //Add breakiterator boundary check for primary strength search.
1.2133 + if (!strsrch->search->breakIter && strsrch->strength == UCOL_PRIMARY) {
1.2134 + checkBreakBoundary(strsrch, textoffset, &end);
1.2135 + }
1.2136 +
1.2137 + strsrch->search->matchedIndex = *textoffset;
1.2138 + strsrch->search->matchedLength = end - *textoffset;
1.2139 + return TRUE;
1.2140 +}
1.2141 +
1.2142 +/**
1.2143 +* Rearranges the end accents to try matching.
1.2144 +* Suffix accents in the text will be grouped according to their combining
1.2145 +* class and the groups will be mixed and matched to try find the perfect
1.2146 +* match with the pattern.
1.2147 +* So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1.2148 +* step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1.2149 +* "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1.2150 +* "\u0301\u0325".
1.2151 +* step 2: check if any of the generated substrings matches the pattern.
1.2152 +* Internal method, status assumed to be success, user has to check status
1.2153 +* before calling this method.
1.2154 +* @param strsrch string search match
1.2155 +* @param start offset of the first base character
1.2156 +* @param end start of the last accent set
1.2157 +* @param status only error status if any
1.2158 +* @return USEARCH_DONE if a match is not found, otherwise return the ending
1.2159 +* offset of the match. Note this start includes all following accents.
1.2160 +*/
1.2161 +static
1.2162 +int32_t doPreviousCanonicalSuffixMatch(UStringSearch *strsrch,
1.2163 + int32_t start,
1.2164 + int32_t end,
1.2165 + UErrorCode *status)
1.2166 +{
1.2167 + const UChar *text = strsrch->search->text;
1.2168 + int32_t tempend = end;
1.2169 +
1.2170 + U16_BACK_1(text, 0, tempend);
1.2171 + if (!(getFCD(text, &tempend, strsrch->search->textLength) &
1.2172 + LAST_BYTE_MASK_)) {
1.2173 + // die... failed at a base character
1.2174 + return USEARCH_DONE;
1.2175 + }
1.2176 + end = getNextBaseOffset(text, end, strsrch->search->textLength);
1.2177 +
1.2178 + if (U_SUCCESS(*status)) {
1.2179 + UChar accents[INITIAL_ARRAY_SIZE_];
1.2180 + int32_t offset = getPreviousBaseOffset(text, end);
1.2181 + // normalizing the offensive string
1.2182 + unorm_normalize(text + offset, end - offset, UNORM_NFD, 0, accents,
1.2183 + INITIAL_ARRAY_SIZE_, status);
1.2184 +
1.2185 + int32_t accentsindex[INITIAL_ARRAY_SIZE_];
1.2186 + int32_t accentsize = getUnblockedAccentIndex(accents,
1.2187 + accentsindex);
1.2188 + int32_t count = (2 << (accentsize - 1)) - 1;
1.2189 + UChar buffer[INITIAL_ARRAY_SIZE_];
1.2190 + UCollationElements *coleiter = strsrch->utilIter;
1.2191 + while (U_SUCCESS(*status) && count > 0) {
1.2192 + UChar *rearrange = strsrch->canonicalSuffixAccents;
1.2193 + // copy the base characters
1.2194 + for (int k = 0; k < accentsindex[0]; k ++) {
1.2195 + *rearrange ++ = accents[k];
1.2196 + }
1.2197 + // forming all possible canonical rearrangement by dropping
1.2198 + // sets of accents
1.2199 + for (int i = 0; i <= accentsize - 1; i ++) {
1.2200 + int32_t mask = 1 << (accentsize - i - 1);
1.2201 + if (count & mask) {
1.2202 + for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1.2203 + *rearrange ++ = accents[j];
1.2204 + }
1.2205 + }
1.2206 + }
1.2207 + *rearrange = 0;
1.2208 + int32_t matchsize = INITIAL_ARRAY_SIZE_;
1.2209 + UChar *match = addToUCharArray(buffer, &matchsize,
1.2210 + strsrch->canonicalPrefixAccents,
1.2211 + strsrch->search->text + start,
1.2212 + offset - start,
1.2213 + strsrch->canonicalSuffixAccents,
1.2214 + status);
1.2215 +
1.2216 + // run the collator iterator through this match
1.2217 + // if status is a failure ucol_setText does nothing
1.2218 + ucol_setText(coleiter, match, matchsize, status);
1.2219 + if (U_SUCCESS(*status)) {
1.2220 + if (checkCollationMatch(strsrch, coleiter)) {
1.2221 + if (match != buffer) {
1.2222 + uprv_free(match);
1.2223 + }
1.2224 + return end;
1.2225 + }
1.2226 + }
1.2227 + count --;
1.2228 + }
1.2229 + }
1.2230 + return USEARCH_DONE;
1.2231 +}
1.2232 +
1.2233 +/**
1.2234 +* Take the rearranged start accents and tries matching. If match failed at
1.2235 +* a seperate following set of accents (seperated from the rearranged on by
1.2236 +* at least a base character) then we rearrange the preceding accents and
1.2237 +* tries matching again.
1.2238 +* We allow skipping of the ends of the accent set if the ces do not match.
1.2239 +* However if the failure is found before the accent set, it fails.
1.2240 +* Internal method, status assumed to be success, caller has to check status
1.2241 +* before calling this method.
1.2242 +* @param strsrch string search data
1.2243 +* @param textoffset of the ends of the rearranged accent
1.2244 +* @param status output error status if any
1.2245 +* @return USEARCH_DONE if a match is not found, otherwise return the ending
1.2246 +* offset of the match. Note this start includes all following accents.
1.2247 +*/
1.2248 +static
1.2249 +int32_t doPreviousCanonicalPrefixMatch(UStringSearch *strsrch,
1.2250 + int32_t textoffset,
1.2251 + UErrorCode *status)
1.2252 +{
1.2253 + const UChar *text = strsrch->search->text;
1.2254 + const UCollator *collator = strsrch->collator;
1.2255 + int32_t safelength = 0;
1.2256 + UChar *safetext;
1.2257 + int32_t safetextlength;
1.2258 + UChar safebuffer[INITIAL_ARRAY_SIZE_];
1.2259 + int32_t safeoffset = textoffset;
1.2260 +
1.2261 + if (textoffset &&
1.2262 + ucol_unsafeCP(strsrch->canonicalPrefixAccents[
1.2263 + u_strlen(strsrch->canonicalPrefixAccents) - 1
1.2264 + ], collator)) {
1.2265 + safeoffset = getNextSafeOffset(collator, text, textoffset,
1.2266 + strsrch->search->textLength);
1.2267 + safelength = safeoffset - textoffset;
1.2268 + safetextlength = INITIAL_ARRAY_SIZE_;
1.2269 + safetext = addToUCharArray(safebuffer, &safetextlength,
1.2270 + strsrch->canonicalPrefixAccents,
1.2271 + text + textoffset, safelength,
1.2272 + NULL, status);
1.2273 + }
1.2274 + else {
1.2275 + safetextlength = u_strlen(strsrch->canonicalPrefixAccents);
1.2276 + safetext = strsrch->canonicalPrefixAccents;
1.2277 + }
1.2278 +
1.2279 + UCollationElements *coleiter = strsrch->utilIter;
1.2280 + // if status is a failure, ucol_setText does nothing
1.2281 + ucol_setText(coleiter, safetext, safetextlength, status);
1.2282 + // status checked in loop below
1.2283 +
1.2284 + int32_t *ce = strsrch->pattern.CE;
1.2285 + int32_t celength = strsrch->pattern.CELength;
1.2286 + int ceindex = 0;
1.2287 + UBool isSafe = TRUE; // safe zone indication flag for position
1.2288 + int32_t prefixlength = u_strlen(strsrch->canonicalPrefixAccents);
1.2289 +
1.2290 + while (ceindex < celength) {
1.2291 + int32_t textce = ucol_next(coleiter, status);
1.2292 + if (U_FAILURE(*status)) {
1.2293 + if (isSafe) {
1.2294 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2295 + }
1.2296 + return USEARCH_DONE;
1.2297 + }
1.2298 + if (textce == UCOL_NULLORDER) {
1.2299 + // check if we have passed the safe buffer
1.2300 + if (coleiter == strsrch->textIter) {
1.2301 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2302 + return USEARCH_DONE;
1.2303 + }
1.2304 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2305 + safetext = safebuffer;
1.2306 + coleiter = strsrch->textIter;
1.2307 + setColEIterOffset(coleiter, safeoffset);
1.2308 + // status checked at the start of the loop
1.2309 + isSafe = FALSE;
1.2310 + continue;
1.2311 + }
1.2312 + textce = getCE(strsrch, textce);
1.2313 + if (textce != UCOL_IGNORABLE && textce != ce[ceindex]) {
1.2314 + // do the beginning stuff
1.2315 + int32_t failedoffset = ucol_getOffset(coleiter);
1.2316 + if (isSafe && failedoffset <= prefixlength) {
1.2317 + // alas... no hope. failed at rearranged accent set
1.2318 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2319 + return USEARCH_DONE;
1.2320 + }
1.2321 + else {
1.2322 + if (isSafe) {
1.2323 + failedoffset = safeoffset - failedoffset;
1.2324 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2325 + }
1.2326 +
1.2327 + // try rearranging the end accents
1.2328 + int32_t result = doPreviousCanonicalSuffixMatch(strsrch,
1.2329 + textoffset, failedoffset, status);
1.2330 + if (result != USEARCH_DONE) {
1.2331 + // if status is a failure, ucol_setOffset does nothing
1.2332 + setColEIterOffset(strsrch->textIter, result);
1.2333 + }
1.2334 + if (U_FAILURE(*status)) {
1.2335 + return USEARCH_DONE;
1.2336 + }
1.2337 + return result;
1.2338 + }
1.2339 + }
1.2340 + if (textce == ce[ceindex]) {
1.2341 + ceindex ++;
1.2342 + }
1.2343 + }
1.2344 + // set offset here
1.2345 + if (isSafe) {
1.2346 + int32_t result = ucol_getOffset(coleiter);
1.2347 + // sets the text iterator here with the correct expansion and offset
1.2348 + int32_t leftoverces = getExpansionSuffix(coleiter);
1.2349 + cleanUpSafeText(strsrch, safetext, safebuffer);
1.2350 + if (result <= prefixlength) {
1.2351 + result = textoffset;
1.2352 + }
1.2353 + else {
1.2354 + result = textoffset + (safeoffset - result);
1.2355 + }
1.2356 + setColEIterOffset(strsrch->textIter, result);
1.2357 + setExpansionSuffix(strsrch->textIter, leftoverces);
1.2358 + return result;
1.2359 + }
1.2360 +
1.2361 + return ucol_getOffset(coleiter);
1.2362 +}
1.2363 +
1.2364 +/**
1.2365 +* Trying out the substring and sees if it can be a canonical match.
1.2366 +* This will try normalizing the starting accents and arranging them into
1.2367 +* canonical equivalents and check their corresponding ces with the pattern ce.
1.2368 +* Prefix accents in the text will be grouped according to their combining
1.2369 +* class and the groups will be mixed and matched to try find the perfect
1.2370 +* match with the pattern.
1.2371 +* So for instance looking for "\u0301" in "\u030A\u0301\u0325"
1.2372 +* step 1: split "\u030A\u0301" into 6 other type of potential accent substrings
1.2373 +* "\u030A", "\u0301", "\u0325", "\u030A\u0301", "\u030A\u0325",
1.2374 +* "\u0301\u0325".
1.2375 +* step 2: check if any of the generated substrings matches the pattern.
1.2376 +* Internal method, status assumed to be success, caller has to check status
1.2377 +* before calling this method.
1.2378 +* @param strsrch string search data
1.2379 +* @param textoffset start offset in the collation element text that starts
1.2380 +* with the accents to be rearranged
1.2381 +* @param status output error status if any
1.2382 +* @return TRUE if the match is valid, FALSE otherwise
1.2383 +*/
1.2384 +static
1.2385 +UBool doPreviousCanonicalMatch(UStringSearch *strsrch,
1.2386 + int32_t textoffset,
1.2387 + UErrorCode *status)
1.2388 +{
1.2389 + const UChar *text = strsrch->search->text;
1.2390 + int32_t temp = textoffset;
1.2391 + int32_t textlength = strsrch->search->textLength;
1.2392 + if ((getFCD(text, &temp, textlength) >> SECOND_LAST_BYTE_SHIFT_) == 0) {
1.2393 + UCollationElements *coleiter = strsrch->textIter;
1.2394 + int32_t offset = ucol_getOffset(coleiter);
1.2395 + if (strsrch->pattern.hasSuffixAccents) {
1.2396 + offset = doPreviousCanonicalSuffixMatch(strsrch, textoffset,
1.2397 + offset, status);
1.2398 + if (U_SUCCESS(*status) && offset != USEARCH_DONE) {
1.2399 + setColEIterOffset(coleiter, offset);
1.2400 + return TRUE;
1.2401 + }
1.2402 + }
1.2403 + return FALSE;
1.2404 + }
1.2405 +
1.2406 + if (!strsrch->pattern.hasPrefixAccents) {
1.2407 + return FALSE;
1.2408 + }
1.2409 +
1.2410 + UChar accents[INITIAL_ARRAY_SIZE_];
1.2411 + // offset to the last base character in substring to search
1.2412 + int32_t baseoffset = getNextBaseOffset(text, textoffset, textlength);
1.2413 + // normalizing the offensive string
1.2414 + unorm_normalize(text + textoffset, baseoffset - textoffset, UNORM_NFD,
1.2415 + 0, accents, INITIAL_ARRAY_SIZE_, status);
1.2416 + // status checked in loop
1.2417 +
1.2418 + int32_t accentsindex[INITIAL_ARRAY_SIZE_];
1.2419 + int32_t size = getUnblockedAccentIndex(accents, accentsindex);
1.2420 +
1.2421 + // 2 power n - 1 plus the full set of accents
1.2422 + int32_t count = (2 << (size - 1)) - 1;
1.2423 + while (U_SUCCESS(*status) && count > 0) {
1.2424 + UChar *rearrange = strsrch->canonicalPrefixAccents;
1.2425 + // copy the base characters
1.2426 + for (int k = 0; k < accentsindex[0]; k ++) {
1.2427 + *rearrange ++ = accents[k];
1.2428 + }
1.2429 + // forming all possible canonical rearrangement by dropping
1.2430 + // sets of accents
1.2431 + for (int i = 0; i <= size - 1; i ++) {
1.2432 + int32_t mask = 1 << (size - i - 1);
1.2433 + if (count & mask) {
1.2434 + for (int j = accentsindex[i]; j < accentsindex[i + 1]; j ++) {
1.2435 + *rearrange ++ = accents[j];
1.2436 + }
1.2437 + }
1.2438 + }
1.2439 + *rearrange = 0;
1.2440 + int32_t offset = doPreviousCanonicalPrefixMatch(strsrch,
1.2441 + baseoffset, status);
1.2442 + if (offset != USEARCH_DONE) {
1.2443 + return TRUE; // match found
1.2444 + }
1.2445 + count --;
1.2446 + }
1.2447 + return FALSE;
1.2448 +}
1.2449 +
1.2450 +/**
1.2451 +* Checks match for contraction.
1.2452 +* If the match starts with a partial contraction we fail.
1.2453 +* Internal method, status assumed to be success, caller has to check status
1.2454 +* before calling this method.
1.2455 +* @param strsrch string search data
1.2456 +* @param start offset of potential match, to be modified if necessary
1.2457 +* @param end offset of potential match, to be modified if necessary
1.2458 +* @param status only error status if any
1.2459 +* @return TRUE if match passes the contraction test, FALSE otherwise
1.2460 +*/
1.2461 +static
1.2462 +UBool checkPreviousCanonicalContractionMatch(UStringSearch *strsrch,
1.2463 + int32_t *start,
1.2464 + int32_t *end, UErrorCode *status)
1.2465 +{
1.2466 + UCollationElements *coleiter = strsrch->textIter;
1.2467 + int32_t textlength = strsrch->search->textLength;
1.2468 + int32_t temp = *end;
1.2469 + const UCollator *collator = strsrch->collator;
1.2470 + const UChar *text = strsrch->search->text;
1.2471 + // This part checks if either if the start of the match contains potential
1.2472 + // contraction. If so we'll have to iterate through them
1.2473 + // Since we used ucol_next while previously looking for the potential
1.2474 + // match, this guarantees that our end will not be a partial contraction,
1.2475 + // or a partial supplementary character.
1.2476 + if (*start < textlength && ucol_unsafeCP(text[*start], collator)) {
1.2477 + int32_t expansion = getExpansionSuffix(coleiter);
1.2478 + UBool expandflag = expansion > 0;
1.2479 + setColEIterOffset(coleiter, *end);
1.2480 + while (expansion > 0) {
1.2481 + // getting rid of the redundant ce
1.2482 + // since forward contraction/expansion may have extra ces
1.2483 + // if we are in the normalization buffer, hasAccentsBeforeMatch
1.2484 + // would have taken care of it.
1.2485 + // E.g. the character \u01FA will have an expansion of 3, but if
1.2486 + // we are only looking for A ring A\u030A, we'll have to skip the
1.2487 + // last ce in the expansion buffer
1.2488 + ucol_previous(coleiter, status);
1.2489 + if (U_FAILURE(*status)) {
1.2490 + return FALSE;
1.2491 + }
1.2492 + if (ucol_getOffset(coleiter) != temp) {
1.2493 + *end = temp;
1.2494 + temp = ucol_getOffset(coleiter);
1.2495 + }
1.2496 + expansion --;
1.2497 + }
1.2498 +
1.2499 + int32_t *patternce = strsrch->pattern.CE;
1.2500 + int32_t patterncelength = strsrch->pattern.CELength;
1.2501 + int32_t count = patterncelength;
1.2502 + while (count > 0) {
1.2503 + int32_t ce = getCE(strsrch, ucol_previous(coleiter, status));
1.2504 + // status checked below, note that if status is a failure
1.2505 + // ucol_previous returns UCOL_NULLORDER
1.2506 + if (ce == UCOL_IGNORABLE) {
1.2507 + continue;
1.2508 + }
1.2509 + if (expandflag && count == 0 &&
1.2510 + getColElemIterOffset(coleiter, FALSE) != temp) {
1.2511 + *end = temp;
1.2512 + temp = ucol_getOffset(coleiter);
1.2513 + }
1.2514 + if (count == patterncelength &&
1.2515 + ce != patternce[patterncelength - 1]) {
1.2516 + // accents may have extra starting ces, this occurs when a
1.2517 + // pure accent pattern is matched without rearrangement
1.2518 + int32_t expected = patternce[patterncelength - 1];
1.2519 + U16_BACK_1(text, 0, *end);
1.2520 + if (getFCD(text, end, textlength) & LAST_BYTE_MASK_) {
1.2521 + ce = getCE(strsrch, ucol_previous(coleiter, status));
1.2522 + while (U_SUCCESS(*status) && ce != expected &&
1.2523 + ce != UCOL_NULLORDER &&
1.2524 + ucol_getOffset(coleiter) <= *start) {
1.2525 + ce = getCE(strsrch, ucol_previous(coleiter, status));
1.2526 + }
1.2527 + }
1.2528 + }
1.2529 + if (U_FAILURE(*status) || ce != patternce[count - 1]) {
1.2530 + (*start) --;
1.2531 + *start = getPreviousBaseOffset(text, *start);
1.2532 + return FALSE;
1.2533 + }
1.2534 + count --;
1.2535 + }
1.2536 + }
1.2537 + return TRUE;
1.2538 +}
1.2539 +
1.2540 +/**
1.2541 +* Checks and sets the match information if found.
1.2542 +* Checks
1.2543 +* <ul>
1.2544 +* <li> the potential match does not repeat the previous match
1.2545 +* <li> boundaries are correct
1.2546 +* <li> potential match does not end in the middle of a contraction
1.2547 +* <li> identical matches
1.2548 +* <\ul>
1.2549 +* Otherwise the offset will be shifted to the next character.
1.2550 +* Internal method, status assumed to be success, caller has to check status
1.2551 +* before calling this method.
1.2552 +* @param strsrch string search data
1.2553 +* @param textoffset offset in the collation element text. the returned value
1.2554 +* will be the truncated start offset of the match or the new start
1.2555 +* search offset.
1.2556 +* @param status only error status if any
1.2557 +* @return TRUE if the match is valid, FALSE otherwise
1.2558 +*/
1.2559 +static
1.2560 +inline UBool checkPreviousCanonicalMatch(UStringSearch *strsrch,
1.2561 + int32_t *textoffset,
1.2562 + UErrorCode *status)
1.2563 +{
1.2564 + // to ensure that the start and ends are not composite characters
1.2565 + UCollationElements *coleiter = strsrch->textIter;
1.2566 + // if we have a canonical accent match
1.2567 + if ((strsrch->pattern.hasSuffixAccents &&
1.2568 + strsrch->canonicalSuffixAccents[0]) ||
1.2569 + (strsrch->pattern.hasPrefixAccents &&
1.2570 + strsrch->canonicalPrefixAccents[0])) {
1.2571 + strsrch->search->matchedIndex = *textoffset;
1.2572 + strsrch->search->matchedLength =
1.2573 + getNextUStringSearchBaseOffset(strsrch,
1.2574 + getColElemIterOffset(coleiter, FALSE))
1.2575 + - *textoffset;
1.2576 + return TRUE;
1.2577 + }
1.2578 +
1.2579 + int32_t end = ucol_getOffset(coleiter);
1.2580 + if (!checkPreviousCanonicalContractionMatch(strsrch, textoffset, &end,
1.2581 + status) ||
1.2582 + U_FAILURE(*status)) {
1.2583 + return FALSE;
1.2584 + }
1.2585 +
1.2586 + end = getNextUStringSearchBaseOffset(strsrch, end);
1.2587 + // this totally matches, however we need to check if it is repeating
1.2588 + if (checkRepeatedMatch(strsrch, *textoffset, end) ||
1.2589 + !isBreakUnit(strsrch, *textoffset, end) ||
1.2590 + !checkIdentical(strsrch, *textoffset, end)) {
1.2591 + (*textoffset) --;
1.2592 + *textoffset = getPreviousBaseOffset(strsrch->search->text,
1.2593 + *textoffset);
1.2594 + return FALSE;
1.2595 + }
1.2596 +
1.2597 + strsrch->search->matchedIndex = *textoffset;
1.2598 + strsrch->search->matchedLength = end - *textoffset;
1.2599 + return TRUE;
1.2600 +}
1.2601 +#endif // #if BOYER_MOORE
1.2602 +
1.2603 +// constructors and destructor -------------------------------------------
1.2604 +
1.2605 +U_CAPI UStringSearch * U_EXPORT2 usearch_open(const UChar *pattern,
1.2606 + int32_t patternlength,
1.2607 + const UChar *text,
1.2608 + int32_t textlength,
1.2609 + const char *locale,
1.2610 + UBreakIterator *breakiter,
1.2611 + UErrorCode *status)
1.2612 +{
1.2613 + if (U_FAILURE(*status)) {
1.2614 + return NULL;
1.2615 + }
1.2616 +#if UCONFIG_NO_BREAK_ITERATION
1.2617 + if (breakiter != NULL) {
1.2618 + *status = U_UNSUPPORTED_ERROR;
1.2619 + return NULL;
1.2620 + }
1.2621 +#endif
1.2622 + if (locale) {
1.2623 + // ucol_open internally checks for status
1.2624 + UCollator *collator = ucol_open(locale, status);
1.2625 + // pattern, text checks are done in usearch_openFromCollator
1.2626 + UStringSearch *result = usearch_openFromCollator(pattern,
1.2627 + patternlength, text, textlength,
1.2628 + collator, breakiter, status);
1.2629 +
1.2630 + if (result == NULL || U_FAILURE(*status)) {
1.2631 + if (collator) {
1.2632 + ucol_close(collator);
1.2633 + }
1.2634 + return NULL;
1.2635 + }
1.2636 + else {
1.2637 + result->ownCollator = TRUE;
1.2638 + }
1.2639 + return result;
1.2640 + }
1.2641 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2642 + return NULL;
1.2643 +}
1.2644 +
1.2645 +U_CAPI UStringSearch * U_EXPORT2 usearch_openFromCollator(
1.2646 + const UChar *pattern,
1.2647 + int32_t patternlength,
1.2648 + const UChar *text,
1.2649 + int32_t textlength,
1.2650 + const UCollator *collator,
1.2651 + UBreakIterator *breakiter,
1.2652 + UErrorCode *status)
1.2653 +{
1.2654 + if (U_FAILURE(*status)) {
1.2655 + return NULL;
1.2656 + }
1.2657 +#if UCONFIG_NO_BREAK_ITERATION
1.2658 + if (breakiter != NULL) {
1.2659 + *status = U_UNSUPPORTED_ERROR;
1.2660 + return NULL;
1.2661 + }
1.2662 +#endif
1.2663 + if (pattern == NULL || text == NULL || collator == NULL) {
1.2664 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2665 + return NULL;
1.2666 + }
1.2667 +
1.2668 + // string search does not really work when numeric collation is turned on
1.2669 + if(ucol_getAttribute(collator, UCOL_NUMERIC_COLLATION, status) == UCOL_ON) {
1.2670 + *status = U_UNSUPPORTED_ERROR;
1.2671 + return NULL;
1.2672 + }
1.2673 +
1.2674 + if (U_SUCCESS(*status)) {
1.2675 + initializeFCD(status);
1.2676 + if (U_FAILURE(*status)) {
1.2677 + return NULL;
1.2678 + }
1.2679 +
1.2680 + UStringSearch *result;
1.2681 + if (textlength == -1) {
1.2682 + textlength = u_strlen(text);
1.2683 + }
1.2684 + if (patternlength == -1) {
1.2685 + patternlength = u_strlen(pattern);
1.2686 + }
1.2687 + if (textlength <= 0 || patternlength <= 0) {
1.2688 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2689 + return NULL;
1.2690 + }
1.2691 +
1.2692 + result = (UStringSearch *)uprv_malloc(sizeof(UStringSearch));
1.2693 + if (result == NULL) {
1.2694 + *status = U_MEMORY_ALLOCATION_ERROR;
1.2695 + return NULL;
1.2696 + }
1.2697 +
1.2698 + result->collator = collator;
1.2699 + result->strength = ucol_getStrength(collator);
1.2700 + result->ceMask = getMask(result->strength);
1.2701 + result->toShift =
1.2702 + ucol_getAttribute(collator, UCOL_ALTERNATE_HANDLING, status) ==
1.2703 + UCOL_SHIFTED;
1.2704 + result->variableTop = ucol_getVariableTop(collator, status);
1.2705 +
1.2706 + result->nfd = Normalizer2Factory::getNFDInstance(*status);
1.2707 +
1.2708 + if (U_FAILURE(*status)) {
1.2709 + uprv_free(result);
1.2710 + return NULL;
1.2711 + }
1.2712 +
1.2713 + result->search = (USearch *)uprv_malloc(sizeof(USearch));
1.2714 + if (result->search == NULL) {
1.2715 + *status = U_MEMORY_ALLOCATION_ERROR;
1.2716 + uprv_free(result);
1.2717 + return NULL;
1.2718 + }
1.2719 +
1.2720 + result->search->text = text;
1.2721 + result->search->textLength = textlength;
1.2722 +
1.2723 + result->pattern.text = pattern;
1.2724 + result->pattern.textLength = patternlength;
1.2725 + result->pattern.CE = NULL;
1.2726 + result->pattern.PCE = NULL;
1.2727 +
1.2728 + result->search->breakIter = breakiter;
1.2729 +#if !UCONFIG_NO_BREAK_ITERATION
1.2730 + result->search->internalBreakIter = ubrk_open(UBRK_CHARACTER, ucol_getLocaleByType(result->collator, ULOC_VALID_LOCALE, status), text, textlength, status);
1.2731 + if (breakiter) {
1.2732 + ubrk_setText(breakiter, text, textlength, status);
1.2733 + }
1.2734 +#endif
1.2735 +
1.2736 + result->ownCollator = FALSE;
1.2737 + result->search->matchedLength = 0;
1.2738 + result->search->matchedIndex = USEARCH_DONE;
1.2739 + result->utilIter = NULL;
1.2740 + result->textIter = ucol_openElements(collator, text,
1.2741 + textlength, status);
1.2742 + if (U_FAILURE(*status)) {
1.2743 + usearch_close(result);
1.2744 + return NULL;
1.2745 + }
1.2746 +
1.2747 + result->search->isOverlap = FALSE;
1.2748 + result->search->isCanonicalMatch = FALSE;
1.2749 + result->search->elementComparisonType = 0;
1.2750 + result->search->isForwardSearching = TRUE;
1.2751 + result->search->reset = TRUE;
1.2752 +
1.2753 + initialize(result, status);
1.2754 +
1.2755 + if (U_FAILURE(*status)) {
1.2756 + usearch_close(result);
1.2757 + return NULL;
1.2758 + }
1.2759 +
1.2760 + return result;
1.2761 + }
1.2762 + return NULL;
1.2763 +}
1.2764 +
1.2765 +U_CAPI void U_EXPORT2 usearch_close(UStringSearch *strsrch)
1.2766 +{
1.2767 + if (strsrch) {
1.2768 + if (strsrch->pattern.CE != strsrch->pattern.CEBuffer &&
1.2769 + strsrch->pattern.CE) {
1.2770 + uprv_free(strsrch->pattern.CE);
1.2771 + }
1.2772 +
1.2773 + if (strsrch->pattern.PCE != NULL &&
1.2774 + strsrch->pattern.PCE != strsrch->pattern.PCEBuffer) {
1.2775 + uprv_free(strsrch->pattern.PCE);
1.2776 + }
1.2777 +
1.2778 + ucol_closeElements(strsrch->textIter);
1.2779 + ucol_closeElements(strsrch->utilIter);
1.2780 +
1.2781 + if (strsrch->ownCollator && strsrch->collator) {
1.2782 + ucol_close((UCollator *)strsrch->collator);
1.2783 + }
1.2784 +
1.2785 +#if !UCONFIG_NO_BREAK_ITERATION
1.2786 + if (strsrch->search->internalBreakIter) {
1.2787 + ubrk_close(strsrch->search->internalBreakIter);
1.2788 + }
1.2789 +#endif
1.2790 +
1.2791 + uprv_free(strsrch->search);
1.2792 + uprv_free(strsrch);
1.2793 + }
1.2794 +}
1.2795 +
1.2796 +// set and get methods --------------------------------------------------
1.2797 +
1.2798 +U_CAPI void U_EXPORT2 usearch_setOffset(UStringSearch *strsrch,
1.2799 + int32_t position,
1.2800 + UErrorCode *status)
1.2801 +{
1.2802 + if (U_SUCCESS(*status) && strsrch) {
1.2803 + if (isOutOfBounds(strsrch->search->textLength, position)) {
1.2804 + *status = U_INDEX_OUTOFBOUNDS_ERROR;
1.2805 + }
1.2806 + else {
1.2807 + setColEIterOffset(strsrch->textIter, position);
1.2808 + }
1.2809 + strsrch->search->matchedIndex = USEARCH_DONE;
1.2810 + strsrch->search->matchedLength = 0;
1.2811 + strsrch->search->reset = FALSE;
1.2812 + }
1.2813 +}
1.2814 +
1.2815 +U_CAPI int32_t U_EXPORT2 usearch_getOffset(const UStringSearch *strsrch)
1.2816 +{
1.2817 + if (strsrch) {
1.2818 + int32_t result = ucol_getOffset(strsrch->textIter);
1.2819 + if (isOutOfBounds(strsrch->search->textLength, result)) {
1.2820 + return USEARCH_DONE;
1.2821 + }
1.2822 + return result;
1.2823 + }
1.2824 + return USEARCH_DONE;
1.2825 +}
1.2826 +
1.2827 +U_CAPI void U_EXPORT2 usearch_setAttribute(UStringSearch *strsrch,
1.2828 + USearchAttribute attribute,
1.2829 + USearchAttributeValue value,
1.2830 + UErrorCode *status)
1.2831 +{
1.2832 + if (U_SUCCESS(*status) && strsrch) {
1.2833 + switch (attribute)
1.2834 + {
1.2835 + case USEARCH_OVERLAP :
1.2836 + strsrch->search->isOverlap = (value == USEARCH_ON ? TRUE : FALSE);
1.2837 + break;
1.2838 + case USEARCH_CANONICAL_MATCH :
1.2839 + strsrch->search->isCanonicalMatch = (value == USEARCH_ON ? TRUE :
1.2840 + FALSE);
1.2841 + break;
1.2842 + case USEARCH_ELEMENT_COMPARISON :
1.2843 + if (value == USEARCH_PATTERN_BASE_WEIGHT_IS_WILDCARD || value == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD) {
1.2844 + strsrch->search->elementComparisonType = (int16_t)value;
1.2845 + } else {
1.2846 + strsrch->search->elementComparisonType = 0;
1.2847 + }
1.2848 + break;
1.2849 + case USEARCH_ATTRIBUTE_COUNT :
1.2850 + default:
1.2851 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2852 + }
1.2853 + }
1.2854 + if (value == USEARCH_ATTRIBUTE_VALUE_COUNT) {
1.2855 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2856 + }
1.2857 +}
1.2858 +
1.2859 +U_CAPI USearchAttributeValue U_EXPORT2 usearch_getAttribute(
1.2860 + const UStringSearch *strsrch,
1.2861 + USearchAttribute attribute)
1.2862 +{
1.2863 + if (strsrch) {
1.2864 + switch (attribute) {
1.2865 + case USEARCH_OVERLAP :
1.2866 + return (strsrch->search->isOverlap == TRUE ? USEARCH_ON :
1.2867 + USEARCH_OFF);
1.2868 + case USEARCH_CANONICAL_MATCH :
1.2869 + return (strsrch->search->isCanonicalMatch == TRUE ? USEARCH_ON :
1.2870 + USEARCH_OFF);
1.2871 + case USEARCH_ELEMENT_COMPARISON :
1.2872 + {
1.2873 + int16_t value = strsrch->search->elementComparisonType;
1.2874 + if (value == USEARCH_PATTERN_BASE_WEIGHT_IS_WILDCARD || value == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD) {
1.2875 + return (USearchAttributeValue)value;
1.2876 + } else {
1.2877 + return USEARCH_STANDARD_ELEMENT_COMPARISON;
1.2878 + }
1.2879 + }
1.2880 + case USEARCH_ATTRIBUTE_COUNT :
1.2881 + return USEARCH_DEFAULT;
1.2882 + }
1.2883 + }
1.2884 + return USEARCH_DEFAULT;
1.2885 +}
1.2886 +
1.2887 +U_CAPI int32_t U_EXPORT2 usearch_getMatchedStart(
1.2888 + const UStringSearch *strsrch)
1.2889 +{
1.2890 + if (strsrch == NULL) {
1.2891 + return USEARCH_DONE;
1.2892 + }
1.2893 + return strsrch->search->matchedIndex;
1.2894 +}
1.2895 +
1.2896 +
1.2897 +U_CAPI int32_t U_EXPORT2 usearch_getMatchedText(const UStringSearch *strsrch,
1.2898 + UChar *result,
1.2899 + int32_t resultCapacity,
1.2900 + UErrorCode *status)
1.2901 +{
1.2902 + if (U_FAILURE(*status)) {
1.2903 + return USEARCH_DONE;
1.2904 + }
1.2905 + if (strsrch == NULL || resultCapacity < 0 || (resultCapacity > 0 &&
1.2906 + result == NULL)) {
1.2907 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2908 + return USEARCH_DONE;
1.2909 + }
1.2910 +
1.2911 + int32_t copylength = strsrch->search->matchedLength;
1.2912 + int32_t copyindex = strsrch->search->matchedIndex;
1.2913 + if (copyindex == USEARCH_DONE) {
1.2914 + u_terminateUChars(result, resultCapacity, 0, status);
1.2915 + return USEARCH_DONE;
1.2916 + }
1.2917 +
1.2918 + if (resultCapacity < copylength) {
1.2919 + copylength = resultCapacity;
1.2920 + }
1.2921 + if (copylength > 0) {
1.2922 + uprv_memcpy(result, strsrch->search->text + copyindex,
1.2923 + copylength * sizeof(UChar));
1.2924 + }
1.2925 + return u_terminateUChars(result, resultCapacity,
1.2926 + strsrch->search->matchedLength, status);
1.2927 +}
1.2928 +
1.2929 +U_CAPI int32_t U_EXPORT2 usearch_getMatchedLength(
1.2930 + const UStringSearch *strsrch)
1.2931 +{
1.2932 + if (strsrch) {
1.2933 + return strsrch->search->matchedLength;
1.2934 + }
1.2935 + return USEARCH_DONE;
1.2936 +}
1.2937 +
1.2938 +#if !UCONFIG_NO_BREAK_ITERATION
1.2939 +
1.2940 +U_CAPI void U_EXPORT2 usearch_setBreakIterator(UStringSearch *strsrch,
1.2941 + UBreakIterator *breakiter,
1.2942 + UErrorCode *status)
1.2943 +{
1.2944 + if (U_SUCCESS(*status) && strsrch) {
1.2945 + strsrch->search->breakIter = breakiter;
1.2946 + if (breakiter) {
1.2947 + ubrk_setText(breakiter, strsrch->search->text,
1.2948 + strsrch->search->textLength, status);
1.2949 + }
1.2950 + }
1.2951 +}
1.2952 +
1.2953 +U_CAPI const UBreakIterator* U_EXPORT2
1.2954 +usearch_getBreakIterator(const UStringSearch *strsrch)
1.2955 +{
1.2956 + if (strsrch) {
1.2957 + return strsrch->search->breakIter;
1.2958 + }
1.2959 + return NULL;
1.2960 +}
1.2961 +
1.2962 +#endif
1.2963 +
1.2964 +U_CAPI void U_EXPORT2 usearch_setText( UStringSearch *strsrch,
1.2965 + const UChar *text,
1.2966 + int32_t textlength,
1.2967 + UErrorCode *status)
1.2968 +{
1.2969 + if (U_SUCCESS(*status)) {
1.2970 + if (strsrch == NULL || text == NULL || textlength < -1 ||
1.2971 + textlength == 0) {
1.2972 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2973 + }
1.2974 + else {
1.2975 + if (textlength == -1) {
1.2976 + textlength = u_strlen(text);
1.2977 + }
1.2978 + strsrch->search->text = text;
1.2979 + strsrch->search->textLength = textlength;
1.2980 + ucol_setText(strsrch->textIter, text, textlength, status);
1.2981 + strsrch->search->matchedIndex = USEARCH_DONE;
1.2982 + strsrch->search->matchedLength = 0;
1.2983 + strsrch->search->reset = TRUE;
1.2984 +#if !UCONFIG_NO_BREAK_ITERATION
1.2985 + if (strsrch->search->breakIter != NULL) {
1.2986 + ubrk_setText(strsrch->search->breakIter, text,
1.2987 + textlength, status);
1.2988 + }
1.2989 + ubrk_setText(strsrch->search->internalBreakIter, text, textlength, status);
1.2990 +#endif
1.2991 + }
1.2992 + }
1.2993 +}
1.2994 +
1.2995 +U_CAPI const UChar * U_EXPORT2 usearch_getText(const UStringSearch *strsrch,
1.2996 + int32_t *length)
1.2997 +{
1.2998 + if (strsrch) {
1.2999 + *length = strsrch->search->textLength;
1.3000 + return strsrch->search->text;
1.3001 + }
1.3002 + return NULL;
1.3003 +}
1.3004 +
1.3005 +U_CAPI void U_EXPORT2 usearch_setCollator( UStringSearch *strsrch,
1.3006 + const UCollator *collator,
1.3007 + UErrorCode *status)
1.3008 +{
1.3009 + if (U_SUCCESS(*status)) {
1.3010 + if (collator == NULL) {
1.3011 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.3012 + return;
1.3013 + }
1.3014 +
1.3015 + if (strsrch) {
1.3016 + if (strsrch->ownCollator && (strsrch->collator != collator)) {
1.3017 + ucol_close((UCollator *)strsrch->collator);
1.3018 + strsrch->ownCollator = FALSE;
1.3019 + }
1.3020 + strsrch->collator = collator;
1.3021 + strsrch->strength = ucol_getStrength(collator);
1.3022 + strsrch->ceMask = getMask(strsrch->strength);
1.3023 +#if !UCONFIG_NO_BREAK_ITERATION
1.3024 + ubrk_close(strsrch->search->internalBreakIter);
1.3025 + strsrch->search->internalBreakIter = ubrk_open(UBRK_CHARACTER, ucol_getLocaleByType(collator, ULOC_VALID_LOCALE, status),
1.3026 + strsrch->search->text, strsrch->search->textLength, status);
1.3027 +#endif
1.3028 + // if status is a failure, ucol_getAttribute returns UCOL_DEFAULT
1.3029 + strsrch->toShift =
1.3030 + ucol_getAttribute(collator, UCOL_ALTERNATE_HANDLING, status) ==
1.3031 + UCOL_SHIFTED;
1.3032 + // if status is a failure, ucol_getVariableTop returns 0
1.3033 + strsrch->variableTop = ucol_getVariableTop(collator, status);
1.3034 + if (U_SUCCESS(*status)) {
1.3035 + initialize(strsrch, status);
1.3036 + if (U_SUCCESS(*status)) {
1.3037 + /* free offset buffer to avoid memory leak before initializing. */
1.3038 + ucol_freeOffsetBuffer(&(strsrch->textIter->iteratordata_));
1.3039 + uprv_init_collIterate(collator, strsrch->search->text,
1.3040 + strsrch->search->textLength,
1.3041 + &(strsrch->textIter->iteratordata_),
1.3042 + status);
1.3043 + strsrch->utilIter->iteratordata_.coll = collator;
1.3044 + }
1.3045 + }
1.3046 + }
1.3047 +
1.3048 + // **** are these calls needed?
1.3049 + // **** we call uprv_init_pce in initializePatternPCETable
1.3050 + // **** and the CEBuffer constructor...
1.3051 +#if 0
1.3052 + uprv_init_pce(strsrch->textIter);
1.3053 + uprv_init_pce(strsrch->utilIter);
1.3054 +#endif
1.3055 + }
1.3056 +}
1.3057 +
1.3058 +U_CAPI UCollator * U_EXPORT2 usearch_getCollator(const UStringSearch *strsrch)
1.3059 +{
1.3060 + if (strsrch) {
1.3061 + return (UCollator *)strsrch->collator;
1.3062 + }
1.3063 + return NULL;
1.3064 +}
1.3065 +
1.3066 +U_CAPI void U_EXPORT2 usearch_setPattern( UStringSearch *strsrch,
1.3067 + const UChar *pattern,
1.3068 + int32_t patternlength,
1.3069 + UErrorCode *status)
1.3070 +{
1.3071 + if (U_SUCCESS(*status)) {
1.3072 + if (strsrch == NULL || pattern == NULL) {
1.3073 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.3074 + }
1.3075 + else {
1.3076 + if (patternlength == -1) {
1.3077 + patternlength = u_strlen(pattern);
1.3078 + }
1.3079 + if (patternlength == 0) {
1.3080 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.3081 + return;
1.3082 + }
1.3083 + strsrch->pattern.text = pattern;
1.3084 + strsrch->pattern.textLength = patternlength;
1.3085 + initialize(strsrch, status);
1.3086 + }
1.3087 + }
1.3088 +}
1.3089 +
1.3090 +U_CAPI const UChar* U_EXPORT2
1.3091 +usearch_getPattern(const UStringSearch *strsrch,
1.3092 + int32_t *length)
1.3093 +{
1.3094 + if (strsrch) {
1.3095 + *length = strsrch->pattern.textLength;
1.3096 + return strsrch->pattern.text;
1.3097 + }
1.3098 + return NULL;
1.3099 +}
1.3100 +
1.3101 +// miscellanous methods --------------------------------------------------
1.3102 +
1.3103 +U_CAPI int32_t U_EXPORT2 usearch_first(UStringSearch *strsrch,
1.3104 + UErrorCode *status)
1.3105 +{
1.3106 + if (strsrch && U_SUCCESS(*status)) {
1.3107 + strsrch->search->isForwardSearching = TRUE;
1.3108 + usearch_setOffset(strsrch, 0, status);
1.3109 + if (U_SUCCESS(*status)) {
1.3110 + return usearch_next(strsrch, status);
1.3111 + }
1.3112 + }
1.3113 + return USEARCH_DONE;
1.3114 +}
1.3115 +
1.3116 +U_CAPI int32_t U_EXPORT2 usearch_following(UStringSearch *strsrch,
1.3117 + int32_t position,
1.3118 + UErrorCode *status)
1.3119 +{
1.3120 + if (strsrch && U_SUCCESS(*status)) {
1.3121 + strsrch->search->isForwardSearching = TRUE;
1.3122 + // position checked in usearch_setOffset
1.3123 + usearch_setOffset(strsrch, position, status);
1.3124 + if (U_SUCCESS(*status)) {
1.3125 + return usearch_next(strsrch, status);
1.3126 + }
1.3127 + }
1.3128 + return USEARCH_DONE;
1.3129 +}
1.3130 +
1.3131 +U_CAPI int32_t U_EXPORT2 usearch_last(UStringSearch *strsrch,
1.3132 + UErrorCode *status)
1.3133 +{
1.3134 + if (strsrch && U_SUCCESS(*status)) {
1.3135 + strsrch->search->isForwardSearching = FALSE;
1.3136 + usearch_setOffset(strsrch, strsrch->search->textLength, status);
1.3137 + if (U_SUCCESS(*status)) {
1.3138 + return usearch_previous(strsrch, status);
1.3139 + }
1.3140 + }
1.3141 + return USEARCH_DONE;
1.3142 +}
1.3143 +
1.3144 +U_CAPI int32_t U_EXPORT2 usearch_preceding(UStringSearch *strsrch,
1.3145 + int32_t position,
1.3146 + UErrorCode *status)
1.3147 +{
1.3148 + if (strsrch && U_SUCCESS(*status)) {
1.3149 + strsrch->search->isForwardSearching = FALSE;
1.3150 + // position checked in usearch_setOffset
1.3151 + usearch_setOffset(strsrch, position, status);
1.3152 + if (U_SUCCESS(*status)) {
1.3153 + return usearch_previous(strsrch, status);
1.3154 + }
1.3155 + }
1.3156 + return USEARCH_DONE;
1.3157 +}
1.3158 +
1.3159 +/**
1.3160 +* If a direction switch is required, we'll count the number of ces till the
1.3161 +* beginning of the collation element iterator and iterate forwards that
1.3162 +* number of times. This is so that we get to the correct point within the
1.3163 +* string to continue the search in. Imagine when we are in the middle of the
1.3164 +* normalization buffer when the change in direction is request. arrrgghh....
1.3165 +* After searching the offset within the collation element iterator will be
1.3166 +* shifted to the start of the match. If a match is not found, the offset would
1.3167 +* have been set to the end of the text string in the collation element
1.3168 +* iterator.
1.3169 +* Okay, here's my take on normalization buffer. The only time when there can
1.3170 +* be 2 matches within the same normalization is when the pattern is consists
1.3171 +* of all accents. But since the offset returned is from the text string, we
1.3172 +* should not confuse the caller by returning the second match within the
1.3173 +* same normalization buffer. If we do, the 2 results will have the same match
1.3174 +* offsets, and that'll be confusing. I'll return the next match that doesn't
1.3175 +* fall within the same normalization buffer. Note this does not affect the
1.3176 +* results of matches spanning the text and the normalization buffer.
1.3177 +* The position to start searching is taken from the collation element
1.3178 +* iterator. Callers of this API would have to set the offset in the collation
1.3179 +* element iterator before using this method.
1.3180 +*/
1.3181 +U_CAPI int32_t U_EXPORT2 usearch_next(UStringSearch *strsrch,
1.3182 + UErrorCode *status)
1.3183 +{
1.3184 + if (U_SUCCESS(*status) && strsrch) {
1.3185 + // note offset is either equivalent to the start of the previous match
1.3186 + // or is set by the user
1.3187 + int32_t offset = usearch_getOffset(strsrch);
1.3188 + USearch *search = strsrch->search;
1.3189 + search->reset = FALSE;
1.3190 + int32_t textlength = search->textLength;
1.3191 + if (search->isForwardSearching) {
1.3192 +#if BOYER_MOORE
1.3193 + if (offset == textlength
1.3194 + || (!search->isOverlap &&
1.3195 + (offset + strsrch->pattern.defaultShiftSize > textlength ||
1.3196 + (search->matchedIndex != USEARCH_DONE &&
1.3197 + offset + search->matchedLength >= textlength)))) {
1.3198 + // not enough characters to match
1.3199 + setMatchNotFound(strsrch);
1.3200 + return USEARCH_DONE;
1.3201 + }
1.3202 +#else
1.3203 + if (offset == textlength ||
1.3204 + (! search->isOverlap &&
1.3205 + (search->matchedIndex != USEARCH_DONE &&
1.3206 + offset + search->matchedLength > textlength))) {
1.3207 + // not enough characters to match
1.3208 + setMatchNotFound(strsrch);
1.3209 + return USEARCH_DONE;
1.3210 + }
1.3211 +#endif
1.3212 + }
1.3213 + else {
1.3214 + // switching direction.
1.3215 + // if matchedIndex == USEARCH_DONE, it means that either a
1.3216 + // setOffset has been called or that previous ran off the text
1.3217 + // string. the iterator would have been set to offset 0 if a
1.3218 + // match is not found.
1.3219 + search->isForwardSearching = TRUE;
1.3220 + if (search->matchedIndex != USEARCH_DONE) {
1.3221 + // there's no need to set the collation element iterator
1.3222 + // the next call to next will set the offset.
1.3223 + return search->matchedIndex;
1.3224 + }
1.3225 + }
1.3226 +
1.3227 + if (U_SUCCESS(*status)) {
1.3228 + if (strsrch->pattern.CELength == 0) {
1.3229 + if (search->matchedIndex == USEARCH_DONE) {
1.3230 + search->matchedIndex = offset;
1.3231 + }
1.3232 + else { // moves by codepoints
1.3233 + U16_FWD_1(search->text, search->matchedIndex, textlength);
1.3234 + }
1.3235 +
1.3236 + search->matchedLength = 0;
1.3237 + setColEIterOffset(strsrch->textIter, search->matchedIndex);
1.3238 + // status checked below
1.3239 + if (search->matchedIndex == textlength) {
1.3240 + search->matchedIndex = USEARCH_DONE;
1.3241 + }
1.3242 + }
1.3243 + else {
1.3244 + if (search->matchedLength > 0) {
1.3245 + // if matchlength is 0 we are at the start of the iteration
1.3246 + if (search->isOverlap) {
1.3247 + ucol_setOffset(strsrch->textIter, offset + 1, status);
1.3248 + }
1.3249 + else {
1.3250 + ucol_setOffset(strsrch->textIter,
1.3251 + offset + search->matchedLength, status);
1.3252 + }
1.3253 + }
1.3254 + else {
1.3255 + // for boundary check purposes. this will ensure that the
1.3256 + // next match will not preceed the current offset
1.3257 + // note search->matchedIndex will always be set to something
1.3258 + // in the code
1.3259 + search->matchedIndex = offset - 1;
1.3260 + }
1.3261 +
1.3262 + if (search->isCanonicalMatch) {
1.3263 + // can't use exact here since extra accents are allowed.
1.3264 + usearch_handleNextCanonical(strsrch, status);
1.3265 + }
1.3266 + else {
1.3267 + usearch_handleNextExact(strsrch, status);
1.3268 + }
1.3269 + }
1.3270 +
1.3271 + if (U_FAILURE(*status)) {
1.3272 + return USEARCH_DONE;
1.3273 + }
1.3274 +
1.3275 +#if !BOYER_MOORE
1.3276 + if (search->matchedIndex == USEARCH_DONE) {
1.3277 + ucol_setOffset(strsrch->textIter, search->textLength, status);
1.3278 + } else {
1.3279 + ucol_setOffset(strsrch->textIter, search->matchedIndex, status);
1.3280 + }
1.3281 +#endif
1.3282 +
1.3283 + return search->matchedIndex;
1.3284 + }
1.3285 + }
1.3286 + return USEARCH_DONE;
1.3287 +}
1.3288 +
1.3289 +U_CAPI int32_t U_EXPORT2 usearch_previous(UStringSearch *strsrch,
1.3290 + UErrorCode *status)
1.3291 +{
1.3292 + if (U_SUCCESS(*status) && strsrch) {
1.3293 + int32_t offset;
1.3294 + USearch *search = strsrch->search;
1.3295 + if (search->reset) {
1.3296 + offset = search->textLength;
1.3297 + search->isForwardSearching = FALSE;
1.3298 + search->reset = FALSE;
1.3299 + setColEIterOffset(strsrch->textIter, offset);
1.3300 + }
1.3301 + else {
1.3302 + offset = usearch_getOffset(strsrch);
1.3303 + }
1.3304 +
1.3305 + int32_t matchedindex = search->matchedIndex;
1.3306 + if (search->isForwardSearching == TRUE) {
1.3307 + // switching direction.
1.3308 + // if matchedIndex == USEARCH_DONE, it means that either a
1.3309 + // setOffset has been called or that next ran off the text
1.3310 + // string. the iterator would have been set to offset textLength if
1.3311 + // a match is not found.
1.3312 + search->isForwardSearching = FALSE;
1.3313 + if (matchedindex != USEARCH_DONE) {
1.3314 + return matchedindex;
1.3315 + }
1.3316 + }
1.3317 + else {
1.3318 +#if BOYER_MOORE
1.3319 + if (offset == 0 || matchedindex == 0 ||
1.3320 + (!search->isOverlap &&
1.3321 + (offset < strsrch->pattern.defaultShiftSize ||
1.3322 + (matchedindex != USEARCH_DONE &&
1.3323 + matchedindex < strsrch->pattern.defaultShiftSize)))) {
1.3324 + // not enough characters to match
1.3325 + setMatchNotFound(strsrch);
1.3326 + return USEARCH_DONE;
1.3327 + }
1.3328 +#else
1.3329 + // Could check pattern length, but the
1.3330 + // linear search will do the right thing
1.3331 + if (offset == 0 || matchedindex == 0) {
1.3332 + setMatchNotFound(strsrch);
1.3333 + return USEARCH_DONE;
1.3334 + }
1.3335 +#endif
1.3336 + }
1.3337 +
1.3338 + if (U_SUCCESS(*status)) {
1.3339 + if (strsrch->pattern.CELength == 0) {
1.3340 + search->matchedIndex =
1.3341 + (matchedindex == USEARCH_DONE ? offset : matchedindex);
1.3342 + if (search->matchedIndex == 0) {
1.3343 + setMatchNotFound(strsrch);
1.3344 + // status checked below
1.3345 + }
1.3346 + else { // move by codepoints
1.3347 + U16_BACK_1(search->text, 0, search->matchedIndex);
1.3348 + setColEIterOffset(strsrch->textIter, search->matchedIndex);
1.3349 + // status checked below
1.3350 + search->matchedLength = 0;
1.3351 + }
1.3352 + }
1.3353 + else {
1.3354 + if (strsrch->search->isCanonicalMatch) {
1.3355 + // can't use exact here since extra accents are allowed.
1.3356 + usearch_handlePreviousCanonical(strsrch, status);
1.3357 + // status checked below
1.3358 + }
1.3359 + else {
1.3360 + usearch_handlePreviousExact(strsrch, status);
1.3361 + // status checked below
1.3362 + }
1.3363 + }
1.3364 +
1.3365 + if (U_FAILURE(*status)) {
1.3366 + return USEARCH_DONE;
1.3367 + }
1.3368 +
1.3369 + return search->matchedIndex;
1.3370 + }
1.3371 + }
1.3372 + return USEARCH_DONE;
1.3373 +}
1.3374 +
1.3375 +
1.3376 +
1.3377 +U_CAPI void U_EXPORT2 usearch_reset(UStringSearch *strsrch)
1.3378 +{
1.3379 + /*
1.3380 + reset is setting the attributes that are already in
1.3381 + string search, hence all attributes in the collator should
1.3382 + be retrieved without any problems
1.3383 + */
1.3384 + if (strsrch) {
1.3385 + UErrorCode status = U_ZERO_ERROR;
1.3386 + UBool sameCollAttribute = TRUE;
1.3387 + uint32_t ceMask;
1.3388 + UBool shift;
1.3389 + uint32_t varTop;
1.3390 +
1.3391 + // **** hack to deal w/ how processed CEs encode quaternary ****
1.3392 + UCollationStrength newStrength = ucol_getStrength(strsrch->collator);
1.3393 + if ((strsrch->strength < UCOL_QUATERNARY && newStrength >= UCOL_QUATERNARY) ||
1.3394 + (strsrch->strength >= UCOL_QUATERNARY && newStrength < UCOL_QUATERNARY)) {
1.3395 + sameCollAttribute = FALSE;
1.3396 + }
1.3397 +
1.3398 + strsrch->strength = ucol_getStrength(strsrch->collator);
1.3399 + ceMask = getMask(strsrch->strength);
1.3400 + if (strsrch->ceMask != ceMask) {
1.3401 + strsrch->ceMask = ceMask;
1.3402 + sameCollAttribute = FALSE;
1.3403 + }
1.3404 +
1.3405 + // if status is a failure, ucol_getAttribute returns UCOL_DEFAULT
1.3406 + shift = ucol_getAttribute(strsrch->collator, UCOL_ALTERNATE_HANDLING,
1.3407 + &status) == UCOL_SHIFTED;
1.3408 + if (strsrch->toShift != shift) {
1.3409 + strsrch->toShift = shift;
1.3410 + sameCollAttribute = FALSE;
1.3411 + }
1.3412 +
1.3413 + // if status is a failure, ucol_getVariableTop returns 0
1.3414 + varTop = ucol_getVariableTop(strsrch->collator, &status);
1.3415 + if (strsrch->variableTop != varTop) {
1.3416 + strsrch->variableTop = varTop;
1.3417 + sameCollAttribute = FALSE;
1.3418 + }
1.3419 + if (!sameCollAttribute) {
1.3420 + initialize(strsrch, &status);
1.3421 + }
1.3422 + /* free offset buffer to avoid memory leak before initializing. */
1.3423 + ucol_freeOffsetBuffer(&(strsrch->textIter->iteratordata_));
1.3424 + uprv_init_collIterate(strsrch->collator, strsrch->search->text,
1.3425 + strsrch->search->textLength,
1.3426 + &(strsrch->textIter->iteratordata_),
1.3427 + &status);
1.3428 + strsrch->search->matchedLength = 0;
1.3429 + strsrch->search->matchedIndex = USEARCH_DONE;
1.3430 + strsrch->search->isOverlap = FALSE;
1.3431 + strsrch->search->isCanonicalMatch = FALSE;
1.3432 + strsrch->search->elementComparisonType = 0;
1.3433 + strsrch->search->isForwardSearching = TRUE;
1.3434 + strsrch->search->reset = TRUE;
1.3435 + }
1.3436 +}
1.3437 +
1.3438 +//
1.3439 +// CEI Collation Element + source text index.
1.3440 +// These structs are kept in the circular buffer.
1.3441 +//
1.3442 +struct CEI {
1.3443 + int64_t ce;
1.3444 + int32_t lowIndex;
1.3445 + int32_t highIndex;
1.3446 +};
1.3447 +
1.3448 +U_NAMESPACE_BEGIN
1.3449 +
1.3450 +
1.3451 +//
1.3452 +// CEBuffer A circular buffer of CEs from the text being searched.
1.3453 +//
1.3454 +#define DEFAULT_CEBUFFER_SIZE 96
1.3455 +#define CEBUFFER_EXTRA 32
1.3456 +// Some typical max values to make buffer size more reasonable for asymmetric search.
1.3457 +// #8694 is for a better long-term solution to allocation of this buffer.
1.3458 +#define MAX_TARGET_IGNORABLES_PER_PAT_JAMO_L 8
1.3459 +#define MAX_TARGET_IGNORABLES_PER_PAT_OTHER 3
1.3460 +#define MIGHT_BE_JAMO_L(c) ((c >= 0x1100 && c <= 0x115E) || (c >= 0x3131 && c <= 0x314E) || (c >= 0x3165 && c <= 0x3186))
1.3461 +struct CEBuffer {
1.3462 + CEI defBuf[DEFAULT_CEBUFFER_SIZE];
1.3463 + CEI *buf;
1.3464 + int32_t bufSize;
1.3465 + int32_t firstIx;
1.3466 + int32_t limitIx;
1.3467 + UCollationElements *ceIter;
1.3468 + UStringSearch *strSearch;
1.3469 +
1.3470 +
1.3471 +
1.3472 + CEBuffer(UStringSearch *ss, UErrorCode *status);
1.3473 + ~CEBuffer();
1.3474 + const CEI *get(int32_t index);
1.3475 + const CEI *getPrevious(int32_t index);
1.3476 +};
1.3477 +
1.3478 +
1.3479 +CEBuffer::CEBuffer(UStringSearch *ss, UErrorCode *status) {
1.3480 + buf = defBuf;
1.3481 + strSearch = ss;
1.3482 + bufSize = ss->pattern.PCELength + CEBUFFER_EXTRA;
1.3483 + if (ss->search->elementComparisonType != 0) {
1.3484 + const UChar * patText = ss->pattern.text;
1.3485 + if (patText) {
1.3486 + const UChar * patTextLimit = patText + ss->pattern.textLength;
1.3487 + while ( patText < patTextLimit ) {
1.3488 + UChar c = *patText++;
1.3489 + if (MIGHT_BE_JAMO_L(c)) {
1.3490 + bufSize += MAX_TARGET_IGNORABLES_PER_PAT_JAMO_L;
1.3491 + } else {
1.3492 + // No check for surrogates, we might allocate slightly more buffer than necessary.
1.3493 + bufSize += MAX_TARGET_IGNORABLES_PER_PAT_OTHER;
1.3494 + }
1.3495 + }
1.3496 + }
1.3497 + }
1.3498 + ceIter = ss->textIter;
1.3499 + firstIx = 0;
1.3500 + limitIx = 0;
1.3501 +
1.3502 + uprv_init_pce(ceIter);
1.3503 +
1.3504 + if (bufSize>DEFAULT_CEBUFFER_SIZE) {
1.3505 + buf = (CEI *)uprv_malloc(bufSize * sizeof(CEI));
1.3506 + if (buf == NULL) {
1.3507 + *status = U_MEMORY_ALLOCATION_ERROR;
1.3508 + }
1.3509 + }
1.3510 +}
1.3511 +
1.3512 +// TODO: add a reset or init function so that allocated
1.3513 +// buffers can be retained & reused.
1.3514 +
1.3515 +CEBuffer::~CEBuffer() {
1.3516 + if (buf != defBuf) {
1.3517 + uprv_free(buf);
1.3518 + }
1.3519 +}
1.3520 +
1.3521 +
1.3522 +// Get the CE with the specified index.
1.3523 +// Index must be in the range
1.3524 +// n-history_size < index < n+1
1.3525 +// where n is the largest index to have been fetched by some previous call to this function.
1.3526 +// The CE value will be UCOL__PROCESSED_NULLORDER at end of input.
1.3527 +//
1.3528 +const CEI *CEBuffer::get(int32_t index) {
1.3529 + int i = index % bufSize;
1.3530 +
1.3531 + if (index>=firstIx && index<limitIx) {
1.3532 + // The request was for an entry already in our buffer.
1.3533 + // Just return it.
1.3534 + return &buf[i];
1.3535 + }
1.3536 +
1.3537 + // Caller is requesting a new, never accessed before, CE.
1.3538 + // Verify that it is the next one in sequence, which is all
1.3539 + // that is allowed.
1.3540 + if (index != limitIx) {
1.3541 + U_ASSERT(FALSE);
1.3542 +
1.3543 + return NULL;
1.3544 + }
1.3545 +
1.3546 + // Manage the circular CE buffer indexing
1.3547 + limitIx++;
1.3548 +
1.3549 + if (limitIx - firstIx >= bufSize) {
1.3550 + // The buffer is full, knock out the lowest-indexed entry.
1.3551 + firstIx++;
1.3552 + }
1.3553 +
1.3554 + UErrorCode status = U_ZERO_ERROR;
1.3555 +
1.3556 + buf[i].ce = ucol_nextProcessed(ceIter, &buf[i].lowIndex, &buf[i].highIndex, &status);
1.3557 +
1.3558 + return &buf[i];
1.3559 +}
1.3560 +
1.3561 +// Get the CE with the specified index.
1.3562 +// Index must be in the range
1.3563 +// n-history_size < index < n+1
1.3564 +// where n is the largest index to have been fetched by some previous call to this function.
1.3565 +// The CE value will be UCOL__PROCESSED_NULLORDER at end of input.
1.3566 +//
1.3567 +const CEI *CEBuffer::getPrevious(int32_t index) {
1.3568 + int i = index % bufSize;
1.3569 +
1.3570 + if (index>=firstIx && index<limitIx) {
1.3571 + // The request was for an entry already in our buffer.
1.3572 + // Just return it.
1.3573 + return &buf[i];
1.3574 + }
1.3575 +
1.3576 + // Caller is requesting a new, never accessed before, CE.
1.3577 + // Verify that it is the next one in sequence, which is all
1.3578 + // that is allowed.
1.3579 + if (index != limitIx) {
1.3580 + U_ASSERT(FALSE);
1.3581 +
1.3582 + return NULL;
1.3583 + }
1.3584 +
1.3585 + // Manage the circular CE buffer indexing
1.3586 + limitIx++;
1.3587 +
1.3588 + if (limitIx - firstIx >= bufSize) {
1.3589 + // The buffer is full, knock out the lowest-indexed entry.
1.3590 + firstIx++;
1.3591 + }
1.3592 +
1.3593 + UErrorCode status = U_ZERO_ERROR;
1.3594 +
1.3595 + buf[i].ce = ucol_previousProcessed(ceIter, &buf[i].lowIndex, &buf[i].highIndex, &status);
1.3596 +
1.3597 + return &buf[i];
1.3598 +}
1.3599 +
1.3600 +U_NAMESPACE_END
1.3601 +
1.3602 +
1.3603 +// #define USEARCH_DEBUG
1.3604 +
1.3605 +#ifdef USEARCH_DEBUG
1.3606 +#include <stdio.h>
1.3607 +#include <stdlib.h>
1.3608 +#endif
1.3609 +
1.3610 +/*
1.3611 + * Find the next break boundary after startIndex. If the UStringSearch object
1.3612 + * has an external break iterator, use that. Otherwise use the internal character
1.3613 + * break iterator.
1.3614 + */
1.3615 +static int32_t nextBoundaryAfter(UStringSearch *strsrch, int32_t startIndex) {
1.3616 +#if 0
1.3617 + const UChar *text = strsrch->search->text;
1.3618 + int32_t textLen = strsrch->search->textLength;
1.3619 +
1.3620 + U_ASSERT(startIndex>=0);
1.3621 + U_ASSERT(startIndex<=textLen);
1.3622 +
1.3623 + if (startIndex >= textLen) {
1.3624 + return startIndex;
1.3625 + }
1.3626 +
1.3627 + UChar32 c;
1.3628 + int32_t i = startIndex;
1.3629 + U16_NEXT(text, i, textLen, c);
1.3630 +
1.3631 + // If we are on a control character, stop without looking for combining marks.
1.3632 + // Control characters do not combine.
1.3633 + int32_t gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
1.3634 + if (gcProperty==U_GCB_CONTROL || gcProperty==U_GCB_LF || gcProperty==U_GCB_CR) {
1.3635 + return i;
1.3636 + }
1.3637 +
1.3638 + // The initial character was not a control, and can thus accept trailing
1.3639 + // combining characters. Advance over however many of them there are.
1.3640 + int32_t indexOfLastCharChecked;
1.3641 + for (;;) {
1.3642 + indexOfLastCharChecked = i;
1.3643 + if (i>=textLen) {
1.3644 + break;
1.3645 + }
1.3646 + U16_NEXT(text, i, textLen, c);
1.3647 + gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
1.3648 + if (gcProperty != U_GCB_EXTEND && gcProperty != U_GCB_SPACING_MARK) {
1.3649 + break;
1.3650 + }
1.3651 + }
1.3652 + return indexOfLastCharChecked;
1.3653 +#elif !UCONFIG_NO_BREAK_ITERATION
1.3654 + UBreakIterator *breakiterator = strsrch->search->breakIter;
1.3655 +
1.3656 + if (breakiterator == NULL) {
1.3657 + breakiterator = strsrch->search->internalBreakIter;
1.3658 + }
1.3659 +
1.3660 + if (breakiterator != NULL) {
1.3661 + return ubrk_following(breakiterator, startIndex);
1.3662 + }
1.3663 +
1.3664 + return startIndex;
1.3665 +#else
1.3666 + // **** or should we use the original code? ****
1.3667 + return startIndex;
1.3668 +#endif
1.3669 +
1.3670 +}
1.3671 +
1.3672 +/*
1.3673 + * Returns TRUE if index is on a break boundary. If the UStringSearch
1.3674 + * has an external break iterator, test using that, otherwise test
1.3675 + * using the internal character break iterator.
1.3676 + */
1.3677 +static UBool isBreakBoundary(UStringSearch *strsrch, int32_t index) {
1.3678 +#if 0
1.3679 + const UChar *text = strsrch->search->text;
1.3680 + int32_t textLen = strsrch->search->textLength;
1.3681 +
1.3682 + U_ASSERT(index>=0);
1.3683 + U_ASSERT(index<=textLen);
1.3684 +
1.3685 + if (index>=textLen || index<=0) {
1.3686 + return TRUE;
1.3687 + }
1.3688 +
1.3689 + // If the character at the current index is not a GRAPHEME_EXTEND
1.3690 + // then we can not be within a combining sequence.
1.3691 + UChar32 c;
1.3692 + U16_GET(text, 0, index, textLen, c);
1.3693 + int32_t gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
1.3694 + if (gcProperty != U_GCB_EXTEND && gcProperty != U_GCB_SPACING_MARK) {
1.3695 + return TRUE;
1.3696 + }
1.3697 +
1.3698 + // We are at a combining mark. If the preceding character is anything
1.3699 + // except a CONTROL, CR or LF, we are in a combining sequence.
1.3700 + U16_PREV(text, 0, index, c);
1.3701 + gcProperty = u_getIntPropertyValue(c, UCHAR_GRAPHEME_CLUSTER_BREAK);
1.3702 + UBool combining = !(gcProperty==U_GCB_CONTROL || gcProperty==U_GCB_LF || gcProperty==U_GCB_CR);
1.3703 + return !combining;
1.3704 +#elif !UCONFIG_NO_BREAK_ITERATION
1.3705 + UBreakIterator *breakiterator = strsrch->search->breakIter;
1.3706 +
1.3707 + if (breakiterator == NULL) {
1.3708 + breakiterator = strsrch->search->internalBreakIter;
1.3709 + }
1.3710 +
1.3711 + return (breakiterator != NULL && ubrk_isBoundary(breakiterator, index));
1.3712 +#else
1.3713 + // **** or use the original code? ****
1.3714 + return TRUE;
1.3715 +#endif
1.3716 +}
1.3717 +
1.3718 +#if 0
1.3719 +static UBool onBreakBoundaries(const UStringSearch *strsrch, int32_t start, int32_t end)
1.3720 +{
1.3721 +#if !UCONFIG_NO_BREAK_ITERATION
1.3722 + UBreakIterator *breakiterator = strsrch->search->breakIter;
1.3723 +
1.3724 + if (breakiterator != NULL) {
1.3725 + int32_t startindex = ubrk_first(breakiterator);
1.3726 + int32_t endindex = ubrk_last(breakiterator);
1.3727 +
1.3728 + // out-of-range indexes are never boundary positions
1.3729 + if (start < startindex || start > endindex ||
1.3730 + end < startindex || end > endindex) {
1.3731 + return FALSE;
1.3732 + }
1.3733 +
1.3734 + return ubrk_isBoundary(breakiterator, start) &&
1.3735 + ubrk_isBoundary(breakiterator, end);
1.3736 + }
1.3737 +#endif
1.3738 +
1.3739 + return TRUE;
1.3740 +}
1.3741 +#endif
1.3742 +
1.3743 +typedef enum {
1.3744 + U_CE_MATCH = -1,
1.3745 + U_CE_NO_MATCH = 0,
1.3746 + U_CE_SKIP_TARG,
1.3747 + U_CE_SKIP_PATN
1.3748 +} UCompareCEsResult;
1.3749 +#define U_CE_LEVEL2_BASE 0x00000005
1.3750 +#define U_CE_LEVEL3_BASE 0x00050000
1.3751 +
1.3752 +static UCompareCEsResult compareCE64s(int64_t targCE, int64_t patCE, int16_t compareType) {
1.3753 + if (targCE == patCE) {
1.3754 + return U_CE_MATCH;
1.3755 + }
1.3756 + if (compareType == 0) {
1.3757 + return U_CE_NO_MATCH;
1.3758 + }
1.3759 +
1.3760 + int64_t targCEshifted = targCE >> 32;
1.3761 + int64_t patCEshifted = patCE >> 32;
1.3762 + int64_t mask;
1.3763 +
1.3764 + mask = 0xFFFF0000;
1.3765 + int32_t targLev1 = (int32_t)(targCEshifted & mask);
1.3766 + int32_t patLev1 = (int32_t)(patCEshifted & mask);
1.3767 + if ( targLev1 != patLev1 ) {
1.3768 + if ( targLev1 == 0 ) {
1.3769 + return U_CE_SKIP_TARG;
1.3770 + }
1.3771 + if ( patLev1 == 0 && compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD ) {
1.3772 + return U_CE_SKIP_PATN;
1.3773 + }
1.3774 + return U_CE_NO_MATCH;
1.3775 + }
1.3776 +
1.3777 + mask = 0x0000FFFF;
1.3778 + int32_t targLev2 = (int32_t)(targCEshifted & mask);
1.3779 + int32_t patLev2 = (int32_t)(patCEshifted & mask);
1.3780 + if ( targLev2 != patLev2 ) {
1.3781 + if ( targLev2 == 0 ) {
1.3782 + return U_CE_SKIP_TARG;
1.3783 + }
1.3784 + if ( patLev2 == 0 && compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD ) {
1.3785 + return U_CE_SKIP_PATN;
1.3786 + }
1.3787 + return (patLev2 == U_CE_LEVEL2_BASE || (compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD && targLev2 == U_CE_LEVEL2_BASE) )?
1.3788 + U_CE_MATCH: U_CE_NO_MATCH;
1.3789 + }
1.3790 +
1.3791 + mask = 0xFFFF0000;
1.3792 + int32_t targLev3 = (int32_t)(targCE & mask);
1.3793 + int32_t patLev3 = (int32_t)(patCE & mask);
1.3794 + if ( targLev3 != patLev3 ) {
1.3795 + return (patLev3 == U_CE_LEVEL3_BASE || (compareType == USEARCH_ANY_BASE_WEIGHT_IS_WILDCARD && targLev3 == U_CE_LEVEL3_BASE) )?
1.3796 + U_CE_MATCH: U_CE_NO_MATCH;
1.3797 + }
1.3798 +
1.3799 + return U_CE_MATCH;
1.3800 +}
1.3801 +
1.3802 +#if BOYER_MOORE
1.3803 +// TODO: #if BOYER_MOORE, need 32-bit version of compareCE64s
1.3804 +#endif
1.3805 +
1.3806 +U_CAPI UBool U_EXPORT2 usearch_search(UStringSearch *strsrch,
1.3807 + int32_t startIdx,
1.3808 + int32_t *matchStart,
1.3809 + int32_t *matchLimit,
1.3810 + UErrorCode *status)
1.3811 +{
1.3812 + if (U_FAILURE(*status)) {
1.3813 + return FALSE;
1.3814 + }
1.3815 +
1.3816 + // TODO: reject search patterns beginning with a combining char.
1.3817 +
1.3818 +#ifdef USEARCH_DEBUG
1.3819 + if (getenv("USEARCH_DEBUG") != NULL) {
1.3820 + printf("Pattern CEs\n");
1.3821 + for (int ii=0; ii<strsrch->pattern.CELength; ii++) {
1.3822 + printf(" %8x", strsrch->pattern.CE[ii]);
1.3823 + }
1.3824 + printf("\n");
1.3825 + }
1.3826 +
1.3827 +#endif
1.3828 + // Input parameter sanity check.
1.3829 + // TODO: should input indicies clip to the text length
1.3830 + // in the same way that UText does.
1.3831 + if(strsrch->pattern.CELength == 0 ||
1.3832 + startIdx < 0 ||
1.3833 + startIdx > strsrch->search->textLength ||
1.3834 + strsrch->pattern.CE == NULL) {
1.3835 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.3836 + return FALSE;
1.3837 + }
1.3838 +
1.3839 + if (strsrch->pattern.PCE == NULL) {
1.3840 + initializePatternPCETable(strsrch, status);
1.3841 + }
1.3842 +
1.3843 + ucol_setOffset(strsrch->textIter, startIdx, status);
1.3844 + CEBuffer ceb(strsrch, status);
1.3845 +
1.3846 +
1.3847 + int32_t targetIx = 0;
1.3848 + const CEI *targetCEI = NULL;
1.3849 + int32_t patIx;
1.3850 + UBool found;
1.3851 +
1.3852 + int32_t mStart = -1;
1.3853 + int32_t mLimit = -1;
1.3854 + int32_t minLimit;
1.3855 + int32_t maxLimit;
1.3856 +
1.3857 +
1.3858 +
1.3859 + // Outer loop moves over match starting positions in the
1.3860 + // target CE space.
1.3861 + // Here we see the target as a sequence of collation elements, resulting from the following:
1.3862 + // 1. Target characters were decomposed, and (if appropriate) other compressions and expansions are applied
1.3863 + // (for example, digraphs such as IJ may be broken into two characters).
1.3864 + // 2. An int64_t CE weight is determined for each resulting unit (high 16 bits are primary strength, next
1.3865 + // 16 bits are secondary, next 16 (the high 16 bits of the low 32-bit half) are tertiary. Any of these
1.3866 + // fields that are for strengths below that of the collator are set to 0. If this makes the int64_t
1.3867 + // CE weight 0 (as for a combining diacritic with secondary weight when the collator strentgh is primary),
1.3868 + // then the CE is deleted, so the following code sees only CEs that are relevant.
1.3869 + // For each CE, the lowIndex and highIndex correspond to where this CE begins and ends in the original text.
1.3870 + // If lowIndex==highIndex, either the CE resulted from an expansion/decomposition of one of the original text
1.3871 + // characters, or the CE marks the limit of the target text (in which case the CE weight is UCOL_PROCESSED_NULLORDER).
1.3872 + //
1.3873 + for(targetIx=0; ; targetIx++)
1.3874 + {
1.3875 + found = TRUE;
1.3876 + // Inner loop checks for a match beginning at each
1.3877 + // position from the outer loop.
1.3878 + int32_t targetIxOffset = 0;
1.3879 + int64_t patCE = 0;
1.3880 + // For targetIx > 0, this ceb.get gets a CE that is as far back in the ring buffer
1.3881 + // (compared to the last CE fetched for the previous targetIx value) as we need to go
1.3882 + // for this targetIx value, so if it is non-NULL then other ceb.get calls should be OK.
1.3883 + const CEI *firstCEI = ceb.get(targetIx);
1.3884 + if (firstCEI == NULL) {
1.3885 + *status = U_INTERNAL_PROGRAM_ERROR;
1.3886 + found = FALSE;
1.3887 + break;
1.3888 + }
1.3889 +
1.3890 + for (patIx=0; patIx<strsrch->pattern.PCELength; patIx++) {
1.3891 + patCE = strsrch->pattern.PCE[patIx];
1.3892 + targetCEI = ceb.get(targetIx+patIx+targetIxOffset);
1.3893 + // Compare CE from target string with CE from the pattern.
1.3894 + // Note that the target CE will be UCOL_PROCESSED_NULLORDER if we reach the end of input,
1.3895 + // which will fail the compare, below.
1.3896 + UCompareCEsResult ceMatch = compareCE64s(targetCEI->ce, patCE, strsrch->search->elementComparisonType);
1.3897 + if ( ceMatch == U_CE_NO_MATCH ) {
1.3898 + found = FALSE;
1.3899 + break;
1.3900 + } else if ( ceMatch > U_CE_NO_MATCH ) {
1.3901 + if ( ceMatch == U_CE_SKIP_TARG ) {
1.3902 + // redo with same patCE, next targCE
1.3903 + patIx--;
1.3904 + targetIxOffset++;
1.3905 + } else { // ceMatch == U_CE_SKIP_PATN
1.3906 + // redo with same targCE, next patCE
1.3907 + targetIxOffset--;
1.3908 + }
1.3909 + }
1.3910 + }
1.3911 + targetIxOffset += strsrch->pattern.PCELength; // this is now the offset in target CE space to end of the match so far
1.3912 +
1.3913 + if (!found && ((targetCEI == NULL) || (targetCEI->ce != UCOL_PROCESSED_NULLORDER))) {
1.3914 + // No match at this targetIx. Try again at the next.
1.3915 + continue;
1.3916 + }
1.3917 +
1.3918 + if (!found) {
1.3919 + // No match at all, we have run off the end of the target text.
1.3920 + break;
1.3921 + }
1.3922 +
1.3923 +
1.3924 + // We have found a match in CE space.
1.3925 + // Now determine the bounds in string index space.
1.3926 + // There still is a chance of match failure if the CE range not correspond to
1.3927 + // an acceptable character range.
1.3928 + //
1.3929 + const CEI *lastCEI = ceb.get(targetIx + targetIxOffset - 1);
1.3930 +
1.3931 + mStart = firstCEI->lowIndex;
1.3932 + minLimit = lastCEI->lowIndex;
1.3933 +
1.3934 + // Look at the CE following the match. If it is UCOL_NULLORDER the match
1.3935 + // extended to the end of input, and the match is good.
1.3936 +
1.3937 + // Look at the high and low indices of the CE following the match. If
1.3938 + // they are the same it means one of two things:
1.3939 + // 1. The match extended to the last CE from the target text, which is OK, or
1.3940 + // 2. The last CE that was part of the match is in an expansion that extends
1.3941 + // to the first CE after the match. In this case, we reject the match.
1.3942 + const CEI *nextCEI = 0;
1.3943 + if (strsrch->search->elementComparisonType == 0) {
1.3944 + nextCEI = ceb.get(targetIx + targetIxOffset);
1.3945 + maxLimit = nextCEI->lowIndex;
1.3946 + if (nextCEI->lowIndex == nextCEI->highIndex && nextCEI->ce != UCOL_PROCESSED_NULLORDER) {
1.3947 + found = FALSE;
1.3948 + }
1.3949 + } else {
1.3950 + for ( ; ; ++targetIxOffset ) {
1.3951 + nextCEI = ceb.get(targetIx + targetIxOffset);
1.3952 + maxLimit = nextCEI->lowIndex;
1.3953 + // If we are at the end of the target too, match succeeds
1.3954 + if ( nextCEI->ce == UCOL_PROCESSED_NULLORDER ) {
1.3955 + break;
1.3956 + }
1.3957 + // As long as the next CE has primary weight of 0,
1.3958 + // it is part of the last target element matched by the pattern;
1.3959 + // make sure it can be part of a match with the last patCE
1.3960 + if ( (((nextCEI->ce) >> 32) & 0xFFFF0000UL) == 0 ) {
1.3961 + UCompareCEsResult ceMatch = compareCE64s(nextCEI->ce, patCE, strsrch->search->elementComparisonType);
1.3962 + if ( ceMatch == U_CE_NO_MATCH || ceMatch == U_CE_SKIP_PATN ) {
1.3963 + found = FALSE;
1.3964 + break;
1.3965 + }
1.3966 + // If lowIndex == highIndex, this target CE is part of an expansion of the last matched
1.3967 + // target element, but it has non-zero primary weight => match fails
1.3968 + } else if ( nextCEI->lowIndex == nextCEI->highIndex ) {
1.3969 + found = false;
1.3970 + break;
1.3971 + // Else the target CE is not part of an expansion of the last matched element, match succeeds
1.3972 + } else {
1.3973 + break;
1.3974 + }
1.3975 + }
1.3976 + }
1.3977 +
1.3978 +
1.3979 + // Check for the start of the match being within a combining sequence.
1.3980 + // This can happen if the pattern itself begins with a combining char, and
1.3981 + // the match found combining marks in the target text that were attached
1.3982 + // to something else.
1.3983 + // This type of match should be rejected for not completely consuming a
1.3984 + // combining sequence.
1.3985 + if (!isBreakBoundary(strsrch, mStart)) {
1.3986 + found = FALSE;
1.3987 + }
1.3988 +
1.3989 + // Check for the start of the match being within an Collation Element Expansion,
1.3990 + // meaning that the first char of the match is only partially matched.
1.3991 + // With exapnsions, the first CE will report the index of the source
1.3992 + // character, and all subsequent (expansions) CEs will report the source index of the
1.3993 + // _following_ character.
1.3994 + int32_t secondIx = firstCEI->highIndex;
1.3995 + if (mStart == secondIx) {
1.3996 + found = FALSE;
1.3997 + }
1.3998 +
1.3999 + // Advance the match end position to the first acceptable match boundary.
1.4000 + // This advances the index over any combining charcters.
1.4001 + mLimit = maxLimit;
1.4002 + if (minLimit < maxLimit) {
1.4003 + // When the last CE's low index is same with its high index, the CE is likely
1.4004 + // a part of expansion. In this case, the index is located just after the
1.4005 + // character corresponding to the CEs compared above. If the index is right
1.4006 + // at the break boundary, move the position to the next boundary will result
1.4007 + // incorrect match length when there are ignorable characters exist between
1.4008 + // the position and the next character produces CE(s). See ticket#8482.
1.4009 + if (minLimit == lastCEI->highIndex && isBreakBoundary(strsrch, minLimit)) {
1.4010 + mLimit = minLimit;
1.4011 + } else {
1.4012 + int32_t nba = nextBoundaryAfter(strsrch, minLimit);
1.4013 + if (nba >= lastCEI->highIndex) {
1.4014 + mLimit = nba;
1.4015 + }
1.4016 + }
1.4017 + }
1.4018 +
1.4019 + #ifdef USEARCH_DEBUG
1.4020 + if (getenv("USEARCH_DEBUG") != NULL) {
1.4021 + printf("minLimit, maxLimit, mLimit = %d, %d, %d\n", minLimit, maxLimit, mLimit);
1.4022 + }
1.4023 + #endif
1.4024 +
1.4025 + // If advancing to the end of a combining sequence in character indexing space
1.4026 + // advanced us beyond the end of the match in CE space, reject this match.
1.4027 + if (mLimit > maxLimit) {
1.4028 + found = FALSE;
1.4029 + }
1.4030 +
1.4031 + if (!isBreakBoundary(strsrch, mLimit)) {
1.4032 + found = FALSE;
1.4033 + }
1.4034 +
1.4035 + if (! checkIdentical(strsrch, mStart, mLimit)) {
1.4036 + found = FALSE;
1.4037 + }
1.4038 +
1.4039 + if (found) {
1.4040 + break;
1.4041 + }
1.4042 + }
1.4043 +
1.4044 + #ifdef USEARCH_DEBUG
1.4045 + if (getenv("USEARCH_DEBUG") != NULL) {
1.4046 + printf("Target CEs [%d .. %d]\n", ceb.firstIx, ceb.limitIx);
1.4047 + int32_t lastToPrint = ceb.limitIx+2;
1.4048 + for (int ii=ceb.firstIx; ii<lastToPrint; ii++) {
1.4049 + printf("%8x@%d ", ceb.get(ii)->ce, ceb.get(ii)->srcIndex);
1.4050 + }
1.4051 + printf("\n%s\n", found? "match found" : "no match");
1.4052 + }
1.4053 + #endif
1.4054 +
1.4055 + // All Done. Store back the match bounds to the caller.
1.4056 + //
1.4057 + if (found==FALSE) {
1.4058 + mLimit = -1;
1.4059 + mStart = -1;
1.4060 + }
1.4061 +
1.4062 + if (matchStart != NULL) {
1.4063 + *matchStart= mStart;
1.4064 + }
1.4065 +
1.4066 + if (matchLimit != NULL) {
1.4067 + *matchLimit = mLimit;
1.4068 + }
1.4069 +
1.4070 + return found;
1.4071 +}
1.4072 +
1.4073 +U_CAPI UBool U_EXPORT2 usearch_searchBackwards(UStringSearch *strsrch,
1.4074 + int32_t startIdx,
1.4075 + int32_t *matchStart,
1.4076 + int32_t *matchLimit,
1.4077 + UErrorCode *status)
1.4078 +{
1.4079 + if (U_FAILURE(*status)) {
1.4080 + return FALSE;
1.4081 + }
1.4082 +
1.4083 + // TODO: reject search patterns beginning with a combining char.
1.4084 +
1.4085 +#ifdef USEARCH_DEBUG
1.4086 + if (getenv("USEARCH_DEBUG") != NULL) {
1.4087 + printf("Pattern CEs\n");
1.4088 + for (int ii=0; ii<strsrch->pattern.CELength; ii++) {
1.4089 + printf(" %8x", strsrch->pattern.CE[ii]);
1.4090 + }
1.4091 + printf("\n");
1.4092 + }
1.4093 +
1.4094 +#endif
1.4095 + // Input parameter sanity check.
1.4096 + // TODO: should input indicies clip to the text length
1.4097 + // in the same way that UText does.
1.4098 + if(strsrch->pattern.CELength == 0 ||
1.4099 + startIdx < 0 ||
1.4100 + startIdx > strsrch->search->textLength ||
1.4101 + strsrch->pattern.CE == NULL) {
1.4102 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.4103 + return FALSE;
1.4104 + }
1.4105 +
1.4106 + if (strsrch->pattern.PCE == NULL) {
1.4107 + initializePatternPCETable(strsrch, status);
1.4108 + }
1.4109 +
1.4110 + CEBuffer ceb(strsrch, status);
1.4111 + int32_t targetIx = 0;
1.4112 +
1.4113 + /*
1.4114 + * Pre-load the buffer with the CE's for the grapheme
1.4115 + * after our starting position so that we're sure that
1.4116 + * we can look at the CE following the match when we
1.4117 + * check the match boundaries.
1.4118 + *
1.4119 + * This will also pre-fetch the first CE that we'll
1.4120 + * consider for the match.
1.4121 + */
1.4122 + if (startIdx < strsrch->search->textLength) {
1.4123 + UBreakIterator *bi = strsrch->search->internalBreakIter;
1.4124 + int32_t next = ubrk_following(bi, startIdx);
1.4125 +
1.4126 + ucol_setOffset(strsrch->textIter, next, status);
1.4127 +
1.4128 + for (targetIx = 0; ; targetIx += 1) {
1.4129 + if (ceb.getPrevious(targetIx)->lowIndex < startIdx) {
1.4130 + break;
1.4131 + }
1.4132 + }
1.4133 + } else {
1.4134 + ucol_setOffset(strsrch->textIter, startIdx, status);
1.4135 + }
1.4136 +
1.4137 +
1.4138 + const CEI *targetCEI = NULL;
1.4139 + int32_t patIx;
1.4140 + UBool found;
1.4141 +
1.4142 + int32_t limitIx = targetIx;
1.4143 + int32_t mStart = -1;
1.4144 + int32_t mLimit = -1;
1.4145 + int32_t minLimit;
1.4146 + int32_t maxLimit;
1.4147 +
1.4148 +
1.4149 +
1.4150 + // Outer loop moves over match starting positions in the
1.4151 + // target CE space.
1.4152 + // Here, targetIx values increase toward the beginning of the base text (i.e. we get the text CEs in reverse order).
1.4153 + // But patIx is 0 at the beginning of the pattern and increases toward the end.
1.4154 + // So this loop performs a comparison starting with the end of pattern, and prcessd toward the beginning of the pattern
1.4155 + // and the beginning of the base text.
1.4156 + for(targetIx = limitIx; ; targetIx += 1)
1.4157 + {
1.4158 + found = TRUE;
1.4159 + // For targetIx > limitIx, this ceb.getPrevious gets a CE that is as far back in the ring buffer
1.4160 + // (compared to the last CE fetched for the previous targetIx value) as we need to go
1.4161 + // for this targetIx value, so if it is non-NULL then other ceb.getPrevious calls should be OK.
1.4162 + const CEI *lastCEI = ceb.getPrevious(targetIx);
1.4163 + if (lastCEI == NULL) {
1.4164 + *status = U_INTERNAL_PROGRAM_ERROR;
1.4165 + found = FALSE;
1.4166 + break;
1.4167 + }
1.4168 + // Inner loop checks for a match beginning at each
1.4169 + // position from the outer loop.
1.4170 + int32_t targetIxOffset = 0;
1.4171 + for (patIx = strsrch->pattern.PCELength - 1; patIx >= 0; patIx -= 1) {
1.4172 + int64_t patCE = strsrch->pattern.PCE[patIx];
1.4173 +
1.4174 + targetCEI = ceb.getPrevious(targetIx + strsrch->pattern.PCELength - 1 - patIx + targetIxOffset);
1.4175 + // Compare CE from target string with CE from the pattern.
1.4176 + // Note that the target CE will be UCOL_NULLORDER if we reach the end of input,
1.4177 + // which will fail the compare, below.
1.4178 + UCompareCEsResult ceMatch = compareCE64s(targetCEI->ce, patCE, strsrch->search->elementComparisonType);
1.4179 + if ( ceMatch == U_CE_NO_MATCH ) {
1.4180 + found = FALSE;
1.4181 + break;
1.4182 + } else if ( ceMatch > U_CE_NO_MATCH ) {
1.4183 + if ( ceMatch == U_CE_SKIP_TARG ) {
1.4184 + // redo with same patCE, next targCE
1.4185 + patIx++;
1.4186 + targetIxOffset++;
1.4187 + } else { // ceMatch == U_CE_SKIP_PATN
1.4188 + // redo with same targCE, next patCE
1.4189 + targetIxOffset--;
1.4190 + }
1.4191 + }
1.4192 + }
1.4193 +
1.4194 + if (!found && ((targetCEI == NULL) || (targetCEI->ce != UCOL_PROCESSED_NULLORDER))) {
1.4195 + // No match at this targetIx. Try again at the next.
1.4196 + continue;
1.4197 + }
1.4198 +
1.4199 + if (!found) {
1.4200 + // No match at all, we have run off the end of the target text.
1.4201 + break;
1.4202 + }
1.4203 +
1.4204 +
1.4205 + // We have found a match in CE space.
1.4206 + // Now determine the bounds in string index space.
1.4207 + // There still is a chance of match failure if the CE range not correspond to
1.4208 + // an acceptable character range.
1.4209 + //
1.4210 + const CEI *firstCEI = ceb.getPrevious(targetIx + strsrch->pattern.PCELength - 1 + targetIxOffset);
1.4211 + mStart = firstCEI->lowIndex;
1.4212 +
1.4213 + // Check for the start of the match being within a combining sequence.
1.4214 + // This can happen if the pattern itself begins with a combining char, and
1.4215 + // the match found combining marks in the target text that were attached
1.4216 + // to something else.
1.4217 + // This type of match should be rejected for not completely consuming a
1.4218 + // combining sequence.
1.4219 + if (!isBreakBoundary(strsrch, mStart)) {
1.4220 + found = FALSE;
1.4221 + }
1.4222 +
1.4223 + // Look at the high index of the first CE in the match. If it's the same as the
1.4224 + // low index, the first CE in the match is in the middle of an expansion.
1.4225 + if (mStart == firstCEI->highIndex) {
1.4226 + found = FALSE;
1.4227 + }
1.4228 +
1.4229 +
1.4230 + minLimit = lastCEI->lowIndex;
1.4231 +
1.4232 + if (targetIx > 0) {
1.4233 + // Look at the CE following the match. If it is UCOL_NULLORDER the match
1.4234 + // extended to the end of input, and the match is good.
1.4235 +
1.4236 + // Look at the high and low indices of the CE following the match. If
1.4237 + // they are the same it means one of two things:
1.4238 + // 1. The match extended to the last CE from the target text, which is OK, or
1.4239 + // 2. The last CE that was part of the match is in an expansion that extends
1.4240 + // to the first CE after the match. In this case, we reject the match.
1.4241 + const CEI *nextCEI = ceb.getPrevious(targetIx - 1);
1.4242 +
1.4243 + if (nextCEI->lowIndex == nextCEI->highIndex && nextCEI->ce != UCOL_PROCESSED_NULLORDER) {
1.4244 + found = FALSE;
1.4245 + }
1.4246 +
1.4247 + mLimit = maxLimit = nextCEI->lowIndex;
1.4248 +
1.4249 + // Advance the match end position to the first acceptable match boundary.
1.4250 + // This advances the index over any combining charcters.
1.4251 + if (minLimit < maxLimit) {
1.4252 + int32_t nba = nextBoundaryAfter(strsrch, minLimit);
1.4253 +
1.4254 + if (nba >= lastCEI->highIndex) {
1.4255 + mLimit = nba;
1.4256 + }
1.4257 + }
1.4258 +
1.4259 + // If advancing to the end of a combining sequence in character indexing space
1.4260 + // advanced us beyond the end of the match in CE space, reject this match.
1.4261 + if (mLimit > maxLimit) {
1.4262 + found = FALSE;
1.4263 + }
1.4264 +
1.4265 + // Make sure the end of the match is on a break boundary
1.4266 + if (!isBreakBoundary(strsrch, mLimit)) {
1.4267 + found = FALSE;
1.4268 + }
1.4269 +
1.4270 + } else {
1.4271 + // No non-ignorable CEs after this point.
1.4272 + // The maximum position is detected by boundary after
1.4273 + // the last non-ignorable CE. Combining sequence
1.4274 + // across the start index will be truncated.
1.4275 + int32_t nba = nextBoundaryAfter(strsrch, minLimit);
1.4276 + mLimit = maxLimit = (nba > 0) && (startIdx > nba) ? nba : startIdx;
1.4277 + }
1.4278 +
1.4279 + #ifdef USEARCH_DEBUG
1.4280 + if (getenv("USEARCH_DEBUG") != NULL) {
1.4281 + printf("minLimit, maxLimit, mLimit = %d, %d, %d\n", minLimit, maxLimit, mLimit);
1.4282 + }
1.4283 + #endif
1.4284 +
1.4285 +
1.4286 + if (! checkIdentical(strsrch, mStart, mLimit)) {
1.4287 + found = FALSE;
1.4288 + }
1.4289 +
1.4290 + if (found) {
1.4291 + break;
1.4292 + }
1.4293 + }
1.4294 +
1.4295 + #ifdef USEARCH_DEBUG
1.4296 + if (getenv("USEARCH_DEBUG") != NULL) {
1.4297 + printf("Target CEs [%d .. %d]\n", ceb.firstIx, ceb.limitIx);
1.4298 + int32_t lastToPrint = ceb.limitIx+2;
1.4299 + for (int ii=ceb.firstIx; ii<lastToPrint; ii++) {
1.4300 + printf("%8x@%d ", ceb.get(ii)->ce, ceb.get(ii)->srcIndex);
1.4301 + }
1.4302 + printf("\n%s\n", found? "match found" : "no match");
1.4303 + }
1.4304 + #endif
1.4305 +
1.4306 + // All Done. Store back the match bounds to the caller.
1.4307 + //
1.4308 + if (found==FALSE) {
1.4309 + mLimit = -1;
1.4310 + mStart = -1;
1.4311 + }
1.4312 +
1.4313 + if (matchStart != NULL) {
1.4314 + *matchStart= mStart;
1.4315 + }
1.4316 +
1.4317 + if (matchLimit != NULL) {
1.4318 + *matchLimit = mLimit;
1.4319 + }
1.4320 +
1.4321 + return found;
1.4322 +}
1.4323 +
1.4324 +// internal use methods declared in usrchimp.h -----------------------------
1.4325 +
1.4326 +UBool usearch_handleNextExact(UStringSearch *strsrch, UErrorCode *status)
1.4327 +{
1.4328 + if (U_FAILURE(*status)) {
1.4329 + setMatchNotFound(strsrch);
1.4330 + return FALSE;
1.4331 + }
1.4332 +
1.4333 +#if BOYER_MOORE
1.4334 + UCollationElements *coleiter = strsrch->textIter;
1.4335 + int32_t textlength = strsrch->search->textLength;
1.4336 + int32_t *patternce = strsrch->pattern.CE;
1.4337 + int32_t patterncelength = strsrch->pattern.CELength;
1.4338 + int32_t textoffset = ucol_getOffset(coleiter);
1.4339 +
1.4340 + // status used in setting coleiter offset, since offset is checked in
1.4341 + // shiftForward before setting the coleiter offset, status never
1.4342 + // a failure
1.4343 + textoffset = shiftForward(strsrch, textoffset, UCOL_NULLORDER,
1.4344 + patterncelength);
1.4345 + while (textoffset <= textlength)
1.4346 + {
1.4347 + uint32_t patternceindex = patterncelength - 1;
1.4348 + int32_t targetce;
1.4349 + UBool found = FALSE;
1.4350 + int32_t lastce = UCOL_NULLORDER;
1.4351 +
1.4352 + setColEIterOffset(coleiter, textoffset);
1.4353 +
1.4354 + for (;;) {
1.4355 + // finding the last pattern ce match, imagine composite characters
1.4356 + // for example: search for pattern A in text \u00C0
1.4357 + // we'll have to skip \u0300 the grave first before we get to A
1.4358 + targetce = ucol_previous(coleiter, status);
1.4359 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4360 + found = FALSE;
1.4361 + break;
1.4362 + }
1.4363 + targetce = getCE(strsrch, targetce);
1.4364 + if (targetce == UCOL_IGNORABLE && inNormBuf(coleiter)) {
1.4365 + // this is for the text \u0315\u0300 that requires
1.4366 + // normalization and pattern \u0300, where \u0315 is ignorable
1.4367 + continue;
1.4368 + }
1.4369 + if (lastce == UCOL_NULLORDER || lastce == UCOL_IGNORABLE) {
1.4370 + lastce = targetce;
1.4371 + }
1.4372 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4373 + if (targetce == patternce[patternceindex]) {
1.4374 + // the first ce can be a contraction
1.4375 + found = TRUE;
1.4376 + break;
1.4377 + }
1.4378 + if (!hasExpansion(coleiter)) {
1.4379 + found = FALSE;
1.4380 + break;
1.4381 + }
1.4382 + }
1.4383 +
1.4384 + //targetce = lastce;
1.4385 +
1.4386 + while (found && patternceindex > 0) {
1.4387 + lastce = targetce;
1.4388 + targetce = ucol_previous(coleiter, status);
1.4389 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4390 + found = FALSE;
1.4391 + break;
1.4392 + }
1.4393 + targetce = getCE(strsrch, targetce);
1.4394 + if (targetce == UCOL_IGNORABLE) {
1.4395 + continue;
1.4396 + }
1.4397 +
1.4398 + patternceindex --;
1.4399 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4400 + found = found && targetce == patternce[patternceindex];
1.4401 + }
1.4402 +
1.4403 + targetce = lastce;
1.4404 +
1.4405 + if (!found) {
1.4406 + if (U_FAILURE(*status)) {
1.4407 + break;
1.4408 + }
1.4409 + textoffset = shiftForward(strsrch, textoffset, lastce,
1.4410 + patternceindex);
1.4411 + // status checked at loop.
1.4412 + patternceindex = patterncelength;
1.4413 + continue;
1.4414 + }
1.4415 +
1.4416 + if (checkNextExactMatch(strsrch, &textoffset, status)) {
1.4417 + // status checked in ucol_setOffset
1.4418 + setColEIterOffset(coleiter, strsrch->search->matchedIndex);
1.4419 + return TRUE;
1.4420 + }
1.4421 + }
1.4422 + setMatchNotFound(strsrch);
1.4423 + return FALSE;
1.4424 +#else
1.4425 + int32_t textOffset = ucol_getOffset(strsrch->textIter);
1.4426 + int32_t start = -1;
1.4427 + int32_t end = -1;
1.4428 +
1.4429 + if (usearch_search(strsrch, textOffset, &start, &end, status)) {
1.4430 + strsrch->search->matchedIndex = start;
1.4431 + strsrch->search->matchedLength = end - start;
1.4432 + return TRUE;
1.4433 + } else {
1.4434 + setMatchNotFound(strsrch);
1.4435 + return FALSE;
1.4436 + }
1.4437 +#endif
1.4438 +}
1.4439 +
1.4440 +UBool usearch_handleNextCanonical(UStringSearch *strsrch, UErrorCode *status)
1.4441 +{
1.4442 + if (U_FAILURE(*status)) {
1.4443 + setMatchNotFound(strsrch);
1.4444 + return FALSE;
1.4445 + }
1.4446 +
1.4447 +#if BOYER_MOORE
1.4448 + UCollationElements *coleiter = strsrch->textIter;
1.4449 + int32_t textlength = strsrch->search->textLength;
1.4450 + int32_t *patternce = strsrch->pattern.CE;
1.4451 + int32_t patterncelength = strsrch->pattern.CELength;
1.4452 + int32_t textoffset = ucol_getOffset(coleiter);
1.4453 + UBool hasPatternAccents =
1.4454 + strsrch->pattern.hasSuffixAccents || strsrch->pattern.hasPrefixAccents;
1.4455 +
1.4456 + textoffset = shiftForward(strsrch, textoffset, UCOL_NULLORDER,
1.4457 + patterncelength);
1.4458 + strsrch->canonicalPrefixAccents[0] = 0;
1.4459 + strsrch->canonicalSuffixAccents[0] = 0;
1.4460 +
1.4461 + while (textoffset <= textlength)
1.4462 + {
1.4463 + int32_t patternceindex = patterncelength - 1;
1.4464 + int32_t targetce;
1.4465 + UBool found = FALSE;
1.4466 + int32_t lastce = UCOL_NULLORDER;
1.4467 +
1.4468 + setColEIterOffset(coleiter, textoffset);
1.4469 +
1.4470 + for (;;) {
1.4471 + // finding the last pattern ce match, imagine composite characters
1.4472 + // for example: search for pattern A in text \u00C0
1.4473 + // we'll have to skip \u0300 the grave first before we get to A
1.4474 + targetce = ucol_previous(coleiter, status);
1.4475 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4476 + found = FALSE;
1.4477 + break;
1.4478 + }
1.4479 + targetce = getCE(strsrch, targetce);
1.4480 + if (lastce == UCOL_NULLORDER || lastce == UCOL_IGNORABLE) {
1.4481 + lastce = targetce;
1.4482 + }
1.4483 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4484 + if (targetce == patternce[patternceindex]) {
1.4485 + // the first ce can be a contraction
1.4486 + found = TRUE;
1.4487 + break;
1.4488 + }
1.4489 + if (!hasExpansion(coleiter)) {
1.4490 + found = FALSE;
1.4491 + break;
1.4492 + }
1.4493 + }
1.4494 +
1.4495 + while (found && patternceindex > 0) {
1.4496 + targetce = ucol_previous(coleiter, status);
1.4497 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4498 + found = FALSE;
1.4499 + break;
1.4500 + }
1.4501 + targetce = getCE(strsrch, targetce);
1.4502 + if (targetce == UCOL_IGNORABLE) {
1.4503 + continue;
1.4504 + }
1.4505 +
1.4506 + patternceindex --;
1.4507 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4508 + found = found && targetce == patternce[patternceindex];
1.4509 + }
1.4510 +
1.4511 + // initializing the rearranged accent array
1.4512 + if (hasPatternAccents && !found) {
1.4513 + strsrch->canonicalPrefixAccents[0] = 0;
1.4514 + strsrch->canonicalSuffixAccents[0] = 0;
1.4515 + if (U_FAILURE(*status)) {
1.4516 + break;
1.4517 + }
1.4518 + found = doNextCanonicalMatch(strsrch, textoffset, status);
1.4519 + }
1.4520 +
1.4521 + if (!found) {
1.4522 + if (U_FAILURE(*status)) {
1.4523 + break;
1.4524 + }
1.4525 + textoffset = shiftForward(strsrch, textoffset, lastce,
1.4526 + patternceindex);
1.4527 + // status checked at loop
1.4528 + patternceindex = patterncelength;
1.4529 + continue;
1.4530 + }
1.4531 +
1.4532 + if (checkNextCanonicalMatch(strsrch, &textoffset, status)) {
1.4533 + setColEIterOffset(coleiter, strsrch->search->matchedIndex);
1.4534 + return TRUE;
1.4535 + }
1.4536 + }
1.4537 + setMatchNotFound(strsrch);
1.4538 + return FALSE;
1.4539 +#else
1.4540 + int32_t textOffset = ucol_getOffset(strsrch->textIter);
1.4541 + int32_t start = -1;
1.4542 + int32_t end = -1;
1.4543 +
1.4544 + if (usearch_search(strsrch, textOffset, &start, &end, status)) {
1.4545 + strsrch->search->matchedIndex = start;
1.4546 + strsrch->search->matchedLength = end - start;
1.4547 + return TRUE;
1.4548 + } else {
1.4549 + setMatchNotFound(strsrch);
1.4550 + return FALSE;
1.4551 + }
1.4552 +#endif
1.4553 +}
1.4554 +
1.4555 +UBool usearch_handlePreviousExact(UStringSearch *strsrch, UErrorCode *status)
1.4556 +{
1.4557 + if (U_FAILURE(*status)) {
1.4558 + setMatchNotFound(strsrch);
1.4559 + return FALSE;
1.4560 + }
1.4561 +
1.4562 +#if BOYER_MOORE
1.4563 + UCollationElements *coleiter = strsrch->textIter;
1.4564 + int32_t *patternce = strsrch->pattern.CE;
1.4565 + int32_t patterncelength = strsrch->pattern.CELength;
1.4566 + int32_t textoffset = ucol_getOffset(coleiter);
1.4567 +
1.4568 + // shifting it check for setting offset
1.4569 + // if setOffset is called previously or there was no previous match, we
1.4570 + // leave the offset as it is.
1.4571 + if (strsrch->search->matchedIndex != USEARCH_DONE) {
1.4572 + textoffset = strsrch->search->matchedIndex;
1.4573 + }
1.4574 +
1.4575 + textoffset = reverseShift(strsrch, textoffset, UCOL_NULLORDER,
1.4576 + patterncelength);
1.4577 +
1.4578 + while (textoffset >= 0)
1.4579 + {
1.4580 + int32_t patternceindex = 1;
1.4581 + int32_t targetce;
1.4582 + UBool found = FALSE;
1.4583 + int32_t firstce = UCOL_NULLORDER;
1.4584 +
1.4585 + // if status is a failure, ucol_setOffset does nothing
1.4586 + setColEIterOffset(coleiter, textoffset);
1.4587 +
1.4588 + for (;;) {
1.4589 + // finding the first pattern ce match, imagine composite
1.4590 + // characters. for example: search for pattern \u0300 in text
1.4591 + // \u00C0, we'll have to skip A first before we get to
1.4592 + // \u0300 the grave accent
1.4593 + targetce = ucol_next(coleiter, status);
1.4594 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4595 + found = FALSE;
1.4596 + break;
1.4597 + }
1.4598 + targetce = getCE(strsrch, targetce);
1.4599 + if (firstce == UCOL_NULLORDER || firstce == UCOL_IGNORABLE) {
1.4600 + firstce = targetce;
1.4601 + }
1.4602 + if (targetce == UCOL_IGNORABLE && strsrch->strength != UCOL_PRIMARY) {
1.4603 + continue;
1.4604 + }
1.4605 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4606 + if (targetce == patternce[0]) {
1.4607 + found = TRUE;
1.4608 + break;
1.4609 + }
1.4610 + if (!hasExpansion(coleiter)) {
1.4611 + // checking for accents in composite character
1.4612 + found = FALSE;
1.4613 + break;
1.4614 + }
1.4615 + }
1.4616 +
1.4617 + //targetce = firstce;
1.4618 +
1.4619 + while (found && (patternceindex < patterncelength)) {
1.4620 + firstce = targetce;
1.4621 + targetce = ucol_next(coleiter, status);
1.4622 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4623 + found = FALSE;
1.4624 + break;
1.4625 + }
1.4626 + targetce = getCE(strsrch, targetce);
1.4627 + if (targetce == UCOL_IGNORABLE) {
1.4628 + continue;
1.4629 + }
1.4630 +
1.4631 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4632 + found = found && targetce == patternce[patternceindex];
1.4633 + patternceindex ++;
1.4634 + }
1.4635 +
1.4636 + targetce = firstce;
1.4637 +
1.4638 + if (!found) {
1.4639 + if (U_FAILURE(*status)) {
1.4640 + break;
1.4641 + }
1.4642 +
1.4643 + textoffset = reverseShift(strsrch, textoffset, targetce,
1.4644 + patternceindex);
1.4645 + patternceindex = 0;
1.4646 + continue;
1.4647 + }
1.4648 +
1.4649 + if (checkPreviousExactMatch(strsrch, &textoffset, status)) {
1.4650 + setColEIterOffset(coleiter, textoffset);
1.4651 + return TRUE;
1.4652 + }
1.4653 + }
1.4654 + setMatchNotFound(strsrch);
1.4655 + return FALSE;
1.4656 +#else
1.4657 + int32_t textOffset;
1.4658 +
1.4659 + if (strsrch->search->isOverlap) {
1.4660 + if (strsrch->search->matchedIndex != USEARCH_DONE) {
1.4661 + textOffset = strsrch->search->matchedIndex + strsrch->search->matchedLength - 1;
1.4662 + } else {
1.4663 + // move the start position at the end of possible match
1.4664 + initializePatternPCETable(strsrch, status);
1.4665 + for (int32_t nPCEs = 0; nPCEs < strsrch->pattern.PCELength - 1; nPCEs++) {
1.4666 + int64_t pce = ucol_nextProcessed(strsrch->textIter, NULL, NULL, status);
1.4667 + if (pce == UCOL_PROCESSED_NULLORDER) {
1.4668 + // at the end of the text
1.4669 + break;
1.4670 + }
1.4671 + }
1.4672 + if (U_FAILURE(*status)) {
1.4673 + setMatchNotFound(strsrch);
1.4674 + return FALSE;
1.4675 + }
1.4676 + textOffset = ucol_getOffset(strsrch->textIter);
1.4677 + }
1.4678 + } else {
1.4679 + textOffset = ucol_getOffset(strsrch->textIter);
1.4680 + }
1.4681 +
1.4682 + int32_t start = -1;
1.4683 + int32_t end = -1;
1.4684 +
1.4685 + if (usearch_searchBackwards(strsrch, textOffset, &start, &end, status)) {
1.4686 + strsrch->search->matchedIndex = start;
1.4687 + strsrch->search->matchedLength = end - start;
1.4688 + return TRUE;
1.4689 + } else {
1.4690 + setMatchNotFound(strsrch);
1.4691 + return FALSE;
1.4692 + }
1.4693 +#endif
1.4694 +}
1.4695 +
1.4696 +UBool usearch_handlePreviousCanonical(UStringSearch *strsrch,
1.4697 + UErrorCode *status)
1.4698 +{
1.4699 + if (U_FAILURE(*status)) {
1.4700 + setMatchNotFound(strsrch);
1.4701 + return FALSE;
1.4702 + }
1.4703 +
1.4704 +#if BOYER_MOORE
1.4705 + UCollationElements *coleiter = strsrch->textIter;
1.4706 + int32_t *patternce = strsrch->pattern.CE;
1.4707 + int32_t patterncelength = strsrch->pattern.CELength;
1.4708 + int32_t textoffset = ucol_getOffset(coleiter);
1.4709 + UBool hasPatternAccents =
1.4710 + strsrch->pattern.hasSuffixAccents || strsrch->pattern.hasPrefixAccents;
1.4711 +
1.4712 + // shifting it check for setting offset
1.4713 + // if setOffset is called previously or there was no previous match, we
1.4714 + // leave the offset as it is.
1.4715 + if (strsrch->search->matchedIndex != USEARCH_DONE) {
1.4716 + textoffset = strsrch->search->matchedIndex;
1.4717 + }
1.4718 +
1.4719 + textoffset = reverseShift(strsrch, textoffset, UCOL_NULLORDER,
1.4720 + patterncelength);
1.4721 + strsrch->canonicalPrefixAccents[0] = 0;
1.4722 + strsrch->canonicalSuffixAccents[0] = 0;
1.4723 +
1.4724 + while (textoffset >= 0)
1.4725 + {
1.4726 + int32_t patternceindex = 1;
1.4727 + int32_t targetce;
1.4728 + UBool found = FALSE;
1.4729 + int32_t firstce = UCOL_NULLORDER;
1.4730 +
1.4731 + setColEIterOffset(coleiter, textoffset);
1.4732 + for (;;) {
1.4733 + // finding the first pattern ce match, imagine composite
1.4734 + // characters. for example: search for pattern \u0300 in text
1.4735 + // \u00C0, we'll have to skip A first before we get to
1.4736 + // \u0300 the grave accent
1.4737 + targetce = ucol_next(coleiter, status);
1.4738 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4739 + found = FALSE;
1.4740 + break;
1.4741 + }
1.4742 + targetce = getCE(strsrch, targetce);
1.4743 + if (firstce == UCOL_NULLORDER || firstce == UCOL_IGNORABLE) {
1.4744 + firstce = targetce;
1.4745 + }
1.4746 +
1.4747 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4748 + if (targetce == patternce[0]) {
1.4749 + // the first ce can be a contraction
1.4750 + found = TRUE;
1.4751 + break;
1.4752 + }
1.4753 + if (!hasExpansion(coleiter)) {
1.4754 + // checking for accents in composite character
1.4755 + found = FALSE;
1.4756 + break;
1.4757 + }
1.4758 + }
1.4759 +
1.4760 + targetce = firstce;
1.4761 +
1.4762 + while (found && patternceindex < patterncelength) {
1.4763 + targetce = ucol_next(coleiter, status);
1.4764 + if (U_FAILURE(*status) || targetce == UCOL_NULLORDER) {
1.4765 + found = FALSE;
1.4766 + break;
1.4767 + }
1.4768 + targetce = getCE(strsrch, targetce);
1.4769 + if (targetce == UCOL_IGNORABLE) {
1.4770 + continue;
1.4771 + }
1.4772 +
1.4773 + // TODO: #if BOYER_MOORE, replace with code using 32-bit version of compareCE64s
1.4774 + found = found && targetce == patternce[patternceindex];
1.4775 + patternceindex ++;
1.4776 + }
1.4777 +
1.4778 + // initializing the rearranged accent array
1.4779 + if (hasPatternAccents && !found) {
1.4780 + strsrch->canonicalPrefixAccents[0] = 0;
1.4781 + strsrch->canonicalSuffixAccents[0] = 0;
1.4782 + if (U_FAILURE(*status)) {
1.4783 + break;
1.4784 + }
1.4785 + found = doPreviousCanonicalMatch(strsrch, textoffset, status);
1.4786 + }
1.4787 +
1.4788 + if (!found) {
1.4789 + if (U_FAILURE(*status)) {
1.4790 + break;
1.4791 + }
1.4792 + textoffset = reverseShift(strsrch, textoffset, targetce,
1.4793 + patternceindex);
1.4794 + patternceindex = 0;
1.4795 + continue;
1.4796 + }
1.4797 +
1.4798 + if (checkPreviousCanonicalMatch(strsrch, &textoffset, status)) {
1.4799 + setColEIterOffset(coleiter, textoffset);
1.4800 + return TRUE;
1.4801 + }
1.4802 + }
1.4803 + setMatchNotFound(strsrch);
1.4804 + return FALSE;
1.4805 +#else
1.4806 + int32_t textOffset;
1.4807 +
1.4808 + if (strsrch->search->isOverlap) {
1.4809 + if (strsrch->search->matchedIndex != USEARCH_DONE) {
1.4810 + textOffset = strsrch->search->matchedIndex + strsrch->search->matchedLength - 1;
1.4811 + } else {
1.4812 + // move the start position at the end of possible match
1.4813 + initializePatternPCETable(strsrch, status);
1.4814 + for (int32_t nPCEs = 0; nPCEs < strsrch->pattern.PCELength - 1; nPCEs++) {
1.4815 + int64_t pce = ucol_nextProcessed(strsrch->textIter, NULL, NULL, status);
1.4816 + if (pce == UCOL_PROCESSED_NULLORDER) {
1.4817 + // at the end of the text
1.4818 + break;
1.4819 + }
1.4820 + }
1.4821 + if (U_FAILURE(*status)) {
1.4822 + setMatchNotFound(strsrch);
1.4823 + return FALSE;
1.4824 + }
1.4825 + textOffset = ucol_getOffset(strsrch->textIter);
1.4826 + }
1.4827 + } else {
1.4828 + textOffset = ucol_getOffset(strsrch->textIter);
1.4829 + }
1.4830 +
1.4831 + int32_t start = -1;
1.4832 + int32_t end = -1;
1.4833 +
1.4834 + if (usearch_searchBackwards(strsrch, textOffset, &start, &end, status)) {
1.4835 + strsrch->search->matchedIndex = start;
1.4836 + strsrch->search->matchedLength = end - start;
1.4837 + return TRUE;
1.4838 + } else {
1.4839 + setMatchNotFound(strsrch);
1.4840 + return FALSE;
1.4841 + }
1.4842 +#endif
1.4843 +}
1.4844 +
1.4845 +#endif /* #if !UCONFIG_NO_COLLATION */
