icuSources/common/caniter.cpp

   1 /*
   2  *****************************************************************************
   3  * Copyright (C) 1996-2006, International Business Machines Corporation and  *
   4  * others. All Rights Reserved.                                              *
   5  *****************************************************************************
   6  */
   7
   8 #include "unicode/utypes.h"
   9
  10 #if !UCONFIG_NO_NORMALIZATION
  11
  12 #include "unicode/uset.h"
  13 #include "unicode/ustring.h"
  14 #include "hash.h"
  15 #include "unormimp.h"
  16 #include "unicode/caniter.h"
  17 #include "unicode/normlzr.h"
  18 #include "unicode/uchar.h"
  19 #include "cmemory.h"
  20
  21 /**
  22  * This class allows one to iterate through all the strings that are canonically equivalent to a given
  23  * string. For example, here are some sample results:
  24 Results for: {LATIN CAPITAL LETTER A WITH RING ABOVE}{LATIN SMALL LETTER D}{COMBINING DOT ABOVE}{COMBINING CEDILLA}
  25 1: \u0041\u030A\u0064\u0307\u0327
  26  = {LATIN CAPITAL LETTER A}{COMBINING RING ABOVE}{LATIN SMALL LETTER D}{COMBINING DOT ABOVE}{COMBINING CEDILLA}
  27 2: \u0041\u030A\u0064\u0327\u0307
  28  = {LATIN CAPITAL LETTER A}{COMBINING RING ABOVE}{LATIN SMALL LETTER D}{COMBINING CEDILLA}{COMBINING DOT ABOVE}
  29 3: \u0041\u030A\u1E0B\u0327
  30  = {LATIN CAPITAL LETTER A}{COMBINING RING ABOVE}{LATIN SMALL LETTER D WITH DOT ABOVE}{COMBINING CEDILLA}
  31 4: \u0041\u030A\u1E11\u0307
  32  = {LATIN CAPITAL LETTER A}{COMBINING RING ABOVE}{LATIN SMALL LETTER D WITH CEDILLA}{COMBINING DOT ABOVE}
  33 5: \u00C5\u0064\u0307\u0327
  34  = {LATIN CAPITAL LETTER A WITH RING ABOVE}{LATIN SMALL LETTER D}{COMBINING DOT ABOVE}{COMBINING CEDILLA}
  35 6: \u00C5\u0064\u0327\u0307
  36  = {LATIN CAPITAL LETTER A WITH RING ABOVE}{LATIN SMALL LETTER D}{COMBINING CEDILLA}{COMBINING DOT ABOVE}
  37 7: \u00C5\u1E0B\u0327
  38  = {LATIN CAPITAL LETTER A WITH RING ABOVE}{LATIN SMALL LETTER D WITH DOT ABOVE}{COMBINING CEDILLA}
  39 8: \u00C5\u1E11\u0307
  40  = {LATIN CAPITAL LETTER A WITH RING ABOVE}{LATIN SMALL LETTER D WITH CEDILLA}{COMBINING DOT ABOVE}
  41 9: \u212B\u0064\u0307\u0327
  42  = {ANGSTROM SIGN}{LATIN SMALL LETTER D}{COMBINING DOT ABOVE}{COMBINING CEDILLA}
  43 10: \u212B\u0064\u0327\u0307
  44  = {ANGSTROM SIGN}{LATIN SMALL LETTER D}{COMBINING CEDILLA}{COMBINING DOT ABOVE}
  45 11: \u212B\u1E0B\u0327
  46  = {ANGSTROM SIGN}{LATIN SMALL LETTER D WITH DOT ABOVE}{COMBINING CEDILLA}
  47 12: \u212B\u1E11\u0307
  48  = {ANGSTROM SIGN}{LATIN SMALL LETTER D WITH CEDILLA}{COMBINING DOT ABOVE}
  49  *<br>Note: the code is intended for use with small strings, and is not suitable for larger ones,
  50  * since it has not been optimized for that situation.
  51  *@author M. Davis
  52  *@draft
  53  */
  54
  55 // public
  56
  57 U_NAMESPACE_BEGIN
  58
  59 // TODO: add boilerplate methods.
  60
  61 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(CanonicalIterator)
  62
  63 /**
  64  *@param source string to get results for
  65  */
  66 CanonicalIterator::CanonicalIterator(const UnicodeString &sourceStr, UErrorCode &status) :
  67     pieces(NULL),
  68     pieces_length(0),
  69     pieces_lengths(NULL),
  70     current(NULL),
  71     current_length(0)
  72 {
  73     if(U_SUCCESS(status)) {
  74       setSource(sourceStr, status);
  75     }
  76 }
  77
  78 CanonicalIterator::~CanonicalIterator() {
  79   cleanPieces();
  80 }
  81
  82 void CanonicalIterator::cleanPieces() {
  83     int32_t i = 0;
  84     if(pieces != NULL) {
  85         for(i = 0; i < pieces_length; i++) {
  86             if(pieces[i] != NULL) {
  87                 delete[] pieces[i];
  88             }
  89         }
  90         uprv_free(pieces);
  91         pieces = NULL;
  92         pieces_length = 0;
  93     }
  94     if(pieces_lengths != NULL) {
  95         uprv_free(pieces_lengths);
  96         pieces_lengths = NULL;
  97     }
  98     if(current != NULL) {
  99         uprv_free(current);
 100         current = NULL;
 101         current_length = 0;
 102     }
 103 }
 104
 105 /**
 106  *@return gets the source: NOTE: it is the NFD form of source
 107  */
 108 UnicodeString CanonicalIterator::getSource() {
 109   return source;
 110 }
 111
 112 /**
 113  * Resets the iterator so that one can start again from the beginning.
 114  */
 115 void CanonicalIterator::reset() {
 116     done = FALSE;
 117     for (int i = 0; i < current_length; ++i) {
 118         current[i] = 0;
 119     }
 120 }
 121
 122 /**
 123  *@return the next string that is canonically equivalent. The value null is returned when
 124  * the iteration is done.
 125  */
 126 UnicodeString CanonicalIterator::next() {
 127     int32_t i = 0;
 128
 129     if (done) {
 130       buffer.setToBogus();
 131       return buffer;
 132     }
 133
 134     // delete old contents
 135     buffer.remove();
 136
 137     // construct return value
 138
 139     for (i = 0; i < pieces_length; ++i) {
 140         buffer.append(pieces[i][current[i]]);
 141     }
 142     //String result = buffer.toString(); // not needed
 143
 144     // find next value for next time
 145
 146     for (i = current_length - 1; ; --i) {
 147         if (i < 0) {
 148             done = TRUE;
 149             break;
 150         }
 151         current[i]++;
 152         if (current[i] < pieces_lengths[i]) break; // got sequence
 153         current[i] = 0;
 154     }
 155     return buffer;
 156 }
 157
 158 /**
 159  *@param set the source string to iterate against. This allows the same iterator to be used
 160  * while changing the source string, saving object creation.
 161  */
 162 void CanonicalIterator::setSource(const UnicodeString &newSource, UErrorCode &status) {
 163     int32_t list_length = 0;
 164     UChar32 cp = 0;
 165     int32_t start = 0;
 166     int32_t i = 0;
 167     UnicodeString *list = NULL;
 168
 169     Normalizer::normalize(newSource, UNORM_NFD, 0, source, status);
 170     if(U_FAILURE(status)) {
 171       return;
 172     }
 173     done = FALSE;
 174
 175     cleanPieces();
 176
 177     // catch degenerate case
 178     if (newSource.length() == 0) {
 179         pieces = (UnicodeString **)uprv_malloc(sizeof(UnicodeString *));
 180         pieces_lengths = (int32_t*)uprv_malloc(1 * sizeof(int32_t));
 181         pieces_length = 1;
 182         current = (int32_t*)uprv_malloc(1 * sizeof(int32_t));
 183         current_length = 1;
 184         if (pieces == NULL || pieces_lengths == NULL || current == NULL) {
 185             status = U_MEMORY_ALLOCATION_ERROR;
 186             goto CleanPartialInitialization;
 187         }
 188         current[0] = 0;
 189         pieces[0] = new UnicodeString[1];
 190         pieces_lengths[0] = 1;
 191         if (pieces[0] == 0) {
 192             status = U_MEMORY_ALLOCATION_ERROR;
 193             goto CleanPartialInitialization;
 194         }
 195         return;
 196     }
 197
 198
 199     list = new UnicodeString[source.length()];
 200     if (list == 0) {
 201         status = U_MEMORY_ALLOCATION_ERROR;
 202         goto CleanPartialInitialization;
 203     }
 204
 205     // i should initialy be the number of code units at the
 206     // start of the string
 207     i = UTF16_CHAR_LENGTH(source.char32At(0));
 208     //int32_t i = 1;
 209     // find the segments
 210     // This code iterates through the source string and
 211     // extracts segments that end up on a codepoint that
 212     // doesn't start any decompositions. (Analysis is done
 213     // on the NFD form - see above).
 214     for (; i < source.length(); i += UTF16_CHAR_LENGTH(cp)) {
 215         cp = source.char32At(i);
 216         if (unorm_isCanonSafeStart(cp)) {
 217             source.extract(start, i-start, list[list_length++]); // add up to i
 218             start = i;
 219         }
 220     }
 221     source.extract(start, i-start, list[list_length++]); // add last one
 222
 223
 224     // allocate the arrays, and find the strings that are CE to each segment
 225     pieces = (UnicodeString **)uprv_malloc(list_length * sizeof(UnicodeString *));
 226     pieces_length = list_length;
 227     pieces_lengths = (int32_t*)uprv_malloc(list_length * sizeof(int32_t));
 228     current = (int32_t*)uprv_malloc(list_length * sizeof(int32_t));
 229     current_length = list_length;
 230     if (pieces == NULL || pieces_lengths == NULL || current == NULL) {
 231         status = U_MEMORY_ALLOCATION_ERROR;
 232         goto CleanPartialInitialization;
 233     }
 234
 235     for (i = 0; i < current_length; i++) {
 236         current[i] = 0;
 237     }
 238     // for each segment, get all the combinations that can produce
 239     // it after NFD normalization
 240     for (i = 0; i < pieces_length; ++i) {
 241         //if (PROGRESS) printf("SEGMENT\n");
 242         pieces[i] = getEquivalents(list[i], pieces_lengths[i], status);
 243     }
 244
 245     delete[] list;
 246     return;
 247 // Common section to cleanup all local variables and reset object variables.
 248 CleanPartialInitialization:
 249     if (list != NULL) {
 250         delete[] list;
 251     }
 252     cleanPieces();
 253 }
 254
 255 /**
 256  * Dumb recursive implementation of permutation.
 257  * TODO: optimize
 258  * @param source the string to find permutations for
 259  * @return the results in a set.
 260  */
 261 void U_EXPORT2 CanonicalIterator::permute(UnicodeString &source, UBool skipZeros, Hashtable *result, UErrorCode &status) {
 262     if(U_FAILURE(status)) {
 263         return;
 264     }
 265     //if (PROGRESS) printf("Permute: %s\n", UToS(Tr(source)));
 266     int32_t i = 0;
 267
 268     // optimization:
 269     // if zero or one character, just return a set with it
 270     // we check for length < 2 to keep from counting code points all the time
 271     if (source.length() <= 2 && source.countChar32() <= 1) {
 272         UnicodeString *toPut = new UnicodeString(source);
 273         /* test for NULL */
 274         if (toPut == 0) {
 275             status = U_MEMORY_ALLOCATION_ERROR;
 276             return;
 277         }
 278         result->put(source, toPut, status);
 279         return;
 280     }
 281
 282     // otherwise iterate through the string, and recursively permute all the other characters
 283     UChar32 cp;
 284     Hashtable subpermute(status);
 285     if(U_FAILURE(status)) {
 286         return;
 287     }
 288     subpermute.setValueDeleter(uhash_deleteUnicodeString);
 289
 290     for (i = 0; i < source.length(); i += UTF16_CHAR_LENGTH(cp)) {
 291         cp = source.char32At(i);
 292         const UHashElement *ne = NULL;
 293         int32_t el = -1;
 294         UnicodeString subPermuteString = source;
 295
 296         // optimization:
 297         // if the character is canonical combining class zero,
 298         // don't permute it
 299         if (skipZeros && i != 0 && u_getCombiningClass(cp) == 0) {
 300             //System.out.println("Skipping " + Utility.hex(UTF16.valueOf(source, i)));
 301             continue;
 302         }
 303
 304         subpermute.removeAll();
 305
 306         // see what the permutations of the characters before and after this one are
 307         //Hashtable *subpermute = permute(source.substring(0,i) + source.substring(i + UTF16.getCharCount(cp)));
 308         permute(subPermuteString.replace(i, UTF16_CHAR_LENGTH(cp), NULL, 0), skipZeros, &subpermute, status);
 309         /* Test for buffer overflows */
 310         if(U_FAILURE(status)) {
 311             return;
 312         }
 313         // The upper replace is destructive. The question is do we have to make a copy, or we don't care about the contents
 314         // of source at this point.
 315
 316         // prefix this character to all of them
 317         ne = subpermute.nextElement(el);
 318         while (ne != NULL) {
 319             UnicodeString *permRes = (UnicodeString *)(ne->value.pointer);
 320             UnicodeString *chStr = new UnicodeString(cp);
 321             //test for  NULL
 322             if (chStr == NULL) {
 323                 status = U_MEMORY_ALLOCATION_ERROR;
 324                 return;
 325             }
 326             chStr->append(*permRes); //*((UnicodeString *)(ne->value.pointer));
 327             //if (PROGRESS) printf("  Piece: %s\n", UToS(*chStr));
 328             result->put(*chStr, chStr, status);
 329             ne = subpermute.nextElement(el);
 330         }
 331     }
 332     //return result;
 333 }
 334
 335 // privates
 336
 337 // we have a segment, in NFD. Find all the strings that are canonically equivalent to it.
 338 UnicodeString* CanonicalIterator::getEquivalents(const UnicodeString &segment, int32_t &result_len, UErrorCode &status) {
 339     Hashtable result(status);
 340     Hashtable permutations(status);
 341     Hashtable basic(status);
 342     if (U_FAILURE(status)) {
 343         return 0;
 344     }
 345     result.setValueDeleter(uhash_deleteUnicodeString);
 346     permutations.setValueDeleter(uhash_deleteUnicodeString);
 347     basic.setValueDeleter(uhash_deleteUnicodeString);
 348
 349     UChar USeg[256];
 350     int32_t segLen = segment.extract(USeg, 256, status);
 351     getEquivalents2(&basic, USeg, segLen, status);
 352
 353     // now get all the permutations
 354     // add only the ones that are canonically equivalent
 355     // TODO: optimize by not permuting any class zero.
 356
 357     const UHashElement *ne = NULL;
 358     int32_t el = -1;
 359     //Iterator it = basic.iterator();
 360     ne = basic.nextElement(el);
 361     //while (it.hasNext())
 362     while (ne != NULL) {
 363         //String item = (String) it.next();
 364         UnicodeString item = *((UnicodeString *)(ne->value.pointer));
 365
 366         permutations.removeAll();
 367         permute(item, CANITER_SKIP_ZEROES, &permutations, status);
 368         const UHashElement *ne2 = NULL;
 369         int32_t el2 = -1;
 370         //Iterator it2 = permutations.iterator();
 371         ne2 = permutations.nextElement(el2);
 372         //while (it2.hasNext())
 373         while (ne2 != NULL) {
 374             //String possible = (String) it2.next();
 375             //UnicodeString *possible = new UnicodeString(*((UnicodeString *)(ne2->value.pointer)));
 376             UnicodeString possible(*((UnicodeString *)(ne2->value.pointer)));
 377             UnicodeString attempt;
 378             Normalizer::normalize(possible, UNORM_NFD, 0, attempt, status);
 379
 380             // TODO: check if operator == is semanticaly the same as attempt.equals(segment)
 381             if (attempt==segment) {
 382                 //if (PROGRESS) printf("Adding Permutation: %s\n", UToS(Tr(*possible)));
 383                 // TODO: use the hashtable just to catch duplicates - store strings directly (somehow).
 384                 result.put(possible, new UnicodeString(possible), status); //add(possible);
 385             } else {
 386                 //if (PROGRESS) printf("-Skipping Permutation: %s\n", UToS(Tr(*possible)));
 387             }
 388
 389             ne2 = permutations.nextElement(el2);
 390         }
 391         ne = basic.nextElement(el);
 392     }
 393
 394     /* Test for buffer overflows */
 395     if(U_FAILURE(status)) {
 396         return 0;
 397     }
 398     // convert into a String[] to clean up storage
 399     //String[] finalResult = new String[result.size()];
 400     UnicodeString *finalResult = NULL;
 401     int32_t resultCount;
 402     if((resultCount = result.count())) {
 403         finalResult = new UnicodeString[resultCount];
 404         if (finalResult == 0) {
 405             status = U_MEMORY_ALLOCATION_ERROR;
 406             return NULL;
 407         }
 408     }
 409     else {
 410         status = U_ILLEGAL_ARGUMENT_ERROR;
 411         return NULL;
 412     }
 413     //result.toArray(finalResult);
 414     result_len = 0;
 415     el = -1;
 416     ne = result.nextElement(el);
 417     while(ne != NULL) {
 418         finalResult[result_len++] = *((UnicodeString *)(ne->value.pointer));
 419         ne = result.nextElement(el);
 420     }
 421
 422
 423     return finalResult;
 424 }
 425
 426 Hashtable *CanonicalIterator::getEquivalents2(Hashtable *fillinResult, const UChar *segment, int32_t segLen, UErrorCode &status) {
 427
 428     if (U_FAILURE(status)) {
 429         return NULL;
 430     }
 431
 432     //if (PROGRESS) printf("Adding: %s\n", UToS(Tr(segment)));
 433
 434     UnicodeString toPut(segment, segLen);
 435
 436     fillinResult->put(toPut, new UnicodeString(toPut), status);
 437
 438     USerializedSet starts;
 439
 440     // cycle through all the characters
 441     UChar32 cp, end = 0;
 442     int32_t i = 0, j;
 443     for (i = 0; i < segLen; i += UTF16_CHAR_LENGTH(cp)) {
 444         // see if any character is at the start of some decomposition
 445         UTF_GET_CHAR(segment, 0, i, segLen, cp);
 446         if (!unorm_getCanonStartSet(cp, &starts)) {
 447             continue;
 448         }
 449         // if so, see which decompositions match
 450         for(j = 0, cp = end+1; cp <= end || uset_getSerializedRange(&starts, j++, &cp, &end); ++cp) {
 451             Hashtable remainder(status);
 452             remainder.setValueDeleter(uhash_deleteUnicodeString);
 453             if (extract(&remainder, cp, segment, segLen, i, status) == NULL) {
 454                 continue;
 455             }
 456
 457             // there were some matches, so add all the possibilities to the set.
 458             UnicodeString prefix(segment, i);
 459             prefix += cp;
 460
 461             int32_t el = -1;
 462             const UHashElement *ne = remainder.nextElement(el);
 463             while (ne != NULL) {
 464                 UnicodeString item = *((UnicodeString *)(ne->value.pointer));
 465                 UnicodeString *toAdd = new UnicodeString(prefix);
 466                 /* test for NULL */
 467                 if (toAdd == 0) {
 468                     status = U_MEMORY_ALLOCATION_ERROR;
 469                     return NULL;
 470                 }
 471                 *toAdd += item;
 472                 fillinResult->put(*toAdd, toAdd, status);
 473
 474                 //if (PROGRESS) printf("Adding: %s\n", UToS(Tr(*toAdd)));
 475
 476                 ne = remainder.nextElement(el);
 477             }
 478         }
 479     }
 480
 481     /* Test for buffer overflows */
 482     if(U_FAILURE(status)) {
 483         return NULL;
 484     }
 485     return fillinResult;
 486 }
 487
 488 /**
 489  * See if the decomposition of cp2 is at segment starting at segmentPos
 490  * (with canonical rearrangment!)
 491  * If so, take the remainder, and return the equivalents
 492  */
 493 Hashtable *CanonicalIterator::extract(Hashtable *fillinResult, UChar32 comp, const UChar *segment, int32_t segLen, int32_t segmentPos, UErrorCode &status) {
 494 //Hashtable *CanonicalIterator::extract(UChar32 comp, const UnicodeString &segment, int32_t segLen, int32_t segmentPos, UErrorCode &status) {
 495     //if (PROGRESS) printf(" extract: %s, ", UToS(Tr(UnicodeString(comp))));
 496     //if (PROGRESS) printf("%s, %i\n", UToS(Tr(segment)), segmentPos);
 497
 498     if (U_FAILURE(status)) {
 499         return NULL;
 500     }
 501
 502     const int32_t bufSize = 256;
 503     int32_t bufLen = 0;
 504     UChar temp[bufSize];
 505
 506     int32_t inputLen = 0, decompLen;
 507     UChar stackBuffer[4];
 508     const UChar *decomp;
 509
 510     U16_APPEND_UNSAFE(temp, inputLen, comp);
 511     decomp = unorm_getCanonicalDecomposition(comp, stackBuffer, &decompLen);
 512     if(decomp == NULL) {
 513         /* copy temp */
 514         stackBuffer[0] = temp[0];
 515         if(inputLen > 1) {
 516             stackBuffer[1] = temp[1];
 517         }
 518         decomp = stackBuffer;
 519         decompLen = inputLen;
 520     }
 521
 522     UChar *buff = temp+inputLen;
 523
 524     // See if it matches the start of segment (at segmentPos)
 525     UBool ok = FALSE;
 526     UChar32 cp;
 527     int32_t decompPos = 0;
 528     UChar32 decompCp;
 529     UTF_NEXT_CHAR(decomp, decompPos, decompLen, decompCp);
 530
 531     int32_t i;
 532     UBool overflow = FALSE;
 533
 534     i = segmentPos;
 535     while(i < segLen) {
 536         UTF_NEXT_CHAR(segment, i, segLen, cp);
 537
 538         if (cp == decompCp) { // if equal, eat another cp from decomp
 539
 540             //if (PROGRESS) printf("  matches: %s\n", UToS(Tr(UnicodeString(cp))));
 541
 542             if (decompPos == decompLen) { // done, have all decomp characters!
 543                 //u_strcat(buff+bufLen, segment+i);
 544                 uprv_memcpy(buff+bufLen, segment+i, (segLen-i)*sizeof(UChar));
 545                 bufLen+=segLen-i;
 546
 547                 ok = TRUE;
 548                 break;
 549             }
 550             UTF_NEXT_CHAR(decomp, decompPos, decompLen, decompCp);
 551         } else {
 552             //if (PROGRESS) printf("  buffer: %s\n", UToS(Tr(UnicodeString(cp))));
 553
 554             // brute force approach
 555
 556             U16_APPEND(buff, bufLen, bufSize, cp, overflow);
 557
 558             if(overflow) {
 559                 /*
 560                  * ### TODO handle buffer overflow
 561                  * The buffer is large, but an overflow may still happen with
 562                  * unusual input (many combining marks?).
 563                  * Reallocate buffer and continue.
 564                  * markus 20020929
 565                  */
 566
 567                 overflow = FALSE;
 568             }
 569
 570             /* TODO: optimize
 571             // since we know that the classes are monotonically increasing, after zero
 572             // e.g. 0 5 7 9 0 3
 573             // we can do an optimization
 574             // there are only a few cases that work: zero, less, same, greater
 575             // if both classes are the same, we fail
 576             // if the decomp class < the segment class, we fail
 577
 578             segClass = getClass(cp);
 579             if (decompClass <= segClass) return null;
 580             */
 581         }
 582     }
 583     if (!ok)
 584         return NULL; // we failed, characters left over
 585
 586     //if (PROGRESS) printf("Matches\n");
 587
 588     if (bufLen == 0) {
 589         fillinResult->put(UnicodeString(), new UnicodeString(), status);
 590         return fillinResult; // succeed, but no remainder
 591     }
 592
 593     // brute force approach
 594     // check to make sure result is canonically equivalent
 595     int32_t tempLen = inputLen + bufLen;
 596
 597     UChar trial[bufSize];
 598     unorm_decompose(trial, bufSize, temp, tempLen, FALSE, 0, &status);
 599
 600     if(U_FAILURE(status)
 601         || uprv_memcmp(segment+segmentPos, trial, (segLen - segmentPos)*sizeof(UChar)) != 0)
 602     {
 603         return NULL;
 604     }
 605
 606     return getEquivalents2(fillinResult, buff, bufLen, status);
 607 }
 608
 609 U_NAMESPACE_END
 610
 611 #endif /* #if !UCONFIG_NO_NORMALIZATION */