icuSources/common/unormcmp.cpp

   1 /*
   2 *******************************************************************************
   3 *
   4 *   Copyright (C) 2001-2004, International Business Machines
   5 *   Corporation and others.  All Rights Reserved.
   6 *
   7 *******************************************************************************
   8 *   file name:  unormcmp.cpp
   9 *   encoding:   US-ASCII
  10 *   tab size:   8 (not used)
  11 *   indentation:4
  12 *
  13 *   created on: 2004sep13
  14 *   created by: Markus W. Scherer
  15 *
  16 *   unorm_compare() function moved here from unorm.cpp for better modularization.
  17 *   Depends on both normalization and case folding.
  18 *   Allows unorm.cpp to not depend on any character properties code.
  19 */
  20
  21 #include "unicode/utypes.h"
  22
  23 #if !UCONFIG_NO_NORMALIZATION
  24
  25 #include "unicode/ustring.h"
  26 #include "unicode/unorm.h"
  27 #include "unicode/uniset.h"
  28 #include "unormimp.h"
  29 #include "ucase.h"
  30 #include "cmemory.h"
  31
  32 #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
  33
  34 /* compare canonically equivalent ------------------------------------------- */
  35
  36 /*
  37  * Compare two strings for canonical equivalence.
  38  * Further options include case-insensitive comparison and
  39  * code point order (as opposed to code unit order).
  40  *
  41  * In this function, canonical equivalence is optional as well.
  42  * If canonical equivalence is tested, then both strings must fulfill
  43  * the FCD check.
  44  *
  45  * Semantically, this is equivalent to
  46  *   strcmp[CodePointOrder](NFD(foldCase(s1)), NFD(foldCase(s2)))
  47  * where code point order, NFD and foldCase are all optional.
  48  *
  49  * String comparisons almost always yield results before processing both strings
  50  * completely.
  51  * They are generally more efficient working incrementally instead of
  52  * performing the sub-processing (strlen, normalization, case-folding)
  53  * on the entire strings first.
  54  *
  55  * It is also unnecessary to not normalize identical characters.
  56  *
  57  * This function works in principle as follows:
  58  *
  59  * loop {
  60  *   get one code unit c1 from s1 (-1 if end of source)
  61  *   get one code unit c2 from s2 (-1 if end of source)
  62  *
  63  *   if(either string finished) {
  64  *     return result;
  65  *   }
  66  *   if(c1==c2) {
  67  *     continue;
  68  *   }
  69  *
  70  *   // c1!=c2
  71  *   try to decompose/case-fold c1/c2, and continue if one does;
  72  *
  73  *   // still c1!=c2 and neither decomposes/case-folds, return result
  74  *   return c1-c2;
  75  * }
  76  *
  77  * When a character decomposes, then the pointer for that source changes to
  78  * the decomposition, pushing the previous pointer onto a stack.
  79  * When the end of the decomposition is reached, then the code unit reader
  80  * pops the previous source from the stack.
  81  * (Same for case-folding.)
  82  *
  83  * This is complicated further by operating on variable-width UTF-16.
  84  * The top part of the loop works on code units, while lookups for decomposition
  85  * and case-folding need code points.
  86  * Code points are assembled after the equality/end-of-source part.
  87  * The source pointer is only advanced beyond all code units when the code point
  88  * actually decomposes/case-folds.
  89  *
  90  * If we were on a trail surrogate unit when assembling a code point,
  91  * and the code point decomposes/case-folds, then the decomposition/folding
  92  * result must be compared with the part of the other string that corresponds to
  93  * this string's lead surrogate.
  94  * Since we only assemble a code point when hitting a trail unit when the
  95  * preceding lead units were identical, we back up the other string by one unit
  96  * in such a case.
  97  *
  98  * The optional code point order comparison at the end works with
  99  * the same fix-up as the other code point order comparison functions.
 100  * See ustring.c and the comment near the end of this function.
 101  *
 102  * Assumption: A decomposition or case-folding result string never contains
 103  * a single surrogate. This is a safe assumption in the Unicode Standard.
 104  * Therefore, we do not need to check for surrogate pairs across
 105  * decomposition/case-folding boundaries.
 106  *
 107  * Further assumptions (see verifications tstnorm.cpp):
 108  * The API function checks for FCD first, while the core function
 109  * first case-folds and then decomposes. This requires that case-folding does not
 110  * un-FCD any strings.
 111  *
 112  * The API function may also NFD the input and turn off decomposition.
 113  * This requires that case-folding does not un-NFD strings either.
 114  *
 115  * TODO If any of the above two assumptions is violated,
 116  * then this entire code must be re-thought.
 117  * If this happens, then a simple solution is to case-fold both strings up front
 118  * and to turn off UNORM_INPUT_IS_FCD.
 119  * We already do this when not both strings are in FCD because makeFCD
 120  * would be a partial NFD before the case folding, which does not work.
 121  * Note that all of this is only a problem when case-folding _and_
 122  * canonical equivalence come together.
 123  * (Comments in unorm_compare() are more up to date than this TODO.)
 124  *
 125  * This function could be moved to a different source file, at increased cost
 126  * for calling the decomposition access function.
 127  */
 128
 129 /* stack element for previous-level source/decomposition pointers */
 130 struct CmpEquivLevel {
 131     const UChar *start, *s, *limit;
 132 };
 133 typedef struct CmpEquivLevel CmpEquivLevel;
 134
 135 /* internal function */
 136 static int32_t
 137 unorm_cmpEquivFold(const UChar *s1, int32_t length1,
 138                    const UChar *s2, int32_t length2,
 139                    uint32_t options,
 140                    UErrorCode *pErrorCode) {
 141     UCaseProps *csp;
 142
 143     /* current-level start/limit - s1/s2 as current */
 144     const UChar *start1, *start2, *limit1, *limit2;
 145
 146     /* decomposition and case folding variables */
 147     const UChar *p;
 148     int32_t length;
 149
 150     /* stacks of previous-level start/current/limit */
 151     CmpEquivLevel stack1[2], stack2[2];
 152
 153     /* decomposition buffers for Hangul */
 154     UChar decomp1[4], decomp2[4];
 155
 156     /* case folding buffers, only use current-level start/limit */
 157     UChar fold1[UCASE_MAX_STRING_LENGTH+1], fold2[UCASE_MAX_STRING_LENGTH+1];
 158
 159     /* track which is the current level per string */
 160     int32_t level1, level2;
 161
 162     /* current code units, and code points for lookups */
 163     UChar32 c1, c2, cp1, cp2;
 164
 165     /* no argument error checking because this itself is not an API */
 166
 167     /*
 168      * assume that at least one of the options _COMPARE_EQUIV and U_COMPARE_IGNORE_CASE is set
 169      * otherwise this function must behave exactly as uprv_strCompare()
 170      * not checking for that here makes testing this function easier
 171      */
 172
 173     /* normalization/properties data loaded? */
 174     if( ((options&_COMPARE_EQUIV)!=0 && !unorm_haveData(pErrorCode)) ||
 175         U_FAILURE(*pErrorCode)
 176     ) {
 177         return 0;
 178     }
 179     if((options&U_COMPARE_IGNORE_CASE)!=0) {
 180         csp=ucase_getSingleton(pErrorCode);
 181         if(U_FAILURE(*pErrorCode)) {
 182             return 0;
 183         }
 184     } else {
 185         csp=NULL;
 186     }
 187
 188     /* initialize */
 189     start1=s1;
 190     if(length1==-1) {
 191         limit1=NULL;
 192     } else {
 193         limit1=s1+length1;
 194     }
 195
 196     start2=s2;
 197     if(length2==-1) {
 198         limit2=NULL;
 199     } else {
 200         limit2=s2+length2;
 201     }
 202
 203     level1=level2=0;
 204     c1=c2=-1;
 205
 206     /* comparison loop */
 207     for(;;) {
 208         /*
 209          * here a code unit value of -1 means "get another code unit"
 210          * below it will mean "this source is finished"
 211          */
 212
 213         if(c1<0) {
 214             /* get next code unit from string 1, post-increment */
 215             for(;;) {
 216                 if(s1==limit1 || ((c1=*s1)==0 && (limit1==NULL || (options&_STRNCMP_STYLE)))) {
 217                     if(level1==0) {
 218                         c1=-1;
 219                         break;
 220                     }
 221                 } else {
 222                     ++s1;
 223                     break;
 224                 }
 225
 226                 /* reached end of level buffer, pop one level */
 227                 do {
 228                     --level1;
 229                     start1=stack1[level1].start;
 230                 } while(start1==NULL);
 231                 s1=stack1[level1].s;
 232                 limit1=stack1[level1].limit;
 233             }
 234         }
 235
 236         if(c2<0) {
 237             /* get next code unit from string 2, post-increment */
 238             for(;;) {
 239                 if(s2==limit2 || ((c2=*s2)==0 && (limit2==NULL || (options&_STRNCMP_STYLE)))) {
 240                     if(level2==0) {
 241                         c2=-1;
 242                         break;
 243                     }
 244                 } else {
 245                     ++s2;
 246                     break;
 247                 }
 248
 249                 /* reached end of level buffer, pop one level */
 250                 do {
 251                     --level2;
 252                     start2=stack2[level2].start;
 253                 } while(start2==NULL);
 254                 s2=stack2[level2].s;
 255                 limit2=stack2[level2].limit;
 256             }
 257         }
 258
 259         /*
 260          * compare c1 and c2
 261          * either variable c1, c2 is -1 only if the corresponding string is finished
 262          */
 263         if(c1==c2) {
 264             if(c1<0) {
 265                 return 0;   /* c1==c2==-1 indicating end of strings */
 266             }
 267             c1=c2=-1;       /* make us fetch new code units */
 268             continue;
 269         } else if(c1<0) {
 270             return -1;      /* string 1 ends before string 2 */
 271         } else if(c2<0) {
 272             return 1;       /* string 2 ends before string 1 */
 273         }
 274         /* c1!=c2 && c1>=0 && c2>=0 */
 275
 276         /* get complete code points for c1, c2 for lookups if either is a surrogate */
 277         cp1=c1;
 278         if(U_IS_SURROGATE(c1)) {
 279             UChar c;
 280
 281             if(U_IS_SURROGATE_LEAD(c1)) {
 282                 if(s1!=limit1 && U16_IS_TRAIL(c=*s1)) {
 283                     /* advance ++s1; only below if cp1 decomposes/case-folds */
 284                     cp1=U16_GET_SUPPLEMENTARY(c1, c);
 285                 }
 286             } else /* isTrail(c1) */ {
 287                 if(start1<=(s1-2) && U16_IS_LEAD(c=*(s1-2))) {
 288                     cp1=U16_GET_SUPPLEMENTARY(c, c1);
 289                 }
 290             }
 291         }
 292
 293         cp2=c2;
 294         if(U_IS_SURROGATE(c2)) {
 295             UChar c;
 296
 297             if(U_IS_SURROGATE_LEAD(c2)) {
 298                 if(s2!=limit2 && U16_IS_TRAIL(c=*s2)) {
 299                     /* advance ++s2; only below if cp2 decomposes/case-folds */
 300                     cp2=U16_GET_SUPPLEMENTARY(c2, c);
 301                 }
 302             } else /* isTrail(c2) */ {
 303                 if(start2<=(s2-2) && U16_IS_LEAD(c=*(s2-2))) {
 304                     cp2=U16_GET_SUPPLEMENTARY(c, c2);
 305                 }
 306             }
 307         }
 308
 309         /*
 310          * go down one level for each string
 311          * continue with the main loop as soon as there is a real change
 312          */
 313
 314         if( level1==0 && (options&U_COMPARE_IGNORE_CASE) &&
 315             (length=ucase_toFullFolding(csp, (UChar32)cp1, &p, options))>=0
 316         ) {
 317             /* cp1 case-folds to the code point "length" or to p[length] */
 318             if(U_IS_SURROGATE(c1)) {
 319                 if(U_IS_SURROGATE_LEAD(c1)) {
 320                     /* advance beyond source surrogate pair if it case-folds */
 321                     ++s1;
 322                 } else /* isTrail(c1) */ {
 323                     /*
 324                      * we got a supplementary code point when hitting its trail surrogate,
 325                      * therefore the lead surrogate must have been the same as in the other string;
 326                      * compare this decomposition with the lead surrogate in the other string
 327                      * remember that this simulates bulk text replacement:
 328                      * the decomposition would replace the entire code point
 329                      */
 330                     --s2;
 331                     c2=*(s2-1);
 332                 }
 333             }
 334
 335             /* push current level pointers */
 336             stack1[0].start=start1;
 337             stack1[0].s=s1;
 338             stack1[0].limit=limit1;
 339             ++level1;
 340
 341             /* copy the folding result to fold1[] */
 342             if(length<=UCASE_MAX_STRING_LENGTH) {
 343                 u_memcpy(fold1, p, length);
 344             } else {
 345                 int32_t i=0;
 346                 U16_APPEND_UNSAFE(fold1, i, length);
 347                 length=i;
 348             }
 349
 350             /* set next level pointers to case folding */
 351             start1=s1=fold1;
 352             limit1=fold1+length;
 353
 354             /* get ready to read from decomposition, continue with loop */
 355             c1=-1;
 356             continue;
 357         }
 358
 359         if( level2==0 && (options&U_COMPARE_IGNORE_CASE) &&
 360             (length=ucase_toFullFolding(csp, (UChar32)cp2, &p, options))>=0
 361         ) {
 362             /* cp2 case-folds to the code point "length" or to p[length] */
 363             if(U_IS_SURROGATE(c2)) {
 364                 if(U_IS_SURROGATE_LEAD(c2)) {
 365                     /* advance beyond source surrogate pair if it case-folds */
 366                     ++s2;
 367                 } else /* isTrail(c2) */ {
 368                     /*
 369                      * we got a supplementary code point when hitting its trail surrogate,
 370                      * therefore the lead surrogate must have been the same as in the other string;
 371                      * compare this decomposition with the lead surrogate in the other string
 372                      * remember that this simulates bulk text replacement:
 373                      * the decomposition would replace the entire code point
 374                      */
 375                     --s1;
 376                     c1=*(s1-1);
 377                 }
 378             }
 379
 380             /* push current level pointers */
 381             stack2[0].start=start2;
 382             stack2[0].s=s2;
 383             stack2[0].limit=limit2;
 384             ++level2;
 385
 386             /* copy the folding result to fold2[] */
 387             if(length<=UCASE_MAX_STRING_LENGTH) {
 388                 u_memcpy(fold2, p, length);
 389             } else {
 390                 int32_t i=0;
 391                 U16_APPEND_UNSAFE(fold2, i, length);
 392                 length=i;
 393             }
 394
 395             /* set next level pointers to case folding */
 396             start2=s2=fold2;
 397             limit2=fold2+length;
 398
 399             /* get ready to read from decomposition, continue with loop */
 400             c2=-1;
 401             continue;
 402         }
 403
 404         if( level1<2 && (options&_COMPARE_EQUIV) &&
 405             0!=(p=unorm_getCanonicalDecomposition((UChar32)cp1, decomp1, &length))
 406         ) {
 407             /* cp1 decomposes into p[length] */
 408             if(U_IS_SURROGATE(c1)) {
 409                 if(U_IS_SURROGATE_LEAD(c1)) {
 410                     /* advance beyond source surrogate pair if it decomposes */
 411                     ++s1;
 412                 } else /* isTrail(c1) */ {
 413                     /*
 414                      * we got a supplementary code point when hitting its trail surrogate,
 415                      * therefore the lead surrogate must have been the same as in the other string;
 416                      * compare this decomposition with the lead surrogate in the other string
 417                      * remember that this simulates bulk text replacement:
 418                      * the decomposition would replace the entire code point
 419                      */
 420                     --s2;
 421                     c2=*(s2-1);
 422                 }
 423             }
 424
 425             /* push current level pointers */
 426             stack1[level1].start=start1;
 427             stack1[level1].s=s1;
 428             stack1[level1].limit=limit1;
 429             ++level1;
 430
 431             /* set empty intermediate level if skipped */
 432             if(level1<2) {
 433                 stack1[level1++].start=NULL;
 434             }
 435
 436             /* set next level pointers to decomposition */
 437             start1=s1=p;
 438             limit1=p+length;
 439
 440             /* get ready to read from decomposition, continue with loop */
 441             c1=-1;
 442             continue;
 443         }
 444
 445         if( level2<2 && (options&_COMPARE_EQUIV) &&
 446             0!=(p=unorm_getCanonicalDecomposition((UChar32)cp2, decomp2, &length))
 447         ) {
 448             /* cp2 decomposes into p[length] */
 449             if(U_IS_SURROGATE(c2)) {
 450                 if(U_IS_SURROGATE_LEAD(c2)) {
 451                     /* advance beyond source surrogate pair if it decomposes */
 452                     ++s2;
 453                 } else /* isTrail(c2) */ {
 454                     /*
 455                      * we got a supplementary code point when hitting its trail surrogate,
 456                      * therefore the lead surrogate must have been the same as in the other string;
 457                      * compare this decomposition with the lead surrogate in the other string
 458                      * remember that this simulates bulk text replacement:
 459                      * the decomposition would replace the entire code point
 460                      */
 461                     --s1;
 462                     c1=*(s1-1);
 463                 }
 464             }
 465
 466             /* push current level pointers */
 467             stack2[level2].start=start2;
 468             stack2[level2].s=s2;
 469             stack2[level2].limit=limit2;
 470             ++level2;
 471
 472             /* set empty intermediate level if skipped */
 473             if(level2<2) {
 474                 stack2[level2++].start=NULL;
 475             }
 476
 477             /* set next level pointers to decomposition */
 478             start2=s2=p;
 479             limit2=p+length;
 480
 481             /* get ready to read from decomposition, continue with loop */
 482             c2=-1;
 483             continue;
 484         }
 485
 486         /*
 487          * no decomposition/case folding, max level for both sides:
 488          * return difference result
 489          *
 490          * code point order comparison must not just return cp1-cp2
 491          * because when single surrogates are present then the surrogate pairs
 492          * that formed cp1 and cp2 may be from different string indexes
 493          *
 494          * example: { d800 d800 dc01 } vs. { d800 dc00 }, compare at second code units
 495          * c1=d800 cp1=10001 c2=dc00 cp2=10000
 496          * cp1-cp2>0 but c1-c2<0 and in fact in UTF-32 it is { d800 10001 } < { 10000 }
 497          *
 498          * therefore, use same fix-up as in ustring.c/uprv_strCompare()
 499          * except: uprv_strCompare() fetches c=*s while this functions fetches c=*s++
 500          * so we have slightly different pointer/start/limit comparisons here
 501          */
 502
 503         if(c1>=0xd800 && c2>=0xd800 && (options&U_COMPARE_CODE_POINT_ORDER)) {
 504             /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
 505             if(
 506                 (c1<=0xdbff && s1!=limit1 && U16_IS_TRAIL(*s1)) ||
 507                 (U16_IS_TRAIL(c1) && start1!=(s1-1) && U16_IS_LEAD(*(s1-2)))
 508             ) {
 509                 /* part of a surrogate pair, leave >=d800 */
 510             } else {
 511                 /* BMP code point - may be surrogate code point - make <d800 */
 512                 c1-=0x2800;
 513             }
 514
 515             if(
 516                 (c2<=0xdbff && s2!=limit2 && U16_IS_TRAIL(*s2)) ||
 517                 (U16_IS_TRAIL(c2) && start2!=(s2-1) && U16_IS_LEAD(*(s2-2)))
 518             ) {
 519                 /* part of a surrogate pair, leave >=d800 */
 520             } else {
 521                 /* BMP code point - may be surrogate code point - make <d800 */
 522                 c2-=0x2800;
 523             }
 524         }
 525
 526         return c1-c2;
 527     }
 528 }
 529
 530 U_CAPI int32_t U_EXPORT2
 531 unorm_compare(const UChar *s1, int32_t length1,
 532               const UChar *s2, int32_t length2,
 533               uint32_t options,
 534               UErrorCode *pErrorCode) {
 535     UChar fcd1[300], fcd2[300];
 536     UChar *d1, *d2;
 537     const UnicodeSet *nx;
 538     UNormalizationMode mode;
 539     int32_t normOptions;
 540     int32_t result;
 541
 542     /* argument checking */
 543     if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
 544         return 0;
 545     }
 546     if(s1==0 || length1<-1 || s2==0 || length2<-1) {
 547         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
 548         return 0;
 549     }
 550
 551     if(!unorm_haveData(pErrorCode)) {
 552         return 0;
 553     }
 554     if(!uprv_haveProperties(pErrorCode)) {
 555         return 0;
 556     }
 557
 558     normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT);
 559     nx=unorm_getNX(normOptions, pErrorCode);
 560     if(U_FAILURE(*pErrorCode)) {
 561         return 0;
 562     }
 563
 564     d1=d2=0;
 565     options|=_COMPARE_EQUIV;
 566     result=0;
 567
 568     /*
 569      * UAX #21 Case Mappings, as fixed for Unicode version 4
 570      * (see Jitterbug 2021), defines a canonical caseless match as
 571      *
 572      * A string X is a canonical caseless match
 573      * for a string Y if and only if
 574      * NFD(toCasefold(NFD(X))) = NFD(toCasefold(NFD(Y)))
 575      *
 576      * For better performance, we check for FCD (or let the caller tell us that
 577      * both strings are in FCD) for the inner normalization.
 578      * BasicNormalizerTest::FindFoldFCDExceptions() makes sure that
 579      * case-folding preserves the FCD-ness of a string.
 580      * The outer normalization is then only performed by unorm_cmpEquivFold()
 581      * when there is a difference.
 582      *
 583      * Exception: When using the Turkic case-folding option, we do perform
 584      * full NFD first. This is because in the Turkic case precomposed characters
 585      * with 0049 capital I or 0069 small i fold differently whether they
 586      * are first decomposed or not, so an FCD check - a check only for
 587      * canonical order - is not sufficient.
 588      */
 589     if(options&U_FOLD_CASE_EXCLUDE_SPECIAL_I) {
 590         mode=UNORM_NFD;
 591         options&=~UNORM_INPUT_IS_FCD;
 592     } else {
 593         mode=UNORM_FCD;
 594     }
 595
 596     if(!(options&UNORM_INPUT_IS_FCD)) {
 597         int32_t _len1, _len2;
 598         UBool isFCD1, isFCD2;
 599
 600         // check if s1 and/or s2 fulfill the FCD conditions
 601         isFCD1= UNORM_YES==unorm_internalQuickCheck(s1, length1, mode, TRUE, nx, pErrorCode);
 602         isFCD2= UNORM_YES==unorm_internalQuickCheck(s2, length2, mode, TRUE, nx, pErrorCode);
 603         if(U_FAILURE(*pErrorCode)) {
 604             return 0;
 605         }
 606
 607         /*
 608          * ICU 2.4 had a further optimization:
 609          * If both strings were not in FCD, then they were both NFD'ed,
 610          * and the _COMPARE_EQUIV option was turned off.
 611          * It is not entirely clear that this is valid with the current
 612          * definition of the canonical caseless match.
 613          * Therefore, ICU 2.6 removes that optimization.
 614          */
 615
 616         if(!isFCD1) {
 617             _len1=unorm_internalNormalizeWithNX(fcd1, LENGTHOF(fcd1),
 618                                                 s1, length1,
 619                                                 mode, normOptions, nx,
 620                                                 pErrorCode);
 621             if(*pErrorCode!=U_BUFFER_OVERFLOW_ERROR) {
 622                 s1=fcd1;
 623             } else {
 624                 d1=(UChar *)uprv_malloc(_len1*U_SIZEOF_UCHAR);
 625                 if(d1==0) {
 626                     *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
 627                     goto cleanup;
 628                 }
 629
 630                 *pErrorCode=U_ZERO_ERROR;
 631                 _len1=unorm_internalNormalizeWithNX(d1, _len1,
 632                                                     s1, length1,
 633                                                     mode, normOptions, nx,
 634                                                     pErrorCode);
 635                 if(U_FAILURE(*pErrorCode)) {
 636                     goto cleanup;
 637                 }
 638
 639                 s1=d1;
 640             }
 641             length1=_len1;
 642         }
 643
 644         if(!isFCD2) {
 645             _len2=unorm_internalNormalizeWithNX(fcd2, LENGTHOF(fcd2),
 646                                                 s2, length2,
 647                                                 mode, normOptions, nx,
 648                                                 pErrorCode);
 649             if(*pErrorCode!=U_BUFFER_OVERFLOW_ERROR) {
 650                 s2=fcd2;
 651             } else {
 652                 d2=(UChar *)uprv_malloc(_len2*U_SIZEOF_UCHAR);
 653                 if(d2==0) {
 654                     *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
 655                     goto cleanup;
 656                 }
 657
 658                 *pErrorCode=U_ZERO_ERROR;
 659                 _len2=unorm_internalNormalizeWithNX(d2, _len2,
 660                                                     s2, length2,
 661                                                     mode, normOptions, nx,
 662                                                     pErrorCode);
 663                 if(U_FAILURE(*pErrorCode)) {
 664                     goto cleanup;
 665                 }
 666
 667                 s2=d2;
 668             }
 669             length2=_len2;
 670         }
 671     }
 672
 673     if(U_SUCCESS(*pErrorCode)) {
 674         result=unorm_cmpEquivFold(s1, length1, s2, length2, options, pErrorCode);
 675     }
 676
 677 cleanup:
 678     if(d1!=0) {
 679         uprv_free(d1);
 680     }
 681     if(d2!=0) {
 682         uprv_free(d2);
 683     }
 684
 685     return result;
 686 }
 687
 688 #endif /* #if !UCONFIG_NO_NORMALIZATION */