src/third_party/icu/source/common/ustrcase.cpp

   1 /*
   2 *******************************************************************************
   3 *
   4 *   Copyright (C) 2001-2011, International Business Machines
   5 *   Corporation and others.  All Rights Reserved.
   6 *
   7 *******************************************************************************
   8 *   file name:  ustrcase.cpp
   9 *   encoding:   US-ASCII
  10 *   tab size:   8 (not used)
  11 *   indentation:4
  12 *
  13 *   created on: 2002feb20
  14 *   created by: Markus W. Scherer
  15 *
  16 *   Implementation file for string casing C API functions.
  17 *   Uses functions from uchar.c for basic functionality that requires access
  18 *   to the Unicode Character Database (uprops.dat).
  19 */
  20
  21 #include "unicode/utypes.h"
  22 #include "unicode/brkiter.h"
  23 #include "unicode/ustring.h"
  24 #include "unicode/ucasemap.h"
  25 #include "unicode/ubrk.h"
  26 #include "unicode/utf.h"
  27 #include "unicode/utf16.h"
  28 #include "cmemory.h"
  29 #include "ucase.h"
  30 #include "ustr_imp.h"
  31
  32 #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
  33
  34 U_NAMESPACE_USE
  35
  36 /* string casing ------------------------------------------------------------ */
  37
  38 /* Appends a full case mapping result, see UCASE_MAX_STRING_LENGTH. */
  39 static inline int32_t
  40 appendResult(UChar *dest, int32_t destIndex, int32_t destCapacity,
  41              int32_t result, const UChar *s) {
  42     UChar32 c;
  43     int32_t length;
  44
  45     /* decode the result */
  46     if(result<0) {
  47         /* (not) original code point */
  48         c=~result;
  49         length=-1;
  50     } else if(result<=UCASE_MAX_STRING_LENGTH) {
  51         c=U_SENTINEL;
  52         length=result;
  53     } else {
  54         c=result;
  55         length=-1;
  56     }
  57
  58     if(destIndex<destCapacity) {
  59         /* append the result */
  60         if(length<0) {
  61             /* code point */
  62             UBool isError=FALSE;
  63             U16_APPEND(dest, destIndex, destCapacity, c, isError);
  64             if(isError) {
  65                 /* overflow, nothing written */
  66                 destIndex+=U16_LENGTH(c);
  67             }
  68         } else {
  69             /* string */
  70             if((destIndex+length)<=destCapacity) {
  71                 while(length>0) {
  72                     dest[destIndex++]=*s++;
  73                     --length;
  74                 }
  75             } else {
  76                 /* overflow */
  77                 destIndex+=length;
  78             }
  79         }
  80     } else {
  81         /* preflight */
  82         if(length<0) {
  83             destIndex+=U16_LENGTH(c);
  84         } else {
  85             destIndex+=length;
  86         }
  87     }
  88     return destIndex;
  89 }
  90
  91 static UChar32 U_CALLCONV
  92 utf16_caseContextIterator(void *context, int8_t dir) {
  93     UCaseContext *csc=(UCaseContext *)context;
  94     UChar32 c;
  95
  96     if(dir<0) {
  97         /* reset for backward iteration */
  98         csc->index=csc->cpStart;
  99         csc->dir=dir;
 100     } else if(dir>0) {
 101         /* reset for forward iteration */
 102         csc->index=csc->cpLimit;
 103         csc->dir=dir;
 104     } else {
 105         /* continue current iteration direction */
 106         dir=csc->dir;
 107     }
 108
 109     if(dir<0) {
 110         if(csc->start<csc->index) {
 111             U16_PREV((const UChar *)csc->p, csc->start, csc->index, c);
 112             return c;
 113         }
 114     } else {
 115         if(csc->index<csc->limit) {
 116             U16_NEXT((const UChar *)csc->p, csc->index, csc->limit, c);
 117             return c;
 118         }
 119     }
 120     return U_SENTINEL;
 121 }
 122
 123 /*
 124  * Case-maps [srcStart..srcLimit[ but takes
 125  * context [0..srcLength[ into account.
 126  */
 127 static int32_t
 128 _caseMap(const UCaseMap *csm, UCaseMapFull *map,
 129          UChar *dest, int32_t destCapacity,
 130          const UChar *src, UCaseContext *csc,
 131          int32_t srcStart, int32_t srcLimit,
 132          UErrorCode *pErrorCode) {
 133     const UChar *s;
 134     UChar32 c, c2 = 0;
 135     int32_t srcIndex, destIndex;
 136     int32_t locCache;
 137
 138     locCache=csm->locCache;
 139
 140     /* case mapping loop */
 141     srcIndex=srcStart;
 142     destIndex=0;
 143     while(srcIndex<srcLimit) {
 144         csc->cpStart=srcIndex;
 145         U16_NEXT(src, srcIndex, srcLimit, c);
 146         csc->cpLimit=srcIndex;
 147         c=map(csm->csp, c, utf16_caseContextIterator, csc, &s, csm->locale, &locCache);
 148         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
 149             /* fast path version of appendResult() for BMP results */
 150             dest[destIndex++]=(UChar)c2;
 151         } else {
 152             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 153         }
 154     }
 155
 156     if(destIndex>destCapacity) {
 157         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 158     }
 159     return destIndex;
 160 }
 161
 162 #if !UCONFIG_NO_BREAK_ITERATION
 163
 164 U_CFUNC int32_t U_CALLCONV
 165 ustrcase_internalToTitle(const UCaseMap *csm,
 166                          UChar *dest, int32_t destCapacity,
 167                          const UChar *src, int32_t srcLength,
 168                          UErrorCode *pErrorCode) {
 169     const UChar *s;
 170     UChar32 c;
 171     int32_t prev, titleStart, titleLimit, idx, destIndex, length;
 172     UBool isFirstIndex;
 173
 174     if(U_FAILURE(*pErrorCode)) {
 175         return 0;
 176     }
 177
 178     // Use the C++ abstract base class to minimize dependencies.
 179     // TODO: Change UCaseMap.iter to store a BreakIterator directly.
 180     BreakIterator *bi=reinterpret_cast<BreakIterator *>(csm->iter);
 181
 182     /* set up local variables */
 183     int32_t locCache=csm->locCache;
 184     UCaseContext csc=UCASECONTEXT_INITIALIZER;
 185     csc.p=(void *)src;
 186     csc.limit=srcLength;
 187     destIndex=0;
 188     prev=0;
 189     isFirstIndex=TRUE;
 190
 191     /* titlecasing loop */
 192     while(prev<srcLength) {
 193         /* find next index where to titlecase */
 194         if(isFirstIndex) {
 195             isFirstIndex=FALSE;
 196             idx=bi->first();
 197         } else {
 198             idx=bi->next();
 199         }
 200         if(idx==UBRK_DONE || idx>srcLength) {
 201             idx=srcLength;
 202         }
 203
 204         /*
 205          * Unicode 4 & 5 section 3.13 Default Case Operations:
 206          *
 207          * R3  toTitlecase(X): Find the word boundaries based on Unicode Standard Annex
 208          * #29, "Text Boundaries." Between each pair of word boundaries, find the first
 209          * cased character F. If F exists, map F to default_title(F); then map each
 210          * subsequent character C to default_lower(C).
 211          *
 212          * In this implementation, segment [prev..index[ into 3 parts:
 213          * a) uncased characters (copy as-is) [prev..titleStart[
 214          * b) first case letter (titlecase)         [titleStart..titleLimit[
 215          * c) subsequent characters (lowercase)                 [titleLimit..index[
 216          */
 217         if(prev<idx) {
 218             /* find and copy uncased characters [prev..titleStart[ */
 219             titleStart=titleLimit=prev;
 220             U16_NEXT(src, titleLimit, idx, c);
 221             if((csm->options&U_TITLECASE_NO_BREAK_ADJUSTMENT)==0 && UCASE_NONE==ucase_getType(csm->csp, c)) {
 222                 /* Adjust the titlecasing index (titleStart) to the next cased character. */
 223                 for(;;) {
 224                     titleStart=titleLimit;
 225                     if(titleLimit==idx) {
 226                         /*
 227                          * only uncased characters in [prev..index[
 228                          * stop with titleStart==titleLimit==index
 229                          */
 230                         break;
 231                     }
 232                     U16_NEXT(src, titleLimit, idx, c);
 233                     if(UCASE_NONE!=ucase_getType(csm->csp, c)) {
 234                         break; /* cased letter at [titleStart..titleLimit[ */
 235                     }
 236                 }
 237                 length=titleStart-prev;
 238                 if(length>0) {
 239                     if((destIndex+length)<=destCapacity) {
 240                         uprv_memcpy(dest+destIndex, src+prev, length*U_SIZEOF_UCHAR);
 241                     }
 242                     destIndex+=length;
 243                 }
 244             }
 245
 246             if(titleStart<titleLimit) {
 247                 /* titlecase c which is from [titleStart..titleLimit[ */
 248                 csc.cpStart=titleStart;
 249                 csc.cpLimit=titleLimit;
 250                 c=ucase_toFullTitle(csm->csp, c, utf16_caseContextIterator, &csc, &s, csm->locale, &locCache);
 251                 destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 252
 253                 /* Special case Dutch IJ titlecasing */
 254                 if ( titleStart+1 < idx &&
 255                      ucase_getCaseLocale(csm->locale,&locCache) == UCASE_LOC_DUTCH &&
 256                      ( src[titleStart] == (UChar32) 0x0049 || src[titleStart] == (UChar32) 0x0069 ) &&
 257                      ( src[titleStart+1] == (UChar32) 0x004A || src[titleStart+1] == (UChar32) 0x006A )) {
 258                             c=(UChar32) 0x004A;
 259                             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 260                             titleLimit++;
 261                 }
 262
 263                 /* lowercase [titleLimit..index[ */
 264                 if(titleLimit<idx) {
 265                     if((csm->options&U_TITLECASE_NO_LOWERCASE)==0) {
 266                         /* Normal operation: Lowercase the rest of the word. */
 267                         destIndex+=
 268                             _caseMap(
 269                                 csm, ucase_toFullLower,
 270                                 dest+destIndex, destCapacity-destIndex,
 271                                 src, &csc,
 272                                 titleLimit, idx,
 273                                 pErrorCode);
 274                     } else {
 275                         /* Optionally just copy the rest of the word unchanged. */
 276                         length=idx-titleLimit;
 277                         if((destIndex+length)<=destCapacity) {
 278                             uprv_memcpy(dest+destIndex, src+titleLimit, length*U_SIZEOF_UCHAR);
 279                         }
 280                         destIndex+=length;
 281                     }
 282                 }
 283             }
 284         }
 285
 286         prev=idx;
 287     }
 288
 289     if(destIndex>destCapacity) {
 290         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 291     }
 292     return destIndex;
 293 }
 294
 295 #endif  // !UCONFIG_NO_BREAK_ITERATION
 296
 297 /* functions available in the common library (for unistr_case.cpp) */
 298
 299 U_CFUNC int32_t U_CALLCONV
 300 ustrcase_internalToLower(const UCaseMap *csm,
 301                          UChar *dest, int32_t destCapacity,
 302                          const UChar *src, int32_t srcLength,
 303                          UErrorCode *pErrorCode) {
 304     UCaseContext csc=UCASECONTEXT_INITIALIZER;
 305     csc.p=(void *)src;
 306     csc.limit=srcLength;
 307     return _caseMap(
 308         csm, ucase_toFullLower,
 309         dest, destCapacity,
 310         src, &csc, 0, srcLength,
 311         pErrorCode);
 312 }
 313
 314 U_CFUNC int32_t U_CALLCONV
 315 ustrcase_internalToUpper(const UCaseMap *csm,
 316                          UChar *dest, int32_t destCapacity,
 317                          const UChar *src, int32_t srcLength,
 318                          UErrorCode *pErrorCode) {
 319     UCaseContext csc=UCASECONTEXT_INITIALIZER;
 320     csc.p=(void *)src;
 321     csc.limit=srcLength;
 322     return _caseMap(
 323         csm, ucase_toFullUpper,
 324         dest, destCapacity,
 325         src, &csc, 0, srcLength,
 326         pErrorCode);
 327 }
 328
 329 static int32_t
 330 ustr_foldCase(const UCaseProps *csp,
 331               UChar *dest, int32_t destCapacity,
 332               const UChar *src, int32_t srcLength,
 333               uint32_t options,
 334               UErrorCode *pErrorCode) {
 335     int32_t srcIndex, destIndex;
 336
 337     const UChar *s;
 338     UChar32 c, c2 = 0;
 339
 340     /* case mapping loop */
 341     srcIndex=destIndex=0;
 342     while(srcIndex<srcLength) {
 343         U16_NEXT(src, srcIndex, srcLength, c);
 344         c=ucase_toFullFolding(csp, c, &s, options);
 345         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
 346             /* fast path version of appendResult() for BMP results */
 347             dest[destIndex++]=(UChar)c2;
 348         } else {
 349             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 350         }
 351     }
 352
 353     if(destIndex>destCapacity) {
 354         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 355     }
 356     return destIndex;
 357 }
 358
 359 U_CFUNC int32_t U_CALLCONV
 360 ustrcase_internalFold(const UCaseMap *csm,
 361                       UChar *dest, int32_t destCapacity,
 362                       const UChar *src, int32_t srcLength,
 363                       UErrorCode *pErrorCode) {
 364     return ustr_foldCase(csm->csp, dest, destCapacity, src, srcLength, csm->options, pErrorCode);
 365 }
 366
 367 U_CFUNC int32_t
 368 ustrcase_map(const UCaseMap *csm,
 369              UChar *dest, int32_t destCapacity,
 370              const UChar *src, int32_t srcLength,
 371              UStringCaseMapper *stringCaseMapper,
 372              UErrorCode *pErrorCode) {
 373     UChar buffer[300];
 374     UChar *temp;
 375
 376     int32_t destLength;
 377
 378     /* check argument values */
 379     if(U_FAILURE(*pErrorCode)) {
 380         return 0;
 381     }
 382     if( destCapacity<0 ||
 383         (dest==NULL && destCapacity>0) ||
 384         src==NULL ||
 385         srcLength<-1
 386     ) {
 387         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
 388         return 0;
 389     }
 390
 391     /* get the string length */
 392     if(srcLength==-1) {
 393         srcLength=u_strlen(src);
 394     }
 395
 396     /* check for overlapping source and destination */
 397     if( dest!=NULL &&
 398         ((src>=dest && src<(dest+destCapacity)) ||
 399          (dest>=src && dest<(src+srcLength)))
 400     ) {
 401         /* overlap: provide a temporary destination buffer and later copy the result */
 402         if(destCapacity<=LENGTHOF(buffer)) {
 403             /* the stack buffer is large enough */
 404             temp=buffer;
 405         } else {
 406             /* allocate a buffer */
 407             temp=(UChar *)uprv_malloc(destCapacity*U_SIZEOF_UCHAR);
 408             if(temp==NULL) {
 409                 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
 410                 return 0;
 411             }
 412         }
 413     } else {
 414         temp=dest;
 415     }
 416
 417     destLength=stringCaseMapper(csm, temp, destCapacity, src, srcLength, pErrorCode);
 418     if(temp!=dest) {
 419         /* copy the result string to the destination buffer */
 420         if(destLength>0) {
 421             int32_t copyLength= destLength<=destCapacity ? destLength : destCapacity;
 422             if(copyLength>0) {
 423                 uprv_memmove(dest, temp, copyLength*U_SIZEOF_UCHAR);
 424             }
 425         }
 426         if(temp!=buffer) {
 427             uprv_free(temp);
 428         }
 429     }
 430
 431     return u_terminateUChars(dest, destCapacity, destLength, pErrorCode);
 432 }
 433
 434 /* public API functions */
 435
 436 U_CAPI int32_t U_EXPORT2
 437 u_strFoldCase(UChar *dest, int32_t destCapacity,
 438               const UChar *src, int32_t srcLength,
 439               uint32_t options,
 440               UErrorCode *pErrorCode) {
 441     UCaseMap csm=UCASEMAP_INITIALIZER;
 442     csm.csp=ucase_getSingleton();
 443     csm.options=options;
 444     return ustrcase_map(
 445         &csm,
 446         dest, destCapacity,
 447         src, srcLength,
 448         ustrcase_internalFold, pErrorCode);
 449 }
 450
 451 /* case-insensitive string comparisons -------------------------------------- */
 452
 453 /*
 454  * This function is a copy of unorm_cmpEquivFold() minus the parts for
 455  * canonical equivalence.
 456  * Keep the functions in sync, and see there for how this works.
 457  * The duplication is for modularization:
 458  * It makes caseless (but not canonical caseless) matches independent of
 459  * the normalization code.
 460  */
 461
 462 /* stack element for previous-level source/decomposition pointers */
 463 struct CmpEquivLevel {
 464     const UChar *start, *s, *limit;
 465 };
 466 typedef struct CmpEquivLevel CmpEquivLevel;
 467
 468 /* internal function */
 469 U_CFUNC int32_t
 470 u_strcmpFold(const UChar *s1, int32_t length1,
 471              const UChar *s2, int32_t length2,
 472              uint32_t options,
 473              UErrorCode *pErrorCode) {
 474     const UCaseProps *csp;
 475
 476     /* current-level start/limit - s1/s2 as current */
 477     const UChar *start1, *start2, *limit1, *limit2;
 478
 479     /* case folding variables */
 480     const UChar *p;
 481     int32_t length;
 482
 483     /* stacks of previous-level start/current/limit */
 484     CmpEquivLevel stack1[2], stack2[2];
 485
 486     /* case folding buffers, only use current-level start/limit */
 487     UChar fold1[UCASE_MAX_STRING_LENGTH+1], fold2[UCASE_MAX_STRING_LENGTH+1];
 488
 489     /* track which is the current level per string */
 490     int32_t level1, level2;
 491
 492     /* current code units, and code points for lookups */
 493     UChar32 c1, c2, cp1, cp2;
 494
 495     /* no argument error checking because this itself is not an API */
 496
 497     /*
 498      * assume that at least the option U_COMPARE_IGNORE_CASE is set
 499      * otherwise this function would have to behave exactly as uprv_strCompare()
 500      */
 501     csp=ucase_getSingleton();
 502     if(U_FAILURE(*pErrorCode)) {
 503         return 0;
 504     }
 505
 506     /* initialize */
 507     start1=s1;
 508     if(length1==-1) {
 509         limit1=NULL;
 510     } else {
 511         limit1=s1+length1;
 512     }
 513
 514     start2=s2;
 515     if(length2==-1) {
 516         limit2=NULL;
 517     } else {
 518         limit2=s2+length2;
 519     }
 520
 521     level1=level2=0;
 522     c1=c2=-1;
 523
 524     /* comparison loop */
 525     for(;;) {
 526         /*
 527          * here a code unit value of -1 means "get another code unit"
 528          * below it will mean "this source is finished"
 529          */
 530
 531         if(c1<0) {
 532             /* get next code unit from string 1, post-increment */
 533             for(;;) {
 534                 if(s1==limit1 || ((c1=*s1)==0 && (limit1==NULL || (options&_STRNCMP_STYLE)))) {
 535                     if(level1==0) {
 536                         c1=-1;
 537                         break;
 538                     }
 539                 } else {
 540                     ++s1;
 541                     break;
 542                 }
 543
 544                 /* reached end of level buffer, pop one level */
 545                 do {
 546                     --level1;
 547                     start1=stack1[level1].start;    /*Not uninitialized*/
 548                 } while(start1==NULL);
 549                 s1=stack1[level1].s;                /*Not uninitialized*/
 550                 limit1=stack1[level1].limit;        /*Not uninitialized*/
 551             }
 552         }
 553
 554         if(c2<0) {
 555             /* get next code unit from string 2, post-increment */
 556             for(;;) {
 557                 if(s2==limit2 || ((c2=*s2)==0 && (limit2==NULL || (options&_STRNCMP_STYLE)))) {
 558                     if(level2==0) {
 559                         c2=-1;
 560                         break;
 561                     }
 562                 } else {
 563                     ++s2;
 564                     break;
 565                 }
 566
 567                 /* reached end of level buffer, pop one level */
 568                 do {
 569                     --level2;
 570                     start2=stack2[level2].start;    /*Not uninitialized*/
 571                 } while(start2==NULL);
 572                 s2=stack2[level2].s;                /*Not uninitialized*/
 573                 limit2=stack2[level2].limit;        /*Not uninitialized*/
 574             }
 575         }
 576
 577         /*
 578          * compare c1 and c2
 579          * either variable c1, c2 is -1 only if the corresponding string is finished
 580          */
 581         if(c1==c2) {
 582             if(c1<0) {
 583                 return 0;   /* c1==c2==-1 indicating end of strings */
 584             }
 585             c1=c2=-1;       /* make us fetch new code units */
 586             continue;
 587         } else if(c1<0) {
 588             return -1;      /* string 1 ends before string 2 */
 589         } else if(c2<0) {
 590             return 1;       /* string 2 ends before string 1 */
 591         }
 592         /* c1!=c2 && c1>=0 && c2>=0 */
 593
 594         /* get complete code points for c1, c2 for lookups if either is a surrogate */
 595         cp1=c1;
 596         if(U_IS_SURROGATE(c1)) {
 597             UChar c;
 598
 599             if(U_IS_SURROGATE_LEAD(c1)) {
 600                 if(s1!=limit1 && U16_IS_TRAIL(c=*s1)) {
 601                     /* advance ++s1; only below if cp1 decomposes/case-folds */
 602                     cp1=U16_GET_SUPPLEMENTARY(c1, c);
 603                 }
 604             } else /* isTrail(c1) */ {
 605                 if(start1<=(s1-2) && U16_IS_LEAD(c=*(s1-2))) {
 606                     cp1=U16_GET_SUPPLEMENTARY(c, c1);
 607                 }
 608             }
 609         }
 610
 611         cp2=c2;
 612         if(U_IS_SURROGATE(c2)) {
 613             UChar c;
 614
 615             if(U_IS_SURROGATE_LEAD(c2)) {
 616                 if(s2!=limit2 && U16_IS_TRAIL(c=*s2)) {
 617                     /* advance ++s2; only below if cp2 decomposes/case-folds */
 618                     cp2=U16_GET_SUPPLEMENTARY(c2, c);
 619                 }
 620             } else /* isTrail(c2) */ {
 621                 if(start2<=(s2-2) && U16_IS_LEAD(c=*(s2-2))) {
 622                     cp2=U16_GET_SUPPLEMENTARY(c, c2);
 623                 }
 624             }
 625         }
 626
 627         /*
 628          * go down one level for each string
 629          * continue with the main loop as soon as there is a real change
 630          */
 631
 632         if( level1==0 &&
 633             (length=ucase_toFullFolding(csp, (UChar32)cp1, &p, options))>=0
 634         ) {
 635             /* cp1 case-folds to the code point "length" or to p[length] */
 636             if(U_IS_SURROGATE(c1)) {
 637                 if(U_IS_SURROGATE_LEAD(c1)) {
 638                     /* advance beyond source surrogate pair if it case-folds */
 639                     ++s1;
 640                 } else /* isTrail(c1) */ {
 641                     /*
 642                      * we got a supplementary code point when hitting its trail surrogate,
 643                      * therefore the lead surrogate must have been the same as in the other string;
 644                      * compare this decomposition with the lead surrogate in the other string
 645                      * remember that this simulates bulk text replacement:
 646                      * the decomposition would replace the entire code point
 647                      */
 648                     --s2;
 649                     c2=*(s2-1);
 650                 }
 651             }
 652
 653             /* push current level pointers */
 654             stack1[0].start=start1;
 655             stack1[0].s=s1;
 656             stack1[0].limit=limit1;
 657             ++level1;
 658
 659             /* copy the folding result to fold1[] */
 660             if(length<=UCASE_MAX_STRING_LENGTH) {
 661                 u_memcpy(fold1, p, length);
 662             } else {
 663                 int32_t i=0;
 664                 U16_APPEND_UNSAFE(fold1, i, length);
 665                 length=i;
 666             }
 667
 668             /* set next level pointers to case folding */
 669             start1=s1=fold1;
 670             limit1=fold1+length;
 671
 672             /* get ready to read from decomposition, continue with loop */
 673             c1=-1;
 674             continue;
 675         }
 676
 677         if( level2==0 &&
 678             (length=ucase_toFullFolding(csp, (UChar32)cp2, &p, options))>=0
 679         ) {
 680             /* cp2 case-folds to the code point "length" or to p[length] */
 681             if(U_IS_SURROGATE(c2)) {
 682                 if(U_IS_SURROGATE_LEAD(c2)) {
 683                     /* advance beyond source surrogate pair if it case-folds */
 684                     ++s2;
 685                 } else /* isTrail(c2) */ {
 686                     /*
 687                      * we got a supplementary code point when hitting its trail surrogate,
 688                      * therefore the lead surrogate must have been the same as in the other string;
 689                      * compare this decomposition with the lead surrogate in the other string
 690                      * remember that this simulates bulk text replacement:
 691                      * the decomposition would replace the entire code point
 692                      */
 693                     --s1;
 694                     c1=*(s1-1);
 695                 }
 696             }
 697
 698             /* push current level pointers */
 699             stack2[0].start=start2;
 700             stack2[0].s=s2;
 701             stack2[0].limit=limit2;
 702             ++level2;
 703
 704             /* copy the folding result to fold2[] */
 705             if(length<=UCASE_MAX_STRING_LENGTH) {
 706                 u_memcpy(fold2, p, length);
 707             } else {
 708                 int32_t i=0;
 709                 U16_APPEND_UNSAFE(fold2, i, length);
 710                 length=i;
 711             }
 712
 713             /* set next level pointers to case folding */
 714             start2=s2=fold2;
 715             limit2=fold2+length;
 716
 717             /* get ready to read from decomposition, continue with loop */
 718             c2=-1;
 719             continue;
 720         }
 721
 722         /*
 723          * no decomposition/case folding, max level for both sides:
 724          * return difference result
 725          *
 726          * code point order comparison must not just return cp1-cp2
 727          * because when single surrogates are present then the surrogate pairs
 728          * that formed cp1 and cp2 may be from different string indexes
 729          *
 730          * example: { d800 d800 dc01 } vs. { d800 dc00 }, compare at second code units
 731          * c1=d800 cp1=10001 c2=dc00 cp2=10000
 732          * cp1-cp2>0 but c1-c2<0 and in fact in UTF-32 it is { d800 10001 } < { 10000 }
 733          *
 734          * therefore, use same fix-up as in ustring.c/uprv_strCompare()
 735          * except: uprv_strCompare() fetches c=*s while this functions fetches c=*s++
 736          * so we have slightly different pointer/start/limit comparisons here
 737          */
 738
 739         if(c1>=0xd800 && c2>=0xd800 && (options&U_COMPARE_CODE_POINT_ORDER)) {
 740             /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
 741             if(
 742                 (c1<=0xdbff && s1!=limit1 && U16_IS_TRAIL(*s1)) ||
 743                 (U16_IS_TRAIL(c1) && start1!=(s1-1) && U16_IS_LEAD(*(s1-2)))
 744             ) {
 745                 /* part of a surrogate pair, leave >=d800 */
 746             } else {
 747                 /* BMP code point - may be surrogate code point - make <d800 */
 748                 c1-=0x2800;
 749             }
 750
 751             if(
 752                 (c2<=0xdbff && s2!=limit2 && U16_IS_TRAIL(*s2)) ||
 753                 (U16_IS_TRAIL(c2) && start2!=(s2-1) && U16_IS_LEAD(*(s2-2)))
 754             ) {
 755                 /* part of a surrogate pair, leave >=d800 */
 756             } else {
 757                 /* BMP code point - may be surrogate code point - make <d800 */
 758                 c2-=0x2800;
 759             }
 760         }
 761
 762         return c1-c2;
 763     }
 764 }
 765
 766 /* public API functions */
 767
 768 U_CAPI int32_t U_EXPORT2
 769 u_strCaseCompare(const UChar *s1, int32_t length1,
 770                  const UChar *s2, int32_t length2,
 771                  uint32_t options,
 772                  UErrorCode *pErrorCode) {
 773     /* argument checking */
 774     if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
 775         return 0;
 776     }
 777     if(s1==NULL || length1<-1 || s2==NULL || length2<-1) {
 778         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
 779         return 0;
 780     }
 781     return u_strcmpFold(s1, length1, s2, length2,
 782                         options|U_COMPARE_IGNORE_CASE,
 783                         pErrorCode);
 784 }
 785
 786 U_CAPI int32_t U_EXPORT2
 787 u_strcasecmp(const UChar *s1, const UChar *s2, uint32_t options) {
 788     UErrorCode errorCode=U_ZERO_ERROR;
 789     return u_strcmpFold(s1, -1, s2, -1,
 790                         options|U_COMPARE_IGNORE_CASE,
 791                         &errorCode);
 792 }
 793
 794 U_CAPI int32_t U_EXPORT2
 795 u_memcasecmp(const UChar *s1, const UChar *s2, int32_t length, uint32_t options) {
 796     UErrorCode errorCode=U_ZERO_ERROR;
 797     return u_strcmpFold(s1, length, s2, length,
 798                         options|U_COMPARE_IGNORE_CASE,
 799                         &errorCode);
 800 }
 801
 802 U_CAPI int32_t U_EXPORT2
 803 u_strncasecmp(const UChar *s1, const UChar *s2, int32_t n, uint32_t options) {
 804     UErrorCode errorCode=U_ZERO_ERROR;
 805     return u_strcmpFold(s1, n, s2, n,
 806                         options|(U_COMPARE_IGNORE_CASE|_STRNCMP_STYLE),
 807                         &errorCode);
 808 }