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

   1 /*
   2 *******************************************************************************
   3 *
   4 *   Copyright (C) 2001-2010, International Business Machines
   5 *   Corporation and others.  All Rights Reserved.
   6 *
   7 *******************************************************************************
   8 *   file name:  ustrcase.c
   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/uloc.h"
  23 #include "unicode/ustring.h"
  24 #include "unicode/ucasemap.h"
  25 #include "unicode/ubrk.h"
  26 #include "cmemory.h"
  27 #include "ucase.h"
  28 #include "ustr_imp.h"
  29
  30 /* string casing ------------------------------------------------------------ */
  31
  32 /* append a full case mapping result, see UCASE_MAX_STRING_LENGTH */
  33 static U_INLINE int32_t
  34 appendResult(UChar *dest, int32_t destIndex, int32_t destCapacity,
  35              int32_t result, const UChar *s) {
  36     UChar32 c;
  37     int32_t length;
  38
  39     /* decode the result */
  40     if(result<0) {
  41         /* (not) original code point */
  42         c=~result;
  43         length=-1;
  44     } else if(result<=UCASE_MAX_STRING_LENGTH) {
  45         c=U_SENTINEL;
  46         length=result;
  47     } else {
  48         c=result;
  49         length=-1;
  50     }
  51
  52     if(destIndex<destCapacity) {
  53         /* append the result */
  54         if(length<0) {
  55             /* code point */
  56             UBool isError=FALSE;
  57             U16_APPEND(dest, destIndex, destCapacity, c, isError);
  58             if(isError) {
  59                 /* overflow, nothing written */
  60                 destIndex+=U16_LENGTH(c);
  61             }
  62         } else {
  63             /* string */
  64             if((destIndex+length)<=destCapacity) {
  65                 while(length>0) {
  66                     dest[destIndex++]=*s++;
  67                     --length;
  68                 }
  69             } else {
  70                 /* overflow */
  71                 destIndex+=length;
  72             }
  73         }
  74     } else {
  75         /* preflight */
  76         if(length<0) {
  77             destIndex+=U16_LENGTH(c);
  78         } else {
  79             destIndex+=length;
  80         }
  81     }
  82     return destIndex;
  83 }
  84
  85 static UChar32 U_CALLCONV
  86 utf16_caseContextIterator(void *context, int8_t dir) {
  87     UCaseContext *csc=(UCaseContext *)context;
  88     UChar32 c;
  89
  90     if(dir<0) {
  91         /* reset for backward iteration */
  92         csc->index=csc->cpStart;
  93         csc->dir=dir;
  94     } else if(dir>0) {
  95         /* reset for forward iteration */
  96         csc->index=csc->cpLimit;
  97         csc->dir=dir;
  98     } else {
  99         /* continue current iteration direction */
 100         dir=csc->dir;
 101     }
 102
 103     if(dir<0) {
 104         if(csc->start<csc->index) {
 105             U16_PREV((const UChar *)csc->p, csc->start, csc->index, c);
 106             return c;
 107         }
 108     } else {
 109         if(csc->index<csc->limit) {
 110             U16_NEXT((const UChar *)csc->p, csc->index, csc->limit, c);
 111             return c;
 112         }
 113     }
 114     return U_SENTINEL;
 115 }
 116
 117 /*
 118  * Case-maps [srcStart..srcLimit[ but takes
 119  * context [0..srcLength[ into account.
 120  */
 121 static int32_t
 122 _caseMap(const UCaseMap *csm, UCaseMapFull *map,
 123          UChar *dest, int32_t destCapacity,
 124          const UChar *src, UCaseContext *csc,
 125          int32_t srcStart, int32_t srcLimit,
 126          UErrorCode *pErrorCode) {
 127     const UChar *s;
 128     UChar32 c, c2 = 0;
 129     int32_t srcIndex, destIndex;
 130     int32_t locCache;
 131
 132     locCache=csm->locCache;
 133
 134     /* case mapping loop */
 135     srcIndex=srcStart;
 136     destIndex=0;
 137     while(srcIndex<srcLimit) {
 138         csc->cpStart=srcIndex;
 139         U16_NEXT(src, srcIndex, srcLimit, c);
 140         csc->cpLimit=srcIndex;
 141         c=map(csm->csp, c, utf16_caseContextIterator, csc, &s, csm->locale, &locCache);
 142         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
 143             /* fast path version of appendResult() for BMP results */
 144             dest[destIndex++]=(UChar)c2;
 145         } else {
 146             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 147         }
 148     }
 149
 150     if(destIndex>destCapacity) {
 151         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 152     }
 153     return destIndex;
 154 }
 155
 156 static void
 157 setTempCaseMapLocale(UCaseMap *csm, const char *locale, UErrorCode *pErrorCode) {
 158     /*
 159      * We could call ucasemap_setLocale(), but here we really only care about
 160      * the initial language subtag, we need not return the real string via
 161      * ucasemap_getLocale(), and we don't care about only getting "x" from
 162      * "x-some-thing" etc.
 163      *
 164      * We ignore locales with a longer-than-3 initial subtag.
 165      *
 166      * We also do not fill in the locCache because it is rarely used,
 167      * and not worth setting unless we reuse it for many case mapping operations.
 168      * (That's why UCaseMap was created.)
 169      */
 170     int i;
 171     char c;
 172
 173     /* the internal functions require locale!=NULL */
 174     if(locale==NULL) {
 175         locale=uloc_getDefault();
 176     }
 177     for(i=0; i<4 && (c=locale[i])!=0 && c!='-' && c!='_'; ++i) {
 178         csm->locale[i]=c;
 179     }
 180     if(i<=3) {
 181         csm->locale[i]=0;  /* Up to 3 non-separator characters. */
 182     } else {
 183         csm->locale[0]=0;  /* Longer-than-3 initial subtag: Ignore. */
 184     }
 185 }
 186
 187 /*
 188  * Set parameters on an empty UCaseMap, for UCaseMap-less API functions.
 189  * Do this fast because it is called with every function call.
 190  */
 191 static U_INLINE void
 192 setTempCaseMap(UCaseMap *csm, const char *locale, UErrorCode *pErrorCode) {
 193     if(csm->csp==NULL) {
 194         csm->csp=ucase_getSingleton();
 195     }
 196     if(locale!=NULL && locale[0]==0) {
 197         csm->locale[0]=0;
 198     } else {
 199         setTempCaseMapLocale(csm, locale, pErrorCode);
 200     }
 201 }
 202
 203 #if !UCONFIG_NO_BREAK_ITERATION
 204
 205 /*
 206  * Internal titlecasing function.
 207  */
 208 static int32_t
 209 _toTitle(UCaseMap *csm,
 210          UChar *dest, int32_t destCapacity,
 211          const UChar *src, UCaseContext *csc,
 212          int32_t srcLength,
 213          UErrorCode *pErrorCode) {
 214     const UChar *s;
 215     UChar32 c;
 216     int32_t prev, titleStart, titleLimit, idx, destIndex, length;
 217     UBool isFirstIndex;
 218
 219     if(csm->iter!=NULL) {
 220         ubrk_setText(csm->iter, src, srcLength, pErrorCode);
 221     } else {
 222         csm->iter=ubrk_open(UBRK_WORD, csm->locale,
 223                             src, srcLength,
 224                             pErrorCode);
 225     }
 226     if(U_FAILURE(*pErrorCode)) {
 227         return 0;
 228     }
 229
 230     /* set up local variables */
 231     destIndex=0;
 232     prev=0;
 233     isFirstIndex=TRUE;
 234
 235     /* titlecasing loop */
 236     while(prev<srcLength) {
 237         /* find next index where to titlecase */
 238         if(isFirstIndex) {
 239             isFirstIndex=FALSE;
 240             idx=ubrk_first(csm->iter);
 241         } else {
 242             idx=ubrk_next(csm->iter);
 243         }
 244         if(idx==UBRK_DONE || idx>srcLength) {
 245             idx=srcLength;
 246         }
 247
 248         /*
 249          * Unicode 4 & 5 section 3.13 Default Case Operations:
 250          *
 251          * R3  toTitlecase(X): Find the word boundaries based on Unicode Standard Annex
 252          * #29, "Text Boundaries." Between each pair of word boundaries, find the first
 253          * cased character F. If F exists, map F to default_title(F); then map each
 254          * subsequent character C to default_lower(C).
 255          *
 256          * In this implementation, segment [prev..index[ into 3 parts:
 257          * a) uncased characters (copy as-is) [prev..titleStart[
 258          * b) first case letter (titlecase)         [titleStart..titleLimit[
 259          * c) subsequent characters (lowercase)                 [titleLimit..index[
 260          */
 261         if(prev<idx) {
 262             /* find and copy uncased characters [prev..titleStart[ */
 263             titleStart=titleLimit=prev;
 264             U16_NEXT(src, titleLimit, idx, c);
 265             if((csm->options&U_TITLECASE_NO_BREAK_ADJUSTMENT)==0 && UCASE_NONE==ucase_getType(csm->csp, c)) {
 266                 /* Adjust the titlecasing index (titleStart) to the next cased character. */
 267                 for(;;) {
 268                     titleStart=titleLimit;
 269                     if(titleLimit==idx) {
 270                         /*
 271                          * only uncased characters in [prev..index[
 272                          * stop with titleStart==titleLimit==index
 273                          */
 274                         break;
 275                     }
 276                     U16_NEXT(src, titleLimit, idx, c);
 277                     if(UCASE_NONE!=ucase_getType(csm->csp, c)) {
 278                         break; /* cased letter at [titleStart..titleLimit[ */
 279                     }
 280                 }
 281                 length=titleStart-prev;
 282                 if(length>0) {
 283                     if((destIndex+length)<=destCapacity) {
 284                         uprv_memcpy(dest+destIndex, src+prev, length*U_SIZEOF_UCHAR);
 285                     }
 286                     destIndex+=length;
 287                 }
 288             }
 289
 290             if(titleStart<titleLimit) {
 291                 /* titlecase c which is from [titleStart..titleLimit[ */
 292                 csc->cpStart=titleStart;
 293                 csc->cpLimit=titleLimit;
 294                 c=ucase_toFullTitle(csm->csp, c, utf16_caseContextIterator, csc, &s, csm->locale, &csm->locCache);
 295                 destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 296
 297                 /* Special case Dutch IJ titlecasing */
 298                 if ( titleStart+1 < idx &&
 299                      ucase_getCaseLocale(csm->locale,&csm->locCache) == UCASE_LOC_DUTCH &&
 300                      ( src[titleStart] == (UChar32) 0x0049 || src[titleStart] == (UChar32) 0x0069 ) &&
 301                      ( src[titleStart+1] == (UChar32) 0x004A || src[titleStart+1] == (UChar32) 0x006A )) {
 302                             c=(UChar32) 0x004A;
 303                             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 304                             titleLimit++;
 305                 }
 306
 307                 /* lowercase [titleLimit..index[ */
 308                 if(titleLimit<idx) {
 309                     if((csm->options&U_TITLECASE_NO_LOWERCASE)==0) {
 310                         /* Normal operation: Lowercase the rest of the word. */
 311                         destIndex+=
 312                             _caseMap(
 313                                 csm, ucase_toFullLower,
 314                                 dest+destIndex, destCapacity-destIndex,
 315                                 src, csc,
 316                                 titleLimit, idx,
 317                                 pErrorCode);
 318                     } else {
 319                         /* Optionally just copy the rest of the word unchanged. */
 320                         length=idx-titleLimit;
 321                         if((destIndex+length)<=destCapacity) {
 322                             uprv_memcpy(dest+destIndex, src+titleLimit, length*U_SIZEOF_UCHAR);
 323                         }
 324                         destIndex+=length;
 325                     }
 326                 }
 327             }
 328         }
 329
 330         prev=idx;
 331     }
 332
 333     if(destIndex>destCapacity) {
 334         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 335     }
 336     return destIndex;
 337 }
 338
 339 #endif
 340
 341 /* functions available in the common library (for unistr_case.cpp) */
 342
 343 U_CFUNC int32_t
 344 ustr_toLower(const UCaseProps *csp,
 345              UChar *dest, int32_t destCapacity,
 346              const UChar *src, int32_t srcLength,
 347              const char *locale,
 348              UErrorCode *pErrorCode) {
 349     UCaseMap csm={ NULL };
 350     UCaseContext csc={ NULL };
 351
 352     csm.csp=csp;
 353     setTempCaseMap(&csm, locale, pErrorCode);
 354     csc.p=(void *)src;
 355     csc.limit=srcLength;
 356
 357     return _caseMap(&csm, ucase_toFullLower,
 358                     dest, destCapacity,
 359                     src, &csc, 0, srcLength,
 360                     pErrorCode);
 361 }
 362
 363 U_CFUNC int32_t
 364 ustr_toUpper(const UCaseProps *csp,
 365              UChar *dest, int32_t destCapacity,
 366              const UChar *src, int32_t srcLength,
 367              const char *locale,
 368              UErrorCode *pErrorCode) {
 369     UCaseMap csm={ NULL };
 370     UCaseContext csc={ NULL };
 371
 372     csm.csp=csp;
 373     setTempCaseMap(&csm, locale, pErrorCode);
 374     csc.p=(void *)src;
 375     csc.limit=srcLength;
 376
 377     return _caseMap(&csm, ucase_toFullUpper,
 378                     dest, destCapacity,
 379                     src, &csc, 0, srcLength,
 380                     pErrorCode);
 381 }
 382
 383 #if !UCONFIG_NO_BREAK_ITERATION
 384
 385 U_CFUNC int32_t
 386 ustr_toTitle(const UCaseProps *csp,
 387              UChar *dest, int32_t destCapacity,
 388              const UChar *src, int32_t srcLength,
 389              UBreakIterator *titleIter,
 390              const char *locale, uint32_t options,
 391              UErrorCode *pErrorCode) {
 392     UCaseMap csm={ NULL };
 393     UCaseContext csc={ NULL };
 394     int32_t length;
 395
 396     csm.csp=csp;
 397     csm.iter=titleIter;
 398     csm.options=options;
 399     setTempCaseMap(&csm, locale, pErrorCode);
 400     csc.p=(void *)src;
 401     csc.limit=srcLength;
 402
 403     length=_toTitle(&csm,
 404                     dest, destCapacity,
 405                     src, &csc, srcLength,
 406                     pErrorCode);
 407     if(titleIter==NULL && csm.iter!=NULL) {
 408         ubrk_close(csm.iter);
 409     }
 410     return length;
 411 }
 412
 413 #endif
 414
 415 U_CFUNC int32_t
 416 ustr_foldCase(const UCaseProps *csp,
 417               UChar *dest, int32_t destCapacity,
 418               const UChar *src, int32_t srcLength,
 419               uint32_t options,
 420               UErrorCode *pErrorCode) {
 421     int32_t srcIndex, destIndex;
 422
 423     const UChar *s;
 424     UChar32 c, c2 = 0;
 425
 426     /* case mapping loop */
 427     srcIndex=destIndex=0;
 428     while(srcIndex<srcLength) {
 429         U16_NEXT(src, srcIndex, srcLength, c);
 430         c=ucase_toFullFolding(csp, c, &s, options);
 431         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
 432             /* fast path version of appendResult() for BMP results */
 433             dest[destIndex++]=(UChar)c2;
 434         } else {
 435             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
 436         }
 437     }
 438
 439     if(destIndex>destCapacity) {
 440         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
 441     }
 442     return destIndex;
 443 }
 444
 445 /*
 446  * Implement argument checking and buffer handling
 447  * for string case mapping as a common function.
 448  */
 449
 450 /* common internal function for public API functions */
 451
 452 static int32_t
 453 caseMap(const UCaseMap *csm,
 454         UChar *dest, int32_t destCapacity,
 455         const UChar *src, int32_t srcLength,
 456         int32_t toWhichCase,
 457         UErrorCode *pErrorCode) {
 458     UChar buffer[300];
 459     UChar *temp;
 460
 461     int32_t destLength;
 462
 463     /* check argument values */
 464     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
 465         return 0;
 466     }
 467     if( destCapacity<0 ||
 468         (dest==NULL && destCapacity>0) ||
 469         src==NULL ||
 470         srcLength<-1
 471     ) {
 472         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
 473         return 0;
 474     }
 475
 476     /* get the string length */
 477     if(srcLength==-1) {
 478         srcLength=u_strlen(src);
 479     }
 480
 481     /* check for overlapping source and destination */
 482     if( dest!=NULL &&
 483         ((src>=dest && src<(dest+destCapacity)) ||
 484          (dest>=src && dest<(src+srcLength)))
 485     ) {
 486         /* overlap: provide a temporary destination buffer and later copy the result */
 487         if(destCapacity<=(sizeof(buffer)/U_SIZEOF_UCHAR)) {
 488             /* the stack buffer is large enough */
 489             temp=buffer;
 490         } else {
 491             /* allocate a buffer */
 492             temp=(UChar *)uprv_malloc(destCapacity*U_SIZEOF_UCHAR);
 493             if(temp==NULL) {
 494                 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
 495                 return 0;
 496             }
 497         }
 498     } else {
 499         temp=dest;
 500     }
 501
 502     destLength=0;
 503
 504     if(toWhichCase==FOLD_CASE) {
 505         destLength=ustr_foldCase(csm->csp, temp, destCapacity, src, srcLength,
 506                                  csm->options, pErrorCode);
 507     } else {
 508         UCaseContext csc={ NULL };
 509
 510         csc.p=(void *)src;
 511         csc.limit=srcLength;
 512
 513         if(toWhichCase==TO_LOWER) {
 514             destLength=_caseMap(csm, ucase_toFullLower,
 515                                 temp, destCapacity,
 516                                 src, &csc,
 517                                 0, srcLength,
 518                                 pErrorCode);
 519         } else if(toWhichCase==TO_UPPER) {
 520             destLength=_caseMap(csm, ucase_toFullUpper,
 521                                 temp, destCapacity,
 522                                 src, &csc,
 523                                 0, srcLength,
 524                                 pErrorCode);
 525         } else /* if(toWhichCase==TO_TITLE) */ {
 526 #if UCONFIG_NO_BREAK_ITERATION
 527             *pErrorCode=U_UNSUPPORTED_ERROR;
 528 #else
 529             /* UCaseMap is actually non-const in toTitle() APIs. */
 530             destLength=_toTitle((UCaseMap *)csm, temp, destCapacity,
 531                                 src, &csc, srcLength,
 532                                 pErrorCode);
 533 #endif
 534         }
 535     }
 536     if(temp!=dest) {
 537         /* copy the result string to the destination buffer */
 538         if(destLength>0) {
 539             int32_t copyLength= destLength<=destCapacity ? destLength : destCapacity;
 540             if(copyLength>0) {
 541                 uprv_memmove(dest, temp, copyLength*U_SIZEOF_UCHAR);
 542             }
 543         }
 544         if(temp!=buffer) {
 545             uprv_free(temp);
 546         }
 547     }
 548
 549     return u_terminateUChars(dest, destCapacity, destLength, pErrorCode);
 550 }
 551
 552 /* public API functions */
 553
 554 U_CAPI int32_t U_EXPORT2
 555 u_strToLower(UChar *dest, int32_t destCapacity,
 556              const UChar *src, int32_t srcLength,
 557              const char *locale,
 558              UErrorCode *pErrorCode) {
 559     UCaseMap csm={ NULL };
 560     setTempCaseMap(&csm, locale, pErrorCode);
 561     return caseMap(&csm,
 562                    dest, destCapacity,
 563                    src, srcLength,
 564                    TO_LOWER, pErrorCode);
 565 }
 566
 567 U_CAPI int32_t U_EXPORT2
 568 u_strToUpper(UChar *dest, int32_t destCapacity,
 569              const UChar *src, int32_t srcLength,
 570              const char *locale,
 571              UErrorCode *pErrorCode) {
 572     UCaseMap csm={ NULL };
 573     setTempCaseMap(&csm, locale, pErrorCode);
 574     return caseMap(&csm,
 575                    dest, destCapacity,
 576                    src, srcLength,
 577                    TO_UPPER, pErrorCode);
 578 }
 579
 580 #if !UCONFIG_NO_BREAK_ITERATION
 581
 582 U_CAPI int32_t U_EXPORT2
 583 u_strToTitle(UChar *dest, int32_t destCapacity,
 584              const UChar *src, int32_t srcLength,
 585              UBreakIterator *titleIter,
 586              const char *locale,
 587              UErrorCode *pErrorCode) {
 588     UCaseMap csm={ NULL };
 589     int32_t length;
 590
 591     csm.iter=titleIter;
 592     setTempCaseMap(&csm, locale, pErrorCode);
 593     length=caseMap(&csm,
 594                    dest, destCapacity,
 595                    src, srcLength,
 596                    TO_TITLE, pErrorCode);
 597     if(titleIter==NULL && csm.iter!=NULL) {
 598         ubrk_close(csm.iter);
 599     }
 600     return length;
 601 }
 602
 603 U_CAPI int32_t U_EXPORT2
 604 ucasemap_toTitle(UCaseMap *csm,
 605                  UChar *dest, int32_t destCapacity,
 606                  const UChar *src, int32_t srcLength,
 607                  UErrorCode *pErrorCode) {
 608     return caseMap(csm,
 609                    dest, destCapacity,
 610                    src, srcLength,
 611                    TO_TITLE, pErrorCode);
 612 }
 613
 614 #endif
 615
 616 U_CAPI int32_t U_EXPORT2
 617 u_strFoldCase(UChar *dest, int32_t destCapacity,
 618               const UChar *src, int32_t srcLength,
 619               uint32_t options,
 620               UErrorCode *pErrorCode) {
 621     UCaseMap csm={ NULL };
 622     csm.csp=ucase_getSingleton();
 623     csm.options=options;
 624     return caseMap(&csm,
 625                    dest, destCapacity,
 626                    src, srcLength,
 627                    FOLD_CASE, pErrorCode);
 628 }
 629
 630 /* case-insensitive string comparisons -------------------------------------- */
 631
 632 /*
 633  * This function is a copy of unorm_cmpEquivFold() minus the parts for
 634  * canonical equivalence.
 635  * Keep the functions in sync, and see there for how this works.
 636  * The duplication is for modularization:
 637  * It makes caseless (but not canonical caseless) matches independent of
 638  * the normalization code.
 639  */
 640
 641 /* stack element for previous-level source/decomposition pointers */
 642 struct CmpEquivLevel {
 643     const UChar *start, *s, *limit;
 644 };
 645 typedef struct CmpEquivLevel CmpEquivLevel;
 646
 647 /* internal function */
 648 U_CFUNC int32_t
 649 u_strcmpFold(const UChar *s1, int32_t length1,
 650              const UChar *s2, int32_t length2,
 651              uint32_t options,
 652              UErrorCode *pErrorCode) {
 653     const UCaseProps *csp;
 654
 655     /* current-level start/limit - s1/s2 as current */
 656     const UChar *start1, *start2, *limit1, *limit2;
 657
 658     /* case folding variables */
 659     const UChar *p;
 660     int32_t length;
 661
 662     /* stacks of previous-level start/current/limit */
 663     CmpEquivLevel stack1[2], stack2[2];
 664
 665     /* case folding buffers, only use current-level start/limit */
 666     UChar fold1[UCASE_MAX_STRING_LENGTH+1], fold2[UCASE_MAX_STRING_LENGTH+1];
 667
 668     /* track which is the current level per string */
 669     int32_t level1, level2;
 670
 671     /* current code units, and code points for lookups */
 672     UChar32 c1, c2, cp1, cp2;
 673
 674     /* no argument error checking because this itself is not an API */
 675
 676     /*
 677      * assume that at least the option U_COMPARE_IGNORE_CASE is set
 678      * otherwise this function would have to behave exactly as uprv_strCompare()
 679      */
 680     csp=ucase_getSingleton();
 681     if(U_FAILURE(*pErrorCode)) {
 682         return 0;
 683     }
 684
 685     /* initialize */
 686     start1=s1;
 687     if(length1==-1) {
 688         limit1=NULL;
 689     } else {
 690         limit1=s1+length1;
 691     }
 692
 693     start2=s2;
 694     if(length2==-1) {
 695         limit2=NULL;
 696     } else {
 697         limit2=s2+length2;
 698     }
 699
 700     level1=level2=0;
 701     c1=c2=-1;
 702
 703     /* comparison loop */
 704     for(;;) {
 705         /*
 706          * here a code unit value of -1 means "get another code unit"
 707          * below it will mean "this source is finished"
 708          */
 709
 710         if(c1<0) {
 711             /* get next code unit from string 1, post-increment */
 712             for(;;) {
 713                 if(s1==limit1 || ((c1=*s1)==0 && (limit1==NULL || (options&_STRNCMP_STYLE)))) {
 714                     if(level1==0) {
 715                         c1=-1;
 716                         break;
 717                     }
 718                 } else {
 719                     ++s1;
 720                     break;
 721                 }
 722
 723                 /* reached end of level buffer, pop one level */
 724                 do {
 725                     --level1;
 726                     start1=stack1[level1].start;
 727                 } while(start1==NULL);
 728                 s1=stack1[level1].s;
 729                 limit1=stack1[level1].limit;
 730             }
 731         }
 732
 733         if(c2<0) {
 734             /* get next code unit from string 2, post-increment */
 735             for(;;) {
 736                 if(s2==limit2 || ((c2=*s2)==0 && (limit2==NULL || (options&_STRNCMP_STYLE)))) {
 737                     if(level2==0) {
 738                         c2=-1;
 739                         break;
 740                     }
 741                 } else {
 742                     ++s2;
 743                     break;
 744                 }
 745
 746                 /* reached end of level buffer, pop one level */
 747                 do {
 748                     --level2;
 749                     start2=stack2[level2].start;
 750                 } while(start2==NULL);
 751                 s2=stack2[level2].s;
 752                 limit2=stack2[level2].limit;
 753             }
 754         }
 755
 756         /*
 757          * compare c1 and c2
 758          * either variable c1, c2 is -1 only if the corresponding string is finished
 759          */
 760         if(c1==c2) {
 761             if(c1<0) {
 762                 return 0;   /* c1==c2==-1 indicating end of strings */
 763             }
 764             c1=c2=-1;       /* make us fetch new code units */
 765             continue;
 766         } else if(c1<0) {
 767             return -1;      /* string 1 ends before string 2 */
 768         } else if(c2<0) {
 769             return 1;       /* string 2 ends before string 1 */
 770         }
 771         /* c1!=c2 && c1>=0 && c2>=0 */
 772
 773         /* get complete code points for c1, c2 for lookups if either is a surrogate */
 774         cp1=c1;
 775         if(U_IS_SURROGATE(c1)) {
 776             UChar c;
 777
 778             if(U_IS_SURROGATE_LEAD(c1)) {
 779                 if(s1!=limit1 && U16_IS_TRAIL(c=*s1)) {
 780                     /* advance ++s1; only below if cp1 decomposes/case-folds */
 781                     cp1=U16_GET_SUPPLEMENTARY(c1, c);
 782                 }
 783             } else /* isTrail(c1) */ {
 784                 if(start1<=(s1-2) && U16_IS_LEAD(c=*(s1-2))) {
 785                     cp1=U16_GET_SUPPLEMENTARY(c, c1);
 786                 }
 787             }
 788         }
 789
 790         cp2=c2;
 791         if(U_IS_SURROGATE(c2)) {
 792             UChar c;
 793
 794             if(U_IS_SURROGATE_LEAD(c2)) {
 795                 if(s2!=limit2 && U16_IS_TRAIL(c=*s2)) {
 796                     /* advance ++s2; only below if cp2 decomposes/case-folds */
 797                     cp2=U16_GET_SUPPLEMENTARY(c2, c);
 798                 }
 799             } else /* isTrail(c2) */ {
 800                 if(start2<=(s2-2) && U16_IS_LEAD(c=*(s2-2))) {
 801                     cp2=U16_GET_SUPPLEMENTARY(c, c2);
 802                 }
 803             }
 804         }
 805
 806         /*
 807          * go down one level for each string
 808          * continue with the main loop as soon as there is a real change
 809          */
 810
 811         if( level1==0 &&
 812             (length=ucase_toFullFolding(csp, (UChar32)cp1, &p, options))>=0
 813         ) {
 814             /* cp1 case-folds to the code point "length" or to p[length] */
 815             if(U_IS_SURROGATE(c1)) {
 816                 if(U_IS_SURROGATE_LEAD(c1)) {
 817                     /* advance beyond source surrogate pair if it case-folds */
 818                     ++s1;
 819                 } else /* isTrail(c1) */ {
 820                     /*
 821                      * we got a supplementary code point when hitting its trail surrogate,
 822                      * therefore the lead surrogate must have been the same as in the other string;
 823                      * compare this decomposition with the lead surrogate in the other string
 824                      * remember that this simulates bulk text replacement:
 825                      * the decomposition would replace the entire code point
 826                      */
 827                     --s2;
 828                     c2=*(s2-1);
 829                 }
 830             }
 831
 832             /* push current level pointers */
 833             stack1[0].start=start1;
 834             stack1[0].s=s1;
 835             stack1[0].limit=limit1;
 836             ++level1;
 837
 838             /* copy the folding result to fold1[] */
 839             if(length<=UCASE_MAX_STRING_LENGTH) {
 840                 u_memcpy(fold1, p, length);
 841             } else {
 842                 int32_t i=0;
 843                 U16_APPEND_UNSAFE(fold1, i, length);
 844                 length=i;
 845             }
 846
 847             /* set next level pointers to case folding */
 848             start1=s1=fold1;
 849             limit1=fold1+length;
 850
 851             /* get ready to read from decomposition, continue with loop */
 852             c1=-1;
 853             continue;
 854         }
 855
 856         if( level2==0 &&
 857             (length=ucase_toFullFolding(csp, (UChar32)cp2, &p, options))>=0
 858         ) {
 859             /* cp2 case-folds to the code point "length" or to p[length] */
 860             if(U_IS_SURROGATE(c2)) {
 861                 if(U_IS_SURROGATE_LEAD(c2)) {
 862                     /* advance beyond source surrogate pair if it case-folds */
 863                     ++s2;
 864                 } else /* isTrail(c2) */ {
 865                     /*
 866                      * we got a supplementary code point when hitting its trail surrogate,
 867                      * therefore the lead surrogate must have been the same as in the other string;
 868                      * compare this decomposition with the lead surrogate in the other string
 869                      * remember that this simulates bulk text replacement:
 870                      * the decomposition would replace the entire code point
 871                      */
 872                     --s1;
 873                     c1=*(s1-1);
 874                 }
 875             }
 876
 877             /* push current level pointers */
 878             stack2[0].start=start2;
 879             stack2[0].s=s2;
 880             stack2[0].limit=limit2;
 881             ++level2;
 882
 883             /* copy the folding result to fold2[] */
 884             if(length<=UCASE_MAX_STRING_LENGTH) {
 885                 u_memcpy(fold2, p, length);
 886             } else {
 887                 int32_t i=0;
 888                 U16_APPEND_UNSAFE(fold2, i, length);
 889                 length=i;
 890             }
 891
 892             /* set next level pointers to case folding */
 893             start2=s2=fold2;
 894             limit2=fold2+length;
 895
 896             /* get ready to read from decomposition, continue with loop */
 897             c2=-1;
 898             continue;
 899         }
 900
 901         /*
 902          * no decomposition/case folding, max level for both sides:
 903          * return difference result
 904          *
 905          * code point order comparison must not just return cp1-cp2
 906          * because when single surrogates are present then the surrogate pairs
 907          * that formed cp1 and cp2 may be from different string indexes
 908          *
 909          * example: { d800 d800 dc01 } vs. { d800 dc00 }, compare at second code units
 910          * c1=d800 cp1=10001 c2=dc00 cp2=10000
 911          * cp1-cp2>0 but c1-c2<0 and in fact in UTF-32 it is { d800 10001 } < { 10000 }
 912          *
 913          * therefore, use same fix-up as in ustring.c/uprv_strCompare()
 914          * except: uprv_strCompare() fetches c=*s while this functions fetches c=*s++
 915          * so we have slightly different pointer/start/limit comparisons here
 916          */
 917
 918         if(c1>=0xd800 && c2>=0xd800 && (options&U_COMPARE_CODE_POINT_ORDER)) {
 919             /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
 920             if(
 921                 (c1<=0xdbff && s1!=limit1 && U16_IS_TRAIL(*s1)) ||
 922                 (U16_IS_TRAIL(c1) && start1!=(s1-1) && U16_IS_LEAD(*(s1-2)))
 923             ) {
 924                 /* part of a surrogate pair, leave >=d800 */
 925             } else {
 926                 /* BMP code point - may be surrogate code point - make <d800 */
 927                 c1-=0x2800;
 928             }
 929
 930             if(
 931                 (c2<=0xdbff && s2!=limit2 && U16_IS_TRAIL(*s2)) ||
 932                 (U16_IS_TRAIL(c2) && start2!=(s2-1) && U16_IS_LEAD(*(s2-2)))
 933             ) {
 934                 /* part of a surrogate pair, leave >=d800 */
 935             } else {
 936                 /* BMP code point - may be surrogate code point - make <d800 */
 937                 c2-=0x2800;
 938             }
 939         }
 940
 941         return c1-c2;
 942     }
 943 }
 944
 945 /* public API functions */
 946
 947 U_CAPI int32_t U_EXPORT2
 948 u_strCaseCompare(const UChar *s1, int32_t length1,
 949                  const UChar *s2, int32_t length2,
 950                  uint32_t options,
 951                  UErrorCode *pErrorCode) {
 952     /* argument checking */
 953     if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
 954         return 0;
 955     }
 956     if(s1==NULL || length1<-1 || s2==NULL || length2<-1) {
 957         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
 958         return 0;
 959     }
 960     return u_strcmpFold(s1, length1, s2, length2,
 961                         options|U_COMPARE_IGNORE_CASE,
 962                         pErrorCode);
 963 }
 964
 965 U_CAPI int32_t U_EXPORT2
 966 u_strcasecmp(const UChar *s1, const UChar *s2, uint32_t options) {
 967     UErrorCode errorCode=U_ZERO_ERROR;
 968     return u_strcmpFold(s1, -1, s2, -1,
 969                         options|U_COMPARE_IGNORE_CASE,
 970                         &errorCode);
 971 }
 972
 973 U_CAPI int32_t U_EXPORT2
 974 u_memcasecmp(const UChar *s1, const UChar *s2, int32_t length, uint32_t options) {
 975     UErrorCode errorCode=U_ZERO_ERROR;
 976     return u_strcmpFold(s1, length, s2, length,
 977                         options|U_COMPARE_IGNORE_CASE,
 978                         &errorCode);
 979 }
 980
 981 U_CAPI int32_t U_EXPORT2
 982 u_strncasecmp(const UChar *s1, const UChar *s2, int32_t n, uint32_t options) {
 983     UErrorCode errorCode=U_ZERO_ERROR;
 984     return u_strcmpFold(s1, n, s2, n,
 985                         options|(U_COMPARE_IGNORE_CASE|_STRNCMP_STYLE),
 986                         &errorCode);
 987 }