michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2001-2011, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  ustrcase.cpp
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2002feb20
michael@0: *   created by: Markus W. Scherer
michael@0: *
michael@0: *   Implementation file for string casing C API functions.
michael@0: *   Uses functions from uchar.c for basic functionality that requires access
michael@0: *   to the Unicode Character Database (uprops.dat).
michael@0: */
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "unicode/brkiter.h"
michael@0: #include "unicode/ustring.h"
michael@0: #include "unicode/ucasemap.h"
michael@0: #include "unicode/ubrk.h"
michael@0: #include "unicode/utf.h"
michael@0: #include "unicode/utf16.h"
michael@0: #include "cmemory.h"
michael@0: #include "ucase.h"
michael@0: #include "ustr_imp.h"
michael@0: 
michael@0: #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
michael@0: 
michael@0: U_NAMESPACE_USE
michael@0: 
michael@0: /* string casing ------------------------------------------------------------ */
michael@0: 
michael@0: /* Appends a full case mapping result, see UCASE_MAX_STRING_LENGTH. */
michael@0: static inline int32_t
michael@0: appendResult(UChar *dest, int32_t destIndex, int32_t destCapacity,
michael@0:              int32_t result, const UChar *s) {
michael@0:     UChar32 c;
michael@0:     int32_t length;
michael@0: 
michael@0:     /* decode the result */
michael@0:     if(result<0) {
michael@0:         /* (not) original code point */
michael@0:         c=~result;
michael@0:         length=-1;
michael@0:     } else if(result<=UCASE_MAX_STRING_LENGTH) {
michael@0:         c=U_SENTINEL;
michael@0:         length=result;
michael@0:     } else {
michael@0:         c=result;
michael@0:         length=-1;
michael@0:     }
michael@0: 
michael@0:     if(destIndex<destCapacity) {
michael@0:         /* append the result */
michael@0:         if(length<0) {
michael@0:             /* code point */
michael@0:             UBool isError=FALSE;
michael@0:             U16_APPEND(dest, destIndex, destCapacity, c, isError);
michael@0:             if(isError) {
michael@0:                 /* overflow, nothing written */
michael@0:                 destIndex+=U16_LENGTH(c);
michael@0:             }
michael@0:         } else {
michael@0:             /* string */
michael@0:             if((destIndex+length)<=destCapacity) {
michael@0:                 while(length>0) {
michael@0:                     dest[destIndex++]=*s++;
michael@0:                     --length;
michael@0:                 }
michael@0:             } else {
michael@0:                 /* overflow */
michael@0:                 destIndex+=length;
michael@0:             }
michael@0:         }
michael@0:     } else {
michael@0:         /* preflight */
michael@0:         if(length<0) {
michael@0:             destIndex+=U16_LENGTH(c);
michael@0:         } else {
michael@0:             destIndex+=length;
michael@0:         }
michael@0:     }
michael@0:     return destIndex;
michael@0: }
michael@0: 
michael@0: static UChar32 U_CALLCONV
michael@0: utf16_caseContextIterator(void *context, int8_t dir) {
michael@0:     UCaseContext *csc=(UCaseContext *)context;
michael@0:     UChar32 c;
michael@0: 
michael@0:     if(dir<0) {
michael@0:         /* reset for backward iteration */
michael@0:         csc->index=csc->cpStart;
michael@0:         csc->dir=dir;
michael@0:     } else if(dir>0) {
michael@0:         /* reset for forward iteration */
michael@0:         csc->index=csc->cpLimit;
michael@0:         csc->dir=dir;
michael@0:     } else {
michael@0:         /* continue current iteration direction */
michael@0:         dir=csc->dir;
michael@0:     }
michael@0: 
michael@0:     if(dir<0) {
michael@0:         if(csc->start<csc->index) {
michael@0:             U16_PREV((const UChar *)csc->p, csc->start, csc->index, c);
michael@0:             return c;
michael@0:         }
michael@0:     } else {
michael@0:         if(csc->index<csc->limit) {
michael@0:             U16_NEXT((const UChar *)csc->p, csc->index, csc->limit, c);
michael@0:             return c;
michael@0:         }
michael@0:     }
michael@0:     return U_SENTINEL;
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Case-maps [srcStart..srcLimit[ but takes
michael@0:  * context [0..srcLength[ into account.
michael@0:  */
michael@0: static int32_t
michael@0: _caseMap(const UCaseMap *csm, UCaseMapFull *map,
michael@0:          UChar *dest, int32_t destCapacity,
michael@0:          const UChar *src, UCaseContext *csc,
michael@0:          int32_t srcStart, int32_t srcLimit,
michael@0:          UErrorCode *pErrorCode) {
michael@0:     const UChar *s;
michael@0:     UChar32 c, c2 = 0;
michael@0:     int32_t srcIndex, destIndex;
michael@0:     int32_t locCache;
michael@0: 
michael@0:     locCache=csm->locCache;
michael@0: 
michael@0:     /* case mapping loop */
michael@0:     srcIndex=srcStart;
michael@0:     destIndex=0;
michael@0:     while(srcIndex<srcLimit) {
michael@0:         csc->cpStart=srcIndex;
michael@0:         U16_NEXT(src, srcIndex, srcLimit, c);
michael@0:         csc->cpLimit=srcIndex;
michael@0:         c=map(csm->csp, c, utf16_caseContextIterator, csc, &s, csm->locale, &locCache);
michael@0:         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
michael@0:             /* fast path version of appendResult() for BMP results */
michael@0:             dest[destIndex++]=(UChar)c2;
michael@0:         } else {
michael@0:             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     if(destIndex>destCapacity) {
michael@0:         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
michael@0:     }
michael@0:     return destIndex;
michael@0: }
michael@0: 
michael@0: #if !UCONFIG_NO_BREAK_ITERATION
michael@0: 
michael@0: U_CFUNC int32_t U_CALLCONV
michael@0: ustrcase_internalToTitle(const UCaseMap *csm,
michael@0:                          UChar *dest, int32_t destCapacity,
michael@0:                          const UChar *src, int32_t srcLength,
michael@0:                          UErrorCode *pErrorCode) {
michael@0:     const UChar *s;
michael@0:     UChar32 c;
michael@0:     int32_t prev, titleStart, titleLimit, idx, destIndex, length;
michael@0:     UBool isFirstIndex;
michael@0: 
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     // Use the C++ abstract base class to minimize dependencies.
michael@0:     // TODO: Change UCaseMap.iter to store a BreakIterator directly.
michael@0:     BreakIterator *bi=reinterpret_cast<BreakIterator *>(csm->iter);
michael@0: 
michael@0:     /* set up local variables */
michael@0:     int32_t locCache=csm->locCache;
michael@0:     UCaseContext csc=UCASECONTEXT_INITIALIZER;
michael@0:     csc.p=(void *)src;
michael@0:     csc.limit=srcLength;
michael@0:     destIndex=0;
michael@0:     prev=0;
michael@0:     isFirstIndex=TRUE;
michael@0: 
michael@0:     /* titlecasing loop */
michael@0:     while(prev<srcLength) {
michael@0:         /* find next index where to titlecase */
michael@0:         if(isFirstIndex) {
michael@0:             isFirstIndex=FALSE;
michael@0:             idx=bi->first();
michael@0:         } else {
michael@0:             idx=bi->next();
michael@0:         }
michael@0:         if(idx==UBRK_DONE || idx>srcLength) {
michael@0:             idx=srcLength;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Unicode 4 & 5 section 3.13 Default Case Operations:
michael@0:          *
michael@0:          * R3  toTitlecase(X): Find the word boundaries based on Unicode Standard Annex
michael@0:          * #29, "Text Boundaries." Between each pair of word boundaries, find the first
michael@0:          * cased character F. If F exists, map F to default_title(F); then map each
michael@0:          * subsequent character C to default_lower(C).
michael@0:          *
michael@0:          * In this implementation, segment [prev..index[ into 3 parts:
michael@0:          * a) uncased characters (copy as-is) [prev..titleStart[
michael@0:          * b) first case letter (titlecase)         [titleStart..titleLimit[
michael@0:          * c) subsequent characters (lowercase)                 [titleLimit..index[
michael@0:          */
michael@0:         if(prev<idx) {
michael@0:             /* find and copy uncased characters [prev..titleStart[ */
michael@0:             titleStart=titleLimit=prev;
michael@0:             U16_NEXT(src, titleLimit, idx, c);
michael@0:             if((csm->options&U_TITLECASE_NO_BREAK_ADJUSTMENT)==0 && UCASE_NONE==ucase_getType(csm->csp, c)) {
michael@0:                 /* Adjust the titlecasing index (titleStart) to the next cased character. */
michael@0:                 for(;;) {
michael@0:                     titleStart=titleLimit;
michael@0:                     if(titleLimit==idx) {
michael@0:                         /*
michael@0:                          * only uncased characters in [prev..index[
michael@0:                          * stop with titleStart==titleLimit==index
michael@0:                          */
michael@0:                         break;
michael@0:                     }
michael@0:                     U16_NEXT(src, titleLimit, idx, c);
michael@0:                     if(UCASE_NONE!=ucase_getType(csm->csp, c)) {
michael@0:                         break; /* cased letter at [titleStart..titleLimit[ */
michael@0:                     }
michael@0:                 }
michael@0:                 length=titleStart-prev;
michael@0:                 if(length>0) {
michael@0:                     if((destIndex+length)<=destCapacity) {
michael@0:                         uprv_memcpy(dest+destIndex, src+prev, length*U_SIZEOF_UCHAR);
michael@0:                     }
michael@0:                     destIndex+=length;
michael@0:                 }
michael@0:             }
michael@0: 
michael@0:             if(titleStart<titleLimit) {
michael@0:                 /* titlecase c which is from [titleStart..titleLimit[ */
michael@0:                 csc.cpStart=titleStart;
michael@0:                 csc.cpLimit=titleLimit;
michael@0:                 c=ucase_toFullTitle(csm->csp, c, utf16_caseContextIterator, &csc, &s, csm->locale, &locCache);
michael@0:                 destIndex=appendResult(dest, destIndex, destCapacity, c, s); 
michael@0: 
michael@0:                 /* Special case Dutch IJ titlecasing */
michael@0:                 if ( titleStart+1 < idx && 
michael@0:                      ucase_getCaseLocale(csm->locale,&locCache) == UCASE_LOC_DUTCH &&
michael@0:                      ( src[titleStart] == (UChar32) 0x0049 || src[titleStart] == (UChar32) 0x0069 ) &&
michael@0:                      ( src[titleStart+1] == (UChar32) 0x004A || src[titleStart+1] == (UChar32) 0x006A )) { 
michael@0:                             c=(UChar32) 0x004A;
michael@0:                             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
michael@0:                             titleLimit++;
michael@0:                 }
michael@0: 
michael@0:                 /* lowercase [titleLimit..index[ */
michael@0:                 if(titleLimit<idx) {
michael@0:                     if((csm->options&U_TITLECASE_NO_LOWERCASE)==0) {
michael@0:                         /* Normal operation: Lowercase the rest of the word. */
michael@0:                         destIndex+=
michael@0:                             _caseMap(
michael@0:                                 csm, ucase_toFullLower,
michael@0:                                 dest+destIndex, destCapacity-destIndex,
michael@0:                                 src, &csc,
michael@0:                                 titleLimit, idx,
michael@0:                                 pErrorCode);
michael@0:                     } else {
michael@0:                         /* Optionally just copy the rest of the word unchanged. */
michael@0:                         length=idx-titleLimit;
michael@0:                         if((destIndex+length)<=destCapacity) {
michael@0:                             uprv_memcpy(dest+destIndex, src+titleLimit, length*U_SIZEOF_UCHAR);
michael@0:                         }
michael@0:                         destIndex+=length;
michael@0:                     }
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         prev=idx;
michael@0:     }
michael@0: 
michael@0:     if(destIndex>destCapacity) {
michael@0:         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
michael@0:     }
michael@0:     return destIndex;
michael@0: }
michael@0: 
michael@0: #endif  // !UCONFIG_NO_BREAK_ITERATION
michael@0: 
michael@0: /* functions available in the common library (for unistr_case.cpp) */
michael@0: 
michael@0: U_CFUNC int32_t U_CALLCONV
michael@0: ustrcase_internalToLower(const UCaseMap *csm,
michael@0:                          UChar *dest, int32_t destCapacity,
michael@0:                          const UChar *src, int32_t srcLength,
michael@0:                          UErrorCode *pErrorCode) {
michael@0:     UCaseContext csc=UCASECONTEXT_INITIALIZER;
michael@0:     csc.p=(void *)src;
michael@0:     csc.limit=srcLength;
michael@0:     return _caseMap(
michael@0:         csm, ucase_toFullLower,
michael@0:         dest, destCapacity,
michael@0:         src, &csc, 0, srcLength,
michael@0:         pErrorCode);
michael@0: }
michael@0: 
michael@0: U_CFUNC int32_t U_CALLCONV
michael@0: ustrcase_internalToUpper(const UCaseMap *csm,
michael@0:                          UChar *dest, int32_t destCapacity,
michael@0:                          const UChar *src, int32_t srcLength,
michael@0:                          UErrorCode *pErrorCode) {
michael@0:     UCaseContext csc=UCASECONTEXT_INITIALIZER;
michael@0:     csc.p=(void *)src;
michael@0:     csc.limit=srcLength;
michael@0:     return _caseMap(
michael@0:         csm, ucase_toFullUpper,
michael@0:         dest, destCapacity,
michael@0:         src, &csc, 0, srcLength,
michael@0:         pErrorCode);
michael@0: }
michael@0: 
michael@0: static int32_t
michael@0: ustr_foldCase(const UCaseProps *csp,
michael@0:               UChar *dest, int32_t destCapacity,
michael@0:               const UChar *src, int32_t srcLength,
michael@0:               uint32_t options,
michael@0:               UErrorCode *pErrorCode) {
michael@0:     int32_t srcIndex, destIndex;
michael@0: 
michael@0:     const UChar *s;
michael@0:     UChar32 c, c2 = 0;
michael@0: 
michael@0:     /* case mapping loop */
michael@0:     srcIndex=destIndex=0;
michael@0:     while(srcIndex<srcLength) {
michael@0:         U16_NEXT(src, srcIndex, srcLength, c);
michael@0:         c=ucase_toFullFolding(csp, c, &s, options);
michael@0:         if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
michael@0:             /* fast path version of appendResult() for BMP results */
michael@0:             dest[destIndex++]=(UChar)c2;
michael@0:         } else {
michael@0:             destIndex=appendResult(dest, destIndex, destCapacity, c, s);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     if(destIndex>destCapacity) {
michael@0:         *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
michael@0:     }
michael@0:     return destIndex;
michael@0: }
michael@0: 
michael@0: U_CFUNC int32_t U_CALLCONV
michael@0: ustrcase_internalFold(const UCaseMap *csm,
michael@0:                       UChar *dest, int32_t destCapacity,
michael@0:                       const UChar *src, int32_t srcLength,
michael@0:                       UErrorCode *pErrorCode) {
michael@0:     return ustr_foldCase(csm->csp, dest, destCapacity, src, srcLength, csm->options, pErrorCode);
michael@0: }
michael@0: 
michael@0: U_CFUNC int32_t
michael@0: ustrcase_map(const UCaseMap *csm,
michael@0:              UChar *dest, int32_t destCapacity,
michael@0:              const UChar *src, int32_t srcLength,
michael@0:              UStringCaseMapper *stringCaseMapper,
michael@0:              UErrorCode *pErrorCode) {
michael@0:     UChar buffer[300];
michael@0:     UChar *temp;
michael@0: 
michael@0:     int32_t destLength;
michael@0: 
michael@0:     /* check argument values */
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if( destCapacity<0 ||
michael@0:         (dest==NULL && destCapacity>0) ||
michael@0:         src==NULL ||
michael@0:         srcLength<-1
michael@0:     ) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     /* get the string length */
michael@0:     if(srcLength==-1) {
michael@0:         srcLength=u_strlen(src);
michael@0:     }
michael@0: 
michael@0:     /* check for overlapping source and destination */
michael@0:     if( dest!=NULL &&
michael@0:         ((src>=dest && src<(dest+destCapacity)) ||
michael@0:          (dest>=src && dest<(src+srcLength)))
michael@0:     ) {
michael@0:         /* overlap: provide a temporary destination buffer and later copy the result */
michael@0:         if(destCapacity<=LENGTHOF(buffer)) {
michael@0:             /* the stack buffer is large enough */
michael@0:             temp=buffer;
michael@0:         } else {
michael@0:             /* allocate a buffer */
michael@0:             temp=(UChar *)uprv_malloc(destCapacity*U_SIZEOF_UCHAR);
michael@0:             if(temp==NULL) {
michael@0:                 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:         }
michael@0:     } else {
michael@0:         temp=dest;
michael@0:     }
michael@0: 
michael@0:     destLength=stringCaseMapper(csm, temp, destCapacity, src, srcLength, pErrorCode);
michael@0:     if(temp!=dest) {
michael@0:         /* copy the result string to the destination buffer */
michael@0:         if(destLength>0) {
michael@0:             int32_t copyLength= destLength<=destCapacity ? destLength : destCapacity;
michael@0:             if(copyLength>0) {
michael@0:                 uprv_memmove(dest, temp, copyLength*U_SIZEOF_UCHAR);
michael@0:             }
michael@0:         }
michael@0:         if(temp!=buffer) {
michael@0:             uprv_free(temp);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return u_terminateUChars(dest, destCapacity, destLength, pErrorCode);
michael@0: }
michael@0: 
michael@0: /* public API functions */
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: u_strFoldCase(UChar *dest, int32_t destCapacity,
michael@0:               const UChar *src, int32_t srcLength,
michael@0:               uint32_t options,
michael@0:               UErrorCode *pErrorCode) {
michael@0:     UCaseMap csm=UCASEMAP_INITIALIZER;
michael@0:     csm.csp=ucase_getSingleton();
michael@0:     csm.options=options;
michael@0:     return ustrcase_map(
michael@0:         &csm,
michael@0:         dest, destCapacity,
michael@0:         src, srcLength,
michael@0:         ustrcase_internalFold, pErrorCode);
michael@0: }
michael@0: 
michael@0: /* case-insensitive string comparisons -------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * This function is a copy of unorm_cmpEquivFold() minus the parts for
michael@0:  * canonical equivalence.
michael@0:  * Keep the functions in sync, and see there for how this works.
michael@0:  * The duplication is for modularization:
michael@0:  * It makes caseless (but not canonical caseless) matches independent of
michael@0:  * the normalization code.
michael@0:  */
michael@0: 
michael@0: /* stack element for previous-level source/decomposition pointers */
michael@0: struct CmpEquivLevel {
michael@0:     const UChar *start, *s, *limit;
michael@0: };
michael@0: typedef struct CmpEquivLevel CmpEquivLevel;
michael@0: 
michael@0: /* internal function */
michael@0: U_CFUNC int32_t
michael@0: u_strcmpFold(const UChar *s1, int32_t length1,
michael@0:              const UChar *s2, int32_t length2,
michael@0:              uint32_t options,
michael@0:              UErrorCode *pErrorCode) {
michael@0:     const UCaseProps *csp;
michael@0: 
michael@0:     /* current-level start/limit - s1/s2 as current */
michael@0:     const UChar *start1, *start2, *limit1, *limit2;
michael@0: 
michael@0:     /* case folding variables */
michael@0:     const UChar *p;
michael@0:     int32_t length;
michael@0: 
michael@0:     /* stacks of previous-level start/current/limit */
michael@0:     CmpEquivLevel stack1[2], stack2[2];
michael@0: 
michael@0:     /* case folding buffers, only use current-level start/limit */
michael@0:     UChar fold1[UCASE_MAX_STRING_LENGTH+1], fold2[UCASE_MAX_STRING_LENGTH+1];
michael@0: 
michael@0:     /* track which is the current level per string */
michael@0:     int32_t level1, level2;
michael@0: 
michael@0:     /* current code units, and code points for lookups */
michael@0:     UChar32 c1, c2, cp1, cp2;
michael@0: 
michael@0:     /* no argument error checking because this itself is not an API */
michael@0: 
michael@0:     /*
michael@0:      * assume that at least the option U_COMPARE_IGNORE_CASE is set
michael@0:      * otherwise this function would have to behave exactly as uprv_strCompare()
michael@0:      */
michael@0:     csp=ucase_getSingleton();
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     /* initialize */
michael@0:     start1=s1;
michael@0:     if(length1==-1) {
michael@0:         limit1=NULL;
michael@0:     } else {
michael@0:         limit1=s1+length1;
michael@0:     }
michael@0: 
michael@0:     start2=s2;
michael@0:     if(length2==-1) {
michael@0:         limit2=NULL;
michael@0:     } else {
michael@0:         limit2=s2+length2;
michael@0:     }
michael@0: 
michael@0:     level1=level2=0;
michael@0:     c1=c2=-1;
michael@0: 
michael@0:     /* comparison loop */
michael@0:     for(;;) {
michael@0:         /*
michael@0:          * here a code unit value of -1 means "get another code unit"
michael@0:          * below it will mean "this source is finished"
michael@0:          */
michael@0: 
michael@0:         if(c1<0) {
michael@0:             /* get next code unit from string 1, post-increment */
michael@0:             for(;;) {
michael@0:                 if(s1==limit1 || ((c1=*s1)==0 && (limit1==NULL || (options&_STRNCMP_STYLE)))) {
michael@0:                     if(level1==0) {
michael@0:                         c1=-1;
michael@0:                         break;
michael@0:                     }
michael@0:                 } else {
michael@0:                     ++s1;
michael@0:                     break;
michael@0:                 }
michael@0: 
michael@0:                 /* reached end of level buffer, pop one level */
michael@0:                 do {
michael@0:                     --level1;
michael@0:                     start1=stack1[level1].start;    /*Not uninitialized*/
michael@0:                 } while(start1==NULL);
michael@0:                 s1=stack1[level1].s;                /*Not uninitialized*/
michael@0:                 limit1=stack1[level1].limit;        /*Not uninitialized*/
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         if(c2<0) {
michael@0:             /* get next code unit from string 2, post-increment */
michael@0:             for(;;) {
michael@0:                 if(s2==limit2 || ((c2=*s2)==0 && (limit2==NULL || (options&_STRNCMP_STYLE)))) {
michael@0:                     if(level2==0) {
michael@0:                         c2=-1;
michael@0:                         break;
michael@0:                     }
michael@0:                 } else {
michael@0:                     ++s2;
michael@0:                     break;
michael@0:                 }
michael@0: 
michael@0:                 /* reached end of level buffer, pop one level */
michael@0:                 do {
michael@0:                     --level2;
michael@0:                     start2=stack2[level2].start;    /*Not uninitialized*/
michael@0:                 } while(start2==NULL);
michael@0:                 s2=stack2[level2].s;                /*Not uninitialized*/
michael@0:                 limit2=stack2[level2].limit;        /*Not uninitialized*/
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * compare c1 and c2
michael@0:          * either variable c1, c2 is -1 only if the corresponding string is finished
michael@0:          */
michael@0:         if(c1==c2) {
michael@0:             if(c1<0) {
michael@0:                 return 0;   /* c1==c2==-1 indicating end of strings */
michael@0:             }
michael@0:             c1=c2=-1;       /* make us fetch new code units */
michael@0:             continue;
michael@0:         } else if(c1<0) {
michael@0:             return -1;      /* string 1 ends before string 2 */
michael@0:         } else if(c2<0) {
michael@0:             return 1;       /* string 2 ends before string 1 */
michael@0:         }
michael@0:         /* c1!=c2 && c1>=0 && c2>=0 */
michael@0: 
michael@0:         /* get complete code points for c1, c2 for lookups if either is a surrogate */
michael@0:         cp1=c1;
michael@0:         if(U_IS_SURROGATE(c1)) {
michael@0:             UChar c;
michael@0: 
michael@0:             if(U_IS_SURROGATE_LEAD(c1)) {
michael@0:                 if(s1!=limit1 && U16_IS_TRAIL(c=*s1)) {
michael@0:                     /* advance ++s1; only below if cp1 decomposes/case-folds */
michael@0:                     cp1=U16_GET_SUPPLEMENTARY(c1, c);
michael@0:                 }
michael@0:             } else /* isTrail(c1) */ {
michael@0:                 if(start1<=(s1-2) && U16_IS_LEAD(c=*(s1-2))) {
michael@0:                     cp1=U16_GET_SUPPLEMENTARY(c, c1);
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         cp2=c2;
michael@0:         if(U_IS_SURROGATE(c2)) {
michael@0:             UChar c;
michael@0: 
michael@0:             if(U_IS_SURROGATE_LEAD(c2)) {
michael@0:                 if(s2!=limit2 && U16_IS_TRAIL(c=*s2)) {
michael@0:                     /* advance ++s2; only below if cp2 decomposes/case-folds */
michael@0:                     cp2=U16_GET_SUPPLEMENTARY(c2, c);
michael@0:                 }
michael@0:             } else /* isTrail(c2) */ {
michael@0:                 if(start2<=(s2-2) && U16_IS_LEAD(c=*(s2-2))) {
michael@0:                     cp2=U16_GET_SUPPLEMENTARY(c, c2);
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * go down one level for each string
michael@0:          * continue with the main loop as soon as there is a real change
michael@0:          */
michael@0: 
michael@0:         if( level1==0 &&
michael@0:             (length=ucase_toFullFolding(csp, (UChar32)cp1, &p, options))>=0
michael@0:         ) {
michael@0:             /* cp1 case-folds to the code point "length" or to p[length] */
michael@0:             if(U_IS_SURROGATE(c1)) {
michael@0:                 if(U_IS_SURROGATE_LEAD(c1)) {
michael@0:                     /* advance beyond source surrogate pair if it case-folds */
michael@0:                     ++s1;
michael@0:                 } else /* isTrail(c1) */ {
michael@0:                     /*
michael@0:                      * we got a supplementary code point when hitting its trail surrogate,
michael@0:                      * therefore the lead surrogate must have been the same as in the other string;
michael@0:                      * compare this decomposition with the lead surrogate in the other string
michael@0:                      * remember that this simulates bulk text replacement:
michael@0:                      * the decomposition would replace the entire code point
michael@0:                      */
michael@0:                     --s2;
michael@0:                     c2=*(s2-1);
michael@0:                 }
michael@0:             }
michael@0: 
michael@0:             /* push current level pointers */
michael@0:             stack1[0].start=start1;
michael@0:             stack1[0].s=s1;
michael@0:             stack1[0].limit=limit1;
michael@0:             ++level1;
michael@0: 
michael@0:             /* copy the folding result to fold1[] */
michael@0:             if(length<=UCASE_MAX_STRING_LENGTH) {
michael@0:                 u_memcpy(fold1, p, length);
michael@0:             } else {
michael@0:                 int32_t i=0;
michael@0:                 U16_APPEND_UNSAFE(fold1, i, length);
michael@0:                 length=i;
michael@0:             }
michael@0: 
michael@0:             /* set next level pointers to case folding */
michael@0:             start1=s1=fold1;
michael@0:             limit1=fold1+length;
michael@0: 
michael@0:             /* get ready to read from decomposition, continue with loop */
michael@0:             c1=-1;
michael@0:             continue;
michael@0:         }
michael@0: 
michael@0:         if( level2==0 &&
michael@0:             (length=ucase_toFullFolding(csp, (UChar32)cp2, &p, options))>=0
michael@0:         ) {
michael@0:             /* cp2 case-folds to the code point "length" or to p[length] */
michael@0:             if(U_IS_SURROGATE(c2)) {
michael@0:                 if(U_IS_SURROGATE_LEAD(c2)) {
michael@0:                     /* advance beyond source surrogate pair if it case-folds */
michael@0:                     ++s2;
michael@0:                 } else /* isTrail(c2) */ {
michael@0:                     /*
michael@0:                      * we got a supplementary code point when hitting its trail surrogate,
michael@0:                      * therefore the lead surrogate must have been the same as in the other string;
michael@0:                      * compare this decomposition with the lead surrogate in the other string
michael@0:                      * remember that this simulates bulk text replacement:
michael@0:                      * the decomposition would replace the entire code point
michael@0:                      */
michael@0:                     --s1;
michael@0:                     c1=*(s1-1);
michael@0:                 }
michael@0:             }
michael@0: 
michael@0:             /* push current level pointers */
michael@0:             stack2[0].start=start2;
michael@0:             stack2[0].s=s2;
michael@0:             stack2[0].limit=limit2;
michael@0:             ++level2;
michael@0: 
michael@0:             /* copy the folding result to fold2[] */
michael@0:             if(length<=UCASE_MAX_STRING_LENGTH) {
michael@0:                 u_memcpy(fold2, p, length);
michael@0:             } else {
michael@0:                 int32_t i=0;
michael@0:                 U16_APPEND_UNSAFE(fold2, i, length);
michael@0:                 length=i;
michael@0:             }
michael@0: 
michael@0:             /* set next level pointers to case folding */
michael@0:             start2=s2=fold2;
michael@0:             limit2=fold2+length;
michael@0: 
michael@0:             /* get ready to read from decomposition, continue with loop */
michael@0:             c2=-1;
michael@0:             continue;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * no decomposition/case folding, max level for both sides:
michael@0:          * return difference result
michael@0:          *
michael@0:          * code point order comparison must not just return cp1-cp2
michael@0:          * because when single surrogates are present then the surrogate pairs
michael@0:          * that formed cp1 and cp2 may be from different string indexes
michael@0:          *
michael@0:          * example: { d800 d800 dc01 } vs. { d800 dc00 }, compare at second code units
michael@0:          * c1=d800 cp1=10001 c2=dc00 cp2=10000
michael@0:          * cp1-cp2>0 but c1-c2<0 and in fact in UTF-32 it is { d800 10001 } < { 10000 }
michael@0:          *
michael@0:          * therefore, use same fix-up as in ustring.c/uprv_strCompare()
michael@0:          * except: uprv_strCompare() fetches c=*s while this functions fetches c=*s++
michael@0:          * so we have slightly different pointer/start/limit comparisons here
michael@0:          */
michael@0: 
michael@0:         if(c1>=0xd800 && c2>=0xd800 && (options&U_COMPARE_CODE_POINT_ORDER)) {
michael@0:             /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
michael@0:             if(
michael@0:                 (c1<=0xdbff && s1!=limit1 && U16_IS_TRAIL(*s1)) ||
michael@0:                 (U16_IS_TRAIL(c1) && start1!=(s1-1) && U16_IS_LEAD(*(s1-2)))
michael@0:             ) {
michael@0:                 /* part of a surrogate pair, leave >=d800 */
michael@0:             } else {
michael@0:                 /* BMP code point - may be surrogate code point - make <d800 */
michael@0:                 c1-=0x2800;
michael@0:             }
michael@0: 
michael@0:             if(
michael@0:                 (c2<=0xdbff && s2!=limit2 && U16_IS_TRAIL(*s2)) ||
michael@0:                 (U16_IS_TRAIL(c2) && start2!=(s2-1) && U16_IS_LEAD(*(s2-2)))
michael@0:             ) {
michael@0:                 /* part of a surrogate pair, leave >=d800 */
michael@0:             } else {
michael@0:                 /* BMP code point - may be surrogate code point - make <d800 */
michael@0:                 c2-=0x2800;
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         return c1-c2;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* public API functions */
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: u_strCaseCompare(const UChar *s1, int32_t length1,
michael@0:                  const UChar *s2, int32_t length2,
michael@0:                  uint32_t options,
michael@0:                  UErrorCode *pErrorCode) {
michael@0:     /* argument checking */
michael@0:     if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(s1==NULL || length1<-1 || s2==NULL || length2<-1) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     return u_strcmpFold(s1, length1, s2, length2,
michael@0:                         options|U_COMPARE_IGNORE_CASE,
michael@0:                         pErrorCode);
michael@0: }
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: u_strcasecmp(const UChar *s1, const UChar *s2, uint32_t options) {
michael@0:     UErrorCode errorCode=U_ZERO_ERROR;
michael@0:     return u_strcmpFold(s1, -1, s2, -1,
michael@0:                         options|U_COMPARE_IGNORE_CASE,
michael@0:                         &errorCode);
michael@0: }
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: u_memcasecmp(const UChar *s1, const UChar *s2, int32_t length, uint32_t options) {
michael@0:     UErrorCode errorCode=U_ZERO_ERROR;
michael@0:     return u_strcmpFold(s1, length, s2, length,
michael@0:                         options|U_COMPARE_IGNORE_CASE,
michael@0:                         &errorCode);
michael@0: }
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: u_strncasecmp(const UChar *s1, const UChar *s2, int32_t n, uint32_t options) {
michael@0:     UErrorCode errorCode=U_ZERO_ERROR;
michael@0:     return u_strcmpFold(s1, n, s2, n,
michael@0:                         options|(U_COMPARE_IGNORE_CASE|_STRNCMP_STYLE),
michael@0:                         &errorCode);
michael@0: }