U_CAPI int32_t U_EXPORT2
unorm_compare(const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
UErrorCode *pErrorCode) {
UChar fcd1[300], fcd2[300];
UChar *d1, *d2;
const UnicodeSet *nx;
UNormalizationMode mode;
int32_t normOptions;
int32_t result;
/* argument checking */
if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
return 0;
}
if(s1==0 || length1<-1 || s2==0 || length2<-1) {
*pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
if(!unorm_haveData(pErrorCode)) {
return 0;
}
if(!uprv_haveProperties(pErrorCode)) {
return 0;
}
normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT);
nx=unorm_getNX(normOptions, pErrorCode);
if(U_FAILURE(*pErrorCode)) {
return 0;
}
d1=d2=0;
options|=_COMPARE_EQUIV;
result=0;
/*
* UAX #21 Case Mappings, as fixed for Unicode version 4
* (see Jitterbug 2021), defines a canonical caseless match as
*
* A string X is a canonical caseless match
* for a string Y if and only if
* NFD(toCasefold(NFD(X))) = NFD(toCasefold(NFD(Y)))
*
* For better performance, we check for FCD (or let the caller tell us that
* both strings are in FCD) for the inner normalization.
* BasicNormalizerTest::FindFoldFCDExceptions() makes sure that
* case-folding preserves the FCD-ness of a string.
* The outer normalization is then only performed by unorm_cmpEquivFold()
* when there is a difference.
*
* Exception: When using the Turkic case-folding option, we do perform
* full NFD first. This is because in the Turkic case precomposed characters
* with 0049 capital I or 0069 small i fold differently whether they
* are first decomposed or not, so an FCD check - a check only for
* canonical order - is not sufficient.
*/
if(options&U_FOLD_CASE_EXCLUDE_SPECIAL_I) {
mode=UNORM_NFD;
options&=~UNORM_INPUT_IS_FCD;
} else {
mode=UNORM_FCD;
}
if(!(options&UNORM_INPUT_IS_FCD)) {
int32_t _len1, _len2;
UBool isFCD1, isFCD2;
// check if s1 and/or s2 fulfill the FCD conditions
isFCD1= UNORM_YES==unorm_internalQuickCheck(s1, length1, mode, TRUE, nx, pErrorCode);
isFCD2= UNORM_YES==unorm_internalQuickCheck(s2, length2, mode, TRUE, nx, pErrorCode);
if(U_FAILURE(*pErrorCode)) {
return 0;
}
/*
* ICU 2.4 had a further optimization:
* If both strings were not in FCD, then they were both NFD'ed,
* and the _COMPARE_EQUIV option was turned off.
* It is not entirely clear that this is valid with the current
* definition of the canonical caseless match.
* Therefore, ICU 2.6 removes that optimization.
*/
if(!isFCD1) {
_len1=unorm_internalNormalizeWithNX(fcd1, LENGTHOF(fcd1),
s1, length1,
mode, normOptions, nx,
pErrorCode);
if(*pErrorCode!=U_BUFFER_OVERFLOW_ERROR) {
s1=fcd1;
} else {
d1=(UChar *)uprv_malloc(_len1*U_SIZEOF_UCHAR);
if(d1==0) {
*pErrorCode=U_MEMORY_ALLOCATION_ERROR;
goto cleanup;
}
*pErrorCode=U_ZERO_ERROR;
_len1=unorm_internalNormalizeWithNX(d1, _len1,
s1, length1,
mode, normOptions, nx,
pErrorCode);
if(U_FAILURE(*pErrorCode)) {
goto cleanup;
}
s1=d1;
}
length1=_len1;
}
if(!isFCD2) {
_len2=unorm_internalNormalizeWithNX(fcd2, LENGTHOF(fcd2),
s2, length2,
mode, normOptions, nx,
pErrorCode);
if(*pErrorCode!=U_BUFFER_OVERFLOW_ERROR) {
s2=fcd2;
} else {
d2=(UChar *)uprv_malloc(_len2*U_SIZEOF_UCHAR);
if(d2==0) {
*pErrorCode=U_MEMORY_ALLOCATION_ERROR;
goto cleanup;
}
*pErrorCode=U_ZERO_ERROR;
_len2=unorm_internalNormalizeWithNX(d2, _len2,
s2, length2,
mode, normOptions, nx,
pErrorCode);
if(U_FAILURE(*pErrorCode)) {
goto cleanup;
}
s2=d2;
}
length2=_len2;
}
}
if(U_SUCCESS(*pErrorCode)) {
result=unorm_cmpEquivFold(s1, length1, s2, length2, options, pErrorCode);
}
cleanup:
if(d1!=0) {
uprv_free(d1);
}
if(d2!=0) {
uprv_free(d2);
}
return result;
}
U_CAPI int32_t U_EXPORT2
unorm_compare(const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
UErrorCode *pErrorCode) {
/* argument checking */
if(U_FAILURE(*pErrorCode)) {
return 0;
}
if(s1==0 || length1<-1 || s2==0 || length2<-1) {
*pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
UnicodeString fcd1, fcd2;
int32_t normOptions=(int32_t)(options>>UNORM_COMPARE_NORM_OPTIONS_SHIFT);
options|=_COMPARE_EQUIV;
/*
* UAX #21 Case Mappings, as fixed for Unicode version 4
* (see Jitterbug 2021), defines a canonical caseless match as
*
* A string X is a canonical caseless match
* for a string Y if and only if
* NFD(toCasefold(NFD(X))) = NFD(toCasefold(NFD(Y)))
*
* For better performance, we check for FCD (or let the caller tell us that
* both strings are in FCD) for the inner normalization.
* BasicNormalizerTest::FindFoldFCDExceptions() makes sure that
* case-folding preserves the FCD-ness of a string.
* The outer normalization is then only performed by unorm_cmpEquivFold()
* when there is a difference.
*
* Exception: When using the Turkic case-folding option, we do perform
* full NFD first. This is because in the Turkic case precomposed characters
* with 0049 capital I or 0069 small i fold differently whether they
* are first decomposed or not, so an FCD check - a check only for
* canonical order - is not sufficient.
*/
if(!(options&UNORM_INPUT_IS_FCD) || (options&U_FOLD_CASE_EXCLUDE_SPECIAL_I)) {
const Normalizer2 *n2;
if(options&U_FOLD_CASE_EXCLUDE_SPECIAL_I) {
n2=Normalizer2Factory::getNFDInstance(*pErrorCode);
} else {
n2=Normalizer2Factory::getFCDInstance(*pErrorCode);
}
if (U_FAILURE(*pErrorCode)) {
return 0;
}
// check if s1 and/or s2 fulfill the FCD conditions
const UnicodeSet *uni32;
if(normOptions&UNORM_UNICODE_3_2) {
uni32=uniset_getUnicode32Instance(*pErrorCode);
} else {
uni32=NULL; // unused
}
FilteredNormalizer2 fn2(*n2, *uni32);
if(normOptions&UNORM_UNICODE_3_2) {
n2=&fn2;
}
UnicodeString str1(length1<0, s1, length1);
UnicodeString str2(length2<0, s2, length2);
int32_t spanQCYes1=n2->spanQuickCheckYes(str1, *pErrorCode);
int32_t spanQCYes2=n2->spanQuickCheckYes(str2, *pErrorCode);
if(U_FAILURE(*pErrorCode)) {
return 0;
}
/*
* ICU 2.4 had a further optimization:
* If both strings were not in FCD, then they were both NFD'ed,
* and the _COMPARE_EQUIV option was turned off.
* It is not entirely clear that this is valid with the current
* definition of the canonical caseless match.
* Therefore, ICU 2.6 removes that optimization.
*/
if(spanQCYes1<str1.length()) {
UnicodeString unnormalized=str1.tempSubString(spanQCYes1);
fcd1.setTo(FALSE, str1.getBuffer(), spanQCYes1);
n2->normalizeSecondAndAppend(fcd1, unnormalized, *pErrorCode);
s1=fcd1.getBuffer();
length1=fcd1.length();
}
if(spanQCYes2<str2.length()) {
UnicodeString unnormalized=str2.tempSubString(spanQCYes2);
fcd2.setTo(FALSE, str2.getBuffer(), spanQCYes2);
n2->normalizeSecondAndAppend(fcd2, unnormalized, *pErrorCode);
s2=fcd2.getBuffer();
length2=fcd2.length();
}
}
if(U_SUCCESS(*pErrorCode)) {
return unorm_cmpEquivFold(s1, length1, s2, length2, options, pErrorCode);
} else {
return 0;
}
}
