godot/thirdparty/icu4c/common/ustrcase.cpp

1901 lines
62 KiB
C++

// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
*******************************************************************************
*
* Copyright (C) 2001-2015, International Business Machines
* Corporation and others. All Rights Reserved.
*
*******************************************************************************
* file name: ustrcase.cpp
* encoding: UTF-8
* tab size: 8 (not used)
* indentation:4
*
* created on: 2002feb20
* created by: Markus W. Scherer
*
* Implementation file for string casing C API functions.
* Uses functions from uchar.c for basic functionality that requires access
* to the Unicode Character Database (uprops.dat).
*/
#include "unicode/utypes.h"
#include "unicode/brkiter.h"
#include "unicode/casemap.h"
#include "unicode/edits.h"
#include "unicode/stringoptions.h"
#include "unicode/ustring.h"
#include "unicode/ucasemap.h"
#include "unicode/ubrk.h"
#include "unicode/utf.h"
#include "unicode/utf16.h"
#include "cmemory.h"
#include "ucase.h"
#include "ucasemap_imp.h"
#include "ustr_imp.h"
#include "uassert.h"
/**
* Code point for COMBINING ACUTE ACCENT
* @internal
*/
#define ACUTE u'\u0301'
U_NAMESPACE_BEGIN
namespace {
int32_t checkOverflowAndEditsError(int32_t destIndex, int32_t destCapacity,
Edits *edits, UErrorCode &errorCode) {
if (U_SUCCESS(errorCode)) {
if (destIndex > destCapacity) {
errorCode = U_BUFFER_OVERFLOW_ERROR;
} else if (edits != NULL) {
edits->copyErrorTo(errorCode);
}
}
return destIndex;
}
/* Appends a full case mapping result, see UCASE_MAX_STRING_LENGTH. */
inline int32_t
appendResult(UChar *dest, int32_t destIndex, int32_t destCapacity,
int32_t result, const UChar *s,
int32_t cpLength, uint32_t options, icu::Edits *edits) {
UChar32 c;
int32_t length;
/* decode the result */
if(result<0) {
/* (not) original code point */
if(edits!=NULL) {
edits->addUnchanged(cpLength);
}
if(options & U_OMIT_UNCHANGED_TEXT) {
return destIndex;
}
c=~result;
if(destIndex<destCapacity && c<=0xffff) { // BMP slightly-fastpath
dest[destIndex++]=(UChar)c;
return destIndex;
}
length=cpLength;
} else {
if(result<=UCASE_MAX_STRING_LENGTH) {
c=U_SENTINEL;
length=result;
} else if(destIndex<destCapacity && result<=0xffff) { // BMP slightly-fastpath
dest[destIndex++]=(UChar)result;
if(edits!=NULL) {
edits->addReplace(cpLength, 1);
}
return destIndex;
} else {
c=result;
length=U16_LENGTH(c);
}
if(edits!=NULL) {
edits->addReplace(cpLength, length);
}
}
if(length>(INT32_MAX-destIndex)) {
return -1; // integer overflow
}
if(destIndex<destCapacity) {
/* append the result */
if(c>=0) {
/* code point */
UBool isError=false;
U16_APPEND(dest, destIndex, destCapacity, c, isError);
if(isError) {
/* overflow, nothing written */
destIndex+=length;
}
} else {
/* string */
if((destIndex+length)<=destCapacity) {
while(length>0) {
dest[destIndex++]=*s++;
--length;
}
} else {
/* overflow */
destIndex+=length;
}
}
} else {
/* preflight */
destIndex+=length;
}
return destIndex;
}
inline int32_t
appendUChar(UChar *dest, int32_t destIndex, int32_t destCapacity, UChar c) {
if(destIndex<destCapacity) {
dest[destIndex]=c;
} else if(destIndex==INT32_MAX) {
return -1; // integer overflow
}
return destIndex+1;
}
int32_t
appendNonEmptyUnchanged(UChar *dest, int32_t destIndex, int32_t destCapacity,
const UChar *s, int32_t length, uint32_t options, icu::Edits *edits) {
if(edits!=NULL) {
edits->addUnchanged(length);
}
if(options & U_OMIT_UNCHANGED_TEXT) {
return destIndex;
}
if(length>(INT32_MAX-destIndex)) {
return -1; // integer overflow
}
if((destIndex+length)<=destCapacity) {
u_memcpy(dest+destIndex, s, length);
}
return destIndex + length;
}
inline int32_t
appendUnchanged(UChar *dest, int32_t destIndex, int32_t destCapacity,
const UChar *s, int32_t length, uint32_t options, icu::Edits *edits) {
if (length <= 0) {
return destIndex;
}
return appendNonEmptyUnchanged(dest, destIndex, destCapacity, s, length, options, edits);
}
UChar32 U_CALLCONV
utf16_caseContextIterator(void *context, int8_t dir) {
UCaseContext *csc=(UCaseContext *)context;
UChar32 c;
if(dir<0) {
/* reset for backward iteration */
csc->index=csc->cpStart;
csc->dir=dir;
} else if(dir>0) {
/* reset for forward iteration */
csc->index=csc->cpLimit;
csc->dir=dir;
} else {
/* continue current iteration direction */
dir=csc->dir;
}
if(dir<0) {
if(csc->start<csc->index) {
U16_PREV((const UChar *)csc->p, csc->start, csc->index, c);
return c;
}
} else {
if(csc->index<csc->limit) {
U16_NEXT((const UChar *)csc->p, csc->index, csc->limit, c);
return c;
}
}
return U_SENTINEL;
}
/**
* caseLocale >= 0: Lowercases [srcStart..srcLimit[ but takes context [0..srcLength[ into account.
* caseLocale < 0: Case-folds [srcStart..srcLimit[.
*/
int32_t toLower(int32_t caseLocale, uint32_t options,
UChar *dest, int32_t destCapacity,
const UChar *src, UCaseContext *csc, int32_t srcStart, int32_t srcLimit,
icu::Edits *edits, UErrorCode &errorCode) {
const int8_t *latinToLower;
if (caseLocale == UCASE_LOC_ROOT ||
(caseLocale >= 0 ?
!(caseLocale == UCASE_LOC_TURKISH || caseLocale == UCASE_LOC_LITHUANIAN) :
(options & _FOLD_CASE_OPTIONS_MASK) == U_FOLD_CASE_DEFAULT)) {
latinToLower = LatinCase::TO_LOWER_NORMAL;
} else {
latinToLower = LatinCase::TO_LOWER_TR_LT;
}
const UTrie2 *trie = ucase_getTrie();
int32_t destIndex = 0;
int32_t prev = srcStart;
int32_t srcIndex = srcStart;
for (;;) {
// fast path for simple cases
UChar lead = 0;
while (srcIndex < srcLimit) {
lead = src[srcIndex];
int32_t delta;
if (lead < LatinCase::LONG_S) {
int8_t d = latinToLower[lead];
if (d == LatinCase::EXC) { break; }
++srcIndex;
if (d == 0) { continue; }
delta = d;
} else if (lead >= 0xd800) {
break; // surrogate or higher
} else {
uint16_t props = UTRIE2_GET16_FROM_U16_SINGLE_LEAD(trie, lead);
if (UCASE_HAS_EXCEPTION(props)) { break; }
++srcIndex;
if (!UCASE_IS_UPPER_OR_TITLE(props) || (delta = UCASE_GET_DELTA(props)) == 0) {
continue;
}
}
lead += static_cast<UChar>(delta);
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, srcIndex - 1 - prev, options, edits);
if (destIndex >= 0) {
destIndex = appendUChar(dest, destIndex, destCapacity, lead);
if (edits != nullptr) {
edits->addReplace(1, 1);
}
}
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
prev = srcIndex;
}
if (srcIndex >= srcLimit) {
break;
}
// slow path
int32_t cpStart = srcIndex++;
UChar trail;
UChar32 c;
if (U16_IS_LEAD(lead) && srcIndex < srcLimit && U16_IS_TRAIL(trail = src[srcIndex])) {
c = U16_GET_SUPPLEMENTARY(lead, trail);
++srcIndex;
} else {
c = lead;
}
const UChar *s;
if (caseLocale >= 0) {
csc->cpStart = cpStart;
csc->cpLimit = srcIndex;
c = ucase_toFullLower(c, utf16_caseContextIterator, csc, &s, caseLocale);
} else {
c = ucase_toFullFolding(c, &s, options);
}
if (c >= 0) {
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, cpStart - prev, options, edits);
if (destIndex >= 0) {
destIndex = appendResult(dest, destIndex, destCapacity, c, s,
srcIndex - cpStart, options, edits);
}
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
prev = srcIndex;
}
}
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, srcIndex - prev, options, edits);
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
return destIndex;
}
int32_t toUpper(int32_t caseLocale, uint32_t options,
UChar *dest, int32_t destCapacity,
const UChar *src, UCaseContext *csc, int32_t srcLength,
icu::Edits *edits, UErrorCode &errorCode) {
const int8_t *latinToUpper;
if (caseLocale == UCASE_LOC_TURKISH) {
latinToUpper = LatinCase::TO_UPPER_TR;
} else {
latinToUpper = LatinCase::TO_UPPER_NORMAL;
}
const UTrie2 *trie = ucase_getTrie();
int32_t destIndex = 0;
int32_t prev = 0;
int32_t srcIndex = 0;
for (;;) {
// fast path for simple cases
UChar lead = 0;
while (srcIndex < srcLength) {
lead = src[srcIndex];
int32_t delta;
if (lead < LatinCase::LONG_S) {
int8_t d = latinToUpper[lead];
if (d == LatinCase::EXC) { break; }
++srcIndex;
if (d == 0) { continue; }
delta = d;
} else if (lead >= 0xd800) {
break; // surrogate or higher
} else {
uint16_t props = UTRIE2_GET16_FROM_U16_SINGLE_LEAD(trie, lead);
if (UCASE_HAS_EXCEPTION(props)) { break; }
++srcIndex;
if (UCASE_GET_TYPE(props) != UCASE_LOWER || (delta = UCASE_GET_DELTA(props)) == 0) {
continue;
}
}
lead += static_cast<UChar>(delta);
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, srcIndex - 1 - prev, options, edits);
if (destIndex >= 0) {
destIndex = appendUChar(dest, destIndex, destCapacity, lead);
if (edits != nullptr) {
edits->addReplace(1, 1);
}
}
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
prev = srcIndex;
}
if (srcIndex >= srcLength) {
break;
}
// slow path
int32_t cpStart;
csc->cpStart = cpStart = srcIndex++;
UChar trail;
UChar32 c;
if (U16_IS_LEAD(lead) && srcIndex < srcLength && U16_IS_TRAIL(trail = src[srcIndex])) {
c = U16_GET_SUPPLEMENTARY(lead, trail);
++srcIndex;
} else {
c = lead;
}
csc->cpLimit = srcIndex;
const UChar *s;
c = ucase_toFullUpper(c, utf16_caseContextIterator, csc, &s, caseLocale);
if (c >= 0) {
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, cpStart - prev, options, edits);
if (destIndex >= 0) {
destIndex = appendResult(dest, destIndex, destCapacity, c, s,
srcIndex - cpStart, options, edits);
}
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
prev = srcIndex;
}
}
destIndex = appendUnchanged(dest, destIndex, destCapacity,
src + prev, srcIndex - prev, options, edits);
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
return destIndex;
}
} // namespace
U_NAMESPACE_END
U_NAMESPACE_USE
#if !UCONFIG_NO_BREAK_ITERATION
namespace {
/**
* Input: c is a letter I with or without acute accent.
* start is the index in src after c, and is less than segmentLimit.
* If a plain i/I is followed by a plain j/J,
* or an i/I with acute (precomposed or decomposed) is followed by a j/J with acute,
* then we output accordingly.
*
* @return the src index after the titlecased sequence, or the start index if no Dutch IJ
*/
int32_t maybeTitleDutchIJ(const UChar *src, UChar32 c, int32_t start, int32_t segmentLimit,
UChar *dest, int32_t &destIndex, int32_t destCapacity, uint32_t options,
icu::Edits *edits) {
U_ASSERT(start < segmentLimit);
int32_t index = start;
bool withAcute = false;
// If the conditions are met, then the following variables tell us what to output.
int32_t unchanged1 = 0; // code units before the j, or the whole sequence (0..3)
bool doTitleJ = false; // true if the j needs to be titlecased
int32_t unchanged2 = 0; // after the j (0 or 1)
// next character after the first letter
UChar c2 = src[index++];
// Is the first letter an i/I with accent?
if (c == u'I') {
if (c2 == ACUTE) {
withAcute = true;
unchanged1 = 1;
if (index == segmentLimit) { return start; }
c2 = src[index++];
}
} else { // Í
withAcute = true;
}
// Is the next character a j/J?
if (c2 == u'j') {
doTitleJ = true;
} else if (c2 == u'J') {
++unchanged1;
} else {
return start;
}
// A plain i/I must be followed by a plain j/J.
// An i/I with acute must be followed by a j/J with acute.
if (withAcute) {
if (index == segmentLimit || src[index++] != ACUTE) { return start; }
if (doTitleJ) {
unchanged2 = 1;
} else {
++unchanged1;
}
}
// There must not be another combining mark.
if (index < segmentLimit) {
int32_t cp;
int32_t i = index;
U16_NEXT(src, i, segmentLimit, cp);
uint32_t typeMask = U_GET_GC_MASK(cp);
if ((typeMask & U_GC_M_MASK) != 0) {
return start;
}
}
// Output the rest of the Dutch IJ.
destIndex = appendUnchanged(dest, destIndex, destCapacity, src + start, unchanged1, options, edits);
start += unchanged1;
if (doTitleJ) {
destIndex = appendUChar(dest, destIndex, destCapacity, u'J');
if (edits != nullptr) {
edits->addReplace(1, 1);
}
++start;
}
destIndex = appendUnchanged(dest, destIndex, destCapacity, src + start, unchanged2, options, edits);
U_ASSERT(start + unchanged2 == index);
return index;
}
} // namespace
U_CFUNC int32_t U_CALLCONV
ustrcase_internalToTitle(int32_t caseLocale, uint32_t options, BreakIterator *iter,
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
icu::Edits *edits,
UErrorCode &errorCode) {
if (!ustrcase_checkTitleAdjustmentOptions(options, errorCode)) {
return 0;
}
/* set up local variables */
UCaseContext csc=UCASECONTEXT_INITIALIZER;
csc.p=(void *)src;
csc.limit=srcLength;
int32_t destIndex=0;
int32_t prev=0;
bool isFirstIndex=true;
/* titlecasing loop */
while(prev<srcLength) {
/* find next index where to titlecase */
int32_t index;
if(isFirstIndex) {
isFirstIndex=false;
index=iter->first();
} else {
index=iter->next();
}
if(index==UBRK_DONE || index>srcLength) {
index=srcLength;
}
/*
* Segment [prev..index[ into 3 parts:
* a) skipped characters (copy as-is) [prev..titleStart[
* b) first letter (titlecase) [titleStart..titleLimit[
* c) subsequent characters (lowercase) [titleLimit..index[
*/
if(prev<index) {
// Find and copy skipped characters [prev..titleStart[
int32_t titleStart=prev;
int32_t titleLimit=prev;
UChar32 c;
U16_NEXT(src, titleLimit, index, c);
if ((options&U_TITLECASE_NO_BREAK_ADJUSTMENT)==0) {
// Adjust the titlecasing index to the next cased character,
// or to the next letter/number/symbol/private use.
// Stop with titleStart<titleLimit<=index
// if there is a character to be titlecased,
// or else stop with titleStart==titleLimit==index.
bool toCased = (options&U_TITLECASE_ADJUST_TO_CASED) != 0;
while (toCased ? UCASE_NONE==ucase_getType(c) : !ustrcase_isLNS(c)) {
titleStart=titleLimit;
if(titleLimit==index) {
break;
}
U16_NEXT(src, titleLimit, index, c);
}
if (prev < titleStart) {
destIndex=appendUnchanged(dest, destIndex, destCapacity,
src+prev, titleStart-prev, options, edits);
if(destIndex<0) {
errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
}
}
if(titleStart<titleLimit) {
/* titlecase c which is from [titleStart..titleLimit[ */
csc.cpStart=titleStart;
csc.cpLimit=titleLimit;
const UChar *s;
c=ucase_toFullTitle(c, utf16_caseContextIterator, &csc, &s, caseLocale);
destIndex=appendResult(dest, destIndex, destCapacity, c, s,
titleLimit-titleStart, options, edits);
if(destIndex<0) {
errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
/* Special case Dutch IJ titlecasing */
if (titleStart+1 < index &&
caseLocale == UCASE_LOC_DUTCH) {
if (c < 0) {
c = ~c;
}
if (c == u'I' || c == u'Í') {
titleLimit = maybeTitleDutchIJ(src, c, titleStart + 1, index,
dest, destIndex, destCapacity, options,
edits);
}
}
/* lowercase [titleLimit..index[ */
if(titleLimit<index) {
if((options&U_TITLECASE_NO_LOWERCASE)==0) {
/* Normal operation: Lowercase the rest of the word. */
destIndex+=
toLower(
caseLocale, options,
dest+destIndex, destCapacity-destIndex,
src, &csc, titleLimit, index,
edits, errorCode);
if(errorCode==U_BUFFER_OVERFLOW_ERROR) {
errorCode=U_ZERO_ERROR;
}
if(U_FAILURE(errorCode)) {
return destIndex;
}
} else {
/* Optionally just copy the rest of the word unchanged. */
destIndex=appendUnchanged(dest, destIndex, destCapacity,
src+titleLimit, index-titleLimit, options, edits);
if(destIndex<0) {
errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
}
}
}
}
prev=index;
}
return checkOverflowAndEditsError(destIndex, destCapacity, edits, errorCode);
}
#endif // !UCONFIG_NO_BREAK_ITERATION
U_NAMESPACE_BEGIN
namespace GreekUpper {
// Data generated by prototype code, see
// https://icu.unicode.org/design/case/greek-upper
// TODO: Move this data into ucase.icu.
static const uint16_t data0370[] = {
// U+0370..03FF
0x0370,
0x0370,
0x0372,
0x0372,
0,
0,
0x0376,
0x0376,
0,
0,
0x037A,
0x03FD,
0x03FE,
0x03FF,
0,
0x037F,
0,
0,
0,
0,
0,
0,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0,
0x039F | HAS_VOWEL | HAS_ACCENT,
0,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x0391 | HAS_VOWEL,
0x0392,
0x0393,
0x0394,
0x0395 | HAS_VOWEL,
0x0396,
0x0397 | HAS_VOWEL,
0x0398,
0x0399 | HAS_VOWEL,
0x039A,
0x039B,
0x039C,
0x039D,
0x039E,
0x039F | HAS_VOWEL,
0x03A0,
0x03A1,
0,
0x03A3,
0x03A4,
0x03A5 | HAS_VOWEL,
0x03A6,
0x03A7,
0x03A8,
0x03A9 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_DIALYTIKA,
0x03A5 | HAS_VOWEL | HAS_DIALYTIKA,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x0391 | HAS_VOWEL,
0x0392,
0x0393,
0x0394,
0x0395 | HAS_VOWEL,
0x0396,
0x0397 | HAS_VOWEL,
0x0398,
0x0399 | HAS_VOWEL,
0x039A,
0x039B,
0x039C,
0x039D,
0x039E,
0x039F | HAS_VOWEL,
0x03A0,
0x03A1,
0x03A3,
0x03A3,
0x03A4,
0x03A5 | HAS_VOWEL,
0x03A6,
0x03A7,
0x03A8,
0x03A9 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_DIALYTIKA,
0x03A5 | HAS_VOWEL | HAS_DIALYTIKA,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03CF,
0x0392,
0x0398,
0x03D2,
0x03D2 | HAS_ACCENT,
0x03D2 | HAS_DIALYTIKA,
0x03A6,
0x03A0,
0x03CF,
0x03D8,
0x03D8,
0x03DA,
0x03DA,
0x03DC,
0x03DC,
0x03DE,
0x03DE,
0x03E0,
0x03E0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0x039A,
0x03A1,
0x03F9,
0x037F,
0x03F4,
0x0395 | HAS_VOWEL,
0,
0x03F7,
0x03F7,
0x03F9,
0x03FA,
0x03FA,
0x03FC,
0x03FD,
0x03FE,
0x03FF,
};
static const uint16_t data1F00[] = {
// U+1F00..1FFF
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL,
0x0395 | HAS_VOWEL,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0,
0,
0x0395 | HAS_VOWEL,
0x0395 | HAS_VOWEL,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0,
0,
0x0397 | HAS_VOWEL,
0x0397 | HAS_VOWEL,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL,
0x0397 | HAS_VOWEL,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL,
0x039F | HAS_VOWEL,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0,
0,
0x039F | HAS_VOWEL,
0x039F | HAS_VOWEL,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0,
0,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0,
0x03A5 | HAS_VOWEL,
0,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL,
0x03A9 | HAS_VOWEL,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL,
0x03A9 | HAS_VOWEL,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0,
0,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_ACCENT,
0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0,
0x0399 | HAS_VOWEL,
0,
0,
0,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0395 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_ACCENT,
0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0,
0,
0,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0,
0,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0x0399 | HAS_VOWEL | HAS_ACCENT,
0,
0,
0,
0,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x03A1,
0x03A1,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A5 | HAS_VOWEL | HAS_ACCENT,
0x03A1,
0,
0,
0,
0,
0,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x039F | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_ACCENT,
0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
0,
0,
0,
};
// U+2126 Ohm sign
static const uint16_t data2126 = 0x03A9 | HAS_VOWEL;
uint32_t getLetterData(UChar32 c) {
if (c < 0x370 || 0x2126 < c || (0x3ff < c && c < 0x1f00)) {
return 0;
} else if (c <= 0x3ff) {
return data0370[c - 0x370];
} else if (c <= 0x1fff) {
return data1F00[c - 0x1f00];
} else if (c == 0x2126) {
return data2126;
} else {
return 0;
}
}
uint32_t getDiacriticData(UChar32 c) {
switch (c) {
case 0x0300: // varia
case 0x0301: // tonos = oxia
case 0x0342: // perispomeni
case 0x0302: // circumflex can look like perispomeni
case 0x0303: // tilde can look like perispomeni
case 0x0311: // inverted breve can look like perispomeni
return HAS_ACCENT;
case 0x0308: // dialytika = diaeresis
return HAS_COMBINING_DIALYTIKA;
case 0x0344: // dialytika tonos
return HAS_COMBINING_DIALYTIKA | HAS_ACCENT;
case 0x0345: // ypogegrammeni = iota subscript
return HAS_YPOGEGRAMMENI;
case 0x0304: // macron
case 0x0306: // breve
case 0x0313: // comma above
case 0x0314: // reversed comma above
case 0x0343: // koronis
return HAS_OTHER_GREEK_DIACRITIC;
default:
return 0;
}
}
UBool isFollowedByCasedLetter(const UChar *s, int32_t i, int32_t length) {
while (i < length) {
UChar32 c;
U16_NEXT(s, i, length, c);
int32_t type = ucase_getTypeOrIgnorable(c);
if ((type & UCASE_IGNORABLE) != 0) {
// Case-ignorable, continue with the loop.
} else if (type != UCASE_NONE) {
return true; // Followed by cased letter.
} else {
return false; // Uncased and not case-ignorable.
}
}
return false; // Not followed by cased letter.
}
/**
* Greek string uppercasing with a state machine.
* Probably simpler than a stateless function that has to figure out complex context-before
* for each character.
* TODO: Try to re-consolidate one way or another with the non-Greek function.
*/
int32_t toUpper(uint32_t options,
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
Edits *edits,
UErrorCode &errorCode) {
int32_t destIndex=0;
uint32_t state = 0;
for (int32_t i = 0; i < srcLength;) {
int32_t nextIndex = i;
UChar32 c;
U16_NEXT(src, nextIndex, srcLength, c);
uint32_t nextState = 0;
int32_t type = ucase_getTypeOrIgnorable(c);
if ((type & UCASE_IGNORABLE) != 0) {
// c is case-ignorable
nextState |= (state & AFTER_CASED);
} else if (type != UCASE_NONE) {
// c is cased
nextState |= AFTER_CASED;
}
uint32_t data = getLetterData(c);
if (data > 0) {
uint32_t upper = data & UPPER_MASK;
// Add a dialytika to this iota or ypsilon vowel
// if we removed a tonos from the previous vowel,
// and that previous vowel did not also have (or gain) a dialytika.
// Adding one only to the final vowel in a longer sequence
// (which does not occur in normal writing) would require lookahead.
// Set the same flag as for preserving an existing dialytika.
if ((data & HAS_VOWEL) != 0 && (state & AFTER_VOWEL_WITH_ACCENT) != 0 &&
(upper == 0x399 || upper == 0x3A5)) {
data |= HAS_DIALYTIKA;
}
int32_t numYpogegrammeni = 0; // Map each one to a trailing, spacing, capital iota.
if ((data & HAS_YPOGEGRAMMENI) != 0) {
numYpogegrammeni = 1;
}
// Skip combining diacritics after this Greek letter.
while (nextIndex < srcLength) {
uint32_t diacriticData = getDiacriticData(src[nextIndex]);
if (diacriticData != 0) {
data |= diacriticData;
if ((diacriticData & HAS_YPOGEGRAMMENI) != 0) {
++numYpogegrammeni;
}
++nextIndex;
} else {
break; // not a Greek diacritic
}
}
if ((data & HAS_VOWEL_AND_ACCENT_AND_DIALYTIKA) == HAS_VOWEL_AND_ACCENT) {
nextState |= AFTER_VOWEL_WITH_ACCENT;
}
// Map according to Greek rules.
UBool addTonos = false;
if (upper == 0x397 &&
(data & HAS_ACCENT) != 0 &&
numYpogegrammeni == 0 &&
(state & AFTER_CASED) == 0 &&
!isFollowedByCasedLetter(src, nextIndex, srcLength)) {
// Keep disjunctive "or" with (only) a tonos.
// We use the same "word boundary" conditions as for the Final_Sigma test.
if (i == nextIndex) {
upper = 0x389; // Preserve the precomposed form.
} else {
addTonos = true;
}
} else if ((data & HAS_DIALYTIKA) != 0) {
// Preserve a vowel with dialytika in precomposed form if it exists.
if (upper == 0x399) {
upper = 0x3AA;
data &= ~HAS_EITHER_DIALYTIKA;
} else if (upper == 0x3A5) {
upper = 0x3AB;
data &= ~HAS_EITHER_DIALYTIKA;
}
}
UBool change;
if (edits == nullptr && (options & U_OMIT_UNCHANGED_TEXT) == 0) {
change = true; // common, simple usage
} else {
// Find out first whether we are changing the text.
change = src[i] != upper || numYpogegrammeni > 0;
int32_t i2 = i + 1;
if ((data & HAS_EITHER_DIALYTIKA) != 0) {
change |= i2 >= nextIndex || src[i2] != 0x308;
++i2;
}
if (addTonos) {
change |= i2 >= nextIndex || src[i2] != 0x301;
++i2;
}
int32_t oldLength = nextIndex - i;
int32_t newLength = (i2 - i) + numYpogegrammeni;
change |= oldLength != newLength;
if (change) {
if (edits != NULL) {
edits->addReplace(oldLength, newLength);
}
} else {
if (edits != NULL) {
edits->addUnchanged(oldLength);
}
// Write unchanged text?
change = (options & U_OMIT_UNCHANGED_TEXT) == 0;
}
}
if (change) {
destIndex=appendUChar(dest, destIndex, destCapacity, (UChar)upper);
if (destIndex >= 0 && (data & HAS_EITHER_DIALYTIKA) != 0) {
destIndex=appendUChar(dest, destIndex, destCapacity, 0x308); // restore or add a dialytika
}
if (destIndex >= 0 && addTonos) {
destIndex=appendUChar(dest, destIndex, destCapacity, 0x301);
}
while (destIndex >= 0 && numYpogegrammeni > 0) {
destIndex=appendUChar(dest, destIndex, destCapacity, 0x399);
--numYpogegrammeni;
}
if(destIndex<0) {
errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
}
} else {
const UChar *s;
c=ucase_toFullUpper(c, NULL, NULL, &s, UCASE_LOC_GREEK);
destIndex = appendResult(dest, destIndex, destCapacity, c, s,
nextIndex - i, options, edits);
if (destIndex < 0) {
errorCode = U_INDEX_OUTOFBOUNDS_ERROR;
return 0;
}
}
i = nextIndex;
state = nextState;
}
return destIndex;
}
} // namespace GreekUpper
U_NAMESPACE_END
/* functions available in the common library (for unistr_case.cpp) */
U_CFUNC int32_t U_CALLCONV
ustrcase_internalToLower(int32_t caseLocale, uint32_t options, UCASEMAP_BREAK_ITERATOR_UNUSED
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
icu::Edits *edits,
UErrorCode &errorCode) {
UCaseContext csc=UCASECONTEXT_INITIALIZER;
csc.p=(void *)src;
csc.limit=srcLength;
int32_t destIndex = toLower(
caseLocale, options,
dest, destCapacity,
src, &csc, 0, srcLength,
edits, errorCode);
return checkOverflowAndEditsError(destIndex, destCapacity, edits, errorCode);
}
U_CFUNC int32_t U_CALLCONV
ustrcase_internalToUpper(int32_t caseLocale, uint32_t options, UCASEMAP_BREAK_ITERATOR_UNUSED
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
icu::Edits *edits,
UErrorCode &errorCode) {
int32_t destIndex;
if (caseLocale == UCASE_LOC_GREEK) {
destIndex = GreekUpper::toUpper(options, dest, destCapacity,
src, srcLength, edits, errorCode);
} else {
UCaseContext csc=UCASECONTEXT_INITIALIZER;
csc.p=(void *)src;
csc.limit=srcLength;
destIndex = toUpper(
caseLocale, options,
dest, destCapacity,
src, &csc, srcLength,
edits, errorCode);
}
return checkOverflowAndEditsError(destIndex, destCapacity, edits, errorCode);
}
U_CFUNC int32_t U_CALLCONV
ustrcase_internalFold(int32_t /* caseLocale */, uint32_t options, UCASEMAP_BREAK_ITERATOR_UNUSED
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
icu::Edits *edits,
UErrorCode &errorCode) {
int32_t destIndex = toLower(
-1, options,
dest, destCapacity,
src, nullptr, 0, srcLength,
edits, errorCode);
return checkOverflowAndEditsError(destIndex, destCapacity, edits, errorCode);
}
U_CFUNC int32_t
ustrcase_map(int32_t caseLocale, uint32_t options, UCASEMAP_BREAK_ITERATOR_PARAM
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
UStringCaseMapper *stringCaseMapper,
icu::Edits *edits,
UErrorCode &errorCode) {
int32_t destLength;
/* check argument values */
if(U_FAILURE(errorCode)) {
return 0;
}
if( destCapacity<0 ||
(dest==NULL && destCapacity>0) ||
src==NULL ||
srcLength<-1
) {
errorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
/* get the string length */
if(srcLength==-1) {
srcLength=u_strlen(src);
}
/* check for overlapping source and destination */
if( dest!=NULL &&
((src>=dest && src<(dest+destCapacity)) ||
(dest>=src && dest<(src+srcLength)))
) {
errorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
if (edits != nullptr && (options & U_EDITS_NO_RESET) == 0) {
edits->reset();
}
destLength=stringCaseMapper(caseLocale, options, UCASEMAP_BREAK_ITERATOR
dest, destCapacity, src, srcLength, edits, errorCode);
return u_terminateUChars(dest, destCapacity, destLength, &errorCode);
}
U_CFUNC int32_t
ustrcase_mapWithOverlap(int32_t caseLocale, uint32_t options, UCASEMAP_BREAK_ITERATOR_PARAM
UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
UStringCaseMapper *stringCaseMapper,
UErrorCode &errorCode) {
UChar buffer[300];
UChar *temp;
int32_t destLength;
/* check argument values */
if(U_FAILURE(errorCode)) {
return 0;
}
if( destCapacity<0 ||
(dest==NULL && destCapacity>0) ||
src==NULL ||
srcLength<-1
) {
errorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
/* get the string length */
if(srcLength==-1) {
srcLength=u_strlen(src);
}
/* check for overlapping source and destination */
if( dest!=NULL &&
((src>=dest && src<(dest+destCapacity)) ||
(dest>=src && dest<(src+srcLength)))
) {
/* overlap: provide a temporary destination buffer and later copy the result */
if(destCapacity<=UPRV_LENGTHOF(buffer)) {
/* the stack buffer is large enough */
temp=buffer;
} else {
/* allocate a buffer */
temp=(UChar *)uprv_malloc(destCapacity*U_SIZEOF_UCHAR);
if(temp==NULL) {
errorCode=U_MEMORY_ALLOCATION_ERROR;
return 0;
}
}
} else {
temp=dest;
}
destLength=stringCaseMapper(caseLocale, options, UCASEMAP_BREAK_ITERATOR
temp, destCapacity, src, srcLength, NULL, errorCode);
if(temp!=dest) {
/* copy the result string to the destination buffer */
if (U_SUCCESS(errorCode) && 0 < destLength && destLength <= destCapacity) {
u_memmove(dest, temp, destLength);
}
if(temp!=buffer) {
uprv_free(temp);
}
}
return u_terminateUChars(dest, destCapacity, destLength, &errorCode);
}
/* public API functions */
U_CAPI int32_t U_EXPORT2
u_strFoldCase(UChar *dest, int32_t destCapacity,
const UChar *src, int32_t srcLength,
uint32_t options,
UErrorCode *pErrorCode) {
return ustrcase_mapWithOverlap(
UCASE_LOC_ROOT, options, UCASEMAP_BREAK_ITERATOR_NULL
dest, destCapacity,
src, srcLength,
ustrcase_internalFold, *pErrorCode);
}
U_NAMESPACE_BEGIN
int32_t CaseMap::fold(
uint32_t options,
const UChar *src, int32_t srcLength,
UChar *dest, int32_t destCapacity, Edits *edits,
UErrorCode &errorCode) {
return ustrcase_map(
UCASE_LOC_ROOT, options, UCASEMAP_BREAK_ITERATOR_NULL
dest, destCapacity,
src, srcLength,
ustrcase_internalFold, edits, errorCode);
}
U_NAMESPACE_END
/* case-insensitive string comparisons -------------------------------------- */
/*
* This function is a copy of unorm_cmpEquivFold() minus the parts for
* canonical equivalence.
* Keep the functions in sync, and see there for how this works.
* The duplication is for modularization:
* It makes caseless (but not canonical caseless) matches independent of
* the normalization code.
*/
/* stack element for previous-level source/decomposition pointers */
struct CmpEquivLevel {
const UChar *start, *s, *limit;
};
typedef struct CmpEquivLevel CmpEquivLevel;
/**
* Internal implementation code comparing string with case fold.
* This function is called from u_strcmpFold() and u_caseInsensitivePrefixMatch().
*
* @param s1 input string 1
* @param length1 length of string 1, or -1 (NULL terminated)
* @param s2 input string 2
* @param length2 length of string 2, or -1 (NULL terminated)
* @param options compare options
* @param matchLen1 (output) length of partial prefix match in s1
* @param matchLen2 (output) length of partial prefix match in s2
* @param pErrorCode receives error status
* @return The result of comparison
*/
static int32_t _cmpFold(
const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
int32_t *matchLen1, int32_t *matchLen2,
UErrorCode *pErrorCode) {
int32_t cmpRes = 0;
/* current-level start/limit - s1/s2 as current */
const UChar *start1, *start2, *limit1, *limit2;
/* points to the original start address */
const UChar *org1, *org2;
/* points to the end of match + 1 */
const UChar *m1, *m2;
/* case folding variables */
const UChar *p;
int32_t length;
/* stacks of previous-level start/current/limit */
CmpEquivLevel stack1[2], stack2[2];
/* case folding buffers, only use current-level start/limit */
UChar fold1[UCASE_MAX_STRING_LENGTH+1], fold2[UCASE_MAX_STRING_LENGTH+1];
/* track which is the current level per string */
int32_t level1, level2;
/* current code units, and code points for lookups */
UChar32 c1, c2, cp1, cp2;
/* no argument error checking because this itself is not an API */
/*
* assume that at least the option U_COMPARE_IGNORE_CASE is set
* otherwise this function would have to behave exactly as uprv_strCompare()
*/
if(U_FAILURE(*pErrorCode)) {
return 0;
}
/* initialize */
if(matchLen1) {
U_ASSERT(matchLen2 !=NULL);
*matchLen1=0;
*matchLen2=0;
}
start1=m1=org1=s1;
if(length1==-1) {
limit1=NULL;
} else {
limit1=s1+length1;
}
start2=m2=org2=s2;
if(length2==-1) {
limit2=NULL;
} else {
limit2=s2+length2;
}
level1=level2=0;
c1=c2=-1;
/* comparison loop */
for(;;) {
/*
* here a code unit value of -1 means "get another code unit"
* below it will mean "this source is finished"
*/
if(c1<0) {
/* get next code unit from string 1, post-increment */
for(;;) {
if(s1==limit1 || ((c1=*s1)==0 && (limit1==NULL || (options&_STRNCMP_STYLE)))) {
if(level1==0) {
c1=-1;
break;
}
} else {
++s1;
break;
}
/* reached end of level buffer, pop one level */
do {
--level1;
start1=stack1[level1].start; /*Not uninitialized*/
} while(start1==NULL);
s1=stack1[level1].s; /*Not uninitialized*/
limit1=stack1[level1].limit; /*Not uninitialized*/
}
}
if(c2<0) {
/* get next code unit from string 2, post-increment */
for(;;) {
if(s2==limit2 || ((c2=*s2)==0 && (limit2==NULL || (options&_STRNCMP_STYLE)))) {
if(level2==0) {
c2=-1;
break;
}
} else {
++s2;
break;
}
/* reached end of level buffer, pop one level */
do {
--level2;
start2=stack2[level2].start; /*Not uninitialized*/
} while(start2==NULL);
s2=stack2[level2].s; /*Not uninitialized*/
limit2=stack2[level2].limit; /*Not uninitialized*/
}
}
/*
* compare c1 and c2
* either variable c1, c2 is -1 only if the corresponding string is finished
*/
if(c1==c2) {
const UChar *next1, *next2;
if(c1<0) {
cmpRes=0; /* c1==c2==-1 indicating end of strings */
break;
}
/*
* Note: Move the match positions in both strings at the same time
* only when corresponding code point(s) in the original strings
* are fully consumed. For example, when comparing s1="Fust" and
* s2="Fu\u00dfball", s2[2] is folded into "ss", and s1[2] matches
* the first code point in the case-folded data. But the second "s"
* has no matching code point in s1, so this implementation returns
* 2 as the prefix match length ("Fu").
*/
next1=next2=NULL;
if(level1==0) {
next1=s1;
} else if(s1==limit1) {
/* Note: This implementation only use a single level of stack.
* If this code needs to be changed to use multiple levels
* of stacks, the code above should check if the current
* code is at the end of all stacks.
*/
U_ASSERT(level1==1);
/* is s1 at the end of the current stack? */
next1=stack1[0].s;
}
if (next1!=NULL) {
if(level2==0) {
next2=s2;
} else if(s2==limit2) {
U_ASSERT(level2==1);
/* is s2 at the end of the current stack? */
next2=stack2[0].s;
}
if(next2!=NULL) {
m1=next1;
m2=next2;
}
}
c1=c2=-1; /* make us fetch new code units */
continue;
} else if(c1<0) {
cmpRes=-1; /* string 1 ends before string 2 */
break;
} else if(c2<0) {
cmpRes=1; /* string 2 ends before string 1 */
break;
}
/* c1!=c2 && c1>=0 && c2>=0 */
/* get complete code points for c1, c2 for lookups if either is a surrogate */
cp1=c1;
if(U_IS_SURROGATE(c1)) {
UChar c;
if(U_IS_SURROGATE_LEAD(c1)) {
if(s1!=limit1 && U16_IS_TRAIL(c=*s1)) {
/* advance ++s1; only below if cp1 decomposes/case-folds */
cp1=U16_GET_SUPPLEMENTARY(c1, c);
}
} else /* isTrail(c1) */ {
if(start1<=(s1-2) && U16_IS_LEAD(c=*(s1-2))) {
cp1=U16_GET_SUPPLEMENTARY(c, c1);
}
}
}
cp2=c2;
if(U_IS_SURROGATE(c2)) {
UChar c;
if(U_IS_SURROGATE_LEAD(c2)) {
if(s2!=limit2 && U16_IS_TRAIL(c=*s2)) {
/* advance ++s2; only below if cp2 decomposes/case-folds */
cp2=U16_GET_SUPPLEMENTARY(c2, c);
}
} else /* isTrail(c2) */ {
if(start2<=(s2-2) && U16_IS_LEAD(c=*(s2-2))) {
cp2=U16_GET_SUPPLEMENTARY(c, c2);
}
}
}
/*
* go down one level for each string
* continue with the main loop as soon as there is a real change
*/
if( level1==0 &&
(length=ucase_toFullFolding((UChar32)cp1, &p, options))>=0
) {
/* cp1 case-folds to the code point "length" or to p[length] */
if(U_IS_SURROGATE(c1)) {
if(U_IS_SURROGATE_LEAD(c1)) {
/* advance beyond source surrogate pair if it case-folds */
++s1;
} else /* isTrail(c1) */ {
/*
* we got a supplementary code point when hitting its trail surrogate,
* therefore the lead surrogate must have been the same as in the other string;
* compare this decomposition with the lead surrogate in the other string
* remember that this simulates bulk text replacement:
* the decomposition would replace the entire code point
*/
--s2;
--m2;
c2=*(s2-1);
}
}
/* push current level pointers */
stack1[0].start=start1;
stack1[0].s=s1;
stack1[0].limit=limit1;
++level1;
/* copy the folding result to fold1[] */
if(length<=UCASE_MAX_STRING_LENGTH) {
u_memcpy(fold1, p, length);
} else {
int32_t i=0;
U16_APPEND_UNSAFE(fold1, i, length);
length=i;
}
/* set next level pointers to case folding */
start1=s1=fold1;
limit1=fold1+length;
/* get ready to read from decomposition, continue with loop */
c1=-1;
continue;
}
if( level2==0 &&
(length=ucase_toFullFolding((UChar32)cp2, &p, options))>=0
) {
/* cp2 case-folds to the code point "length" or to p[length] */
if(U_IS_SURROGATE(c2)) {
if(U_IS_SURROGATE_LEAD(c2)) {
/* advance beyond source surrogate pair if it case-folds */
++s2;
} else /* isTrail(c2) */ {
/*
* we got a supplementary code point when hitting its trail surrogate,
* therefore the lead surrogate must have been the same as in the other string;
* compare this decomposition with the lead surrogate in the other string
* remember that this simulates bulk text replacement:
* the decomposition would replace the entire code point
*/
--s1;
--m2;
c1=*(s1-1);
}
}
/* push current level pointers */
stack2[0].start=start2;
stack2[0].s=s2;
stack2[0].limit=limit2;
++level2;
/* copy the folding result to fold2[] */
if(length<=UCASE_MAX_STRING_LENGTH) {
u_memcpy(fold2, p, length);
} else {
int32_t i=0;
U16_APPEND_UNSAFE(fold2, i, length);
length=i;
}
/* set next level pointers to case folding */
start2=s2=fold2;
limit2=fold2+length;
/* get ready to read from decomposition, continue with loop */
c2=-1;
continue;
}
/*
* no decomposition/case folding, max level for both sides:
* return difference result
*
* code point order comparison must not just return cp1-cp2
* because when single surrogates are present then the surrogate pairs
* that formed cp1 and cp2 may be from different string indexes
*
* example: { d800 d800 dc01 } vs. { d800 dc00 }, compare at second code units
* c1=d800 cp1=10001 c2=dc00 cp2=10000
* cp1-cp2>0 but c1-c2<0 and in fact in UTF-32 it is { d800 10001 } < { 10000 }
*
* therefore, use same fix-up as in ustring.c/uprv_strCompare()
* except: uprv_strCompare() fetches c=*s while this functions fetches c=*s++
* so we have slightly different pointer/start/limit comparisons here
*/
if(c1>=0xd800 && c2>=0xd800 && (options&U_COMPARE_CODE_POINT_ORDER)) {
/* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
if(
(c1<=0xdbff && s1!=limit1 && U16_IS_TRAIL(*s1)) ||
(U16_IS_TRAIL(c1) && start1!=(s1-1) && U16_IS_LEAD(*(s1-2)))
) {
/* part of a surrogate pair, leave >=d800 */
} else {
/* BMP code point - may be surrogate code point - make <d800 */
c1-=0x2800;
}
if(
(c2<=0xdbff && s2!=limit2 && U16_IS_TRAIL(*s2)) ||
(U16_IS_TRAIL(c2) && start2!=(s2-1) && U16_IS_LEAD(*(s2-2)))
) {
/* part of a surrogate pair, leave >=d800 */
} else {
/* BMP code point - may be surrogate code point - make <d800 */
c2-=0x2800;
}
}
cmpRes=c1-c2;
break;
}
if(matchLen1) {
*matchLen1=static_cast<int32_t>(m1-org1);
*matchLen2=static_cast<int32_t>(m2-org2);
}
return cmpRes;
}
/* internal function */
U_CFUNC int32_t
u_strcmpFold(const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
UErrorCode *pErrorCode) {
return _cmpFold(s1, length1, s2, length2, options, NULL, NULL, pErrorCode);
}
/* public API functions */
U_CAPI int32_t U_EXPORT2
u_strCaseCompare(const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
UErrorCode *pErrorCode) {
/* argument checking */
if(pErrorCode==0 || U_FAILURE(*pErrorCode)) {
return 0;
}
if(s1==NULL || length1<-1 || s2==NULL || length2<-1) {
*pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
return u_strcmpFold(s1, length1, s2, length2,
options|U_COMPARE_IGNORE_CASE,
pErrorCode);
}
U_CAPI int32_t U_EXPORT2
u_strcasecmp(const UChar *s1, const UChar *s2, uint32_t options) {
UErrorCode errorCode=U_ZERO_ERROR;
return u_strcmpFold(s1, -1, s2, -1,
options|U_COMPARE_IGNORE_CASE,
&errorCode);
}
U_CAPI int32_t U_EXPORT2
u_memcasecmp(const UChar *s1, const UChar *s2, int32_t length, uint32_t options) {
UErrorCode errorCode=U_ZERO_ERROR;
return u_strcmpFold(s1, length, s2, length,
options|U_COMPARE_IGNORE_CASE,
&errorCode);
}
U_CAPI int32_t U_EXPORT2
u_strncasecmp(const UChar *s1, const UChar *s2, int32_t n, uint32_t options) {
UErrorCode errorCode=U_ZERO_ERROR;
return u_strcmpFold(s1, n, s2, n,
options|(U_COMPARE_IGNORE_CASE|_STRNCMP_STYLE),
&errorCode);
}
/* internal API - detect length of shared prefix */
U_CAPI void
u_caseInsensitivePrefixMatch(const UChar *s1, int32_t length1,
const UChar *s2, int32_t length2,
uint32_t options,
int32_t *matchLen1, int32_t *matchLen2,
UErrorCode *pErrorCode) {
_cmpFold(s1, length1, s2, length2, options,
matchLen1, matchLen2, pErrorCode);
}