static jboolean mirror(JNIEnv* env, jobject obj, jcharArray charArray, int start, int count)
{
jchar* data = env->GetCharArrayElements(charArray, NULL);
bool ret = false;
if (data == NULL) {
jniThrowException(env, "java/lang/NullPointerException", NULL);
goto MIRROR_END;
}
if (start < 0 || start > start + count
|| env->GetArrayLength(charArray) < start + count) {
jniThrowException(env, "java/lang/ArrayIndexOutOfBoundsException", NULL);
goto MIRROR_END;
}
for (int i = start; i < start + count; i++) {
// XXX this thinks it knows that surrogates are never mirrored
int c1 = data[i];
int c2 = u_charMirror(c1);
if (c1 != c2) {
data[i] = c2;
ret = true;
}
}
MIRROR_END:
env->ReleaseCharArrayElements(charArray, data, JNI_ABORT);
return ret;
}
String createStringWithMirroredCharacters(StringView string)
{
unsigned length = string.length();
StringBuilder mirroredCharacters;
mirroredCharacters.reserveCapacity(length);
for (unsigned i = 0; i < length; ) {
UChar32 character;
U16_NEXT(string, i, length, character);
mirroredCharacters.append(u_charMirror(character));
}
return mirroredCharacters.toString();
}
void mirrorCharacters(UChar* destination, const UChar* source, int length) const
{
int position = 0;
bool error = false;
// Iterate characters in source and mirror character if needed.
while (position < length) {
UChar32 character;
int nextPosition = position;
U16_NEXT(source, nextPosition, length, character);
character = u_charMirror(character);
U16_APPEND(destination, position, length, character, error);
ASSERT(!error);
position = nextPosition;
}
}
void ComplexTextController::normalizeSpacesAndMirrorChars(const UChar* source, bool rtl, UChar* destination, int length)
{
int position = 0;
bool error = false;
// Iterate characters in source and mirror character if needed.
while (position < length) {
UChar32 character;
int nextPosition = position;
U16_NEXT(source, nextPosition, length, character);
if (Font::treatAsSpace(character))
character = ' ';
else if (Font::treatAsZeroWidthSpace(character))
character = zeroWidthSpace;
else if (rtl)
character = u_charMirror(character);
U16_APPEND(destination, position, length, character, error);
ASSERT(!error);
position = nextPosition;
}
}
static void normalizeSpacesAndMirrorChars(const UChar* source, UChar* destination, int length, HarfBuzzShaperBase::NormalizeMode normalizeMode)
{
int position = 0;
bool error = false;
// Iterate characters in source and mirror character if needed.
while (position < length) {
UChar32 character;
int nextPosition = position;
U16_NEXT(source, nextPosition, length, character);
if (Font::treatAsSpace(character))
character = ' ';
else if (Font::treatAsZeroWidthSpace(character))
character = zeroWidthSpace;
else if (normalizeMode == HarfBuzzShaperBase::NormalizeMirrorChars)
character = u_charMirror(character);
U16_APPEND(destination, position, length, character, error);
ASSERT_UNUSED(error, !error);
position = nextPosition;
}
}
/**
* Performs character mirroring.
*
* @param pTransform Pointer to the <code>UBiDiTransform</code> structure.
* @param pErrorCode Pointer to the error code value.
*
* @return Whether or not this function modifies the text. Besides the return
* value, the caller should also check <code>U_SUCCESS(*pErrorCode)</code>.
*/
static UBool
action_mirror(UBiDiTransform *pTransform, UErrorCode *pErrorCode)
{
UChar32 c;
uint32_t i = 0, j = 0;
if (0 == (pTransform->reorderingOptions & UBIDI_DO_MIRRORING)) {
return FALSE;
}
if (pTransform->destSize < pTransform->srcLength) {
*pErrorCode = U_BUFFER_OVERFLOW_ERROR;
return FALSE;
}
do {
UBool isOdd = ubidi_getLevelAt(pTransform->pBidi, i) & 1;
U16_NEXT(pTransform->src, i, pTransform->srcLength, c);
U16_APPEND_UNSAFE(pTransform->dest, j, isOdd ? u_charMirror(c) : c);
} while (i < pTransform->srcLength);
*pTransform->pDestLength = pTransform->srcLength;
pTransform->reorderingOptions = UBIDI_REORDER_DEFAULT;
return TRUE;
}
std::pair<GlyphData, GlyphPage*> FontGlyphs::glyphDataAndPageForCharacter(const FontDescription& description, UChar32 c, bool mirror, FontDataVariant variant) const
{
ASSERT(isMainThread());
if (variant == AutoVariant) {
if (description.smallCaps() && !primarySimpleFontData(description)->isSVGFont()) {
UChar32 upperC = u_toupper(c);
if (upperC != c) {
c = upperC;
variant = SmallCapsVariant;
} else
variant = NormalVariant;
} else
variant = NormalVariant;
}
if (mirror)
c = u_charMirror(c);
unsigned pageNumber = (c / GlyphPage::size);
GlyphPageTreeNode* node = pageNumber ? m_pages.get(pageNumber) : m_pageZero;
if (!node) {
node = GlyphPageTreeNode::getRootChild(realizeFontDataAt(description, 0), pageNumber);
if (pageNumber)
m_pages.set(pageNumber, node);
else
m_pageZero = node;
}
GlyphPage* page = 0;
if (variant == NormalVariant) {
// Fastest loop, for the common case (normal variant).
while (true) {
page = node->page();
if (page) {
GlyphData data = page->glyphDataForCharacter(c);
if (data.fontData && (data.fontData->platformData().orientation() == Horizontal || data.fontData->isTextOrientationFallback()))
return std::make_pair(data, page);
if (data.fontData) {
if (Font::isCJKIdeographOrSymbol(c)) {
if (!data.fontData->hasVerticalGlyphs()) {
// Use the broken ideograph font data. The broken ideograph font will use the horizontal width of glyphs
// to make sure you get a square (even for broken glyphs like symbols used for punctuation).
variant = BrokenIdeographVariant;
break;
}
#if PLATFORM(COCOA)
else if (data.fontData->platformData().syntheticOblique())
return glyphDataAndPageForCJKCharacterWithoutSyntheticItalic(c, data, page, pageNumber);
#endif
} else
return glyphDataAndPageForNonCJKCharacterWithGlyphOrientation(c, description.nonCJKGlyphOrientation(), data, page, pageNumber);
return std::make_pair(data, page);
}
if (node->isSystemFallback())
break;
}
node = node->getChild(realizeFontDataAt(description, node->level()), pageNumber);
if (pageNumber)
m_pages.set(pageNumber, node);
else
m_pageZero = node;
}
}
if (variant != NormalVariant) {
while (true) {
page = node->page();
if (page) {
GlyphData data = page->glyphDataForCharacter(c);
if (data.fontData) {
// The variantFontData function should not normally return 0.
// But if it does, we will just render the capital letter big.
RefPtr<SimpleFontData> variantFontData = data.fontData->variantFontData(description, variant);
if (!variantFontData)
return std::make_pair(data, page);
GlyphPageTreeNode* variantNode = GlyphPageTreeNode::getRootChild(variantFontData.get(), pageNumber);
GlyphPage* variantPage = variantNode->page();
if (variantPage) {
GlyphData data = variantPage->glyphDataForCharacter(c);
if (data.fontData)
return std::make_pair(data, variantPage);
}
// Do not attempt system fallback off the variantFontData. This is the very unlikely case that
// a font has the lowercase character but the small caps font does not have its uppercase version.
return std::make_pair(variantFontData->missingGlyphData(), page);
}
if (node->isSystemFallback())
break;
}
node = node->getChild(realizeFontDataAt(description, node->level()), pageNumber);
if (pageNumber)
m_pages.set(pageNumber, node);
else
m_pageZero = node;
}
}
ASSERT(page);
ASSERT(node->isSystemFallback());
// System fallback is character-dependent. When we get here, we
// know that the character in question isn't in the system fallback
// font's glyph page. Try to lazily create it here.
UChar codeUnits[2];
int codeUnitsLength;
if (c <= 0xFFFF) {
codeUnits[0] = Font::normalizeSpaces(c);
codeUnitsLength = 1;
} else {
codeUnits[0] = U16_LEAD(c);
codeUnits[1] = U16_TRAIL(c);
codeUnitsLength = 2;
}
const SimpleFontData* originalFontData = primaryFontData(description)->fontDataForCharacter(c);
RefPtr<SimpleFontData> characterFontData = fontCache().systemFallbackForCharacters(description, originalFontData, m_isForPlatformFont, codeUnits, codeUnitsLength);
if (characterFontData) {
if (characterFontData->platformData().orientation() == Vertical && !characterFontData->hasVerticalGlyphs() && Font::isCJKIdeographOrSymbol(c))
variant = BrokenIdeographVariant;
if (variant != NormalVariant)
characterFontData = characterFontData->variantFontData(description, variant);
}
if (characterFontData) {
// Got the fallback glyph and font.
GlyphPage* fallbackPage = GlyphPageTreeNode::getRootChild(characterFontData.get(), pageNumber)->page();
GlyphData data = fallbackPage && fallbackPage->fontDataForCharacter(c) ? fallbackPage->glyphDataForCharacter(c) : characterFontData->missingGlyphData();
// Cache it so we don't have to do system fallback again next time.
if (variant == NormalVariant) {
#if OS(WINCE)
// missingGlyphData returns a null character, which is not suitable for GDI to display.
// Also, sometimes we cannot map a font for the character on WINCE, but GDI can still
// display the character, probably because the font package is not installed correctly.
// So we just always set the glyph to be same as the character, and let GDI solve it.
page->setGlyphDataForCharacter(c, c, characterFontData.get());
characterFontData->setMaxGlyphPageTreeLevel(std::max(characterFontData->maxGlyphPageTreeLevel(), node->level()));
return std::make_pair(page->glyphDataForCharacter(c), page);
#else
page->setGlyphDataForCharacter(c, data.glyph, data.fontData);
data.fontData->setMaxGlyphPageTreeLevel(std::max(data.fontData->maxGlyphPageTreeLevel(), node->level()));
if (!Font::isCJKIdeographOrSymbol(c) && data.fontData->platformData().orientation() != Horizontal && !data.fontData->isTextOrientationFallback())
return glyphDataAndPageForNonCJKCharacterWithGlyphOrientation(c, description.nonCJKGlyphOrientation(), data, fallbackPage, pageNumber);
#endif
}
return std::make_pair(data, page);
}
// Even system fallback can fail; use the missing glyph in that case.
// FIXME: It would be nicer to use the missing glyph from the last resort font instead.
GlyphData data = primarySimpleFontData(description)->missingGlyphData();
if (variant == NormalVariant) {
#if OS(WINCE)
// See comment about WINCE GDI handling near setGlyphDataForCharacter above.
page->setGlyphDataForCharacter(c, c, data.fontData);
data.fontData->setMaxGlyphPageTreeLevel(std::max(data.fontData->maxGlyphPageTreeLevel(), node->level()));
return std::make_pair(page->glyphDataForCharacter(c), page);
#else
page->setGlyphDataForCharacter(c, data.glyph, data.fontData);
data.fontData->setMaxGlyphPageTreeLevel(std::max(data.fontData->maxGlyphPageTreeLevel(), node->level()));
#endif
}
return std::make_pair(data, page);
}
/*
* When we have UBIDI_OUTPUT_REVERSE set on ubidi_writeReordered(), then we
* semantically write RTL runs in reverse and later reverse them again.
* Instead, we actually write them in forward order to begin with.
* However, if the RTL run was to be mirrored, we need to mirror here now
* since the implicit second reversal must not do it.
* It looks strange to do mirroring in LTR output, but it is only because
* we are writing RTL output in reverse.
*/
static int32_t
doWriteForward(const UChar *src, int32_t srcLength,
UChar *dest, int32_t destSize,
uint16_t options,
UErrorCode *pErrorCode) {
/* optimize for several combinations of options */
switch(options&(UBIDI_REMOVE_BIDI_CONTROLS|UBIDI_DO_MIRRORING)) {
case 0: {
/* simply copy the LTR run to the destination */
int32_t length=srcLength;
if(destSize<length) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
return srcLength;
}
do {
*dest++=*src++;
} while(--length>0);
return srcLength;
}
case UBIDI_DO_MIRRORING: {
/* do mirroring */
int32_t i=0, j=0;
UChar32 c;
if(destSize<srcLength) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
return srcLength;
}
do {
UTF_NEXT_CHAR(src, i, srcLength, c);
c=u_charMirror(c);
UTF_APPEND_CHAR_UNSAFE(dest, j, c);
} while(i<srcLength);
return srcLength;
}
case UBIDI_REMOVE_BIDI_CONTROLS: {
/* copy the LTR run and remove any BiDi control characters */
int32_t remaining=destSize;
UChar c;
do {
c=*src++;
if(!IS_BIDI_CONTROL_CHAR(c)) {
if(--remaining<0) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
/* preflight the length */
while(--srcLength>0) {
c=*src++;
if(!IS_BIDI_CONTROL_CHAR(c)) {
--remaining;
}
}
return destSize-remaining;
}
*dest++=c;
}
} while(--srcLength>0);
return destSize-remaining;
}
default: {
/* remove BiDi control characters and do mirroring */
int32_t remaining=destSize;
int32_t i, j=0;
UChar32 c;
do {
i=0;
UTF_NEXT_CHAR(src, i, srcLength, c);
src+=i;
srcLength-=i;
if(!IS_BIDI_CONTROL_CHAR(c)) {
remaining-=i;
if(remaining<0) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
/* preflight the length */
while(srcLength>0) {
c=*src++;
if(!IS_BIDI_CONTROL_CHAR(c)) {
--remaining;
}
--srcLength;
}
return destSize-remaining;
}
c=u_charMirror(c);
UTF_APPEND_CHAR_UNSAFE(dest, j, c);
}
} while(srcLength>0);
return j;
}
} /* end of switch */
}
static int32_t
doWriteReverse(const UChar *src, int32_t srcLength,
UChar *dest, int32_t destSize,
uint16_t options,
UErrorCode *pErrorCode) {
/*
* RTL run -
*
* RTL runs need to be copied to the destination in reverse order
* of code points, not code units, to keep Unicode characters intact.
*
* The general strategy for this is to read the source text
* in backward order, collect all code units for a code point
* (and optionally following combining characters, see below),
* and copy all these code units in ascending order
* to the destination for this run.
*
* Several options request whether combining characters
* should be kept after their base characters,
* whether BiDi control characters should be removed, and
* whether characters should be replaced by their mirror-image
* equivalent Unicode characters.
*/
int32_t i, j;
UChar32 c;
/* optimize for several combinations of options */
switch(options&(UBIDI_REMOVE_BIDI_CONTROLS|UBIDI_DO_MIRRORING|UBIDI_KEEP_BASE_COMBINING)) {
case 0:
/*
* With none of the "complicated" options set, the destination
* run will have the same length as the source run,
* and there is no mirroring and no keeping combining characters
* with their base characters.
*/
if(destSize<srcLength) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
return srcLength;
}
destSize=srcLength;
/* preserve character integrity */
do {
/* i is always after the last code unit known to need to be kept in this segment */
i=srcLength;
/* collect code units for one base character */
UTF_BACK_1(src, 0, srcLength);
/* copy this base character */
j=srcLength;
do {
*dest++=src[j++];
} while(j<i);
} while(srcLength>0);
break;
case UBIDI_KEEP_BASE_COMBINING:
/*
* Here, too, the destination
* run will have the same length as the source run,
* and there is no mirroring.
* We do need to keep combining characters with their base characters.
*/
if(destSize<srcLength) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
return srcLength;
}
destSize=srcLength;
/* preserve character integrity */
do {
/* i is always after the last code unit known to need to be kept in this segment */
i=srcLength;
/* collect code units and modifier letters for one base character */
do {
UTF_PREV_CHAR(src, 0, srcLength, c);
} while(srcLength>0 && IS_COMBINING(u_charType(c)));
/* copy this "user character" */
j=srcLength;
do {
*dest++=src[j++];
} while(j<i);
} while(srcLength>0);
break;
default:
/*
* With several "complicated" options set, this is the most
* general and the slowest copying of an RTL run.
* We will do mirroring, remove BiDi controls, and
* keep combining characters with their base characters
* as requested.
*/
if(!(options&UBIDI_REMOVE_BIDI_CONTROLS)) {
i=srcLength;
} else {
/* we need to find out the destination length of the run,
which will not include the BiDi control characters */
int32_t length=srcLength;
UChar ch;
i=0;
do {
ch=*src++;
if(!IS_BIDI_CONTROL_CHAR(ch)) {
++i;
}
} while(--length>0);
src-=srcLength;
}
if(destSize<i) {
*pErrorCode=U_BUFFER_OVERFLOW_ERROR;
return i;
}
destSize=i;
/* preserve character integrity */
do {
/* i is always after the last code unit known to need to be kept in this segment */
i=srcLength;
/* collect code units for one base character */
UTF_PREV_CHAR(src, 0, srcLength, c);
if(options&UBIDI_KEEP_BASE_COMBINING) {
/* collect modifier letters for this base character */
while(srcLength>0 && IS_COMBINING(u_charType(c))) {
UTF_PREV_CHAR(src, 0, srcLength, c);
}
}
if(options&UBIDI_REMOVE_BIDI_CONTROLS && IS_BIDI_CONTROL_CHAR(c)) {
/* do not copy this BiDi control character */
continue;
}
/* copy this "user character" */
j=srcLength;
if(options&UBIDI_DO_MIRRORING) {
/* mirror only the base character */
int32_t k=0;
c=u_charMirror(c);
UTF_APPEND_CHAR_UNSAFE(dest, k, c);
dest+=k;
j+=k;
}
while(j<i) {
*dest++=src[j++];
}
} while(srcLength>0);
break;
} /* end of switch */
return destSize;
}
static jchar getMirror(JNIEnv* env, jobject obj, jchar c)
{
return u_charMirror(c);
}
HB_BEGIN_DECLS
static hb_codepoint_t hb_icu_get_mirroring (hb_codepoint_t unicode) { return u_charMirror(unicode); }
// Helper sets the character attribute properties and sets up the script table.
// Does not set tops and bottoms.
void SetupBasicProperties(bool report_errors, bool decompose,
UNICHARSET* unicharset) {
for (int unichar_id = 0; unichar_id < unicharset->size(); ++unichar_id) {
// Convert any custom ligatures.
const char* unichar_str = unicharset->id_to_unichar(unichar_id);
for (int i = 0; UNICHARSET::kCustomLigatures[i][0] != nullptr; ++i) {
if (!strcmp(UNICHARSET::kCustomLigatures[i][1], unichar_str)) {
unichar_str = UNICHARSET::kCustomLigatures[i][0];
break;
}
}
// Convert the unichar to UTF32 representation
std::vector<char32> uni_vector = UNICHAR::UTF8ToUTF32(unichar_str);
// Assume that if the property is true for any character in the string,
// then it holds for the whole "character".
bool unichar_isalpha = false;
bool unichar_islower = false;
bool unichar_isupper = false;
bool unichar_isdigit = false;
bool unichar_ispunct = false;
for (char32 u_ch : uni_vector) {
if (u_isalpha(u_ch)) unichar_isalpha = true;
if (u_islower(u_ch)) unichar_islower = true;
if (u_isupper(u_ch)) unichar_isupper = true;
if (u_isdigit(u_ch)) unichar_isdigit = true;
if (u_ispunct(u_ch)) unichar_ispunct = true;
}
unicharset->set_isalpha(unichar_id, unichar_isalpha);
unicharset->set_islower(unichar_id, unichar_islower);
unicharset->set_isupper(unichar_id, unichar_isupper);
unicharset->set_isdigit(unichar_id, unichar_isdigit);
unicharset->set_ispunctuation(unichar_id, unichar_ispunct);
tesseract::IcuErrorCode err;
unicharset->set_script(unichar_id, uscript_getName(
uscript_getScript(uni_vector[0], err)));
const int num_code_points = uni_vector.size();
// Obtain the lower/upper case if needed and record it in the properties.
unicharset->set_other_case(unichar_id, unichar_id);
if (unichar_islower || unichar_isupper) {
std::vector<char32> other_case(num_code_points, 0);
for (int i = 0; i < num_code_points; ++i) {
// TODO(daria): Ideally u_strToLower()/ustrToUpper() should be used.
// However since they deal with UChars (so need a conversion function
// from char32 or UTF8string) and require a meaningful locale string,
// for now u_tolower()/u_toupper() are used.
other_case[i] = unichar_islower ? u_toupper(uni_vector[i]) :
u_tolower(uni_vector[i]);
}
std::string other_case_uch = UNICHAR::UTF32ToUTF8(other_case);
UNICHAR_ID other_case_id =
unicharset->unichar_to_id(other_case_uch.c_str());
if (other_case_id != INVALID_UNICHAR_ID) {
unicharset->set_other_case(unichar_id, other_case_id);
} else if (unichar_id >= SPECIAL_UNICHAR_CODES_COUNT && report_errors) {
tprintf("Other case %s of %s is not in unicharset\n",
other_case_uch.c_str(), unichar_str);
}
}
// Set RTL property and obtain mirror unichar ID from ICU.
std::vector<char32> mirrors(num_code_points, 0);
for (int i = 0; i < num_code_points; ++i) {
mirrors[i] = u_charMirror(uni_vector[i]);
if (i == 0) { // set directionality to that of the 1st code point
unicharset->set_direction(unichar_id,
static_cast<UNICHARSET::Direction>(
u_charDirection(uni_vector[i])));
}
}
std::string mirror_uch = UNICHAR::UTF32ToUTF8(mirrors);
UNICHAR_ID mirror_uch_id = unicharset->unichar_to_id(mirror_uch.c_str());
if (mirror_uch_id != INVALID_UNICHAR_ID) {
unicharset->set_mirror(unichar_id, mirror_uch_id);
} else if (report_errors) {
tprintf("Mirror %s of %s is not in unicharset\n",
mirror_uch.c_str(), unichar_str);
}
// Record normalized version of this unichar.
std::string normed_str;
if (unichar_id != 0 &&
tesseract::NormalizeUTF8String(
decompose ? tesseract::UnicodeNormMode::kNFKD
: tesseract::UnicodeNormMode::kNFKC,
tesseract::OCRNorm::kNormalize, tesseract::GraphemeNorm::kNone,
unichar_str, &normed_str) &&
!normed_str.empty()) {
unicharset->set_normed(unichar_id, normed_str.c_str());
} else {
unicharset->set_normed(unichar_id, unichar_str);
}
ASSERT_HOST(unicharset->get_other_case(unichar_id) < unicharset->size());
}
unicharset->post_load_setup();
}
UChar32 __hs_u_charMirror(UChar32 c)
{
return u_charMirror(c);
}
