QList<TextShaper::TextRun> TextShaper::itemizeBiDi()
{
QList<TextRun> textRuns;
UBiDi *obj = ubidi_open();
UErrorCode err = U_ZERO_ERROR;
UBiDiLevel parLevel = UBIDI_LTR;
ParagraphStyle style = m_story.paragraphStyle(m_firstChar);
if (style.direction() == ParagraphStyle::RTL)
parLevel = UBIDI_RTL;
ubidi_setPara(obj, (const UChar*) m_text.utf16(), m_text.length(), parLevel, NULL, &err);
if (U_SUCCESS(err))
{
int32_t count = ubidi_countRuns(obj, &err);
if (U_SUCCESS(err))
{
textRuns.reserve(count);
for (int32_t i = 0; i < count; i++)
{
int32_t start, length;
UBiDiDirection dir = ubidi_getVisualRun(obj, i, &start, &length);
textRuns.append(TextRun(start, length, dir));
}
}
}
ubidi_close(obj);
return textRuns;
}
nsresult nsBidi::GetVisualRun(int32_t aRunIndex, int32_t* aLogicalStart,
int32_t* aLength, nsBidiDirection* aDirection)
{
*aDirection = nsBidiDirection(ubidi_getVisualRun(mBiDi, aRunIndex,
aLogicalStart, aLength));
return NS_OK;
}
void CTextRenderer::UpdateTextCache_BiDi(array<CHarfbuzzGlyph>* pGlyphChain, const char* pText)
{
//Use ICU for bidirectional text
//note: bidirectional texts appear for example when a latin username is displayed in a arabic text
UErrorCode ICUError = U_ZERO_ERROR;
UnicodeString UTF16Text = icu::UnicodeString::fromUTF8(pText);
UBiDi* pICUBiDi = ubidi_openSized(UTF16Text.length(), 0, &ICUError);
//Perform the BiDi algorithm
//TODO: change UBIDI_DEFAULT_LTR by some variable dependend of the user config
ubidi_setPara(pICUBiDi, UTF16Text.getBuffer(), UTF16Text.length(), (Localization()->GetWritingDirection() == CLocalization::DIRECTION_RTL ? UBIDI_DEFAULT_RTL : UBIDI_DEFAULT_LTR), 0, &ICUError);
if(U_SUCCESS(ICUError))
{
UBiDiLevel ICULevel = 1&ubidi_getParaLevel(pICUBiDi);
UBiDiDirection Direction = ubidi_getDirection(pICUBiDi);
if(Direction != UBIDI_MIXED)
{
UpdateTextCache_Font(pGlyphChain, UTF16Text.getBuffer(), 0, UTF16Text.length(), (Direction == UBIDI_RTL));
}
else
{
int CharStart = 0;
UBiDiLevel level;
int NumberOfParts = ubidi_countRuns(pICUBiDi, &ICUError);
if(U_SUCCESS(ICUError))
{
for(int i=0; i<NumberOfParts; i++)
{
int Start;
int SubLength;
Direction = ubidi_getVisualRun(pICUBiDi, i, &Start, &SubLength);
UpdateTextCache_Font(pGlyphChain, UTF16Text.getBuffer(), Start, SubLength, (Direction == UBIDI_RTL));
}
}
else
{
dbg_msg("TextRenderer", "BiDi algorithm failed (ubidi_countRuns): %s", u_errorName(ICUError));
return;
}
}
}
else
{
dbg_msg("TextRenderer", "BiDi algorithm failed: %s", u_errorName(ICUError));
return;
}
}
void text_itemizer::itemize_direction(unsigned start, unsigned end)
{
direction_runs_.clear();
UErrorCode error = U_ZERO_ERROR;
int32_t length = end - start;
UBiDi *bidi = ubidi_openSized(length, 0, &error);
if (!bidi || U_FAILURE(error))
{
MAPNIK_LOG_ERROR(text_itemizer) << "Failed to create bidi object: " << u_errorName(error) << "\n";
return;
}
ubidi_setPara(bidi, text_.getBuffer() + start, length, UBIDI_DEFAULT_LTR, 0, &error);
if (U_SUCCESS(error))
{
UBiDiDirection direction = ubidi_getDirection(bidi);
if (direction != UBIDI_MIXED)
{
direction_runs_.emplace_back(direction, start, end);
}
else
{
// mixed-directional
int32_t count = ubidi_countRuns(bidi, &error);
if(U_SUCCESS(error))
{
for(int i=0; i<count; ++i)
{
int32_t vis_length;
int32_t run_start;
direction = ubidi_getVisualRun(bidi, i, &run_start, &vis_length);
run_start += start; //Add offset to compensate offset in setPara
direction_runs_.emplace_back(direction, run_start, run_start+vis_length);
}
}
}
}
else
{
MAPNIK_LOG_ERROR(text_itemizer) << "ICU error: " << u_errorName(error) << "\n"; //TODO: Exception
}
ubidi_close(bidi);
}
TextGroup::TextGroup(const std::string &input, hb_script_t script, const std::string &lang, hb_direction_t overallDirection)
:script_(script)
,lang_(lang)
,overallDirection_(overallDirection)
{
if(hb_script_get_horizontal_direction(script_) == HB_DIRECTION_LTR)
{
addRun(input, HB_DIRECTION_LTR);
}
else
{
auto text = UnicodeString::fromUTF8(input);
auto length = text.length();
printf("Hominlinx-->TextGroup::TextGroup str unicodelen[%d] ====text[0x%x]\n",length, text.charAt(0) );
UErrorCode err = U_ZERO_ERROR;
UBiDi *bidi = ubidi_openSized(length, 0, &err);//Bidrectional text
ubidi_setPara(bidi, text.getBuffer(), length, hbDirectionToUBIDILevel(overallDirection_), 0, &err);
auto direction = ubidi_getDirection(bidi);
if(direction != UBIDI_MIXED)
{
addRun(input, uciDirectionToHB(direction));
}
else
{
auto count = ubidi_countRuns(bidi, &err);
for(int i=0; i < count; ++i)
{
int32_t start, length;
direction = ubidi_getVisualRun(bidi, i, &start, &length);
addRun(text, direction, start, start + length);
}
}
ubidi_close(bidi);
}
}
U_CAPI int32_t U_EXPORT2
ubidi_writeReordered(UBiDi *pBiDi,
UChar *dest, int32_t destSize,
uint16_t options,
UErrorCode *pErrorCode) {
const UChar *text;
UChar *saveDest;
int32_t length, destCapacity;
int32_t run, runCount, logicalStart, runLength;
if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
return 0;
}
/* more error checking */
if( pBiDi==NULL ||
(text=pBiDi->text)==NULL || (length=pBiDi->length)<0 ||
destSize<0 || (destSize>0 && dest==NULL))
{
*pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
/* do input and output overlap? */
if( dest!=NULL &&
((text>=dest && text<dest+destSize) ||
(dest>=text && dest<text+pBiDi->originalLength)))
{
*pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
if(length==0) {
/* nothing to do */
return u_terminateUChars(dest, destSize, 0, pErrorCode);
}
runCount=ubidi_countRuns(pBiDi, pErrorCode);
if(U_FAILURE(*pErrorCode)) {
return 0;
}
/* destSize shrinks, later destination length=destCapacity-destSize */
saveDest=dest;
destCapacity=destSize;
/*
* Option "insert marks" implies UBIDI_INSERT_LRM_FOR_NUMERIC if the
* reordering mode (checked below) is appropriate.
*/
if(pBiDi->reorderingOptions & UBIDI_OPTION_INSERT_MARKS) {
options|=UBIDI_INSERT_LRM_FOR_NUMERIC;
options&=~UBIDI_REMOVE_BIDI_CONTROLS;
}
/*
* Option "remove controls" implies UBIDI_REMOVE_BIDI_CONTROLS
* and cancels UBIDI_INSERT_LRM_FOR_NUMERIC.
*/
if(pBiDi->reorderingOptions & UBIDI_OPTION_REMOVE_CONTROLS) {
options|=UBIDI_REMOVE_BIDI_CONTROLS;
options&=~UBIDI_INSERT_LRM_FOR_NUMERIC;
}
/*
* If we do not perform the "inverse BiDi" algorithm, then we
* don't need to insert any LRMs, and don't need to test for it.
*/
if((pBiDi->reorderingMode != UBIDI_REORDER_INVERSE_NUMBERS_AS_L) &&
(pBiDi->reorderingMode != UBIDI_REORDER_INVERSE_LIKE_DIRECT) &&
(pBiDi->reorderingMode != UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL) &&
(pBiDi->reorderingMode != UBIDI_REORDER_RUNS_ONLY)) {
options&=~UBIDI_INSERT_LRM_FOR_NUMERIC;
}
/*
* Iterate through all visual runs and copy the run text segments to
* the destination, according to the options.
*
* The tests for where to insert LRMs ignore the fact that there may be
* BN codes or non-BMP code points at the beginning and end of a run;
* they may insert LRMs unnecessarily but the tests are faster this way
* (this would have to be improved for UTF-8).
*
* Note that the only errors that are set by doWriteXY() are buffer overflow
* errors. Ignore them until the end, and continue for preflighting.
*/
if(!(options&UBIDI_OUTPUT_REVERSE)) {
/* forward output */
if(!(options&UBIDI_INSERT_LRM_FOR_NUMERIC)) {
/* do not insert BiDi controls */
for(run=0; run<runCount; ++run) {
if(UBIDI_LTR==ubidi_getVisualRun(pBiDi, run, &logicalStart, &runLength)) {
runLength=doWriteForward(text+logicalStart, runLength,
dest, destSize,
(uint16_t)(options&~UBIDI_DO_MIRRORING), pErrorCode);
} else {
runLength=doWriteReverse(text+logicalStart, runLength,
dest, destSize,
options, pErrorCode);
}
dest+=runLength;
destSize-=runLength;
}
} else {
/* insert BiDi controls for "inverse BiDi" */
const DirProp *dirProps=pBiDi->dirProps;
const UChar *src;
UChar uc;
UBiDiDirection dir;
int32_t markFlag;
for(run=0; run<runCount; ++run) {
dir=ubidi_getVisualRun(pBiDi, run, &logicalStart, &runLength);
src=text+logicalStart;
/* check if something relevant in insertPoints */
markFlag=pBiDi->runs[run].insertRemove;
if(markFlag<0) { /* BiDi controls count */
markFlag=0;
}
if(UBIDI_LTR==dir) {
if((pBiDi->isInverse) &&
(/*run>0 &&*/ dirProps[logicalStart]!=L)) {
markFlag |= LRM_BEFORE;
}
if (markFlag & LRM_BEFORE) {
uc=LRM_CHAR;
}
else if (markFlag & RLM_BEFORE) {
uc=RLM_CHAR;
}
else uc=0;
if(uc) {
if(destSize>0) {
*dest++=uc;
}
--destSize;
}
runLength=doWriteForward(src, runLength,
dest, destSize,
(uint16_t)(options&~UBIDI_DO_MIRRORING), pErrorCode);
dest+=runLength;
destSize-=runLength;
if((pBiDi->isInverse) &&
(/*run<runCount-1 &&*/ dirProps[logicalStart+runLength-1]!=L)) {
markFlag |= LRM_AFTER;
}
if (markFlag & LRM_AFTER) {
uc=LRM_CHAR;
}
else if (markFlag & RLM_AFTER) {
uc=RLM_CHAR;
}
else uc=0;
if(uc) {
if(destSize>0) {
*dest++=uc;
}
--destSize;
}
} else { /* RTL run */
if((pBiDi->isInverse) &&
(/*run>0 &&*/ !(MASK_R_AL&DIRPROP_FLAG(dirProps[logicalStart+runLength-1])))) {
markFlag |= RLM_BEFORE;
}
if (markFlag & LRM_BEFORE) {
uc=LRM_CHAR;
}
else if (markFlag & RLM_BEFORE) {
uc=RLM_CHAR;
}
else uc=0;
if(uc) {
if(destSize>0) {
*dest++=uc;
}
--destSize;
}
runLength=doWriteReverse(src, runLength,
dest, destSize,
options, pErrorCode);
dest+=runLength;
destSize-=runLength;
if((pBiDi->isInverse) &&
(/*run<runCount-1 &&*/ !(MASK_R_AL&DIRPROP_FLAG(dirProps[logicalStart])))) {
markFlag |= RLM_AFTER;
}
if (markFlag & LRM_AFTER) {
uc=LRM_CHAR;
}
else if (markFlag & RLM_AFTER) {
uc=RLM_CHAR;
}
else uc=0;
if(uc) {
if(destSize>0) {
*dest++=uc;
}
--destSize;
}
}
}
}
} else {
/* reverse output */
if(!(options&UBIDI_INSERT_LRM_FOR_NUMERIC)) {
/* do not insert BiDi controls */
for(run=runCount; --run>=0;) {
if(UBIDI_LTR==ubidi_getVisualRun(pBiDi, run, &logicalStart, &runLength)) {
runLength=doWriteReverse(text+logicalStart, runLength,
dest, destSize,
(uint16_t)(options&~UBIDI_DO_MIRRORING), pErrorCode);
} else {
runLength=doWriteForward(text+logicalStart, runLength,
dest, destSize,
options, pErrorCode);
}
dest+=runLength;
destSize-=runLength;
}
} else {
/* insert BiDi controls for "inverse BiDi" */
const DirProp *dirProps=pBiDi->dirProps;
const UChar *src;
UBiDiDirection dir;
for(run=runCount; --run>=0;) {
/* reverse output */
dir=ubidi_getVisualRun(pBiDi, run, &logicalStart, &runLength);
src=text+logicalStart;
if(UBIDI_LTR==dir) {
if(/*run<runCount-1 &&*/ dirProps[logicalStart+runLength-1]!=L) {
if(destSize>0) {
*dest++=LRM_CHAR;
}
--destSize;
}
runLength=doWriteReverse(src, runLength,
dest, destSize,
(uint16_t)(options&~UBIDI_DO_MIRRORING), pErrorCode);
dest+=runLength;
destSize-=runLength;
if(/*run>0 &&*/ dirProps[logicalStart]!=L) {
if(destSize>0) {
*dest++=LRM_CHAR;
}
--destSize;
}
} else {
if(/*run<runCount-1 &&*/ !(MASK_R_AL&DIRPROP_FLAG(dirProps[logicalStart]))) {
if(destSize>0) {
*dest++=RLM_CHAR;
}
--destSize;
}
runLength=doWriteForward(src, runLength,
dest, destSize,
options, pErrorCode);
dest+=runLength;
destSize-=runLength;
if(/*run>0 &&*/ !(MASK_R_AL&DIRPROP_FLAG(dirProps[logicalStart+runLength-1]))) {
if(destSize>0) {
*dest++=RLM_CHAR;
}
--destSize;
}
}
}
}
}
return u_terminateUChars(saveDest, destCapacity, destCapacity-destSize, pErrorCode);
}
void font_face_set::get_string_info(string_info & info)
{
unsigned width = 0;
unsigned height = 0;
UErrorCode err = U_ZERO_ERROR;
UnicodeString const& ustr = info.get_string();
const UChar * text = ustr.getBuffer();
UBiDi * bidi = ubidi_openSized(ustr.length(),0,&err);
if (U_SUCCESS(err))
{
ubidi_setPara(bidi,text,ustr.length(), UBIDI_DEFAULT_LTR,0,&err);
if (U_SUCCESS(err))
{
int32_t count = ubidi_countRuns(bidi,&err);
int32_t logicalStart;
int32_t length;
for (int32_t i=0; i< count;++i)
{
if (UBIDI_LTR == ubidi_getVisualRun(bidi,i,&logicalStart,&length))
{
do {
UChar ch = text[logicalStart++];
dimension_t char_dim = character_dimensions(ch);
info.add_info(ch, char_dim.width, char_dim.height);
width += char_dim.width;
height = char_dim.height > height ? char_dim.height : height;
} while (--length > 0);
}
else
{
logicalStart += length;
int32_t j=0,i=length;
UnicodeString arabic;
UChar * buf = arabic.getBuffer(length);
do {
UChar ch = text[--logicalStart];
buf[j++] = ch;
} while (--i > 0);
arabic.releaseBuffer(length);
if ( *arabic.getBuffer() >= 0x0600 && *arabic.getBuffer() <= 0x06ff)
{
UnicodeString shaped;
u_shapeArabic(arabic.getBuffer(),arabic.length(),shaped.getBuffer(arabic.length()),arabic.length(),
U_SHAPE_LETTERS_SHAPE|U_SHAPE_LENGTH_FIXED_SPACES_NEAR|
U_SHAPE_TEXT_DIRECTION_VISUAL_LTR
,&err);
shaped.releaseBuffer(arabic.length());
if (U_SUCCESS(err))
{
for (int j=0;j<shaped.length();++j)
{
dimension_t char_dim = character_dimensions(shaped[j]);
info.add_info(shaped[j], char_dim.width, char_dim.height);
width += char_dim.width;
height = char_dim.height > height ? char_dim.height : height;
}
}
} else {
// Non-Arabic RTL
for (int j=0;j<arabic.length();++j)
{
dimension_t char_dim = character_dimensions(arabic[j]);
info.add_info(arabic[j], char_dim.width, char_dim.height);
width += char_dim.width;
height = char_dim.height > height ? char_dim.height : height;
}
}
}
}
}
ubidi_close(bidi);
}
info.set_dimensions(width, height);
}
int icu_bidi_runs(lua_State *L) {
size_t input_l;
const char* input = luaL_checklstring(L, 1, &input_l);
const char* direction = luaL_checkstring(L, 2);
UChar *input_as_uchar;
int32_t l;
utf8_to_uchar(input, input_l, input_as_uchar, l);
UBiDiLevel paraLevel = 0;
if (strncasecmp(direction, "RTL", 3) == 0) {
paraLevel = 1;
}
/* Now let's bidi! */
UBiDi* bidi = ubidi_open();
UErrorCode err = U_ZERO_ERROR;
ubidi_setPara(bidi, input_as_uchar, l, paraLevel, NULL, &err);
if (!U_SUCCESS(err)) {
free(input_as_uchar);
ubidi_close(bidi);
return luaL_error(L, "Error in bidi %s", u_errorName(err));
}
int count = ubidi_countRuns(bidi,&err);
int start, length;
lua_checkstack(L,count);
for (int i=0; i < count; i++) {
UBiDiDirection dir = ubidi_getVisualRun(bidi, i, &start, &length);
lua_newtable(L);
// Convert back to UTF8...
int32_t l3 = 0;
char* possibleOutbuf = malloc(4*length);
if(!possibleOutbuf) {
return luaL_error(L, "Couldn't malloc");
}
u_strToUTF8(possibleOutbuf, 4 * length, &l3, input_as_uchar+start, length, &err);
if (!U_SUCCESS(err)) {
free(possibleOutbuf);
return luaL_error(L, "Bidi run too big? %s", u_errorName(err));
}
lua_pushstring(L, "run");
lua_pushstring(L, possibleOutbuf);
free(possibleOutbuf);
lua_settable(L, -3);
lua_pushstring(L, "start");
int32_t new_start = start;
// Length/start is given in terms of UTF16 codepoints.
// But we want a count of Unicode characters. This means
// surrogate pairs need to be counted as 1.
for (int j=0; j< start; j++) {
if (U_IS_TRAIL(*(input_as_uchar+j))) new_start--;
}
lua_pushinteger(L, new_start);
lua_settable(L, -3);
lua_pushstring(L, "length");
for (int j=start; j< start+length; j++) {
if (U_IS_TRAIL(*(input_as_uchar+j))) length--;
}
lua_pushinteger(L, length);
lua_settable(L, -3);
lua_pushstring(L, "dir");
lua_pushstring(L, dir == UBIDI_RTL ? "RTL" : "LTR");
lua_settable(L, -3);
lua_pushstring(L, "level");
lua_pushinteger(L, ubidi_getLevelAt(bidi, start));
lua_settable(L, -3);
}
free(input_as_uchar);
ubidi_close(bidi);
return count;
}
std::vector<StyledText> BiDi::processStyledText(const StyledText& input, std::set<std::size_t> lineBreakPoints) {
std::vector<StyledText> lines;
const auto& inputText = input.first;
const auto& styleIndices = input.second;
UErrorCode errorCode = U_ZERO_ERROR;
ubidi_setPara(impl->bidiText, mbgl::utf16char_cast<const UChar*>(inputText.c_str()), static_cast<int32_t>(inputText.size()),
UBIDI_DEFAULT_LTR, nullptr, &errorCode);
if (U_FAILURE(errorCode)) {
throw std::runtime_error(std::string("BiDi::processStyledText: ") + u_errorName(errorCode));
}
mergeParagraphLineBreaks(lineBreakPoints);
std::size_t lineStartIndex = 0;
for (std::size_t lineBreakPoint : lineBreakPoints) {
StyledText line;
line.second.reserve(lineBreakPoint - lineStartIndex);
errorCode = U_ZERO_ERROR;
ubidi_setLine(impl->bidiText, static_cast<int32_t>(lineStartIndex), static_cast<int32_t>(lineBreakPoint), impl->bidiLine, &errorCode);
if (U_FAILURE(errorCode)) {
throw std::runtime_error(std::string("BiDi::processStyledText (setLine): ") + u_errorName(errorCode));
}
errorCode = U_ZERO_ERROR;
uint32_t runCount = ubidi_countRuns(impl->bidiLine, &errorCode);
if (U_FAILURE(errorCode)) {
throw std::runtime_error(std::string("BiDi::processStyledText (countRuns): ") + u_errorName(errorCode));
}
for (uint32_t runIndex = 0; runIndex < runCount; runIndex++) {
int32_t runLogicalStart;
int32_t runLength;
UBiDiDirection direction = ubidi_getVisualRun(impl->bidiLine, runIndex, &runLogicalStart, &runLength);
const bool isReversed = direction == UBIDI_RTL;
std::size_t logicalStart = lineStartIndex + runLogicalStart;
std::size_t logicalEnd = logicalStart + runLength;
if (isReversed) {
// Within this reversed section, iterate logically backwards
// Each time we see a change in style, render a reversed chunk
// of everything since the last change
std::size_t styleRunStart = logicalEnd;
uint8_t currentStyleIndex = styleIndices.at(styleRunStart - 1);
for (std::size_t i = logicalEnd - 1; i >= logicalStart; i--) {
if (currentStyleIndex != styleIndices.at(i) || i == logicalStart) {
std::size_t styleRunEnd = i == logicalStart ? i : i + 1;
std::u16string reversed = writeReverse(inputText, styleRunEnd, styleRunStart);
line.first += reversed;
for (std::size_t j = 0; j < reversed.size(); j++) {
line.second.push_back(currentStyleIndex);
}
currentStyleIndex = styleIndices.at(i);
styleRunStart = styleRunEnd;
}
if (i == 0) {
break;
}
}
} else {
line.first += input.first.substr(logicalStart, runLength);
line.second.insert(line.second.end(), styleIndices.begin() + logicalStart, styleIndices.begin() + logicalStart + runLength);
}
}
lines.push_back(line);
lineStartIndex = lineBreakPoint;
}
return lines;
}
