/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ /* * This file is part of the LibreOffice project. * * This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this * file, You can obtain one at http://mozilla.org/MPL/2.0/. * * This file incorporates work covered by the following license notice: * * Licensed to the Apache Software Foundation (ASF) under one or more * contributor license agreements. See the NOTICE file distributed * with this work for additional information regarding copyright * ownership. The ASF licenses this file to you under the Apache * License, Version 2.0 (the "License"); you may not use this file * except in compliance with the License. You may obtain a copy of * the License at http://www.apache.org/licenses/LICENSE-2.0 . */ #include #include #include "sal/config.h" #include #include #include #include #include #include #include #include #include #include #include #if defined _MSC_VER #pragma warning(push, 1) #endif #include #include #if defined _MSC_VER #pragma warning(pop) #endif #include #include #ifdef DEBUG //#define MULTI_SL_DEBUG #endif #ifdef MULTI_SL_DEBUG #include FILE * mslLogFile = NULL; FILE * mslLog() { #ifdef _MSC_VER std::string logFileName(getenv("TEMP")); logFileName.append("\\msllayout.log"); if (mslLogFile == NULL) mslLogFile = fopen(logFileName.c_str(),"w"); else fflush(mslLogFile); return mslLogFile; #else return stdout; #endif } #endif std::ostream &operator <<(std::ostream& s, ImplLayoutArgs &rArgs) { #ifndef SAL_LOG_INFO (void) rArgs; #else s << "ImplLayoutArgs{"; s << "Flags="; if (rArgs.mnFlags == SalLayoutFlags::NONE) s << 0; else { bool need_or = false; s << "{"; #define TEST(x) if (rArgs.mnFlags & SalLayoutFlags::x) { if (need_or) s << "|"; s << #x; need_or = true; } TEST(BiDiRtl); TEST(BiDiStrong); TEST(RightAlign); TEST(KerningPairs); TEST(KerningAsian); TEST(Vertical); TEST(ComplexDisabled); TEST(EnableLigatures); TEST(SubstituteDigits); TEST(KashidaJustification); TEST(DisableGlyphProcessing); TEST(ForFallback); #undef TEST s << "}"; } const int nLength = rArgs.mrStr.getLength(); s << ",Length=" << nLength; s << ",MinCharPos=" << rArgs.mnMinCharPos; s << ",EndCharPos=" << rArgs.mnEndCharPos; s << ",Str=\""; int lim = nLength; if (lim > 10) lim = 7; for (int i = 0; i < lim; i++) { if (rArgs.mrStr[i] == '\n') s << "\\n"; else if (rArgs.mrStr[i] < ' ' || (rArgs.mrStr[i] >= 0x7F && rArgs.mrStr[i] <= 0xFF)) s << "\\0x" << std::hex << std::setw(2) << std::setfill('0') << (int) rArgs.mrStr[i] << std::setfill(' ') << std::setw(1) << std::dec; else if (rArgs.mrStr[i] < 0x7F) s << (char) rArgs.mrStr[i]; else s << "\\u" << std::hex << std::setw(4) << std::setfill('0') << (int) rArgs.mrStr[i] << std::setfill(' ') << std::setw(1) << std::dec; } if (nLength > lim) s << "..."; s << "\""; s << ",DXArray="; if (rArgs.mpDXArray) { s << "["; int count = rArgs.mnEndCharPos - rArgs.mnMinCharPos; lim = count; if (lim > 10) lim = 7; for (int i = 0; i < lim; i++) { s << rArgs.mpDXArray[i]; if (i < lim-1) s << ","; } if (count > lim) { if (count > lim + 1) s << "..."; s << rArgs.mpDXArray[count-1]; } s << "]"; } else s << "NULL"; s << ",LayoutWidth=" << rArgs.mnLayoutWidth; s << "}"; #endif return s; } int GetVerticalFlags( sal_UCS4 nChar ) { if( (nChar >= 0x1100 && nChar <= 0x11f9) // Hangul Jamo || (nChar == 0x2030 || nChar == 0x2031) // per mille sign || (nChar >= 0x3000 && nChar <= 0xfaff) // unified CJK || (nChar >= 0xfe20 && nChar <= 0xfe6f) // CJK compatibility || (nChar >= 0xff00 && nChar <= 0xfffd) ) // other CJK { /* #i52932# remember: nChar == 0x2010 || nChar == 0x2015 nChar == 0x2016 || nChar == 0x2026 are GF_NONE also, but already handled in the outer if condition */ if((nChar >= 0x3008 && nChar <= 0x301C && nChar != 0x3012) || (nChar == 0xFF3B || nChar == 0xFF3D || nChar==0xFF08 || nChar==0xFF09) || (nChar >= 0xFF5B && nChar <= 0xFF9F) // halfwidth forms || (nChar == 0xFFE3) ) return GF_NONE; // not rotated else if( nChar == 0x30fc ) return GF_ROTR; // right return GF_ROTL; // left } else if( (nChar >= 0x20000) && (nChar <= 0x3FFFF) ) // all SIP/TIP ideographs return GF_ROTL; // left return GF_NONE; // not rotated as default } sal_UCS4 GetMirroredChar( sal_UCS4 nChar ) { nChar = u_charMirror( nChar ); return nChar; } sal_UCS4 GetLocalizedChar( sal_UCS4 nChar, LanguageType eLang ) { // currently only conversion from ASCII digits is interesting if( (nChar < '0') || ('9' < nChar) ) return nChar; int nOffset; // eLang & LANGUAGE_MASK_PRIMARY catches language independent of region. // CAVEAT! To some like Mongolian MS assigned the same primary language // although the script type is different! switch( eLang & LANGUAGE_MASK_PRIMARY ) { default: nOffset = 0; break; case LANGUAGE_ARABIC_SAUDI_ARABIA & LANGUAGE_MASK_PRIMARY: nOffset = 0x0660 - '0'; // arabic-indic digits break; case LANGUAGE_FARSI & LANGUAGE_MASK_PRIMARY: case LANGUAGE_URDU_PAKISTAN & LANGUAGE_MASK_PRIMARY: case LANGUAGE_PUNJABI & LANGUAGE_MASK_PRIMARY: //??? case LANGUAGE_SINDHI & LANGUAGE_MASK_PRIMARY: nOffset = 0x06F0 - '0'; // eastern arabic-indic digits break; case LANGUAGE_BENGALI & LANGUAGE_MASK_PRIMARY: nOffset = 0x09E6 - '0'; // bengali break; case LANGUAGE_HINDI & LANGUAGE_MASK_PRIMARY: nOffset = 0x0966 - '0'; // devanagari break; case LANGUAGE_AMHARIC_ETHIOPIA & LANGUAGE_MASK_PRIMARY: case LANGUAGE_TIGRIGNA_ETHIOPIA & LANGUAGE_MASK_PRIMARY: // TODO case: nOffset = 0x1369 - '0'; // ethiopic break; case LANGUAGE_GUJARATI & LANGUAGE_MASK_PRIMARY: nOffset = 0x0AE6 - '0'; // gujarati break; #ifdef LANGUAGE_GURMUKHI // TODO case: case LANGUAGE_GURMUKHI & LANGUAGE_MASK_PRIMARY: nOffset = 0x0A66 - '0'; // gurmukhi break; #endif case LANGUAGE_KANNADA & LANGUAGE_MASK_PRIMARY: nOffset = 0x0CE6 - '0'; // kannada break; case LANGUAGE_KHMER & LANGUAGE_MASK_PRIMARY: nOffset = 0x17E0 - '0'; // khmer break; case LANGUAGE_LAO & LANGUAGE_MASK_PRIMARY: nOffset = 0x0ED0 - '0'; // lao break; case LANGUAGE_MALAYALAM & LANGUAGE_MASK_PRIMARY: nOffset = 0x0D66 - '0'; // malayalam break; case LANGUAGE_MONGOLIAN_MONGOLIAN_LSO & LANGUAGE_MASK_PRIMARY: switch (eLang) { case LANGUAGE_MONGOLIAN_MONGOLIAN_MONGOLIA: case LANGUAGE_MONGOLIAN_MONGOLIAN_CHINA: case LANGUAGE_MONGOLIAN_MONGOLIAN_LSO: nOffset = 0x1810 - '0'; // mongolian break; default: nOffset = 0; // mongolian cyrillic break; } break; case LANGUAGE_BURMESE & LANGUAGE_MASK_PRIMARY: nOffset = 0x1040 - '0'; // myanmar break; case LANGUAGE_ODIA & LANGUAGE_MASK_PRIMARY: nOffset = 0x0B66 - '0'; // odia break; case LANGUAGE_TAMIL & LANGUAGE_MASK_PRIMARY: nOffset = 0x0BE7 - '0'; // tamil break; case LANGUAGE_TELUGU & LANGUAGE_MASK_PRIMARY: nOffset = 0x0C66 - '0'; // telugu break; case LANGUAGE_THAI & LANGUAGE_MASK_PRIMARY: nOffset = 0x0E50 - '0'; // thai break; case LANGUAGE_TIBETAN & LANGUAGE_MASK_PRIMARY: nOffset = 0x0F20 - '0'; // tibetan break; } nChar += nOffset; return nChar; } inline bool IsControlChar( sal_UCS4 cChar ) { // C0 control characters if( (0x0001 <= cChar) && (cChar <= 0x001F) ) return true; // formatting characters if( (0x200E <= cChar) && (cChar <= 0x200F) ) return true; if( (0x2028 <= cChar) && (cChar <= 0x202E) ) return true; // deprecated formatting characters if( (0x206A <= cChar) && (cChar <= 0x206F) ) return true; if( (0x2060 == cChar) ) return true; // byte order markers and invalid unicode if( (cChar == 0xFEFF) || (cChar == 0xFFFE) || (cChar == 0xFFFF) ) return true; return false; } bool ImplLayoutRuns::AddPos( int nCharPos, bool bRTL ) { // check if charpos could extend current run int nIndex = maRuns.size(); if( nIndex >= 2 ) { int nRunPos0 = maRuns[ nIndex-2 ]; int nRunPos1 = maRuns[ nIndex-1 ]; if( ((nCharPos + int(bRTL)) == nRunPos1) && ((nRunPos0 > nRunPos1) == bRTL) ) { // extend current run by new charpos maRuns[ nIndex-1 ] = nCharPos + int(!bRTL); return false; } // ignore new charpos when it is in current run if( (nRunPos0 <= nCharPos) && (nCharPos < nRunPos1) ) return false; if( (nRunPos1 <= nCharPos) && (nCharPos < nRunPos0) ) return false; } // else append a new run consisting of the new charpos maRuns.push_back( nCharPos + (bRTL ? 1 : 0) ); maRuns.push_back( nCharPos + (bRTL ? 0 : 1) ); return true; } bool ImplLayoutRuns::AddRun( int nCharPos0, int nCharPos1, bool bRTL ) { if( nCharPos0 == nCharPos1 ) return false; // swap if needed if( bRTL == (nCharPos0 < nCharPos1) ) { int nTemp = nCharPos0; nCharPos0 = nCharPos1; nCharPos1 = nTemp; } // append new run maRuns.push_back( nCharPos0 ); maRuns.push_back( nCharPos1 ); return true; } bool ImplLayoutRuns::PosIsInRun( int nCharPos ) const { if( mnRunIndex >= (int)maRuns.size() ) return false; int nMinCharPos = maRuns[ mnRunIndex+0 ]; int nEndCharPos = maRuns[ mnRunIndex+1 ]; if( nMinCharPos > nEndCharPos ) // reversed in RTL case { int nTemp = nMinCharPos; nMinCharPos = nEndCharPos; nEndCharPos = nTemp; } if( nCharPos < nMinCharPos ) return false; if( nCharPos >= nEndCharPos ) return false; return true; } bool ImplLayoutRuns::PosIsInAnyRun( int nCharPos ) const { bool bRet = false; int nRunIndex = mnRunIndex; ImplLayoutRuns *pThis = const_cast(this); pThis->ResetPos(); for (size_t i = 0; i < maRuns.size(); i+=2) { if( (bRet = PosIsInRun( nCharPos )) ) break; pThis->NextRun(); } pThis->mnRunIndex = nRunIndex; return bRet; } bool ImplLayoutRuns::GetNextPos( int* nCharPos, bool* bRightToLeft ) { // negative nCharPos => reset to first run if( *nCharPos < 0 ) mnRunIndex = 0; // return false when all runs completed if( mnRunIndex >= (int)maRuns.size() ) return false; int nRunPos0 = maRuns[ mnRunIndex+0 ]; int nRunPos1 = maRuns[ mnRunIndex+1 ]; *bRightToLeft = (nRunPos0 > nRunPos1); if( *nCharPos < 0 ) { // get first valid nCharPos in run *nCharPos = nRunPos0; } else { // advance to next nCharPos for LTR case if( !*bRightToLeft ) ++(*nCharPos); // advance to next run if current run is completed if( *nCharPos == nRunPos1 ) { if( (mnRunIndex += 2) >= (int)maRuns.size() ) return false; nRunPos0 = maRuns[ mnRunIndex+0 ]; nRunPos1 = maRuns[ mnRunIndex+1 ]; *bRightToLeft = (nRunPos0 > nRunPos1); *nCharPos = nRunPos0; } } // advance to next nCharPos for RTL case if( *bRightToLeft ) --(*nCharPos); return true; } bool ImplLayoutRuns::GetRun( int* nMinRunPos, int* nEndRunPos, bool* bRightToLeft ) const { if( mnRunIndex >= (int)maRuns.size() ) return false; int nRunPos0 = maRuns[ mnRunIndex+0 ]; int nRunPos1 = maRuns[ mnRunIndex+1 ]; *bRightToLeft = (nRunPos1 < nRunPos0) ; if( !*bRightToLeft ) { *nMinRunPos = nRunPos0; *nEndRunPos = nRunPos1; } else { *nMinRunPos = nRunPos1; *nEndRunPos = nRunPos0; } return true; } ImplLayoutArgs::ImplLayoutArgs(const OUString& rStr, int nMinCharPos, int nEndCharPos, SalLayoutFlags nFlags, const LanguageTag& rLanguageTag, vcl::TextLayoutCache const*const pLayoutCache) : maLanguageTag( rLanguageTag ), mnFlags( nFlags ), mrStr( rStr ), mnMinCharPos( nMinCharPos ), mnEndCharPos( nEndCharPos ), m_pTextLayoutCache(pLayoutCache), mpDXArray( nullptr ), mnLayoutWidth( 0 ), mnOrientation( 0 ) { if( mnFlags & SalLayoutFlags::BiDiStrong ) { // handle strong BiDi mode // do not bother to BiDi analyze strong LTR/RTL // TODO: can we assume these strings do not have unicode control chars? // if not remove the control characters from the runs bool bRTL(mnFlags & SalLayoutFlags::BiDiRtl); AddRun( mnMinCharPos, mnEndCharPos, bRTL ); } else { // handle weak BiDi mode UBiDiLevel nLevel = UBIDI_DEFAULT_LTR; if( mnFlags & SalLayoutFlags::BiDiRtl ) nLevel = UBIDI_DEFAULT_RTL; // prepare substring for BiDi analysis // TODO: reuse allocated pParaBidi UErrorCode rcI18n = U_ZERO_ERROR; const int nLength = mrStr.getLength(); UBiDi* pParaBidi = ubidi_openSized(nLength, 0, &rcI18n); if( !pParaBidi ) return; ubidi_setPara(pParaBidi, reinterpret_cast(mrStr.getStr()), nLength, nLevel, nullptr, &rcI18n); // UChar != sal_Unicode in MinGW UBiDi* pLineBidi = pParaBidi; int nSubLength = mnEndCharPos - mnMinCharPos; if (nSubLength != nLength) { pLineBidi = ubidi_openSized( nSubLength, 0, &rcI18n ); ubidi_setLine( pParaBidi, mnMinCharPos, mnEndCharPos, pLineBidi, &rcI18n ); } // run BiDi algorithm const int nRunCount = ubidi_countRuns( pLineBidi, &rcI18n ); //maRuns.resize( 2 * nRunCount ); for( int i = 0; i < nRunCount; ++i ) { int32_t nMinPos, nRunLength; const UBiDiDirection nDir = ubidi_getVisualRun( pLineBidi, i, &nMinPos, &nRunLength ); const int nPos0 = nMinPos + mnMinCharPos; const int nPos1 = nPos0 + nRunLength; const bool bRTL = (nDir == UBIDI_RTL); AddRun( nPos0, nPos1, bRTL ); } // cleanup BiDi engine if( pLineBidi != pParaBidi ) ubidi_close( pLineBidi ); ubidi_close( pParaBidi ); } // prepare calls to GetNextPos/GetNextRun maRuns.ResetPos(); } // add a run after splitting it up to get rid of control chars void ImplLayoutArgs::AddRun( int nCharPos0, int nCharPos1, bool bRTL ) { DBG_ASSERT( nCharPos0 <= nCharPos1, "ImplLayoutArgs::AddRun() nCharPos0>=nCharPos1" ); // remove control characters from runs by splitting them up if( !bRTL ) { for( int i = nCharPos0; i < nCharPos1; ++i ) if( IsControlChar( mrStr[i] ) ) { // add run until control char maRuns.AddRun( nCharPos0, i, bRTL ); nCharPos0 = i + 1; } } else { for( int i = nCharPos1; --i >= nCharPos0; ) if( IsControlChar( mrStr[i] ) ) { // add run until control char maRuns.AddRun( i+1, nCharPos1, bRTL ); nCharPos1 = i; } } // add remainder of run maRuns.AddRun( nCharPos0, nCharPos1, bRTL ); } bool ImplLayoutArgs::PrepareFallback() { // short circuit if no fallback is needed if( maFallbackRuns.IsEmpty() ) { maRuns.Clear(); return false; } // convert the fallback requests to layout requests bool bRTL; int nMin, nEnd; // get the individual fallback requests typedef std::vector IntVector; IntVector aPosVector; aPosVector.reserve(mrStr.getLength()); maFallbackRuns.ResetPos(); for(; maFallbackRuns.GetRun( &nMin, &nEnd, &bRTL ); maFallbackRuns.NextRun() ) for( int i = nMin; i < nEnd; ++i ) aPosVector.push_back( i ); maFallbackRuns.Clear(); // sort the individual fallback requests std::sort( aPosVector.begin(), aPosVector.end() ); // adjust fallback runs to have the same order and limits of the original runs ImplLayoutRuns aNewRuns; maRuns.ResetPos(); for(; maRuns.GetRun( &nMin, &nEnd, &bRTL ); maRuns.NextRun() ) { if( !bRTL) { IntVector::const_iterator it = std::lower_bound( aPosVector.begin(), aPosVector.end(), nMin ); for(; (it != aPosVector.end()) && (*it < nEnd); ++it ) aNewRuns.AddPos( *it, bRTL ); } else { IntVector::const_iterator it = std::upper_bound( aPosVector.begin(), aPosVector.end(), nEnd ); while( (it != aPosVector.begin()) && (*--it >= nMin) ) aNewRuns.AddPos( *it, bRTL ); } } maRuns = aNewRuns; // TODO: use vector<>::swap() maRuns.ResetPos(); return true; } bool ImplLayoutArgs::GetNextRun( int* nMinRunPos, int* nEndRunPos, bool* bRTL ) { bool bValid = maRuns.GetRun( nMinRunPos, nEndRunPos, bRTL ); maRuns.NextRun(); return bValid; } SalLayout::SalLayout() : mnMinCharPos( -1 ), mnEndCharPos( -1 ), mnLayoutFlags( SalLayoutFlags::NONE ), mnUnitsPerPixel( 1 ), mnOrientation( 0 ), mnRefCount( 1 ), maDrawOffset( 0, 0 ) {} SalLayout::~SalLayout() {} void SalLayout::AdjustLayout( ImplLayoutArgs& rArgs ) { mnMinCharPos = rArgs.mnMinCharPos; mnEndCharPos = rArgs.mnEndCharPos; mnLayoutFlags = rArgs.mnFlags; mnOrientation = rArgs.mnOrientation; } void SalLayout::Release() const { // TODO: protect when multiple threads can access this if( --mnRefCount > 0 ) return; // const_cast because some compilers violate ANSI C++ spec delete this; } Point SalLayout::GetDrawPosition( const Point& rRelative ) const { Point aPos = maDrawBase; Point aOfs = rRelative + maDrawOffset; if( mnOrientation == 0 ) aPos += aOfs; else { // cache trigonometric results static int nOldOrientation = 0; static double fCos = 1.0, fSin = 0.0; if( nOldOrientation != mnOrientation ) { nOldOrientation = mnOrientation; double fRad = mnOrientation * (M_PI / 1800.0); fCos = cos( fRad ); fSin = sin( fRad ); } double fX = aOfs.X(); double fY = aOfs.Y(); long nX = static_cast( +fCos * fX + fSin * fY ); long nY = static_cast( +fCos * fY - fSin * fX ); aPos += Point( nX, nY ); } return aPos; } // returns asian kerning values in quarter of character width units // to enable automatic halfwidth substitution for fullwidth punctuation // return value is negative for l, positive for r, zero for neutral // If the range doesn't match in 0x3000 and 0x30FB, please change // also ImplCalcKerning. int SalLayout::CalcAsianKerning( sal_UCS4 c, bool bLeft, bool /*TODO:? bVertical*/ ) { // http://www.asahi-net.or.jp/~sd5a-ucd/freetexts/jis/x4051/1995/appendix.html static const signed char nTable[0x30] = { 0, -2, -2, 0, 0, 0, 0, 0, +2, -2, +2, -2, +2, -2, +2, -2, +2, -2, 0, 0, +2, -2, +2, -2, 0, 0, 0, 0, 0, +2, -2, -2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, -2, -2, +2, +2, -2, -2 }; int nResult = 0; if( (c >= 0x3000) && (c < 0x3030) ) nResult = nTable[ c - 0x3000 ]; else switch( c ) { case 0x30FB: nResult = bLeft ? -1 : +1; // 25% left/right/top/bottom break; case 0x2019: case 0x201D: case 0xFF01: case 0xFF09: case 0xFF0C: case 0xFF1A: case 0xFF1B: nResult = -2; break; case 0x2018: case 0x201C: case 0xFF08: nResult = +2; break; default: break; } return nResult; } bool SalLayout::GetOutline( SalGraphics& rSalGraphics, ::basegfx::B2DPolyPolygonVector& rVector ) const { bool bAllOk = true; bool bOneOk = false; Point aPos; ::basegfx::B2DPolyPolygon aGlyphOutline; for( int nStart = 0;;) { sal_GlyphId nLGlyph; if( !GetNextGlyphs( 1, &nLGlyph, aPos, nStart ) ) break; // get outline of individual glyph, ignoring "empty" glyphs bool bSuccess = rSalGraphics.GetGlyphOutline( nLGlyph, aGlyphOutline ); bAllOk &= bSuccess; bOneOk |= bSuccess; // only add non-empty outlines if( bSuccess && (aGlyphOutline.count() > 0) ) { if( aPos.X() || aPos.Y() ) { aGlyphOutline.transform(basegfx::tools::createTranslateB2DHomMatrix(aPos.X(), aPos.Y())); } // insert outline at correct position rVector.push_back( aGlyphOutline ); } } return (bAllOk && bOneOk); } bool SalLayout::GetBoundRect( SalGraphics& rSalGraphics, Rectangle& rRect ) const { bool bRet = false; rRect.SetEmpty(); Point aPos; Rectangle aRectangle; for( int nStart = 0;;) { sal_GlyphId nLGlyph; if( !GetNextGlyphs( 1, &nLGlyph, aPos, nStart ) ) break; // get bounding rectangle of individual glyph if( rSalGraphics.GetGlyphBoundRect( nLGlyph, aRectangle ) ) { // merge rectangle aRectangle += aPos; if (rRect.IsEmpty()) rRect = aRectangle; else rRect.Union(aRectangle); bRet = true; } } return bRet; } bool SalLayout::IsSpacingGlyph( sal_GlyphId nGlyph ) { bool bRet = false; if( nGlyph & GF_ISCHAR ) { long nChar = nGlyph & GF_IDXMASK; bRet = (nChar <= 0x0020) // blank //|| (nChar == 0x00A0) // non breaking space || (nChar >= 0x2000 && nChar <= 0x200F) // whitespace || (nChar == 0x3000); // ideographic space } else bRet = ((nGlyph & GF_IDXMASK) == 3); return bRet; } GenericSalLayout::GenericSalLayout() {} GenericSalLayout::~GenericSalLayout() {} void GenericSalLayout::AppendGlyph( const GlyphItem& rGlyphItem ) { m_GlyphItems.push_back(rGlyphItem); } bool GenericSalLayout::GetCharWidths( DeviceCoordinate* pCharWidths ) const { // initialize character extents buffer int nCharCount = mnEndCharPos - mnMinCharPos; for( int n = 0; n < nCharCount; ++n ) pCharWidths[n] = 0; // determine cluster extents for( GlyphVector::const_iterator pGlyphIter = m_GlyphItems.begin(), end = m_GlyphItems.end(); pGlyphIter != end ; ++pGlyphIter) { // use cluster start to get char index if( !pGlyphIter->IsClusterStart() ) continue; int n = pGlyphIter->mnCharPos; if( n >= mnEndCharPos ) continue; n -= mnMinCharPos; if( n < 0 ) continue; // left glyph in cluster defines default extent long nXPosMin = pGlyphIter->maLinearPos.X(); long nXPosMax = nXPosMin + pGlyphIter->mnNewWidth; // calculate right x-position for this glyph cluster // break if no more glyphs in layout // break at next glyph cluster start while( (pGlyphIter+1 != end) && !pGlyphIter[1].IsClusterStart() ) { // advance to next glyph in cluster ++pGlyphIter; if( pGlyphIter->IsDiacritic() ) continue; // ignore diacritics // get leftmost x-extent of this glyph long nXPos = pGlyphIter->maLinearPos.X(); if( nXPosMin > nXPos ) nXPosMin = nXPos; // get rightmost x-extent of this glyph nXPos += pGlyphIter->mnNewWidth; if( nXPosMax < nXPos ) nXPosMax = nXPos; } // when the current cluster overlaps with the next one assume // rightmost cluster edge is the leftmost edge of next cluster // for clusters that do not have x-sorted glyphs // TODO: avoid recalculation of left bound in next cluster iteration for( GlyphVector::const_iterator pN = pGlyphIter; ++pN != end; ) { if( pN->IsClusterStart() ) break; if( pN->IsDiacritic() ) continue; // ignore diacritics if( nXPosMax > pN->maLinearPos.X() ) nXPosMax = pN->maLinearPos.X(); } if( nXPosMax < nXPosMin ) nXPosMin = nXPosMax = 0; // character width is sum of glyph cluster widths pCharWidths[n] += nXPosMax - nXPosMin; } // TODO: distribute the cluster width proportionally to the characters // clusters (e.g. ligatures) correspond to more than one char index, // so some character widths are still uninitialized. This is solved // by setting the first charwidth of the cluster to the cluster width return true; } DeviceCoordinate GenericSalLayout::FillDXArray( DeviceCoordinate* pCharWidths ) const { if( pCharWidths ) if( !GetCharWidths( pCharWidths ) ) return 0; return GetTextWidth(); } // the text width is the maximum logical extent of all glyphs DeviceCoordinate GenericSalLayout::GetTextWidth() const { if( m_GlyphItems.empty() ) return 0; // initialize the extent DeviceCoordinate nMinPos = 0; DeviceCoordinate nMaxPos = 0; for( GlyphVector::const_iterator pGlyphIter = m_GlyphItems.begin(), end = m_GlyphItems.end(); pGlyphIter != end ; ++pGlyphIter ) { // update the text extent with the glyph extent DeviceCoordinate nXPos = pGlyphIter->maLinearPos.X(); if( nMinPos > nXPos ) nMinPos = nXPos; nXPos += pGlyphIter->mnNewWidth - pGlyphIter->mnXOffset; if( nMaxPos < nXPos ) nMaxPos = nXPos; } DeviceCoordinate nWidth = nMaxPos - nMinPos; return nWidth; } void GenericSalLayout::AdjustLayout( ImplLayoutArgs& rArgs ) { SalLayout::AdjustLayout( rArgs ); if( rArgs.mpDXArray ) ApplyDXArray( rArgs ); else if( rArgs.mnLayoutWidth ) Justify( rArgs.mnLayoutWidth ); } // This DXArray thing is one of the stupidest ideas I have ever seen (I've been // told that it probably a one-to-one mapping of some Windows 3.1 API, which is // telling). // Note: That would be the EMRTEXT structure, which is part of the EMF spec (see // [MS-EMF] section 2.2.5. Basically the field in question is OutputDx, which is: // // "An array of 32-bit unsigned integers that specify the output spacing // between the origins of adjacent character cells in logical units." // // This obviously makes sense for ASCII text (the EMR_EXTTEXTOUTA record), but it // doesn't make sense for Unicode text (the EMR_EXTTEXTOUTW record) because it is // mapping character codes to output spacing, which obviously can cause problems // with CTL. MANY of our concepts are based around Microsoft data structures, this // is obviously one of them and we probably need a rethink about how we go about // this. // To justify a text string, Writer calls OutputDevice::GetTextArray() // to get an array that maps input characters (not glyphs) to their absolute // position, GetTextArray() in turn calls SalLayout::FillDXArray() to get an // array of character widths that it converts to absolute positions. // Writer would then apply justification adjustments to that array of absolute // character positions and return to OutputDevice, which eventually calls // ApplyDXArray(), which needs to extract the individual adjustments for each // character to apply it to corresponding glyphs, and since that information is // already lost it tries to do some heuristics to guess it again. Those // heuristics often fail, and have always been a source of all sorts of weird // text layout bugs, and instead of fixing the broken design a hack after hack // have been applied on top of it, making it a complete mess that nobody // understands. // As you can see by now, this is utterly stupid, why doesn't Writer just send // us the advance width transformations it wants to apply to each character directly // instead of this whole mess? void GenericSalLayout::ApplyDXArray( ImplLayoutArgs& rArgs ) { if( m_GlyphItems.empty()) return; // determine cluster boundaries and x base offset const int nCharCount = rArgs.mnEndCharPos - rArgs.mnMinCharPos; std::unique_ptr const pLogCluster(new int[nCharCount]); size_t i; int n,p; long nBasePointX = -1; if( mnLayoutFlags & SalLayoutFlags::ForFallback ) nBasePointX = 0; for(p = 0; p < nCharCount; ++p ) pLogCluster[ p ] = -1; for( i = 0; i < m_GlyphItems.size(); ++i) { n = m_GlyphItems[i].mnCharPos - rArgs.mnMinCharPos; if( (n < 0) || (nCharCount <= n) ) continue; if( pLogCluster[ n ] < 0 ) pLogCluster[ n ] = i; if( nBasePointX < 0 ) nBasePointX = m_GlyphItems[i].maLinearPos.X(); } // retarget unresolved pLogCluster[n] to a glyph inside the cluster // TODO: better do it while the deleted-glyph markers are still there for( n = 0; n < nCharCount; ++n ) if( (p = pLogCluster[n]) >= 0 ) break; if( n >= nCharCount ) return; for( n = 0; n < nCharCount; ++n ) { if( pLogCluster[ n ] < 0 ) pLogCluster[ n ] = p; else p = pLogCluster[ n ]; } // calculate adjusted cluster widths std::unique_ptr const pNewGlyphWidths(new long[m_GlyphItems.size()]); for( i = 0; i < m_GlyphItems.size(); ++i ) pNewGlyphWidths[ i ] = 0; bool bRTL; for( int nCharPos = p = -1; rArgs.GetNextPos( &nCharPos, &bRTL ); ) { n = nCharPos - rArgs.mnMinCharPos; if( (n < 0) || (nCharCount <= n) ) continue; if( pLogCluster[ n ] >= 0 ) p = pLogCluster[ n ]; if( p >= 0 ) { long nDelta = rArgs.mpDXArray[ n ] ; if( n > 0 ) nDelta -= rArgs.mpDXArray[ n-1 ]; pNewGlyphWidths[ p ] += nDelta * mnUnitsPerPixel; } } // move cluster positions using the adjusted widths long nDelta = 0; long nNewPos = 0; for( i = 0; i < m_GlyphItems.size(); ++i) { if( m_GlyphItems[i].IsClusterStart() ) { // calculate original and adjusted cluster width int nOldClusterWidth = m_GlyphItems[i].mnNewWidth - m_GlyphItems[i].mnXOffset; int nNewClusterWidth = pNewGlyphWidths[i]; size_t j; for( j = i; ++j < m_GlyphItems.size(); ) { if( m_GlyphItems[j].IsClusterStart() ) break; if( !m_GlyphItems[j].IsDiacritic() ) // #i99367# ignore diacritics nOldClusterWidth += m_GlyphItems[j].mnNewWidth - m_GlyphItems[j].mnXOffset; nNewClusterWidth += pNewGlyphWidths[j]; } const int nDiff = nNewClusterWidth - nOldClusterWidth; // adjust cluster glyph widths and positions nDelta = nBasePointX + (nNewPos - m_GlyphItems[i].maLinearPos.X()); if( !m_GlyphItems[i].IsRTLGlyph() ) { // for LTR case extend rightmost glyph in cluster m_GlyphItems[j - 1].mnNewWidth += nDiff; } else { // right align cluster in new space for RTL case m_GlyphItems[i].mnNewWidth += nDiff; nDelta += nDiff; } nNewPos += nNewClusterWidth; } m_GlyphItems[i].maLinearPos.X() += nDelta; } } void GenericSalLayout::Justify( DeviceCoordinate nNewWidth ) { nNewWidth *= mnUnitsPerPixel; DeviceCoordinate nOldWidth = GetTextWidth(); if( !nOldWidth || nNewWidth==nOldWidth ) return; if(m_GlyphItems.empty()) { return; } // find rightmost glyph, it won't get stretched GlyphVector::iterator pGlyphIterRight = m_GlyphItems.begin(); pGlyphIterRight += m_GlyphItems.size() - 1; GlyphVector::iterator pGlyphIter; // count stretchable glyphs int nStretchable = 0; int nMaxGlyphWidth = 0; for(pGlyphIter = m_GlyphItems.begin(); pGlyphIter != pGlyphIterRight; ++pGlyphIter) { if( !pGlyphIter->IsDiacritic() ) ++nStretchable; if( nMaxGlyphWidth < pGlyphIter->mnOrigWidth ) nMaxGlyphWidth = pGlyphIter->mnOrigWidth; } // move rightmost glyph to requested position nOldWidth -= pGlyphIterRight->mnOrigWidth; if( nOldWidth <= 0 ) return; if( nNewWidth < nMaxGlyphWidth) nNewWidth = nMaxGlyphWidth; nNewWidth -= pGlyphIterRight->mnOrigWidth; pGlyphIterRight->maLinearPos.X() = maBasePoint.X() + nNewWidth; // justify glyph widths and positions int nDiffWidth = nNewWidth - nOldWidth; if( nDiffWidth >= 0) // expanded case { // expand width by distributing space between glyphs evenly int nDeltaSum = 0; for( pGlyphIter = m_GlyphItems.begin(); pGlyphIter != pGlyphIterRight; ++pGlyphIter ) { // move glyph to justified position pGlyphIter->maLinearPos.X() += nDeltaSum; // do not stretch non-stretchable glyphs if( pGlyphIter->IsDiacritic() || (nStretchable <= 0) ) continue; // distribute extra space equally to stretchable glyphs int nDeltaWidth = nDiffWidth / nStretchable--; nDiffWidth -= nDeltaWidth; pGlyphIter->mnNewWidth += nDeltaWidth; nDeltaSum += nDeltaWidth; } } else // condensed case { // squeeze width by moving glyphs proportionally double fSqueeze = (double)nNewWidth / nOldWidth; if(m_GlyphItems.size() > 1) { for( pGlyphIter = m_GlyphItems.begin(); ++pGlyphIter != pGlyphIterRight;) { int nX = pGlyphIter->maLinearPos.X() - maBasePoint.X(); nX = (int)(nX * fSqueeze); pGlyphIter->maLinearPos.X() = nX + maBasePoint.X(); } } // adjust glyph widths to new positions for( pGlyphIter = m_GlyphItems.begin(); pGlyphIter != pGlyphIterRight; ++pGlyphIter ) pGlyphIter->mnNewWidth = pGlyphIter[1].maLinearPos.X() - pGlyphIter[0].maLinearPos.X(); } } void GenericSalLayout::ApplyAsianKerning(const OUString& rStr) { const int nLength = rStr.getLength(); long nOffset = 0; for( GlyphVector::iterator pGlyphIter = m_GlyphItems.begin(), pGlyphIterEnd = m_GlyphItems.end(); pGlyphIter != pGlyphIterEnd; ++pGlyphIter ) { const int n = pGlyphIter->mnCharPos; if( n < nLength - 1) { // ignore code ranges that are not affected by asian punctuation compression const sal_Unicode cHere = rStr[n]; if( ((0x3000 != (cHere & 0xFF00)) && (0x2010 != (cHere & 0xFFF0))) || (0xFF00 != (cHere & 0xFF00)) ) continue; const sal_Unicode cNext = rStr[n+1]; if( ((0x3000 != (cNext & 0xFF00)) && (0x2010 != (cNext & 0xFFF0))) || (0xFF00 != (cNext & 0xFF00)) ) continue; // calculate compression values const bool bVertical = false; long nKernFirst = +CalcAsianKerning( cHere, true, bVertical ); long nKernNext = -CalcAsianKerning( cNext, false, bVertical ); // apply punctuation compression to logical glyph widths long nDelta = (nKernFirst < nKernNext) ? nKernFirst : nKernNext; if( nDelta<0 && nKernFirst!=0 && nKernNext!=0 ) { int nGlyphWidth = pGlyphIter->mnOrigWidth; nDelta = (nDelta * nGlyphWidth + 2) / 4; if( pGlyphIter+1 == pGlyphIterEnd ) pGlyphIter->mnNewWidth += nDelta; nOffset += nDelta; } } // adjust the glyph positions to the new glyph widths if( pGlyphIter+1 != pGlyphIterEnd ) pGlyphIter->maLinearPos.X() += nOffset; } } void GenericSalLayout::KashidaJustify( long nKashidaIndex, int nKashidaWidth ) { // TODO: reimplement method when container type for GlyphItems changes // skip if the kashida glyph in the font looks suspicious if( nKashidaWidth <= 0 ) return; // calculate max number of needed kashidas int nKashidaCount = 0; for (GlyphVector::iterator pGlyphIter = m_GlyphItems.begin(); pGlyphIter != m_GlyphItems.end(); ++pGlyphIter) { // only inject kashidas in RTL contexts if( !pGlyphIter->IsRTLGlyph() ) continue; // no kashida-injection for blank justified expansion either if( IsSpacingGlyph( pGlyphIter->maGlyphId) ) continue; // calculate gap, ignore if too small int nGapWidth = pGlyphIter->mnNewWidth - pGlyphIter->mnOrigWidth; // worst case is one kashida even for mini-gaps if( nGapWidth < nKashidaWidth ) continue; nKashidaCount = 0; Point aPos = pGlyphIter->maLinearPos; aPos.X() -= nGapWidth; // cluster is already right aligned int const nCharPos = pGlyphIter->mnCharPos; GlyphVector::iterator pGlyphIter2 = pGlyphIter; for(; nGapWidth > nKashidaWidth; nGapWidth -= nKashidaWidth, ++nKashidaCount ) { pGlyphIter2 = m_GlyphItems.insert(pGlyphIter2, GlyphItem(nCharPos, nKashidaIndex, aPos, GlyphItem::IS_IN_CLUSTER|GlyphItem::IS_RTL_GLYPH, nKashidaWidth )); ++pGlyphIter2; aPos.X() += nKashidaWidth; } // fixup rightmost kashida for gap remainder if( nGapWidth > 0 ) { pGlyphIter2 = m_GlyphItems.insert(pGlyphIter2, GlyphItem(nCharPos, nKashidaIndex, aPos, GlyphItem::IS_IN_CLUSTER|GlyphItem::IS_RTL_GLYPH, nKashidaCount ? nGapWidth : nGapWidth/2 )); ++pGlyphIter2; aPos.X() += nGapWidth; } pGlyphIter = pGlyphIter2; } } void GenericSalLayout::GetCaretPositions( int nMaxIndex, long* pCaretXArray ) const { // initialize result array for (int i = 0; i < nMaxIndex; ++i) pCaretXArray[i] = -1; // calculate caret positions using glyph array for( GlyphVector::const_iterator pGlyphIter = m_GlyphItems.begin(), pGlyphIterEnd = m_GlyphItems.end(); pGlyphIter != pGlyphIterEnd; ++pGlyphIter ) { long nXPos = pGlyphIter->maLinearPos.X(); long nXRight = nXPos + pGlyphIter->mnOrigWidth; int n = pGlyphIter->mnCharPos; int nCurrIdx = 2 * (n - mnMinCharPos); // tdf#86399 if this is not the start of a cluster, don't overwrite the caret bounds of the cluster start if (!pGlyphIter->IsClusterStart() && pCaretXArray[nCurrIdx] != -1) continue; if( !pGlyphIter->IsRTLGlyph() ) { // normal positions for LTR case pCaretXArray[ nCurrIdx ] = nXPos; pCaretXArray[ nCurrIdx+1 ] = nXRight; } else { // reverse positions for RTL case pCaretXArray[ nCurrIdx ] = nXRight; pCaretXArray[ nCurrIdx+1 ] = nXPos; } } } sal_Int32 GenericSalLayout::GetTextBreak( DeviceCoordinate nMaxWidth, DeviceCoordinate nCharExtra, int nFactor ) const { int nCharCapacity = mnEndCharPos - mnMinCharPos; std::unique_ptr const pCharWidths(new DeviceCoordinate[nCharCapacity]); if (!GetCharWidths(pCharWidths.get())) return -1; DeviceCoordinate nWidth = 0; for( int i = mnMinCharPos; i < mnEndCharPos; ++i ) { nWidth += pCharWidths[ i - mnMinCharPos ] * nFactor; if( nWidth > nMaxWidth ) return i; nWidth += nCharExtra; } return -1; } int GenericSalLayout::GetNextGlyphs( int nLen, sal_GlyphId* pGlyphs, Point& rPos, int& nStart, DeviceCoordinate* pGlyphAdvAry, int* pCharPosAry, const PhysicalFontFace** /*pFallbackFonts*/ ) const { GlyphVector::const_iterator pGlyphIter = m_GlyphItems.begin(); GlyphVector::const_iterator pGlyphIterEnd = m_GlyphItems.end(); pGlyphIter += nStart; // find next glyph in substring for(; pGlyphIter != pGlyphIterEnd; ++nStart, ++pGlyphIter ) { int n = pGlyphIter->mnCharPos; if( (mnMinCharPos <= n) && (n < mnEndCharPos) ) break; } // return zero if no more glyph found if( nStart >= (int)m_GlyphItems.size() ) return 0; if( pGlyphIter == pGlyphIterEnd ) return 0; // calculate absolute position in pixel units Point aRelativePos = pGlyphIter->maLinearPos - maBasePoint; // find more glyphs which can be merged into one drawing instruction int nCount = 0; long nYPos = pGlyphIter->maLinearPos.Y(); long nOldFlags = pGlyphIter->maGlyphId; for(;;) { // update return data with glyph info ++nCount; *(pGlyphs++) = pGlyphIter->maGlyphId; if( pCharPosAry ) *(pCharPosAry++) = pGlyphIter->mnCharPos; if( pGlyphAdvAry ) *pGlyphAdvAry = pGlyphIter->mnNewWidth; // break at end of glyph list if( ++nStart >= (int)m_GlyphItems.size() ) break; // break when enough glyphs if( nCount >= nLen ) break; long nGlyphAdvance = pGlyphIter[1].maLinearPos.X() - pGlyphIter->maLinearPos.X(); if( pGlyphAdvAry ) { // override default advance width with correct value *(pGlyphAdvAry++) = nGlyphAdvance; } else { // stop when next x-position is unexpected if( pGlyphIter->mnOrigWidth != nGlyphAdvance ) break; } // advance to next glyph ++pGlyphIter; // stop when next y-position is unexpected if( nYPos != pGlyphIter->maLinearPos.Y() ) break; // stop when no longer in string int n = pGlyphIter->mnCharPos; if( (n < mnMinCharPos) || (mnEndCharPos <= n) ) break; // stop when glyph flags change if( (nOldFlags ^ pGlyphIter->maGlyphId) & GF_FLAGMASK ) break; nOldFlags = pGlyphIter->maGlyphId; // &GF_FLAGMASK not needed for test above } aRelativePos.X() /= mnUnitsPerPixel; aRelativePos.Y() /= mnUnitsPerPixel; rPos = GetDrawPosition( aRelativePos ); return nCount; } void GenericSalLayout::MoveGlyph( int nStart, long nNewXPos ) { if( nStart >= (int)m_GlyphItems.size() ) return; GlyphVector::iterator pGlyphIter = m_GlyphItems.begin(); pGlyphIter += nStart; // the nNewXPos argument determines the new cell position // as RTL-glyphs are right justified in their cell // the cell position needs to be adjusted to the glyph position if( pGlyphIter->IsRTLGlyph() ) nNewXPos += pGlyphIter->mnNewWidth - pGlyphIter->mnOrigWidth; // calculate the x-offset to the old position long nXDelta = nNewXPos - pGlyphIter->maLinearPos.X(); // adjust all following glyph positions if needed if( nXDelta != 0 ) { for( GlyphVector::iterator pGlyphIterEnd = m_GlyphItems.end(); pGlyphIter != pGlyphIterEnd; ++pGlyphIter ) { pGlyphIter->maLinearPos.X() += nXDelta; } } } void GenericSalLayout::DropGlyph( int nStart ) { if( nStart >= (int)m_GlyphItems.size()) return; GlyphVector::iterator pGlyphIter = m_GlyphItems.begin(); pGlyphIter += nStart; pGlyphIter->maGlyphId = GF_DROPPED; pGlyphIter->mnCharPos = -1; } void GenericSalLayout::Simplify( bool bIsBase ) { const sal_GlyphId nDropMarker = bIsBase ? GF_DROPPED : 0; // remove dropped glyphs inplace size_t j = 0; for(size_t i = 0; i < m_GlyphItems.size(); i++ ) { if( m_GlyphItems[i].maGlyphId == nDropMarker ) continue; if( i != j ) { m_GlyphItems[j] = m_GlyphItems[i]; } j += 1; } m_GlyphItems.erase(m_GlyphItems.begin() + j, m_GlyphItems.end()); } // make sure GlyphItems are sorted left to right void GenericSalLayout::SortGlyphItems() { // move cluster components behind their cluster start (especially for RTL) // using insertion sort because the glyph items are "almost sorted" for( GlyphVector::iterator pGlyphIter = m_GlyphItems.begin(), pGlyphIterEnd = m_GlyphItems.end(); pGlyphIter != pGlyphIterEnd; ++pGlyphIter ) { // find a cluster starting with a diacritic if( !pGlyphIter->IsDiacritic() ) continue; if( !pGlyphIter->IsClusterStart() ) continue; for( GlyphVector::iterator pBaseGlyph = pGlyphIter; ++pBaseGlyph != pGlyphIterEnd; ) { // find the base glyph matching to the misplaced diacritic if( pBaseGlyph->IsClusterStart() ) break; if( pBaseGlyph->IsDiacritic() ) continue; // found the matching base glyph // => this base glyph becomes the new cluster start iter_swap(pGlyphIter, pBaseGlyph); // update glyph flags of swapped glyphitems pGlyphIter->mnFlags &= ~GlyphItem::IS_IN_CLUSTER; pBaseGlyph->mnFlags |= GlyphItem::IS_IN_CLUSTER; // prepare for checking next cluster pGlyphIter = pBaseGlyph; break; } } } MultiSalLayout::MultiSalLayout( SalLayout& rBaseLayout, const PhysicalFontFace* pBaseFont ) : SalLayout() , mnLevel( 1 ) , mbInComplete( false ) { //maFallbackRuns[0].Clear(); mpFallbackFonts[ 0 ] = pBaseFont; mpLayouts[ 0 ] = &rBaseLayout; mnUnitsPerPixel = rBaseLayout.GetUnitsPerPixel(); } void MultiSalLayout::SetInComplete(bool bInComplete) { mbInComplete = bInComplete; maFallbackRuns[mnLevel-1] = ImplLayoutRuns(); } MultiSalLayout::~MultiSalLayout() { for( int i = 0; i < mnLevel; ++i ) mpLayouts[ i ]->Release(); } bool MultiSalLayout::AddFallback( SalLayout& rFallback, ImplLayoutRuns& rFallbackRuns, const PhysicalFontFace* pFallbackFont ) { if( mnLevel >= MAX_FALLBACK ) return false; mpFallbackFonts[ mnLevel ] = pFallbackFont; mpLayouts[ mnLevel ] = &rFallback; maFallbackRuns[ mnLevel-1 ] = rFallbackRuns; ++mnLevel; return true; } bool MultiSalLayout::LayoutText( ImplLayoutArgs& rArgs ) { if( mnLevel <= 1 ) return false; if (!mbInComplete) maFallbackRuns[ mnLevel-1 ] = rArgs.maRuns; return true; } void MultiSalLayout::AdjustLayout( ImplLayoutArgs& rArgs ) { SalLayout::AdjustLayout( rArgs ); ImplLayoutArgs aMultiArgs = rArgs; std::unique_ptr pJustificationArray; if( !rArgs.mpDXArray && rArgs.mnLayoutWidth ) { // for stretched text in a MultiSalLayout the target width needs to be // distributed by individually adjusting its virtual character widths DeviceCoordinate nTargetWidth = aMultiArgs.mnLayoutWidth; nTargetWidth *= mnUnitsPerPixel; // convert target width to base font units aMultiArgs.mnLayoutWidth = 0; // we need to get the original unmodified layouts ready for( int n = 0; n < mnLevel; ++n ) mpLayouts[n]->SalLayout::AdjustLayout( aMultiArgs ); // then we can measure the unmodified metrics int nCharCount = rArgs.mnEndCharPos - rArgs.mnMinCharPos; pJustificationArray.reset(new DeviceCoordinate[nCharCount]); FillDXArray( pJustificationArray.get() ); // #i17359# multilayout is not simplified yet, so calculating the // unjustified width needs handholding; also count the number of // stretchable virtual char widths DeviceCoordinate nOrigWidth = 0; int nStretchable = 0; for( int i = 0; i < nCharCount; ++i ) { // convert array from widths to sum of widths nOrigWidth += pJustificationArray[i]; if( pJustificationArray[i] > 0 ) ++nStretchable; } // now we are able to distribute the extra width over the virtual char widths if( nOrigWidth && (nTargetWidth != nOrigWidth) ) { DeviceCoordinate nDiffWidth = nTargetWidth - nOrigWidth; DeviceCoordinate nWidthSum = 0; for( int i = 0; i < nCharCount; ++i ) { DeviceCoordinate nJustWidth = pJustificationArray[i]; if( (nJustWidth > 0) && (nStretchable > 0) ) { DeviceCoordinate nDeltaWidth = nDiffWidth / nStretchable; nJustWidth += nDeltaWidth; nDiffWidth -= nDeltaWidth; --nStretchable; } nWidthSum += nJustWidth; pJustificationArray[i] = nWidthSum; } if( nWidthSum != nTargetWidth ) pJustificationArray[ nCharCount-1 ] = nTargetWidth; // the justification array is still in base level units // => convert it to pixel units if( mnUnitsPerPixel > 1 ) { for( int i = 0; i < nCharCount; ++i ) { DeviceCoordinate nVal = pJustificationArray[ i ]; nVal += (mnUnitsPerPixel + 1) / 2; pJustificationArray[ i ] = nVal / mnUnitsPerPixel; } } // change the mpDXArray temporarily (just for the justification) aMultiArgs.mpDXArray = pJustificationArray.get(); } } // Compute rtl flags, since in some scripts glyphs/char order can be // reversed for a few character sequencies e.g. Myanmar std::vector vRtl(rArgs.mnEndCharPos - rArgs.mnMinCharPos, false); rArgs.ResetPos(); bool bRtl; int nRunStart, nRunEnd; while (rArgs.GetNextRun(&nRunStart, &nRunEnd, &bRtl)) { if (bRtl) std::fill(vRtl.begin() + (nRunStart - rArgs.mnMinCharPos), vRtl.begin() + (nRunEnd - rArgs.mnMinCharPos), true); } rArgs.ResetPos(); // prepare "merge sort" int nStartOld[ MAX_FALLBACK ]; int nStartNew[ MAX_FALLBACK ]; int nCharPos[ MAX_FALLBACK ]; DeviceCoordinate nGlyphAdv[ MAX_FALLBACK ]; int nValid[ MAX_FALLBACK ] = {0}; sal_GlyphId nDummy; Point aPos; int nLevel = 0, n; for( n = 0; n < mnLevel; ++n ) { // now adjust the individual components if( n > 0 ) { aMultiArgs.maRuns = maFallbackRuns[ n-1 ]; aMultiArgs.mnFlags |= SalLayoutFlags::ForFallback; } mpLayouts[n]->AdjustLayout( aMultiArgs ); // disable glyph-injection for glyph-fallback SalLayout iteration SalLayout::DisableGlyphInjection( true ); // remove unused parts of component if( n > 0 ) { if (mbInComplete && (n == mnLevel-1)) mpLayouts[n]->Simplify( true ); else mpLayouts[n]->Simplify( false ); } // prepare merging components nStartNew[ nLevel ] = nStartOld[ nLevel ] = 0; nValid[ nLevel ] = mpLayouts[n]->GetNextGlyphs( 1, &nDummy, aPos, nStartNew[ nLevel ], &nGlyphAdv[ nLevel ], &nCharPos[ nLevel ] ); #ifdef MULTI_SL_DEBUG if (nValid[nLevel]) fprintf(mslLog(), "layout[%d]->GetNextGlyphs %d,%d x%d a%d c%ld %x\n", n, nStartOld[nLevel], nStartNew[nLevel], aPos.X(), (long)nGlyphAdv[nLevel], nCharPos[nLevel], rArgs.mrStr[nCharPos[nLevel]]); #endif if( (n > 0) && !nValid[ nLevel ] ) { // an empty fallback layout can be released mpLayouts[n]->Release(); } else { // reshuffle used fallbacks if needed if( nLevel != n ) { mpLayouts[ nLevel ] = mpLayouts[ n ]; mpFallbackFonts[ nLevel ] = mpFallbackFonts[ n ]; maFallbackRuns[ nLevel ] = maFallbackRuns[ n ]; } ++nLevel; } } mnLevel = nLevel; // prepare merge the fallback levels long nXPos = 0; double fUnitMul = 1.0; for( n = 0; n < nLevel; ++n ) maFallbackRuns[n].ResetPos(); // get the next codepoint index that needs fallback int nActiveCharPos = nCharPos[0]; int nActiveCharIndex = nActiveCharPos - mnMinCharPos; // get the end index of the active run int nLastRunEndChar = (nActiveCharIndex >= 0 && vRtl[nActiveCharIndex]) ? rArgs.mnEndCharPos : rArgs.mnMinCharPos - 1; int nRunVisibleEndChar = nCharPos[0]; // merge the fallback levels while( nValid[0] && (nLevel > 0)) { // find best fallback level for( n = 0; n < nLevel; ++n ) if( nValid[n] && !maFallbackRuns[n].PosIsInAnyRun( nActiveCharPos ) ) // fallback level n wins when it requested no further fallback break; int nFBLevel = n; if( n < nLevel ) { // use base(n==0) or fallback(n>=1) level fUnitMul = mnUnitsPerPixel; fUnitMul /= mpLayouts[n]->GetUnitsPerPixel(); long nNewPos = static_cast(nXPos/fUnitMul + 0.5); mpLayouts[n]->MoveGlyph( nStartOld[n], nNewPos ); } else { n = 0; // keep NotDef in base level fUnitMul = 1.0; } if( n > 0 ) { // drop the NotDef glyphs in the base layout run if a fallback run exists while ( (maFallbackRuns[ n-1 ].PosIsInRun( nCharPos[0] ) ) && (!maFallbackRuns[ n ].PosIsInAnyRun( nCharPos[0] ) ) ) { mpLayouts[0]->DropGlyph( nStartOld[0] ); nStartOld[0] = nStartNew[0]; nValid[0] = mpLayouts[0]->GetNextGlyphs( 1, &nDummy, aPos, nStartNew[0], &nGlyphAdv[0], &nCharPos[0] ); #ifdef MULTI_SL_DEBUG if (nValid[0]) fprintf(mslLog(), "layout[0]->GetNextGlyphs %d,%d x%d a%ld c%d %x\n", nStartOld[0], nStartNew[0], aPos.X(), (long)nGlyphAdv[0], nCharPos[0], rArgs.mrStr[nCharPos[0]]); #endif if( !nValid[0] ) break; } } // skip to end of layout run and calculate its advance width DeviceCoordinate nRunAdvance = 0; bool bKeepNotDef = (nFBLevel >= nLevel); for(;;) { nRunAdvance += nGlyphAdv[n]; // proceed to next glyph nStartOld[n] = nStartNew[n]; int nOrigCharPos = nCharPos[n]; nValid[n] = mpLayouts[n]->GetNextGlyphs( 1, &nDummy, aPos, nStartNew[n], &nGlyphAdv[n], &nCharPos[n] ); #ifdef MULTI_SL_DEBUG if (nValid[n]) fprintf(mslLog(), "layout[%d]->GetNextGlyphs %d,%d a%ld c%d %x\n", n, nStartOld[n], nStartNew[n], (long)nGlyphAdv[n], nCharPos[n], rArgs.mrStr[nCharPos[n]]); #endif // break after last glyph of active layout if( !nValid[n] ) { // performance optimization (when a fallback layout is no longer needed) if( n >= nLevel-1 ) --nLevel; break; } //If the next character is one which belongs to the next level, then we //are finished here for now, and we'll pick up after the next level has //been processed if ((n+1 < nLevel) && (nCharPos[n] != nOrigCharPos)) { if (nOrigCharPos < nCharPos[n]) { if (nCharPos[n+1] > nOrigCharPos && (nCharPos[n+1] < nCharPos[n])) break; } else if (nOrigCharPos > nCharPos[n]) { if (nCharPos[n+1] > nCharPos[n] && (nCharPos[n+1] < nOrigCharPos)) break; } } // break at end of layout run if( n > 0 ) { // skip until end of fallback run if( !maFallbackRuns[n-1].PosIsInRun( nCharPos[n] ) ) break; } else { // break when a fallback is needed and available bool bNeedFallback = maFallbackRuns[0].PosIsInRun( nCharPos[0] ); if( bNeedFallback ) if( !maFallbackRuns[ nLevel-1 ].PosIsInRun( nCharPos[0] ) ) break; // break when change from resolved to unresolved base layout run if( bKeepNotDef && !bNeedFallback ) { maFallbackRuns[0].NextRun(); break; } bKeepNotDef = bNeedFallback; } // check for reordered glyphs if (aMultiArgs.mpDXArray && nRunVisibleEndChar < mnEndCharPos && nRunVisibleEndChar >= mnMinCharPos && nCharPos[n] < mnEndCharPos && nCharPos[n] >= mnMinCharPos) { if (vRtl[nActiveCharPos - mnMinCharPos]) { if (aMultiArgs.mpDXArray[nRunVisibleEndChar-mnMinCharPos] >= aMultiArgs.mpDXArray[nCharPos[n] - mnMinCharPos]) { nRunVisibleEndChar = nCharPos[n]; } } else if (aMultiArgs.mpDXArray[nRunVisibleEndChar-mnMinCharPos] <= aMultiArgs.mpDXArray[nCharPos[n] - mnMinCharPos]) { nRunVisibleEndChar = nCharPos[n]; } } } // if a justification array is available // => use it directly to calculate the corresponding run width if( aMultiArgs.mpDXArray ) { // the run advance is the width from the first char // in the run to the first char in the next run nRunAdvance = 0; #ifdef MULTI_SL_DEBUG const bool bLTR = !(vRtl[nActiveCharPos - mnMinCharPos]);//(nActiveCharPos < nCharPos[0]); int nOldRunAdv = 0; int nDXIndex = nCharPos[0] - mnMinCharPos - bLTR; if( nDXIndex >= 0 ) nOldRunAdv += aMultiArgs.mpDXArray[ nDXIndex ]; nDXIndex = nActiveCharPos - mnMinCharPos - bLTR; if( nDXIndex >= 0 ) nOldRunAdv -= aMultiArgs.mpDXArray[ nDXIndex ]; if( !bLTR ) nOldRunAdv = -nOldRunAdv; #endif nActiveCharIndex = nActiveCharPos - mnMinCharPos; if (nActiveCharIndex >= 0 && vRtl[nActiveCharIndex]) { if (nRunVisibleEndChar > mnMinCharPos && nRunVisibleEndChar <= mnEndCharPos) nRunAdvance -= aMultiArgs.mpDXArray[nRunVisibleEndChar - 1 - mnMinCharPos]; if (nLastRunEndChar > mnMinCharPos && nLastRunEndChar <= mnEndCharPos) nRunAdvance += aMultiArgs.mpDXArray[nLastRunEndChar - 1 - mnMinCharPos]; #ifdef MULTI_SL_DEBUG fprintf(mslLog(), "rtl visible %d-%d,%d-%d adv%d(%d)\n", nLastRunEndChar-1, nRunVisibleEndChar-1, nActiveCharPos - bLTR, nCharPos[0] - bLTR, nRunAdvance, nOldRunAdv); #endif } else { if (nRunVisibleEndChar >= mnMinCharPos) nRunAdvance += aMultiArgs.mpDXArray[nRunVisibleEndChar - mnMinCharPos]; if (nLastRunEndChar >= mnMinCharPos) nRunAdvance -= aMultiArgs.mpDXArray[nLastRunEndChar - mnMinCharPos]; #ifdef MULTI_SL_DEBUG fprintf(mslLog(), "visible %d-%d,%d-%d adv%d(%d)\n", nLastRunEndChar, nRunVisibleEndChar, nActiveCharPos - bLTR, nCharPos[0] - bLTR, nRunAdvance, nOldRunAdv); #endif } nLastRunEndChar = nRunVisibleEndChar; nRunVisibleEndChar = nCharPos[0]; // the requested width is still in pixel units // => convert it to base level font units nRunAdvance *= mnUnitsPerPixel; } else { // the measured width is still in fallback font units // => convert it to base level font units if( n > 0 ) // optimization: because (fUnitMul==1.0) for (n==0) nRunAdvance = static_cast(nRunAdvance*fUnitMul + 0.5); } // calculate new x position (in base level units) nXPos += nRunAdvance; // prepare for next fallback run nActiveCharPos = nCharPos[0]; // it essential that the runs don't get ahead of themselves and in the // if( bKeepNotDef && !bNeedFallback ) statement above, the next run may // have already been reached on the base level for( int i = nFBLevel; --i >= 0;) { if (maFallbackRuns[i].GetRun(&nRunStart, &nRunEnd, &bRtl)) { if (bRtl) { if (nRunStart > nActiveCharPos) maFallbackRuns[i].NextRun(); } else { if (nRunEnd <= nActiveCharPos) maFallbackRuns[i].NextRun(); } } } } mpLayouts[0]->Simplify( true ); // reenable glyph-injection for( n = 0; n < mnLevel; ++n ) SalLayout::DisableGlyphInjection( false ); } void MultiSalLayout::InitFont() const { if( mnLevel > 0 ) mpLayouts[0]->InitFont(); } void MultiSalLayout::DrawText( SalGraphics& rGraphics ) const { for( int i = mnLevel; --i >= 0; ) { SalLayout& rLayout = *mpLayouts[ i ]; rLayout.DrawBase() += maDrawBase; rLayout.DrawOffset() += maDrawOffset; rLayout.InitFont(); rLayout.DrawText( rGraphics ); rLayout.DrawOffset() -= maDrawOffset; rLayout.DrawBase() -= maDrawBase; } // NOTE: now the baselevel font is active again } sal_Int32 MultiSalLayout::GetTextBreak( DeviceCoordinate nMaxWidth, DeviceCoordinate nCharExtra, int nFactor ) const { if( mnLevel <= 0 ) return -1; if( mnLevel == 1 ) return mpLayouts[0]->GetTextBreak( nMaxWidth, nCharExtra, nFactor ); int nCharCount = mnEndCharPos - mnMinCharPos; std::unique_ptr const pCharWidths(new DeviceCoordinate[2 * nCharCount]); mpLayouts[0]->FillDXArray( pCharWidths.get() ); for( int n = 1; n < mnLevel; ++n ) { SalLayout& rLayout = *mpLayouts[ n ]; rLayout.FillDXArray( &pCharWidths[nCharCount] ); double fUnitMul = mnUnitsPerPixel; fUnitMul /= rLayout.GetUnitsPerPixel(); for( int i = 0; i < nCharCount; ++i ) { DeviceCoordinate w = pCharWidths[ i + nCharCount ]; w = (DeviceCoordinate)(w * fUnitMul + 0.5); pCharWidths[ i ] += w; } } DeviceCoordinate nWidth = 0; for( int i = 0; i < nCharCount; ++i ) { nWidth += pCharWidths[ i ] * nFactor; if( nWidth > nMaxWidth ) return (i + mnMinCharPos); nWidth += nCharExtra; } return -1; } DeviceCoordinate MultiSalLayout::FillDXArray( DeviceCoordinate* pCharWidths ) const { DeviceCoordinate nMaxWidth = 0; // prepare merging of fallback levels std::unique_ptr pTempWidths; const int nCharCount = mnEndCharPos - mnMinCharPos; if( pCharWidths ) { for( int i = 0; i < nCharCount; ++i ) pCharWidths[i] = 0; pTempWidths.reset(new DeviceCoordinate[nCharCount]); } for( int n = mnLevel; --n >= 0; ) { // query every fallback level DeviceCoordinate nTextWidth = mpLayouts[n]->FillDXArray( pTempWidths.get() ); if( !nTextWidth ) continue; // merge results from current level double fUnitMul = mnUnitsPerPixel; fUnitMul /= mpLayouts[n]->GetUnitsPerPixel(); nTextWidth = (DeviceCoordinate)(nTextWidth * fUnitMul + 0.5); if( nMaxWidth < nTextWidth ) nMaxWidth = nTextWidth; if( !pCharWidths ) continue; // calculate virtual char widths using most probable fallback layout for( int i = 0; i < nCharCount; ++i ) { // #i17359# restriction: // one char cannot be resolved from different fallbacks if( pCharWidths[i] != 0 ) continue; DeviceCoordinate nCharWidth = pTempWidths[i]; if( !nCharWidth ) continue; nCharWidth = (DeviceCoordinate)(nCharWidth * fUnitMul + 0.5); pCharWidths[i] = nCharWidth; } } return nMaxWidth; } void MultiSalLayout::GetCaretPositions( int nMaxIndex, long* pCaretXArray ) const { SalLayout& rLayout = *mpLayouts[ 0 ]; rLayout.GetCaretPositions( nMaxIndex, pCaretXArray ); if( mnLevel > 1 ) { std::unique_ptr const pTempPos(new long[nMaxIndex]); for( int n = 1; n < mnLevel; ++n ) { mpLayouts[ n ]->GetCaretPositions( nMaxIndex, pTempPos.get() ); double fUnitMul = mnUnitsPerPixel; fUnitMul /= mpLayouts[n]->GetUnitsPerPixel(); for( int i = 0; i < nMaxIndex; ++i ) if( pTempPos[i] >= 0 ) { long w = pTempPos[i]; w = static_cast(w*fUnitMul + 0.5); pCaretXArray[i] = w; } } } } int MultiSalLayout::GetNextGlyphs( int nLen, sal_GlyphId* pGlyphIdxAry, Point& rPos, int& nStart, DeviceCoordinate* pGlyphAdvAry, int* pCharPosAry, const PhysicalFontFace** pFallbackFonts ) const { // for multi-level fallback only single glyphs should be used if( mnLevel > 1 && nLen > 1 ) nLen = 1; // NOTE: nStart is tagged with current font index int nLevel = static_cast(nStart) >> GF_FONTSHIFT; nStart &= ~GF_FONTMASK; for(; nLevel < mnLevel; ++nLevel, nStart=0 ) { SalLayout& rLayout = *mpLayouts[ nLevel ]; rLayout.InitFont(); int nRetVal = rLayout.GetNextGlyphs( nLen, pGlyphIdxAry, rPos, nStart, pGlyphAdvAry, pCharPosAry ); if( nRetVal ) { int nFontTag = nLevel << GF_FONTSHIFT; nStart |= nFontTag; double fUnitMul = mnUnitsPerPixel; fUnitMul /= mpLayouts[nLevel]->GetUnitsPerPixel(); for( int i = 0; i < nRetVal; ++i ) { if( pGlyphAdvAry ) { DeviceCoordinate w = pGlyphAdvAry[i]; w = static_cast(w * fUnitMul + 0.5); pGlyphAdvAry[i] = w; } pGlyphIdxAry[ i ] |= nFontTag; if( pFallbackFonts ) { pFallbackFonts[ i ] = mpFallbackFonts[ nLevel ]; } } rPos += maDrawBase; rPos += maDrawOffset; return nRetVal; } } // #111016# reset to base level font when done mpLayouts[0]->InitFont(); return 0; } bool MultiSalLayout::GetOutline( SalGraphics& rGraphics, ::basegfx::B2DPolyPolygonVector& rPPV ) const { bool bRet = false; for( int i = mnLevel; --i >= 0; ) { SalLayout& rLayout = *mpLayouts[ i ]; rLayout.DrawBase() = maDrawBase; rLayout.DrawOffset() += maDrawOffset; rLayout.InitFont(); bRet |= rLayout.GetOutline( rGraphics, rPPV ); rLayout.DrawOffset() -= maDrawOffset; } return bRet; } std::shared_ptr SalLayout::CreateTextLayoutCache( OUString const&) const { return nullptr; // by default, nothing to cache } /* vim:set shiftwidth=4 softtabstop=4 expandtab: */