1 /*
2  * Copyright (C) 2007, 2008, 2009, 2010 Apple Inc. All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS''
14  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
15  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
16  * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS
17  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
18  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
19  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
20  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
21  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
22  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
23  * THE POSSIBILITY OF SUCH DAMAGE.
24  */
25 
26 #include "config.h"
27 #include "UniscribeController.h"
28 #include "Font.h"
29 #include "SimpleFontData.h"
30 #include "TextRun.h"
31 #include <wtf/MathExtras.h>
32 
33 using namespace std;
34 
35 namespace WebCore {
36 
37 // FIXME: Rearchitect this to be more like WidthIterator in Font.cpp.  Have an advance() method
38 // that does stuff in that method instead of doing everything in the constructor.  Have advance()
39 // take the GlyphBuffer as an arg so that we don't have to populate the glyph buffer when
40 // measuring.
UniscribeController(const Font * font,const TextRun & run,HashSet<const SimpleFontData * > * fallbackFonts)41 UniscribeController::UniscribeController(const Font* font, const TextRun& run, HashSet<const SimpleFontData*>* fallbackFonts)
42     : m_font(*font)
43     , m_run(run)
44     , m_fallbackFonts(fallbackFonts)
45     , m_minGlyphBoundingBoxX(numeric_limits<float>::max())
46     , m_maxGlyphBoundingBoxX(numeric_limits<float>::min())
47     , m_minGlyphBoundingBoxY(numeric_limits<float>::max())
48     , m_maxGlyphBoundingBoxY(numeric_limits<float>::min())
49     , m_end(run.length())
50     , m_currentCharacter(0)
51     , m_runWidthSoFar(0)
52     , m_padding(run.expansion())
53     , m_computingOffsetPosition(false)
54     , m_includePartialGlyphs(false)
55     , m_offsetX(0)
56     , m_offsetPosition(0)
57 {
58     if (!m_padding)
59         m_padPerSpace = 0;
60     else {
61         float numSpaces = 0;
62         for (int s = 0; s < m_run.length(); s++) {
63             if (Font::treatAsSpace(m_run[s]))
64                 numSpaces++;
65         }
66 
67         if (numSpaces == 0)
68             m_padPerSpace = 0;
69         else
70             m_padPerSpace = m_padding / numSpaces;
71     }
72 
73     // Null out our uniscribe structs
74     resetControlAndState();
75 }
76 
offsetForPosition(int x,bool includePartialGlyphs)77 int UniscribeController::offsetForPosition(int x, bool includePartialGlyphs)
78 {
79     m_computingOffsetPosition = true;
80     m_includePartialGlyphs = includePartialGlyphs;
81     m_offsetX = x;
82     m_offsetPosition = 0;
83     advance(m_run.length());
84     if (m_computingOffsetPosition) {
85         // The point is to the left or to the right of the entire run.
86         if (m_offsetX >= m_runWidthSoFar && m_run.ltr() || m_offsetX < 0 && m_run.rtl())
87             m_offsetPosition = m_end;
88     }
89     m_computingOffsetPosition = false;
90     return m_offsetPosition;
91 }
92 
advance(unsigned offset,GlyphBuffer * glyphBuffer)93 void UniscribeController::advance(unsigned offset, GlyphBuffer* glyphBuffer)
94 {
95     // FIXME: We really want to be using a newer version of Uniscribe that supports the new OpenType
96     // functions.  Those functions would allow us to turn off kerning and ligatures.  Without being able
97     // to do that, we will have buggy line breaking and metrics when simple and complex text are close
98     // together (the complex code path will narrow the text because of kerning and ligatures and then
99     // when bidi processing splits into multiple runs, the simple portions will get wider and cause us to
100     // spill off the edge of a line).
101     if (static_cast<int>(offset) > m_end)
102         offset = m_end;
103 
104     int length = offset - m_currentCharacter;
105     if (length <= 0)
106         return;
107 
108     // Itemize the string.
109     const UChar* cp = m_run.data(m_currentCharacter);
110     unsigned baseCharacter = m_currentCharacter;
111 
112     // We break up itemization of the string by fontData and (if needed) the use of small caps.
113 
114     // FIXME: It's inconsistent that we use logical order when itemizing, since this
115     // does not match normal RTL.
116 
117     // FIXME: This function should decode surrogate pairs. Currently it makes little difference that
118     // it does not because the font cache on Windows does not support non-BMP characters.
119     Vector<UChar, 256> smallCapsBuffer;
120     if (m_font.isSmallCaps())
121         smallCapsBuffer.resize(length);
122 
123     unsigned indexOfFontTransition = m_run.rtl() ? length - 1 : 0;
124     const UChar* curr = m_run.rtl() ? cp + length  - 1 : cp;
125     const UChar* end = m_run.rtl() ? cp - 1 : cp + length;
126 
127     const SimpleFontData* fontData;
128     const SimpleFontData* nextFontData = m_font.glyphDataForCharacter(*curr, false).fontData;
129 
130     UChar newC = 0;
131 
132     bool isSmallCaps;
133     bool nextIsSmallCaps = m_font.isSmallCaps() && !(U_GET_GC_MASK(*curr) & U_GC_M_MASK) && (newC = u_toupper(*curr)) != *curr;
134 
135     if (nextIsSmallCaps)
136         smallCapsBuffer[curr - cp] = newC;
137 
138     while (true) {
139         curr = m_run.rtl() ? curr - 1 : curr + 1;
140         if (curr == end)
141             break;
142 
143         fontData = nextFontData;
144         isSmallCaps = nextIsSmallCaps;
145         int index = curr - cp;
146         UChar c = *curr;
147 
148         bool forceSmallCaps = isSmallCaps && (U_GET_GC_MASK(c) & U_GC_M_MASK);
149         nextFontData = m_font.glyphDataForCharacter(*curr, false, forceSmallCaps ? SmallCapsVariant : AutoVariant).fontData;
150         if (m_font.isSmallCaps()) {
151             nextIsSmallCaps = forceSmallCaps || (newC = u_toupper(c)) != c;
152             if (nextIsSmallCaps)
153                 smallCapsBuffer[index] = forceSmallCaps ? c : newC;
154         }
155 
156         if (m_fallbackFonts && nextFontData != fontData && fontData != m_font.primaryFont())
157             m_fallbackFonts->add(fontData);
158 
159         if (nextFontData != fontData || nextIsSmallCaps != isSmallCaps) {
160             int itemStart = m_run.rtl() ? index + 1 : indexOfFontTransition;
161             int itemLength = m_run.rtl() ? indexOfFontTransition - index : index - indexOfFontTransition;
162             m_currentCharacter = baseCharacter + itemStart;
163             itemizeShapeAndPlace((isSmallCaps ? smallCapsBuffer.data() : cp) + itemStart, itemLength, fontData, glyphBuffer);
164             indexOfFontTransition = index;
165         }
166     }
167 
168     int itemLength = m_run.rtl() ? indexOfFontTransition + 1 : length - indexOfFontTransition;
169     if (itemLength) {
170         if (m_fallbackFonts && nextFontData != m_font.primaryFont())
171             m_fallbackFonts->add(nextFontData);
172 
173         int itemStart = m_run.rtl() ? 0 : indexOfFontTransition;
174         m_currentCharacter = baseCharacter + itemStart;
175         itemizeShapeAndPlace((nextIsSmallCaps ? smallCapsBuffer.data() : cp) + itemStart, itemLength, nextFontData, glyphBuffer);
176     }
177 
178     m_currentCharacter = baseCharacter + length;
179 }
180 
itemizeShapeAndPlace(const UChar * cp,unsigned length,const SimpleFontData * fontData,GlyphBuffer * glyphBuffer)181 void UniscribeController::itemizeShapeAndPlace(const UChar* cp, unsigned length, const SimpleFontData* fontData, GlyphBuffer* glyphBuffer)
182 {
183     // ScriptItemize (in Windows XP versions prior to SP2) can overflow by 1.  This is why there is an extra empty item
184     // hanging out at the end of the array
185     m_items.resize(6);
186     int numItems = 0;
187     while (ScriptItemize(cp, length, m_items.size() - 1, &m_control, &m_state, m_items.data(), &numItems) == E_OUTOFMEMORY) {
188         m_items.resize(m_items.size() * 2);
189         resetControlAndState();
190     }
191     m_items.resize(numItems + 1);
192 
193     if (m_run.rtl()) {
194         for (int i = m_items.size() - 2; i >= 0; i--) {
195             if (!shapeAndPlaceItem(cp, i, fontData, glyphBuffer))
196                 return;
197         }
198     } else {
199         for (unsigned i = 0; i < m_items.size() - 1; i++) {
200             if (!shapeAndPlaceItem(cp, i, fontData, glyphBuffer))
201                 return;
202         }
203     }
204 }
205 
resetControlAndState()206 void UniscribeController::resetControlAndState()
207 {
208     memset(&m_control, 0, sizeof(SCRIPT_CONTROL));
209     memset(&m_state, 0, sizeof(SCRIPT_STATE));
210 
211     // Set up the correct direction for the run.
212     m_state.uBidiLevel = m_run.rtl();
213 
214     // Lock the correct directional override.
215     m_state.fOverrideDirection = m_run.directionalOverride();
216 }
217 
shapeAndPlaceItem(const UChar * cp,unsigned i,const SimpleFontData * fontData,GlyphBuffer * glyphBuffer)218 bool UniscribeController::shapeAndPlaceItem(const UChar* cp, unsigned i, const SimpleFontData* fontData, GlyphBuffer* glyphBuffer)
219 {
220     // Determine the string for this item.
221     const UChar* str = cp + m_items[i].iCharPos;
222     int len = m_items[i+1].iCharPos - m_items[i].iCharPos;
223     SCRIPT_ITEM item = m_items[i];
224 
225     // Set up buffers to hold the results of shaping the item.
226     Vector<WORD> glyphs;
227     Vector<WORD> clusters;
228     Vector<SCRIPT_VISATTR> visualAttributes;
229     clusters.resize(len);
230 
231     // Shape the item.
232     // The recommended size for the glyph buffer is 1.5 * the character length + 16 in the uniscribe docs.
233     // Apparently this is a good size to avoid having to make repeated calls to ScriptShape.
234     glyphs.resize(1.5 * len + 16);
235     visualAttributes.resize(glyphs.size());
236 
237     if (!shape(str, len, item, fontData, glyphs, clusters, visualAttributes))
238         return true;
239 
240     // We now have a collection of glyphs.
241     Vector<GOFFSET> offsets;
242     Vector<int> advances;
243     offsets.resize(glyphs.size());
244     advances.resize(glyphs.size());
245     int glyphCount = 0;
246     HRESULT placeResult = ScriptPlace(0, fontData->scriptCache(), glyphs.data(), glyphs.size(), visualAttributes.data(),
247                                       &item.a, advances.data(), offsets.data(), 0);
248     if (placeResult == E_PENDING) {
249         // The script cache isn't primed with enough info yet.  We need to select our HFONT into
250         // a DC and pass the DC in to ScriptPlace.
251         HDC hdc = GetDC(0);
252         HFONT hfont = fontData->platformData().hfont();
253         HFONT oldFont = (HFONT)SelectObject(hdc, hfont);
254         placeResult = ScriptPlace(hdc, fontData->scriptCache(), glyphs.data(), glyphs.size(), visualAttributes.data(),
255                                   &item.a, advances.data(), offsets.data(), 0);
256         SelectObject(hdc, oldFont);
257         ReleaseDC(0, hdc);
258     }
259 
260     if (FAILED(placeResult) || glyphs.isEmpty())
261         return true;
262 
263     // Convert all chars that should be treated as spaces to use the space glyph.
264     // We also create a map that allows us to quickly go from space glyphs back to their corresponding characters.
265     Vector<int> spaceCharacters(glyphs.size());
266     spaceCharacters.fill(-1);
267 
268     const float cLogicalScale = fontData->platformData().useGDI() ? 1.0f : 32.0f;
269     unsigned logicalSpaceWidth = fontData->spaceWidth() * cLogicalScale;
270     float spaceWidth = fontData->spaceWidth();
271 
272     for (int k = 0; k < len; k++) {
273         UChar ch = *(str + k);
274         bool treatAsSpace = Font::treatAsSpace(ch);
275         bool treatAsZeroWidthSpace = ch == zeroWidthSpace || Font::treatAsZeroWidthSpace(ch);
276         if (treatAsSpace || treatAsZeroWidthSpace) {
277             // Substitute in the space glyph at the appropriate place in the glyphs
278             // array.
279             glyphs[clusters[k]] = fontData->spaceGlyph();
280             advances[clusters[k]] = treatAsSpace ? logicalSpaceWidth : 0;
281             if (treatAsSpace)
282                 spaceCharacters[clusters[k]] = m_currentCharacter + k + item.iCharPos;
283         }
284     }
285 
286     // Populate our glyph buffer with this information.
287     bool hasExtraSpacing = m_font.letterSpacing() || m_font.wordSpacing() || m_padding;
288 
289     float leftEdge = m_runWidthSoFar;
290 
291     for (unsigned k = 0; k < glyphs.size(); k++) {
292         Glyph glyph = glyphs[k];
293         float advance = advances[k] / cLogicalScale;
294         float offsetX = offsets[k].du / cLogicalScale;
295         float offsetY = offsets[k].dv / cLogicalScale;
296 
297         // Match AppKit's rules for the integer vs. non-integer rendering modes.
298         float roundedAdvance = roundf(advance);
299         if (!m_font.isPrinterFont() && !fontData->isSystemFont()) {
300             advance = roundedAdvance;
301             offsetX = roundf(offsetX);
302             offsetY = roundf(offsetY);
303         }
304 
305         advance += fontData->syntheticBoldOffset();
306 
307         if (hasExtraSpacing) {
308             // If we're a glyph with an advance, go ahead and add in letter-spacing.
309             // That way we weed out zero width lurkers.  This behavior matches the fast text code path.
310             if (advance && m_font.letterSpacing())
311                 advance += m_font.letterSpacing();
312 
313             // Handle justification and word-spacing.
314             int characterIndex = spaceCharacters[k];
315             // characterIndex is left at the initial value of -1 for glyphs that do not map back to treated-as-space characters.
316             if (characterIndex != -1) {
317                 // Account for padding. WebCore uses space padding to justify text.
318                 // We distribute the specified padding over the available spaces in the run.
319                 if (m_padding) {
320                     // Use leftover padding if not evenly divisible by number of spaces.
321                     if (m_padding < m_padPerSpace) {
322                         advance += m_padding;
323                         m_padding = 0;
324                     } else {
325                         m_padding -= m_padPerSpace;
326                         advance += m_padPerSpace;
327                     }
328                 }
329 
330                 // Account for word-spacing.
331                 if (characterIndex > 0 && !Font::treatAsSpace(*m_run.data(characterIndex - 1)) && m_font.wordSpacing())
332                     advance += m_font.wordSpacing();
333             }
334         }
335 
336         m_runWidthSoFar += advance;
337 
338         // FIXME: We need to take the GOFFSETS for combining glyphs and store them in the glyph buffer
339         // as well, so that when the time comes to draw those glyphs, we can apply the appropriate
340         // translation.
341         if (glyphBuffer) {
342             FloatSize size(offsetX, -offsetY);
343             glyphBuffer->add(glyph, fontData, advance, &size);
344         }
345 
346         FloatRect glyphBounds = fontData->boundsForGlyph(glyph);
347         glyphBounds.move(m_glyphOrigin.x(), m_glyphOrigin.y());
348         m_minGlyphBoundingBoxX = min(m_minGlyphBoundingBoxX, glyphBounds.x());
349         m_maxGlyphBoundingBoxX = max(m_maxGlyphBoundingBoxX, glyphBounds.maxX());
350         m_minGlyphBoundingBoxY = min(m_minGlyphBoundingBoxY, glyphBounds.y());
351         m_maxGlyphBoundingBoxY = max(m_maxGlyphBoundingBoxY, glyphBounds.maxY());
352         m_glyphOrigin.move(advance + offsetX, -offsetY);
353 
354         // Mutate the glyph array to contain our altered advances.
355         if (m_computingOffsetPosition)
356             advances[k] = advance;
357     }
358 
359     while (m_computingOffsetPosition && m_offsetX >= leftEdge && m_offsetX < m_runWidthSoFar) {
360         // The position is somewhere inside this run.
361         int trailing = 0;
362         ScriptXtoCP(m_offsetX - leftEdge, clusters.size(), glyphs.size(), clusters.data(), visualAttributes.data(),
363                     advances.data(), &item.a, &m_offsetPosition, &trailing);
364         if (trailing && m_includePartialGlyphs && m_offsetPosition < len - 1) {
365             m_offsetPosition += m_currentCharacter + m_items[i].iCharPos;
366             m_offsetX += m_run.rtl() ? -trailing : trailing;
367         } else {
368             m_computingOffsetPosition = false;
369             m_offsetPosition += m_currentCharacter + m_items[i].iCharPos;
370             if (trailing && m_includePartialGlyphs)
371                m_offsetPosition++;
372             return false;
373         }
374     }
375 
376     return true;
377 }
378 
shape(const UChar * str,int len,SCRIPT_ITEM item,const SimpleFontData * fontData,Vector<WORD> & glyphs,Vector<WORD> & clusters,Vector<SCRIPT_VISATTR> & visualAttributes)379 bool UniscribeController::shape(const UChar* str, int len, SCRIPT_ITEM item, const SimpleFontData* fontData,
380                                 Vector<WORD>& glyphs, Vector<WORD>& clusters,
381                                 Vector<SCRIPT_VISATTR>& visualAttributes)
382 {
383     HDC hdc = 0;
384     HFONT oldFont = 0;
385     HRESULT shapeResult = E_PENDING;
386     int glyphCount = 0;
387     do {
388         shapeResult = ScriptShape(hdc, fontData->scriptCache(), str, len, glyphs.size(), &item.a,
389                                   glyphs.data(), clusters.data(), visualAttributes.data(), &glyphCount);
390         if (shapeResult == E_PENDING) {
391             // The script cache isn't primed with enough info yet.  We need to select our HFONT into
392             // a DC and pass the DC in to ScriptShape.
393             ASSERT(!hdc);
394             hdc = GetDC(0);
395             HFONT hfont = fontData->platformData().hfont();
396             oldFont = (HFONT)SelectObject(hdc, hfont);
397         } else if (shapeResult == E_OUTOFMEMORY) {
398             // Need to resize our buffers.
399             glyphs.resize(glyphs.size() * 2);
400             visualAttributes.resize(glyphs.size());
401         }
402     } while (shapeResult == E_PENDING || shapeResult == E_OUTOFMEMORY);
403 
404     if (hdc) {
405         SelectObject(hdc, oldFont);
406         ReleaseDC(0, hdc);
407     }
408 
409     if (FAILED(shapeResult))
410         return false;
411 
412     glyphs.shrink(glyphCount);
413     visualAttributes.shrink(glyphCount);
414 
415     return true;
416 }
417 
418 }
419