| /* |
| * Copyright 2016 Google Inc. |
| * |
| * Use of this source code is governed by a BSD-style license that can be |
| * found in the LICENSE file. |
| */ |
| |
| #include "include/core/SkFontMetrics.h" |
| #include "include/core/SkStream.h" |
| #include "include/core/SkTypeface.h" |
| #include "include/private/SkTo.h" |
| #include "modules/skshaper/include/SkShaper.h" |
| #include "src/utils/SkUTF.h" |
| |
| class SkShaperPrimitive : public SkShaper { |
| public: |
| SkShaperPrimitive() {} |
| private: |
| void shape(const char* utf8, size_t utf8Bytes, |
| const SkFont& srcFont, |
| bool leftToRight, |
| SkScalar width, |
| RunHandler*) const override; |
| |
| void shape(const char* utf8, size_t utf8Bytes, |
| FontRunIterator&, |
| BiDiRunIterator&, |
| ScriptRunIterator&, |
| LanguageRunIterator&, |
| SkScalar width, |
| RunHandler*) const override; |
| |
| void shape(const char* utf8, size_t utf8Bytes, |
| FontRunIterator&, |
| BiDiRunIterator&, |
| ScriptRunIterator&, |
| LanguageRunIterator&, |
| const Feature*, size_t featureSize, |
| SkScalar width, |
| RunHandler*) const override; |
| }; |
| |
| std::unique_ptr<SkShaper> SkShaper::MakePrimitive() { |
| return std::make_unique<SkShaperPrimitive>(); |
| } |
| |
| static inline bool is_breaking_whitespace(SkUnichar c) { |
| switch (c) { |
| case 0x0020: // SPACE |
| //case 0x00A0: // NO-BREAK SPACE |
| case 0x1680: // OGHAM SPACE MARK |
| case 0x180E: // MONGOLIAN VOWEL SEPARATOR |
| case 0x2000: // EN QUAD |
| case 0x2001: // EM QUAD |
| case 0x2002: // EN SPACE (nut) |
| case 0x2003: // EM SPACE (mutton) |
| case 0x2004: // THREE-PER-EM SPACE (thick space) |
| case 0x2005: // FOUR-PER-EM SPACE (mid space) |
| case 0x2006: // SIX-PER-EM SPACE |
| case 0x2007: // FIGURE SPACE |
| case 0x2008: // PUNCTUATION SPACE |
| case 0x2009: // THIN SPACE |
| case 0x200A: // HAIR SPACE |
| case 0x200B: // ZERO WIDTH SPACE |
| case 0x202F: // NARROW NO-BREAK SPACE |
| case 0x205F: // MEDIUM MATHEMATICAL SPACE |
| case 0x3000: // IDEOGRAPHIC SPACE |
| //case 0xFEFF: // ZERO WIDTH NO-BREAK SPACE |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| static size_t linebreak(const char text[], const char stop[], |
| const SkFont& font, SkScalar width, |
| SkScalar* advance, |
| size_t* trailing) |
| { |
| SkScalar accumulatedWidth = 0; |
| int glyphIndex = 0; |
| const char* start = text; |
| const char* word_start = text; |
| bool prevWS = true; |
| *trailing = 0; |
| |
| while (text < stop) { |
| const char* prevText = text; |
| SkUnichar uni = SkUTF::NextUTF8(&text, stop); |
| accumulatedWidth += advance[glyphIndex++]; |
| bool currWS = is_breaking_whitespace(uni); |
| |
| if (!currWS && prevWS) { |
| word_start = prevText; |
| } |
| prevWS = currWS; |
| |
| if (width < accumulatedWidth) { |
| if (currWS) { |
| // eat the rest of the whitespace |
| const char* next = text; |
| while (next < stop && is_breaking_whitespace(SkUTF::NextUTF8(&next, stop))) { |
| text = next; |
| } |
| if (trailing) { |
| *trailing = text - prevText; |
| } |
| } else { |
| // backup until a whitespace (or 1 char) |
| if (word_start == start) { |
| if (prevText > start) { |
| text = prevText; |
| } |
| } else { |
| text = word_start; |
| } |
| } |
| break; |
| } |
| } |
| |
| return text - start; |
| } |
| |
| void SkShaperPrimitive::shape(const char* utf8, size_t utf8Bytes, |
| FontRunIterator& font, |
| BiDiRunIterator& bidi, |
| ScriptRunIterator&, |
| LanguageRunIterator&, |
| SkScalar width, |
| RunHandler* handler) const |
| { |
| SkFont skfont; |
| if (!font.atEnd()) { |
| font.consume(); |
| skfont = font.currentFont(); |
| } else { |
| skfont.setTypeface(sk_ref_sp(skfont.getTypefaceOrDefault())); |
| } |
| SkASSERT(skfont.getTypeface()); |
| bool skbidi = 0; |
| if (!bidi.atEnd()) { |
| bidi.consume(); |
| skbidi = (bidi.currentLevel() % 2) == 0; |
| } |
| return this->shape(utf8, utf8Bytes, skfont, skbidi, width, handler); |
| } |
| |
| void SkShaperPrimitive::shape(const char* utf8, size_t utf8Bytes, |
| FontRunIterator& font, |
| BiDiRunIterator& bidi, |
| ScriptRunIterator&, |
| LanguageRunIterator&, |
| const Feature*, size_t, |
| SkScalar width, |
| RunHandler* handler) const { |
| font.consume(); |
| SkASSERT(font.currentFont().getTypeface()); |
| bidi.consume(); |
| return this->shape(utf8, utf8Bytes, font.currentFont(), (bidi.currentLevel() % 2) == 0, |
| width, handler); |
| } |
| |
| void SkShaperPrimitive::shape(const char* utf8, size_t utf8Bytes, |
| const SkFont& font, |
| bool leftToRight, |
| SkScalar width, |
| RunHandler* handler) const { |
| sk_ignore_unused_variable(leftToRight); |
| |
| int glyphCount = font.countText(utf8, utf8Bytes, SkTextEncoding::kUTF8); |
| if (glyphCount <= 0) { |
| return; |
| } |
| |
| std::unique_ptr<SkGlyphID[]> glyphs(new SkGlyphID[glyphCount]); |
| font.textToGlyphs(utf8, utf8Bytes, SkTextEncoding::kUTF8, glyphs.get(), glyphCount); |
| |
| std::unique_ptr<SkScalar[]> advances(new SkScalar[glyphCount]); |
| font.getWidthsBounds(glyphs.get(), glyphCount, advances.get(), nullptr, nullptr); |
| |
| size_t glyphOffset = 0; |
| size_t utf8Offset = 0; |
| while (0 < utf8Bytes) { |
| size_t bytesCollapsed; |
| size_t bytesConsumed = linebreak(utf8, utf8 + utf8Bytes, font, width, |
| advances.get() + glyphOffset, &bytesCollapsed); |
| size_t bytesVisible = bytesConsumed - bytesCollapsed; |
| |
| size_t numGlyphs = SkUTF::CountUTF8(utf8, bytesVisible); |
| const RunHandler::RunInfo info = { |
| font, |
| 0, |
| { font.measureText(utf8, bytesVisible, SkTextEncoding::kUTF8), 0 }, |
| numGlyphs, |
| RunHandler::Range(utf8Offset, bytesVisible) |
| }; |
| handler->beginLine(); |
| handler->runInfo(info); |
| handler->commitRunInfo(); |
| const auto buffer = handler->runBuffer(info); |
| |
| memcpy(buffer.glyphs, glyphs.get() + glyphOffset, numGlyphs * sizeof(SkGlyphID)); |
| SkPoint position = buffer.point; |
| for (size_t i = 0; i < numGlyphs; ++i) { |
| buffer.positions[i] = position; |
| position.fX += advances[i + glyphOffset]; |
| } |
| if (buffer.clusters) { |
| const char* txtPtr = utf8; |
| for (size_t i = 0; i < numGlyphs; ++i) { |
| // Each character maps to exactly one glyph. |
| buffer.clusters[i] = SkToU32(txtPtr - utf8 + utf8Offset); |
| SkUTF::NextUTF8(&txtPtr, utf8 + utf8Bytes); |
| } |
| } |
| handler->commitRunBuffer(info); |
| handler->commitLine(); |
| |
| glyphOffset += SkUTF::CountUTF8(utf8, bytesConsumed); |
| utf8Offset += bytesConsumed; |
| utf8 += bytesConsumed; |
| utf8Bytes -= bytesConsumed; |
| } |
| |
| return; |
| } |