skia2/tools/shape/SkShaper_harfbuzz.cpp
Hal Canary 0e07ad7fe7 ICU on windows
Change-Id: Ib1a2f017d96c5157c60d512332fddfef77c5ae8e
Reviewed-on: https://skia-review.googlesource.com/103001
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
2018-02-08 19:36:49 +00:00

729 lines
26 KiB
C++

/*
* 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 <hb-ot.h>
#include <unicode/brkiter.h>
#include <unicode/locid.h>
#include <unicode/stringpiece.h>
#include <unicode/ubidi.h>
#include <unicode/uchriter.h>
#include <unicode/unistr.h>
#include <unicode/uscript.h>
#include "SkFontMgr.h"
#include "SkLoadICU.h"
#include "SkOnce.h"
#include "SkShaper.h"
#include "SkStream.h"
#include "SkTDPQueue.h"
#include "SkTLazy.h"
#include "SkTemplates.h"
#include "SkTextBlob.h"
#include "SkTypeface.h"
#include "SkUtils.h"
namespace {
template <class T, void(*P)(T*)> using resource = std::unique_ptr<T, SkFunctionWrapper<void, T, P>>;
using HBBlob = resource<hb_blob_t , hb_blob_destroy >;
using HBFace = resource<hb_face_t , hb_face_destroy >;
using HBFont = resource<hb_font_t , hb_font_destroy >;
using HBBuffer = resource<hb_buffer_t, hb_buffer_destroy>;
using ICUBiDi = resource<UBiDi , ubidi_close >;
HBBlob stream_to_blob(std::unique_ptr<SkStreamAsset> asset) {
size_t size = asset->getLength();
HBBlob blob;
if (const void* base = asset->getMemoryBase()) {
blob.reset(hb_blob_create((char*)base, SkToUInt(size),
HB_MEMORY_MODE_READONLY, asset.release(),
[](void* p) { delete (SkStreamAsset*)p; }));
} else {
// SkDebugf("Extra SkStreamAsset copy\n");
void* ptr = size ? sk_malloc_throw(size) : nullptr;
asset->read(ptr, size);
blob.reset(hb_blob_create((char*)ptr, SkToUInt(size),
HB_MEMORY_MODE_READONLY, ptr, sk_free));
}
SkASSERT(blob);
hb_blob_make_immutable(blob.get());
return blob;
}
HBFont create_hb_font(SkTypeface* tf) {
int index;
HBBlob blob(stream_to_blob(std::unique_ptr<SkStreamAsset>(tf->openStream(&index))));
HBFace face(hb_face_create(blob.get(), (unsigned)index));
SkASSERT(face);
if (!face) {
return nullptr;
}
hb_face_set_index(face.get(), (unsigned)index);
hb_face_set_upem(face.get(), tf->getUnitsPerEm());
HBFont font(hb_font_create(face.get()));
SkASSERT(font);
if (!font) {
return nullptr;
}
hb_ot_font_set_funcs(font.get());
int axis_count = tf->getVariationDesignPosition(nullptr, 0);
if (axis_count > 0) {
SkAutoSTMalloc<4, SkFontArguments::VariationPosition::Coordinate> axis_values(axis_count);
if (tf->getVariationDesignPosition(axis_values, axis_count) == axis_count) {
hb_font_set_variations(font.get(),
reinterpret_cast<hb_variation_t*>(axis_values.get()),
axis_count);
}
}
return font;
}
class RunIterator {
public:
virtual ~RunIterator() {}
virtual void consume() = 0;
// Pointer one past the last (utf8) element in the current run.
virtual const char* endOfCurrentRun() const = 0;
virtual bool atEnd() const = 0;
bool operator<(const RunIterator& that) const {
return this->endOfCurrentRun() < that.endOfCurrentRun();
}
};
class BiDiRunIterator : public RunIterator {
public:
static SkTLazy<BiDiRunIterator> Make(const char* utf8, size_t utf8Bytes, UBiDiLevel level) {
SkTLazy<BiDiRunIterator> ret;
// ubidi only accepts utf16 (though internally it basically works on utf32 chars).
// We want an ubidi_setPara(UBiDi*, UText*, UBiDiLevel, UBiDiLevel*, UErrorCode*);
if (!SkTFitsIn<int32_t>(utf8Bytes)) {
SkDebugf("Bidi error: text too long");
return ret;
}
icu::UnicodeString utf16 = icu::UnicodeString::fromUTF8(icu::StringPiece(utf8, utf8Bytes));
UErrorCode status = U_ZERO_ERROR;
ICUBiDi bidi(ubidi_openSized(utf16.length(), 0, &status));
if (U_FAILURE(status)) {
SkDebugf("Bidi error: %s", u_errorName(status));
return ret;
}
SkASSERT(bidi);
// The required lifetime of utf16 isn't well documented.
// It appears it isn't used after ubidi_setPara except through ubidi_getText.
ubidi_setPara(bidi.get(), utf16.getBuffer(), utf16.length(), level, nullptr, &status);
if (U_FAILURE(status)) {
SkDebugf("Bidi error: %s", u_errorName(status));
return ret;
}
ret.init(utf8, std::move(bidi));
return ret;
}
BiDiRunIterator(const char* utf8, ICUBiDi bidi)
: fBidi(std::move(bidi))
, fEndOfCurrentRun(utf8)
, fUTF16LogicalPosition(0)
, fLevel(UBIDI_DEFAULT_LTR)
{}
void consume() override {
SkASSERT(fUTF16LogicalPosition < ubidi_getLength(fBidi.get()));
int32_t endPosition = ubidi_getLength(fBidi.get());
fLevel = ubidi_getLevelAt(fBidi.get(), fUTF16LogicalPosition);
SkUnichar u = SkUTF8_NextUnichar(&fEndOfCurrentRun);
fUTF16LogicalPosition += SkUTF16_FromUnichar(u);
UBiDiLevel level;
while (fUTF16LogicalPosition < endPosition) {
level = ubidi_getLevelAt(fBidi.get(), fUTF16LogicalPosition);
if (level != fLevel) {
break;
}
u = SkUTF8_NextUnichar(&fEndOfCurrentRun);
fUTF16LogicalPosition += SkUTF16_FromUnichar(u);
}
}
const char* endOfCurrentRun() const override {
return fEndOfCurrentRun;
}
bool atEnd() const override {
return fUTF16LogicalPosition == ubidi_getLength(fBidi.get());
}
UBiDiLevel currentLevel() const {
return fLevel;
}
private:
ICUBiDi fBidi;
const char* fEndOfCurrentRun;
int32_t fUTF16LogicalPosition;
UBiDiLevel fLevel;
};
class ScriptRunIterator : public RunIterator {
public:
static SkTLazy<ScriptRunIterator> Make(const char* utf8, size_t utf8Bytes,
hb_unicode_funcs_t* hbUnicode)
{
SkTLazy<ScriptRunIterator> ret;
ret.init(utf8, utf8Bytes, hbUnicode);
return ret;
}
ScriptRunIterator(const char* utf8, size_t utf8Bytes, hb_unicode_funcs_t* hbUnicode)
: fCurrent(utf8), fEnd(fCurrent + utf8Bytes)
, fHBUnicode(hbUnicode)
, fCurrentScript(HB_SCRIPT_UNKNOWN)
{}
void consume() override {
SkASSERT(fCurrent < fEnd);
SkUnichar u = SkUTF8_NextUnichar(&fCurrent);
fCurrentScript = hb_unicode_script(fHBUnicode, u);
while (fCurrent < fEnd) {
const char* prev = fCurrent;
u = SkUTF8_NextUnichar(&fCurrent);
const hb_script_t script = hb_unicode_script(fHBUnicode, u);
if (script != fCurrentScript) {
if (fCurrentScript == HB_SCRIPT_INHERITED || fCurrentScript == HB_SCRIPT_COMMON) {
fCurrentScript = script;
} else if (script == HB_SCRIPT_INHERITED || script == HB_SCRIPT_COMMON) {
continue;
} else {
fCurrent = prev;
break;
}
}
}
if (fCurrentScript == HB_SCRIPT_INHERITED) {
fCurrentScript = HB_SCRIPT_COMMON;
}
}
const char* endOfCurrentRun() const override {
return fCurrent;
}
bool atEnd() const override {
return fCurrent == fEnd;
}
hb_script_t currentScript() const {
return fCurrentScript;
}
private:
const char* fCurrent;
const char* fEnd;
hb_unicode_funcs_t* fHBUnicode;
hb_script_t fCurrentScript;
};
class FontRunIterator : public RunIterator {
public:
static SkTLazy<FontRunIterator> Make(const char* utf8, size_t utf8Bytes,
sk_sp<SkTypeface> typeface,
hb_font_t* hbFace,
sk_sp<SkFontMgr> fallbackMgr)
{
SkTLazy<FontRunIterator> ret;
ret.init(utf8, utf8Bytes, std::move(typeface), hbFace, std::move(fallbackMgr));
return ret;
}
FontRunIterator(const char* utf8, size_t utf8Bytes, sk_sp<SkTypeface> typeface,
hb_font_t* hbFace, sk_sp<SkFontMgr> fallbackMgr)
: fCurrent(utf8), fEnd(fCurrent + utf8Bytes)
, fFallbackMgr(std::move(fallbackMgr))
, fHBFont(hbFace), fTypeface(std::move(typeface))
, fFallbackHBFont(nullptr), fFallbackTypeface(nullptr)
, fCurrentHBFont(fHBFont), fCurrentTypeface(fTypeface.get())
{}
void consume() override {
SkASSERT(fCurrent < fEnd);
SkUnichar u = SkUTF8_NextUnichar(&fCurrent);
// If the starting typeface can handle this character, use it.
if (fTypeface->charsToGlyphs(&u, SkTypeface::kUTF32_Encoding, nullptr, 1)) {
fFallbackTypeface.reset();
// If not, try to find a fallback typeface
} else {
fFallbackTypeface.reset(fFallbackMgr->matchFamilyStyleCharacter(
nullptr, fTypeface->fontStyle(), nullptr, 0, u));
}
if (fFallbackTypeface) {
fFallbackHBFont = create_hb_font(fFallbackTypeface.get());
fCurrentTypeface = fFallbackTypeface.get();
fCurrentHBFont = fFallbackHBFont.get();
} else {
fFallbackHBFont.reset();
fCurrentTypeface = fTypeface.get();
fCurrentHBFont = fHBFont;
}
while (fCurrent < fEnd) {
const char* prev = fCurrent;
u = SkUTF8_NextUnichar(&fCurrent);
// If using a fallback and the initial typeface has this character, stop fallback.
if (fFallbackTypeface &&
fTypeface->charsToGlyphs(&u, SkTypeface::kUTF32_Encoding, nullptr, 1))
{
fCurrent = prev;
return;
}
// If the current typeface cannot handle this character, stop using it.
if (!fCurrentTypeface->charsToGlyphs(&u, SkTypeface::kUTF32_Encoding, nullptr, 1)) {
fCurrent = prev;
return;
}
}
}
const char* endOfCurrentRun() const override {
return fCurrent;
}
bool atEnd() const override {
return fCurrent == fEnd;
}
SkTypeface* currentTypeface() const {
return fCurrentTypeface;
}
hb_font_t* currentHBFont() const {
return fCurrentHBFont;
}
private:
const char* fCurrent;
const char* fEnd;
sk_sp<SkFontMgr> fFallbackMgr;
hb_font_t* fHBFont;
sk_sp<SkTypeface> fTypeface;
HBFont fFallbackHBFont;
sk_sp<SkTypeface> fFallbackTypeface;
hb_font_t* fCurrentHBFont;
SkTypeface* fCurrentTypeface;
};
class RunIteratorQueue {
public:
void insert(RunIterator* runIterator) {
fRunIterators.insert(runIterator);
}
bool advanceRuns() {
const RunIterator* leastRun = fRunIterators.peek();
if (leastRun->atEnd()) {
SkASSERT(this->allRunsAreAtEnd());
return false;
}
const char* leastEnd = leastRun->endOfCurrentRun();
RunIterator* currentRun = nullptr;
SkDEBUGCODE(const char* previousEndOfCurrentRun);
while ((currentRun = fRunIterators.peek())->endOfCurrentRun() <= leastEnd) {
fRunIterators.pop();
SkDEBUGCODE(previousEndOfCurrentRun = currentRun->endOfCurrentRun());
currentRun->consume();
SkASSERT(previousEndOfCurrentRun < currentRun->endOfCurrentRun());
fRunIterators.insert(currentRun);
}
return true;
}
const char* endOfCurrentRun() const {
return fRunIterators.peek()->endOfCurrentRun();
}
private:
bool allRunsAreAtEnd() const {
for (int i = 0; i < fRunIterators.count(); ++i) {
if (!fRunIterators.at(i)->atEnd()) {
return false;
}
}
return true;
}
static bool CompareRunIterator(RunIterator* const& a, RunIterator* const& b) {
return *a < *b;
}
SkTDPQueue<RunIterator*, CompareRunIterator> fRunIterators;
};
struct ShapedGlyph {
SkGlyphID fID;
uint32_t fCluster;
SkPoint fOffset;
SkVector fAdvance;
bool fMayLineBreakBefore;
bool fMustLineBreakBefore;
bool fHasVisual;
};
struct ShapedRun {
ShapedRun(const char* utf8Start, const char* utf8End, int numGlyphs, const SkPaint& paint,
UBiDiLevel level, std::unique_ptr<ShapedGlyph[]> glyphs)
: fUtf8Start(utf8Start), fUtf8End(utf8End), fNumGlyphs(numGlyphs), fPaint(paint)
, fLevel(level), fGlyphs(std::move(glyphs))
{}
const char* fUtf8Start;
const char* fUtf8End;
int fNumGlyphs;
SkPaint fPaint;
UBiDiLevel fLevel;
std::unique_ptr<ShapedGlyph[]> fGlyphs;
};
static constexpr bool is_LTR(UBiDiLevel level) {
return (level & 1) == 0;
}
static void append(SkTextBlobBuilder* b, const ShapedRun& run, int start, int end, SkPoint* p) {
unsigned len = end - start;
auto runBuffer = b->allocRunTextPos(run.fPaint, len, run.fUtf8End - run.fUtf8Start, SkString());
memcpy(runBuffer.utf8text, run.fUtf8Start, run.fUtf8End - run.fUtf8Start);
for (unsigned i = 0; i < len; i++) {
// Glyphs are in logical order, but output ltr since PDF readers seem to expect that.
const ShapedGlyph& glyph = run.fGlyphs[is_LTR(run.fLevel) ? start + i : end - 1 - i];
runBuffer.glyphs[i] = glyph.fID;
runBuffer.clusters[i] = glyph.fCluster;
reinterpret_cast<SkPoint*>(runBuffer.pos)[i] =
SkPoint::Make(p->fX + glyph.fOffset.fX, p->fY - glyph.fOffset.fY);
p->fX += glyph.fAdvance.fX;
p->fY += glyph.fAdvance.fY;
}
}
struct ShapedRunGlyphIterator {
ShapedRunGlyphIterator(const SkTArray<ShapedRun>& origRuns)
: fRuns(&origRuns), fRunIndex(0), fGlyphIndex(0)
{ }
ShapedRunGlyphIterator(const ShapedRunGlyphIterator& that) = default;
ShapedRunGlyphIterator& operator=(const ShapedRunGlyphIterator& that) = default;
bool operator==(const ShapedRunGlyphIterator& that) const {
return fRuns == that.fRuns &&
fRunIndex == that.fRunIndex &&
fGlyphIndex == that.fGlyphIndex;
}
bool operator!=(const ShapedRunGlyphIterator& that) const {
return fRuns != that.fRuns ||
fRunIndex != that.fRunIndex ||
fGlyphIndex != that.fGlyphIndex;
}
ShapedGlyph* next() {
const SkTArray<ShapedRun>& runs = *fRuns;
SkASSERT(fRunIndex < runs.count());
SkASSERT(fGlyphIndex < runs[fRunIndex].fNumGlyphs);
++fGlyphIndex;
if (fGlyphIndex == runs[fRunIndex].fNumGlyphs) {
fGlyphIndex = 0;
++fRunIndex;
if (fRunIndex >= runs.count()) {
return nullptr;
}
}
return &runs[fRunIndex].fGlyphs[fGlyphIndex];
}
ShapedGlyph* current() {
const SkTArray<ShapedRun>& runs = *fRuns;
if (fRunIndex >= runs.count()) {
return nullptr;
}
return &runs[fRunIndex].fGlyphs[fGlyphIndex];
}
const SkTArray<ShapedRun>* fRuns;
int fRunIndex;
int fGlyphIndex;
};
} // namespace
struct SkShaper::Impl {
HBFont fHarfBuzzFont;
HBBuffer fBuffer;
sk_sp<SkTypeface> fTypeface;
std::unique_ptr<icu::BreakIterator> fBreakIterator;
};
SkShaper::SkShaper(sk_sp<SkTypeface> tf) : fImpl(new Impl) {
SkOnce once;
once([] { SkLoadICU(); });
fImpl->fTypeface = tf ? std::move(tf) : SkTypeface::MakeDefault();
fImpl->fHarfBuzzFont = create_hb_font(fImpl->fTypeface.get());
SkASSERT(fImpl->fHarfBuzzFont);
fImpl->fBuffer.reset(hb_buffer_create());
SkASSERT(fImpl->fBuffer);
icu::Locale thai("th");
UErrorCode status = U_ZERO_ERROR;
fImpl->fBreakIterator.reset(icu::BreakIterator::createLineInstance(thai, status));
if (U_FAILURE(status)) {
SkDebugf("Could not create break iterator: %s", u_errorName(status));
SK_ABORT("");
}
}
SkShaper::~SkShaper() {}
bool SkShaper::good() const {
return fImpl->fHarfBuzzFont &&
fImpl->fBuffer &&
fImpl->fTypeface &&
fImpl->fBreakIterator;
}
SkPoint SkShaper::shape(SkTextBlobBuilder* builder,
const SkPaint& srcPaint,
const char* utf8,
size_t utf8Bytes,
bool leftToRight,
SkPoint point,
SkScalar width) const {
sk_sp<SkFontMgr> fontMgr = SkFontMgr::RefDefault();
SkASSERT(builder);
UBiDiLevel defaultLevel = leftToRight ? UBIDI_DEFAULT_LTR : UBIDI_DEFAULT_RTL;
//hb_script_t script = ...
SkTArray<ShapedRun> runs;
{
RunIteratorQueue runSegmenter;
SkTLazy<BiDiRunIterator> maybeBidi(BiDiRunIterator::Make(utf8, utf8Bytes, defaultLevel));
BiDiRunIterator* bidi = maybeBidi.getMaybeNull();
if (!bidi) {
return point;
}
runSegmenter.insert(bidi);
hb_unicode_funcs_t* hbUnicode = hb_buffer_get_unicode_funcs(fImpl->fBuffer.get());
SkTLazy<ScriptRunIterator> maybeScript(ScriptRunIterator::Make(utf8, utf8Bytes, hbUnicode));
ScriptRunIterator* script = maybeScript.getMaybeNull();
if (!script) {
return point;
}
runSegmenter.insert(script);
SkTLazy<FontRunIterator> maybeFont(FontRunIterator::Make(utf8, utf8Bytes,
fImpl->fTypeface,
fImpl->fHarfBuzzFont.get(),
std::move(fontMgr)));
FontRunIterator* font = maybeFont.getMaybeNull();
if (!font) {
return point;
}
runSegmenter.insert(font);
icu::BreakIterator& breakIterator = *fImpl->fBreakIterator;
{
UErrorCode status = U_ZERO_ERROR;
UText utf8UText = UTEXT_INITIALIZER;
utext_openUTF8(&utf8UText, utf8, utf8Bytes, &status);
std::unique_ptr<UText, SkFunctionWrapper<UText*, UText, utext_close>> autoClose(&utf8UText);
if (U_FAILURE(status)) {
SkDebugf("Could not create utf8UText: %s", u_errorName(status));
return point;
}
breakIterator.setText(&utf8UText, status);
//utext_close(&utf8UText);
if (U_FAILURE(status)) {
SkDebugf("Could not setText on break iterator: %s", u_errorName(status));
return point;
}
}
const char* utf8Start = nullptr;
const char* utf8End = utf8;
while (runSegmenter.advanceRuns()) {
utf8Start = utf8End;
utf8End = runSegmenter.endOfCurrentRun();
hb_buffer_t* buffer = fImpl->fBuffer.get();
SkAutoTCallVProc<hb_buffer_t, hb_buffer_clear_contents> autoClearBuffer(buffer);
hb_buffer_set_content_type(buffer, HB_BUFFER_CONTENT_TYPE_UNICODE);
hb_buffer_set_cluster_level(buffer, HB_BUFFER_CLUSTER_LEVEL_MONOTONE_CHARACTERS);
// Populate the hb_buffer directly with utf8 cluster indexes.
const char* utf8Current = utf8Start;
while (utf8Current < utf8End) {
unsigned int cluster = utf8Current - utf8Start;
hb_codepoint_t u = SkUTF8_NextUnichar(&utf8Current);
hb_buffer_add(buffer, u, cluster);
}
size_t utf8runLength = utf8End - utf8Start;
if (!SkTFitsIn<int>(utf8runLength)) {
SkDebugf("Shaping error: utf8 too long");
return point;
}
hb_buffer_set_script(buffer, script->currentScript());
hb_direction_t direction = is_LTR(bidi->currentLevel()) ? HB_DIRECTION_LTR:HB_DIRECTION_RTL;
hb_buffer_set_direction(buffer, direction);
// TODO: language
hb_buffer_guess_segment_properties(buffer);
// TODO: features
hb_shape(font->currentHBFont(), buffer, nullptr, 0);
unsigned len = hb_buffer_get_length(buffer);
if (len == 0) {
continue;
}
if (direction == HB_DIRECTION_RTL) {
// Put the clusters back in logical order.
// Note that the advances remain ltr.
hb_buffer_reverse(buffer);
}
hb_glyph_info_t* info = hb_buffer_get_glyph_infos(buffer, nullptr);
hb_glyph_position_t* pos = hb_buffer_get_glyph_positions(buffer, nullptr);
if (!SkTFitsIn<int>(len)) {
SkDebugf("Shaping error: too many glyphs");
return point;
}
SkPaint paint(srcPaint);
paint.setTextEncoding(SkPaint::kGlyphID_TextEncoding);
paint.setTypeface(sk_ref_sp(font->currentTypeface()));
ShapedRun& run = runs.emplace_back(utf8Start, utf8End, len, paint, bidi->currentLevel(),
std::unique_ptr<ShapedGlyph[]>(new ShapedGlyph[len]));
int scaleX, scaleY;
hb_font_get_scale(font->currentHBFont(), &scaleX, &scaleY);
double textSizeY = run.fPaint.getTextSize() / scaleY;
double textSizeX = run.fPaint.getTextSize() / scaleX * run.fPaint.getTextScaleX();
for (unsigned i = 0; i < len; i++) {
ShapedGlyph& glyph = run.fGlyphs[i];
glyph.fID = info[i].codepoint;
glyph.fCluster = info[i].cluster;
glyph.fOffset.fX = pos[i].x_offset * textSizeX;
glyph.fOffset.fY = pos[i].y_offset * textSizeY;
glyph.fAdvance.fX = pos[i].x_advance * textSizeX;
glyph.fAdvance.fY = pos[i].y_advance * textSizeY;
glyph.fHasVisual = true; //!font->currentTypeface()->glyphBoundsAreZero(glyph.fID);
//info->mask safe_to_break;
glyph.fMustLineBreakBefore = false;
}
int32_t clusterOffset = utf8Start - utf8;
uint32_t previousCluster = 0xFFFFFFFF;
for (unsigned i = 0; i < len; ++i) {
ShapedGlyph& glyph = run.fGlyphs[i];
int32_t glyphCluster = glyph.fCluster + clusterOffset;
int32_t breakIteratorCurrent = breakIterator.current();
while (breakIteratorCurrent != icu::BreakIterator::DONE &&
breakIteratorCurrent < glyphCluster)
{
breakIteratorCurrent = breakIterator.next();
}
glyph.fMayLineBreakBefore = glyph.fCluster != previousCluster &&
breakIteratorCurrent == glyphCluster;
previousCluster = glyph.fCluster;
}
}
}
// Iterate over the glyphs in logical order to mark line endings.
{
SkScalar widthSoFar = 0;
bool previousBreakValid = false; // Set when previousBreak is set to a valid candidate break.
bool canAddBreakNow = false; // Disallow line breaks before the first glyph of a run.
ShapedRunGlyphIterator previousBreak(runs);
ShapedRunGlyphIterator glyphIterator(runs);
while (ShapedGlyph* glyph = glyphIterator.current()) {
if (canAddBreakNow && glyph->fMayLineBreakBefore) {
previousBreakValid = true;
previousBreak = glyphIterator;
}
SkScalar glyphWidth = glyph->fAdvance.fX;
if (widthSoFar + glyphWidth < width) {
widthSoFar += glyphWidth;
glyphIterator.next();
canAddBreakNow = true;
continue;
}
if (widthSoFar == 0) {
// Adding just this glyph is too much, just break with this glyph
glyphIterator.next();
previousBreak = glyphIterator;
} else if (!previousBreakValid) {
// No break opprotunity found yet, just break without this glyph
previousBreak = glyphIterator;
}
glyphIterator = previousBreak;
glyph = glyphIterator.current();
if (glyph) {
glyph->fMustLineBreakBefore = true;
}
widthSoFar = 0;
previousBreakValid = false;
canAddBreakNow = false;
}
}
// Reorder the runs and glyphs per line and write them out.
SkPoint currentPoint = point;
{
ShapedRunGlyphIterator previousBreak(runs);
ShapedRunGlyphIterator glyphIterator(runs);
SkScalar maxAscent = 0;
SkScalar maxDescent = 0;
SkScalar maxLeading = 0;
int previousRunIndex = -1;
while (glyphIterator.current()) {
int runIndex = glyphIterator.fRunIndex;
int glyphIndex = glyphIterator.fGlyphIndex;
ShapedGlyph* nextGlyph = glyphIterator.next();
if (previousRunIndex != runIndex) {
SkPaint::FontMetrics metrics;
runs[runIndex].fPaint.getFontMetrics(&metrics);
maxAscent = SkTMin(maxAscent, metrics.fAscent);
maxDescent = SkTMax(maxDescent, metrics.fDescent);
maxLeading = SkTMax(maxLeading, metrics.fLeading);
previousRunIndex = runIndex;
}
// Nothing can be written until the baseline is known.
if (!(nextGlyph == nullptr || nextGlyph->fMustLineBreakBefore)) {
continue;
}
currentPoint.fY -= maxAscent;
int numRuns = runIndex - previousBreak.fRunIndex + 1;
SkAutoSTMalloc<4, UBiDiLevel> runLevels(numRuns);
for (int i = 0; i < numRuns; ++i) {
runLevels[i] = runs[previousBreak.fRunIndex + i].fLevel;
}
SkAutoSTMalloc<4, int32_t> logicalFromVisual(numRuns);
ubidi_reorderVisual(runLevels, numRuns, logicalFromVisual);
for (int i = 0; i < numRuns; ++i) {
int logicalIndex = previousBreak.fRunIndex + logicalFromVisual[i];
int startGlyphIndex = (logicalIndex == previousBreak.fRunIndex)
? previousBreak.fGlyphIndex
: 0;
int endGlyphIndex = (logicalIndex == runIndex)
? glyphIndex + 1
: runs[logicalIndex].fNumGlyphs;
append(builder, runs[logicalIndex], startGlyphIndex, endGlyphIndex, &currentPoint);
}
currentPoint.fY += maxDescent + maxLeading;
currentPoint.fX = point.fX;
maxAscent = 0;
maxDescent = 0;
maxLeading = 0;
previousRunIndex = -1;
previousBreak = glyphIterator;
}
}
return currentPoint;
}