xref: /external/chromium_org/third_party/WebKit/Source/platform/text/BidiResolver.h

/*
 * Copyright (C) 2000 Lars Knoll (knoll (at) kde.org)
 * Copyright (C) 2003, 2004, 2006, 2007, 2008 Apple Inc.  All right reserved.
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Library General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Library General Public License for more details.
 *
 * You should have received a copy of the GNU Library General Public License
 * along with this library; see the file COPYING.LIB.  If not, write to
 * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
 * Boston, MA 02110-1301, USA.
 *
 */

#ifndef BidiResolver_h
#define BidiResolver_h

#include "platform/text/BidiCharacterRun.h"
#include "platform/text/BidiContext.h"
#include "platform/text/BidiRunList.h"
#include "platform/text/TextDirection.h"
#include "wtf/HashMap.h"
#include "wtf/Noncopyable.h"
#include "wtf/PassRefPtr.h"
#include "wtf/Vector.h"

namespace blink {

class RenderObject;

template <class Iterator> class MidpointState {
public:
    MidpointState()
    {
        reset();
    }

    void reset()
    {
        m_numMidpoints = 0;
        m_currentMidpoint = 0;
        m_betweenMidpoints = false;
    }

    void startIgnoringSpaces(const Iterator& midpoint)
    {
        ASSERT(!(m_numMidpoints % 2));
        addMidpoint(midpoint);
    }

    void stopIgnoringSpaces(const Iterator& midpoint)
    {
        ASSERT(m_numMidpoints % 2);
        addMidpoint(midpoint);
    }

    // When ignoring spaces, this needs to be called for objects that need line boxes such as RenderInlines or
    // hard line breaks to ensure that they're not ignored.
    void ensureLineBoxInsideIgnoredSpaces(RenderObject* renderer)
    {
        Iterator midpoint(0, renderer, 0);
        stopIgnoringSpaces(midpoint);
        startIgnoringSpaces(midpoint);
    }

    // Adding a pair of midpoints before a character will split it out into a new line box.
    void ensureCharacterGetsLineBox(Iterator& textParagraphSeparator)
    {
        startIgnoringSpaces(Iterator(0, textParagraphSeparator.object(), textParagraphSeparator.offset() - 1));
        stopIgnoringSpaces(Iterator(0, textParagraphSeparator.object(), textParagraphSeparator.offset()));
    }

    void checkMidpoints(Iterator& lBreak)
    {
        // Check to see if our last midpoint is a start point beyond the line break. If so,
        // shave it off the list, and shave off a trailing space if the previous end point doesn't
        // preserve whitespace.
        if (lBreak.object() && m_numMidpoints && !(m_numMidpoints % 2)) {
            Iterator* midpointsIterator = m_midpoints.data();
            Iterator& endpoint = midpointsIterator[m_numMidpoints - 2];
            const Iterator& startpoint = midpointsIterator[m_numMidpoints - 1];
            Iterator currpoint = endpoint;
            while (!currpoint.atEnd() && currpoint != startpoint && currpoint != lBreak)
                currpoint.increment();
            if (currpoint == lBreak) {
                // We hit the line break before the start point. Shave off the start point.
                m_numMidpoints--;
                if (endpoint.object()->style()->collapseWhiteSpace() && endpoint.object()->isText())
                    endpoint.setOffset(endpoint.offset() - 1);
            }
        }
    }

    Vector<Iterator>& midpoints() { return m_midpoints; }
    const unsigned& numMidpoints() const { return m_numMidpoints; }
    const unsigned& currentMidpoint() const { return m_currentMidpoint; }
    void incrementCurrentMidpoint() { m_currentMidpoint++; }
    const bool& betweenMidpoints() const { return m_betweenMidpoints; }
    void setBetweenMidpoints(bool betweenMidpoint) { m_betweenMidpoints = betweenMidpoint; }
private:
    // The goal is to reuse the line state across multiple
    // lines so we just keep an array around for midpoints and never clear it across multiple
    // lines. We track the number of items and position using the two other variables.
    Vector<Iterator> m_midpoints;
    unsigned m_numMidpoints;
    unsigned m_currentMidpoint;
    bool m_betweenMidpoints;

    void addMidpoint(const Iterator& midpoint)
    {
        if (m_midpoints.size() <= m_numMidpoints)
            m_midpoints.grow(m_numMidpoints + 10);

        Iterator* midpointsIterator = m_midpoints.data();
        midpointsIterator[m_numMidpoints++] = midpoint;
    }
};

// The BidiStatus at a given position (typically the end of a line) can
// be cached and then used to restart bidi resolution at that position.
struct BidiStatus {
    BidiStatus()
        : eor(WTF::Unicode::OtherNeutral)
        , lastStrong(WTF::Unicode::OtherNeutral)
        , last(WTF::Unicode::OtherNeutral)
    {
    }

    // Creates a BidiStatus representing a new paragraph root with a default direction.
    // Uses TextDirection as it only has two possibilities instead of WTF::Unicode::Direction which has 19.
    BidiStatus(TextDirection textDirection, bool isOverride)
    {
        WTF::Unicode::Direction direction = textDirection == LTR ? WTF::Unicode::LeftToRight : WTF::Unicode::RightToLeft;
        eor = lastStrong = last = direction;
        context = BidiContext::create(textDirection == LTR ? 0 : 1, direction, isOverride);
    }

    BidiStatus(WTF::Unicode::Direction eorDir, WTF::Unicode::Direction lastStrongDir, WTF::Unicode::Direction lastDir, PassRefPtr<BidiContext> bidiContext)
        : eor(eorDir)
        , lastStrong(lastStrongDir)
        , last(lastDir)
        , context(bidiContext)
    {
    }

    WTF::Unicode::Direction eor;
    WTF::Unicode::Direction lastStrong;
    WTF::Unicode::Direction last;
    RefPtr<BidiContext> context;
};

class BidiEmbedding {
public:
    BidiEmbedding(WTF::Unicode::Direction direction, BidiEmbeddingSource source)
    : m_direction(direction)
    , m_source(source)
    {
    }

    WTF::Unicode::Direction direction() const { return m_direction; }
    BidiEmbeddingSource source() const { return m_source; }
private:
    WTF::Unicode::Direction m_direction;
    BidiEmbeddingSource m_source;
};

inline bool operator==(const BidiStatus& status1, const BidiStatus& status2)
{
    return status1.eor == status2.eor && status1.last == status2.last && status1.lastStrong == status2.lastStrong && *(status1.context) == *(status2.context);
}

inline bool operator!=(const BidiStatus& status1, const BidiStatus& status2)
{
    return !(status1 == status2);
}

enum VisualDirectionOverride {
    NoVisualOverride,
    VisualLeftToRightOverride,
    VisualRightToLeftOverride
};

// BidiResolver is WebKit's implementation of the Unicode Bidi Algorithm
// http://unicode.org/reports/tr9
template <class Iterator, class Run> class BidiResolver {
    WTF_MAKE_NONCOPYABLE(BidiResolver);
public:
    BidiResolver()
        : m_direction(WTF::Unicode::OtherNeutral)
        , m_reachedEndOfLine(false)
        , m_emptyRun(true)
        , m_nestedIsolateCount(0)
        , m_trailingSpaceRun(0)
    {
    }

#if ENABLE(ASSERT)
    ~BidiResolver();
#endif

    const Iterator& position() const { return m_current; }
    Iterator& position() { return m_current; }
    void setPositionIgnoringNestedIsolates(const Iterator& position) { m_current = position; }
    void setPosition(const Iterator& position, unsigned nestedIsolatedCount)
    {
        m_current = position;
        m_nestedIsolateCount = nestedIsolatedCount;
    }

    BidiContext* context() const { return m_status.context.get(); }
    void setContext(PassRefPtr<BidiContext> c) { m_status.context = c; }

    void setLastDir(WTF::Unicode::Direction lastDir) { m_status.last = lastDir; }
    void setLastStrongDir(WTF::Unicode::Direction lastStrongDir) { m_status.lastStrong = lastStrongDir; }
    void setEorDir(WTF::Unicode::Direction eorDir) { m_status.eor = eorDir; }

    WTF::Unicode::Direction dir() const { return m_direction; }
    void setDir(WTF::Unicode::Direction d) { m_direction = d; }

    const BidiStatus& status() const { return m_status; }
    void setStatus(const BidiStatus s)
    {
        ASSERT(s.context);
        m_status = s;
        m_paragraphDirectionality = s.context->dir() == WTF::Unicode::LeftToRight ? LTR : RTL;
    }

    MidpointState<Iterator>& midpointState() { return m_midpointState; }

    // The current algorithm handles nested isolates one layer of nesting at a time.
    // But when we layout each isolated span, we will walk into (and ignore) all
    // child isolated spans.
    void enterIsolate() { m_nestedIsolateCount++; }
    void exitIsolate() { ASSERT(m_nestedIsolateCount >= 1); m_nestedIsolateCount--; }
    bool inIsolate() const { return m_nestedIsolateCount; }

    void embed(WTF::Unicode::Direction, BidiEmbeddingSource);
    bool commitExplicitEmbedding(BidiRunList<Run>&);

    void createBidiRunsForLine(const Iterator& end, VisualDirectionOverride = NoVisualOverride, bool hardLineBreak = false, bool reorderRuns = true);

    BidiRunList<Run>& runs() { return m_runs; }

    // FIXME: This used to be part of deleteRuns() but was a layering violation.
    // It's unclear if this is still needed.
    void markCurrentRunEmpty() { m_emptyRun = true; }

    Vector<Run*>& isolatedRuns() { return m_isolatedRuns; }

    bool isEndOfLine(const Iterator& end) { return m_current == end || m_current.atEnd(); }

    TextDirection determineParagraphDirectionality(bool* hasStrongDirectionality = 0);

    void setMidpointStateForIsolatedRun(Run*, const MidpointState<Iterator>&);
    MidpointState<Iterator> midpointStateForIsolatedRun(Run*);

    Iterator endOfLine() const { return m_endOfLine; }

    Run* trailingSpaceRun() const { return m_trailingSpaceRun; }

protected:
    void increment() { m_current.increment(); }
    // FIXME: Instead of InlineBidiResolvers subclassing this method, we should
    // pass in some sort of Traits object which knows how to create runs for appending.
    void appendRun(BidiRunList<Run>&);

    Run* addTrailingRun(BidiRunList<Run>&, int, int, Run*, BidiContext*, TextDirection) const { return 0; }
    Iterator m_current;
    // sor and eor are "start of run" and "end of run" respectively and correpond
    // to abreviations used in UBA spec: http://unicode.org/reports/tr9/#BD7
    Iterator m_sor; // Points to the first character in the current run.
    Iterator m_eor; // Points to the last character in the current run.
    Iterator m_last;
    BidiStatus m_status;
    WTF::Unicode::Direction m_direction;
    // m_endOfRunAtEndOfLine is "the position last eor in the end of line"
    Iterator m_endOfRunAtEndOfLine;
    Iterator m_endOfLine;
    bool m_reachedEndOfLine;
    Iterator m_lastBeforeET; // Before a EuropeanNumberTerminator
    bool m_emptyRun;

    // FIXME: This should not belong to the resolver, but rather be passed
    // into createBidiRunsForLine by the caller.
    BidiRunList<Run> m_runs;

    MidpointState<Iterator> m_midpointState;

    unsigned m_nestedIsolateCount;
    Vector<Run*> m_isolatedRuns;
    Run* m_trailingSpaceRun;
    TextDirection m_paragraphDirectionality;

private:
    void raiseExplicitEmbeddingLevel(BidiRunList<Run>&, WTF::Unicode::Direction from, WTF::Unicode::Direction to);
    void lowerExplicitEmbeddingLevel(BidiRunList<Run>&, WTF::Unicode::Direction from);
    void checkDirectionInLowerRaiseEmbeddingLevel();

    void updateStatusLastFromCurrentDirection(WTF::Unicode::Direction);
    void reorderRunsFromLevels(BidiRunList<Run>&) const;

    bool needsToApplyL1Rule(BidiRunList<Run>&) { return false; }
    int findFirstTrailingSpaceAtRun(Run*) { return 0; }
    // http://www.unicode.org/reports/tr9/#L1
    void applyL1Rule(BidiRunList<Run>&);

    Vector<BidiEmbedding, 8> m_currentExplicitEmbeddingSequence;
    HashMap<Run *, MidpointState<Iterator> > m_midpointStateForIsolatedRun;
};

#if ENABLE(ASSERT)
template <class Iterator, class Run>
BidiResolver<Iterator, Run>::~BidiResolver()
{
    // The owner of this resolver should have handled the isolated runs.
    ASSERT(m_isolatedRuns.isEmpty());
}
#endif

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::appendRun(BidiRunList<Run>& runs)
{
    if (!m_emptyRun && !m_eor.atEnd()) {
        unsigned startOffset = m_sor.offset();
        unsigned endOffset = m_eor.offset();

        if (!m_endOfRunAtEndOfLine.atEnd() && endOffset >= m_endOfRunAtEndOfLine.offset()) {
            m_reachedEndOfLine = true;
            endOffset = m_endOfRunAtEndOfLine.offset();
        }

        if (endOffset >= startOffset)
            runs.addRun(new Run(startOffset, endOffset + 1, context(), m_direction));

        m_eor.increment();
        m_sor = m_eor;
    }

    m_direction = WTF::Unicode::OtherNeutral;
    m_status.eor = WTF::Unicode::OtherNeutral;
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::embed(WTF::Unicode::Direction dir, BidiEmbeddingSource source)
{
    // Isolated spans compute base directionality during their own UBA run.
    // Do not insert fake embed characters once we enter an isolated span.
    ASSERT(!inIsolate());
    using namespace WTF::Unicode;

    ASSERT(dir == PopDirectionalFormat || dir == LeftToRightEmbedding || dir == LeftToRightOverride || dir == RightToLeftEmbedding || dir == RightToLeftOverride);
    m_currentExplicitEmbeddingSequence.append(BidiEmbedding(dir, source));
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::checkDirectionInLowerRaiseEmbeddingLevel()
{
    using namespace WTF::Unicode;

    ASSERT(m_status.eor != OtherNeutral || m_eor.atEnd());
    ASSERT(m_status.last != NonSpacingMark
        && m_status.last != BoundaryNeutral
        && m_status.last != RightToLeftEmbedding
        && m_status.last != LeftToRightEmbedding
        && m_status.last != RightToLeftOverride
        && m_status.last != LeftToRightOverride
        && m_status.last != PopDirectionalFormat);
    if (m_direction == OtherNeutral)
        m_direction = m_status.lastStrong == LeftToRight ? LeftToRight : RightToLeft;
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::lowerExplicitEmbeddingLevel(BidiRunList<Run>& runs, WTF::Unicode::Direction from)
{
    using namespace WTF::Unicode;

    if (!m_emptyRun && m_eor != m_last) {
        checkDirectionInLowerRaiseEmbeddingLevel();
        // bidi.sor ... bidi.eor ... bidi.last eor; need to append the bidi.sor-bidi.eor run or extend it through bidi.last
        if (from == LeftToRight) {
            // bidi.sor ... bidi.eor ... bidi.last L
            if (m_status.eor == EuropeanNumber) {
                if (m_status.lastStrong != LeftToRight) {
                    m_direction = EuropeanNumber;
                    appendRun(runs);
                }
            } else if (m_status.eor == ArabicNumber) {
                m_direction = ArabicNumber;
                appendRun(runs);
            } else if (m_status.lastStrong != LeftToRight) {
                appendRun(runs);
                m_direction = LeftToRight;
            }
        } else if (m_status.eor == EuropeanNumber || m_status.eor == ArabicNumber || m_status.lastStrong == LeftToRight) {
            appendRun(runs);
            m_direction = RightToLeft;
        }
        m_eor = m_last;
    }

    appendRun(runs);
    m_emptyRun = true;

    // sor for the new run is determined by the higher level (rule X10)
    setLastDir(from);
    setLastStrongDir(from);
    m_eor = Iterator();
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::raiseExplicitEmbeddingLevel(BidiRunList<Run>& runs, WTF::Unicode::Direction from, WTF::Unicode::Direction to)
{
    using namespace WTF::Unicode;

    if (!m_emptyRun && m_eor != m_last) {
        checkDirectionInLowerRaiseEmbeddingLevel();
        // bidi.sor ... bidi.eor ... bidi.last eor; need to append the bidi.sor-bidi.eor run or extend it through bidi.last
        if (to == LeftToRight) {
            // bidi.sor ... bidi.eor ... bidi.last L
            if (m_status.eor == EuropeanNumber) {
                if (m_status.lastStrong != LeftToRight) {
                    m_direction = EuropeanNumber;
                    appendRun(runs);
                }
            } else if (m_status.eor == ArabicNumber) {
                m_direction = ArabicNumber;
                appendRun(runs);
            } else if (m_status.lastStrong != LeftToRight && from == LeftToRight) {
                appendRun(runs);
                m_direction = LeftToRight;
            }
        } else if (m_status.eor == ArabicNumber
            || (m_status.eor == EuropeanNumber && (m_status.lastStrong != LeftToRight || from == RightToLeft))
            || (m_status.eor != EuropeanNumber && m_status.lastStrong == LeftToRight && from == RightToLeft)) {
            appendRun(runs);
            m_direction = RightToLeft;
        }
        m_eor = m_last;
    }

    appendRun(runs);
    m_emptyRun = true;

    setLastDir(to);
    setLastStrongDir(to);
    m_eor = Iterator();
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::applyL1Rule(BidiRunList<Run>& runs)
{
    ASSERT(runs.runCount());
    if (!needsToApplyL1Rule(runs))
        return;

    Run* trailingSpaceRun = runs.logicallyLastRun();

    int firstSpace = findFirstTrailingSpaceAtRun(trailingSpaceRun);
    if (firstSpace == trailingSpaceRun->stop())
        return;

    bool shouldReorder = trailingSpaceRun != (m_paragraphDirectionality == LTR ? runs.lastRun() : runs.firstRun());
    if (firstSpace != trailingSpaceRun->start()) {
        BidiContext* baseContext = context();
        while (BidiContext* parent = baseContext->parent())
            baseContext = parent;

        m_trailingSpaceRun = addTrailingRun(runs, firstSpace, trailingSpaceRun->m_stop, trailingSpaceRun, baseContext, m_paragraphDirectionality);
        ASSERT(m_trailingSpaceRun);
        trailingSpaceRun->m_stop = firstSpace;
        return;
    }
    if (!shouldReorder) {
        m_trailingSpaceRun = trailingSpaceRun;
        return;
    }

    if (m_paragraphDirectionality == LTR) {
        runs.moveRunToEnd(trailingSpaceRun);
        trailingSpaceRun->m_level = 0;
    } else {
        runs.moveRunToBeginning(trailingSpaceRun);
        trailingSpaceRun->m_level = 1;
    }
    m_trailingSpaceRun = trailingSpaceRun;
}

template <class Iterator, class Run>
bool BidiResolver<Iterator, Run>::commitExplicitEmbedding(BidiRunList<Run>& runs)
{
    // When we're "inIsolate()" we're resolving the parent context which
    // ignores (skips over) the isolated content, including embedding levels.
    // We should never accrue embedding levels while skipping over isolated content.
    ASSERT(!inIsolate() || m_currentExplicitEmbeddingSequence.isEmpty());

    using namespace WTF::Unicode;

    unsigned char fromLevel = context()->level();
    RefPtr<BidiContext> toContext = context();

    for (size_t i = 0; i < m_currentExplicitEmbeddingSequence.size(); ++i) {
        BidiEmbedding embedding = m_currentExplicitEmbeddingSequence[i];
        if (embedding.direction() == PopDirectionalFormat) {
            if (BidiContext* parentContext = toContext->parent())
                toContext = parentContext;
        } else {
            Direction direction = (embedding.direction() == RightToLeftEmbedding || embedding.direction() == RightToLeftOverride) ? RightToLeft : LeftToRight;
            bool override = embedding.direction() == LeftToRightOverride || embedding.direction() == RightToLeftOverride;
            unsigned char level = toContext->level();
            if (direction == RightToLeft)
                level = nextGreaterOddLevel(level);
            else
                level = nextGreaterEvenLevel(level);
            if (level < BidiContext::kMaxLevel)
                toContext = BidiContext::create(level, direction, override, embedding.source(), toContext.get());
        }
    }

    unsigned char toLevel = toContext->level();

    if (toLevel > fromLevel)
        raiseExplicitEmbeddingLevel(runs, fromLevel % 2 ? RightToLeft : LeftToRight, toLevel % 2 ? RightToLeft : LeftToRight);
    else if (toLevel < fromLevel)
        lowerExplicitEmbeddingLevel(runs, fromLevel % 2 ? RightToLeft : LeftToRight);

    setContext(toContext);

    m_currentExplicitEmbeddingSequence.clear();

    return fromLevel != toLevel;
}

template <class Iterator, class Run>
inline void BidiResolver<Iterator, Run>::updateStatusLastFromCurrentDirection(WTF::Unicode::Direction dirCurrent)
{
    using namespace WTF::Unicode;
    switch (dirCurrent) {
    case EuropeanNumberTerminator:
        if (m_status.last != EuropeanNumber)
            m_status.last = EuropeanNumberTerminator;
        break;
    case EuropeanNumberSeparator:
    case CommonNumberSeparator:
    case SegmentSeparator:
    case WhiteSpaceNeutral:
    case OtherNeutral:
        switch (m_status.last) {
        case LeftToRight:
        case RightToLeft:
        case RightToLeftArabic:
        case EuropeanNumber:
        case ArabicNumber:
            m_status.last = dirCurrent;
            break;
        default:
            m_status.last = OtherNeutral;
        }
        break;
    case NonSpacingMark:
    case BoundaryNeutral:
    case RightToLeftEmbedding:
    case LeftToRightEmbedding:
    case RightToLeftOverride:
    case LeftToRightOverride:
    case PopDirectionalFormat:
        // ignore these
        break;
    case EuropeanNumber:
        // fall through
    default:
        m_status.last = dirCurrent;
    }
}

template <class Iterator, class Run>
inline void BidiResolver<Iterator, Run>::reorderRunsFromLevels(BidiRunList<Run>& runs) const
{
    unsigned char levelLow = BidiContext::kMaxLevel;
    unsigned char levelHigh = 0;
    for (Run* run = runs.firstRun(); run; run = run->next()) {
        levelHigh = std::max(run->level(), levelHigh);
        levelLow = std::min(run->level(), levelLow);
    }

    // This implements reordering of the line (L2 according to Bidi spec):
    // http://unicode.org/reports/tr9/#L2
    // L2. From the highest level found in the text to the lowest odd level on each line,
    // reverse any contiguous sequence of characters that are at that level or higher.

    // Reversing is only done up to the lowest odd level.
    if (!(levelLow % 2))
        levelLow++;

    unsigned count = runs.runCount() - 1;

    while (levelHigh >= levelLow) {
        unsigned i = 0;
        Run* run = runs.firstRun();
        while (i < count) {
            for (;i < count && run && run->level() < levelHigh; i++)
                run = run->next();
            unsigned start = i;
            for (;i <= count && run && run->level() >= levelHigh; i++)
                run = run->next();
            unsigned end = i - 1;
            runs.reverseRuns(start, end);
        }
        levelHigh--;
    }
}

template <class Iterator, class Run>
TextDirection BidiResolver<Iterator, Run>::determineParagraphDirectionality(bool* hasStrongDirectionality)
{
    while (!m_current.atEnd()) {
        if (inIsolate()) {
            increment();
            continue;
        }
        if (m_current.atParagraphSeparator())
            break;
        UChar32 current = m_current.current();
        if (UNLIKELY(U16_IS_SURROGATE(current))) {
            increment();
            // If this not the high part of the surrogate pair, then drop it and move to the next.
            if (!U16_IS_SURROGATE_LEAD(current))
                continue;
            UChar high = static_cast<UChar>(current);
            if (m_current.atEnd())
                continue;
            UChar low = m_current.current();
            // Verify the low part. If invalid, then assume an invalid surrogate pair and retry.
            if (!U16_IS_TRAIL(low))
                continue;
            current = U16_GET_SUPPLEMENTARY(high, low);
        }
        WTF::Unicode::Direction charDirection = WTF::Unicode::direction(current);
        if (charDirection == WTF::Unicode::LeftToRight) {
            if (hasStrongDirectionality)
                *hasStrongDirectionality = true;
            return LTR;
        }
        if (charDirection == WTF::Unicode::RightToLeft || charDirection == WTF::Unicode::RightToLeftArabic) {
            if (hasStrongDirectionality)
                *hasStrongDirectionality = true;
            return RTL;
        }
        increment();
    }
    if (hasStrongDirectionality)
        *hasStrongDirectionality = false;
    return LTR;
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::createBidiRunsForLine(const Iterator& end, VisualDirectionOverride override, bool hardLineBreak, bool reorderRuns)
{
    using namespace WTF::Unicode;

    ASSERT(m_direction == OtherNeutral);
    m_trailingSpaceRun = 0;

    m_endOfLine = end;

    if (override != NoVisualOverride) {
        m_emptyRun = false;
        m_sor = m_current;
        m_eor = Iterator();
        while (m_current != end && !m_current.atEnd()) {
            m_eor = m_current;
            increment();
        }
        m_direction = override == VisualLeftToRightOverride ? LeftToRight : RightToLeft;
        appendRun(m_runs);
        m_runs.setLogicallyLastRun(m_runs.lastRun());
        if (override == VisualRightToLeftOverride && m_runs.runCount())
            m_runs.reverseRuns(0, m_runs.runCount() - 1);
        return;
    }

    m_emptyRun = true;

    m_eor = Iterator();

    m_last = m_current;
    bool lastLineEnded = false;
    BidiResolver<Iterator, Run> stateAtEnd;

    while (true) {
        if (inIsolate() && m_emptyRun) {
            m_sor = m_current;
            m_emptyRun = false;
        }

        if (!lastLineEnded && isEndOfLine(end)) {
            if (m_emptyRun)
                break;

            stateAtEnd.m_status = m_status;
            stateAtEnd.m_sor = m_sor;
            stateAtEnd.m_eor = m_eor;
            stateAtEnd.m_last = m_last;
            stateAtEnd.m_reachedEndOfLine = m_reachedEndOfLine;
            stateAtEnd.m_lastBeforeET = m_lastBeforeET;
            stateAtEnd.m_emptyRun = m_emptyRun;
            m_endOfRunAtEndOfLine = m_last;
            lastLineEnded = true;
        }
        Direction dirCurrent;
        if (lastLineEnded && (hardLineBreak || m_current.atEnd())) {
            BidiContext* c = context();
            if (hardLineBreak) {
                // A deviation from the Unicode Bidi Algorithm in order to match
                // WinIE and user expectations: hard line breaks reset bidi state
                // coming from unicode bidi control characters, but not those from
                // DOM nodes with specified directionality
                stateAtEnd.setContext(c->copyStackRemovingUnicodeEmbeddingContexts());

                dirCurrent = stateAtEnd.context()->dir();
                stateAtEnd.setEorDir(dirCurrent);
                stateAtEnd.setLastDir(dirCurrent);
                stateAtEnd.setLastStrongDir(dirCurrent);
            } else {
                while (c->parent())
                    c = c->parent();
                dirCurrent = c->dir();
            }
        } else {
            dirCurrent = m_current.direction();
            if (context()->override()
                && dirCurrent != RightToLeftEmbedding
                && dirCurrent != LeftToRightEmbedding
                && dirCurrent != RightToLeftOverride
                && dirCurrent != LeftToRightOverride
                && dirCurrent != PopDirectionalFormat)
                dirCurrent = context()->dir();
            else if (dirCurrent == NonSpacingMark)
                dirCurrent = m_status.last;
        }

        // We ignore all character directionality while in unicode-bidi: isolate spans.
        // We'll handle ordering the isolated characters in a second pass.
        if (inIsolate())
            dirCurrent = OtherNeutral;

        ASSERT(m_status.eor != OtherNeutral || m_eor.atEnd());
        switch (dirCurrent) {

        // embedding and overrides (X1-X9 in the Bidi specs)
        case RightToLeftEmbedding:
        case LeftToRightEmbedding:
        case RightToLeftOverride:
        case LeftToRightOverride:
        case PopDirectionalFormat:
            embed(dirCurrent, FromUnicode);
            commitExplicitEmbedding(m_runs);
            break;

        // strong types
        case LeftToRight:
            switch (m_status.last) {
            case RightToLeft:
            case RightToLeftArabic:
            case EuropeanNumber:
            case ArabicNumber:
                if (m_status.last != EuropeanNumber || m_status.lastStrong != LeftToRight)
                    appendRun(m_runs);
                break;
            case LeftToRight:
                break;
            case EuropeanNumberSeparator:
            case EuropeanNumberTerminator:
            case CommonNumberSeparator:
            case BoundaryNeutral:
            case BlockSeparator:
            case SegmentSeparator:
            case WhiteSpaceNeutral:
            case OtherNeutral:
                if (m_status.eor == EuropeanNumber) {
                    if (m_status.lastStrong != LeftToRight) {
                        // the numbers need to be on a higher embedding level, so let's close that run
                        m_direction = EuropeanNumber;
                        appendRun(m_runs);
                        if (context()->dir() != LeftToRight) {
                            // the neutrals take the embedding direction, which is R
                            m_eor = m_last;
                            m_direction = RightToLeft;
                            appendRun(m_runs);
                        }
                    }
                } else if (m_status.eor == ArabicNumber) {
                    // Arabic numbers are always on a higher embedding level, so let's close that run
                    m_direction = ArabicNumber;
                    appendRun(m_runs);
                    if (context()->dir() != LeftToRight) {
                        // the neutrals take the embedding direction, which is R
                        m_eor = m_last;
                        m_direction = RightToLeft;
                        appendRun(m_runs);
                    }
                } else if (m_status.lastStrong != LeftToRight) {
                    // last stuff takes embedding dir
                    if (context()->dir() == RightToLeft) {
                        m_eor = m_last;
                        m_direction = RightToLeft;
                    }
                    appendRun(m_runs);
                }
            default:
                break;
            }
            m_eor = m_current;
            m_status.eor = LeftToRight;
            m_status.lastStrong = LeftToRight;
            m_direction = LeftToRight;
            break;
        case RightToLeftArabic:
        case RightToLeft:
            switch (m_status.last) {
            case LeftToRight:
            case EuropeanNumber:
            case ArabicNumber:
                appendRun(m_runs);
            case RightToLeft:
            case RightToLeftArabic:
                break;
            case EuropeanNumberSeparator:
            case EuropeanNumberTerminator:
            case CommonNumberSeparator:
            case BoundaryNeutral:
            case BlockSeparator:
            case SegmentSeparator:
            case WhiteSpaceNeutral:
            case OtherNeutral:
                if (m_status.eor == EuropeanNumber) {
                    if (m_status.lastStrong == LeftToRight && context()->dir() == LeftToRight)
                        m_eor = m_last;
                    appendRun(m_runs);
                } else if (m_status.eor == ArabicNumber) {
                    appendRun(m_runs);
                } else if (m_status.lastStrong == LeftToRight) {
                    if (context()->dir() == LeftToRight)
                        m_eor = m_last;
                    appendRun(m_runs);
                }
            default:
                break;
            }
            m_eor = m_current;
            m_status.eor = RightToLeft;
            m_status.lastStrong = dirCurrent;
            m_direction = RightToLeft;
            break;

            // weak types:

        case EuropeanNumber:
            if (m_status.lastStrong != RightToLeftArabic) {
                // if last strong was AL change EN to AN
                switch (m_status.last) {
                case EuropeanNumber:
                case LeftToRight:
                    break;
                case RightToLeft:
                case RightToLeftArabic:
                case ArabicNumber:
                    m_eor = m_last;
                    appendRun(m_runs);
                    m_direction = EuropeanNumber;
                    break;
                case EuropeanNumberSeparator:
                case CommonNumberSeparator:
                    if (m_status.eor == EuropeanNumber)
                        break;
                case EuropeanNumberTerminator:
                case BoundaryNeutral:
                case BlockSeparator:
                case SegmentSeparator:
                case WhiteSpaceNeutral:
                case OtherNeutral:
                    if (m_status.eor == EuropeanNumber) {
                        if (m_status.lastStrong == RightToLeft) {
                            // ENs on both sides behave like Rs, so the neutrals should be R.
                            // Terminate the EN run.
                            appendRun(m_runs);
                            // Make an R run.
                            m_eor = m_status.last == EuropeanNumberTerminator ? m_lastBeforeET : m_last;
                            m_direction = RightToLeft;
                            appendRun(m_runs);
                            // Begin a new EN run.
                            m_direction = EuropeanNumber;
                        }
                    } else if (m_status.eor == ArabicNumber) {
                        // Terminate the AN run.
                        appendRun(m_runs);
                        if (m_status.lastStrong == RightToLeft || context()->dir() == RightToLeft) {
                            // Make an R run.
                            m_eor = m_status.last == EuropeanNumberTerminator ? m_lastBeforeET : m_last;
                            m_direction = RightToLeft;
                            appendRun(m_runs);
                            // Begin a new EN run.
                            m_direction = EuropeanNumber;
                        }
                    } else if (m_status.lastStrong == RightToLeft) {
                        // Extend the R run to include the neutrals.
                        m_eor = m_status.last == EuropeanNumberTerminator ? m_lastBeforeET : m_last;
                        m_direction = RightToLeft;
                        appendRun(m_runs);
                        // Begin a new EN run.
                        m_direction = EuropeanNumber;
                    }
                default:
                    break;
                }
                m_eor = m_current;
                m_status.eor = EuropeanNumber;
                if (m_direction == OtherNeutral)
                    m_direction = LeftToRight;
                break;
            }
        case ArabicNumber:
            dirCurrent = ArabicNumber;
            switch (m_status.last) {
            case LeftToRight:
                if (context()->dir() == LeftToRight)
                    appendRun(m_runs);
                break;
            case ArabicNumber:
                break;
            case RightToLeft:
            case RightToLeftArabic:
            case EuropeanNumber:
                m_eor = m_last;
                appendRun(m_runs);
                break;
            case CommonNumberSeparator:
                if (m_status.eor == ArabicNumber)
                    break;
            case EuropeanNumberSeparator:
            case EuropeanNumberTerminator:
            case BoundaryNeutral:
            case BlockSeparator:
            case SegmentSeparator:
            case WhiteSpaceNeutral:
            case OtherNeutral:
                if (m_status.eor == ArabicNumber
                    || (m_status.eor == EuropeanNumber && (m_status.lastStrong == RightToLeft || context()->dir() == RightToLeft))
                    || (m_status.eor != EuropeanNumber && m_status.lastStrong == LeftToRight && context()->dir() == RightToLeft)) {
                    // Terminate the run before the neutrals.
                    appendRun(m_runs);
                    // Begin an R run for the neutrals.
                    m_direction = RightToLeft;
                } else if (m_direction == OtherNeutral) {
                    m_direction = m_status.lastStrong == LeftToRight ? LeftToRight : RightToLeft;
                }
                m_eor = m_last;
                appendRun(m_runs);
            default:
                break;
            }
            m_eor = m_current;
            m_status.eor = ArabicNumber;
            if (m_direction == OtherNeutral)
                m_direction = ArabicNumber;
            break;
        case EuropeanNumberSeparator:
        case CommonNumberSeparator:
            break;
        case EuropeanNumberTerminator:
            if (m_status.last == EuropeanNumber) {
                dirCurrent = EuropeanNumber;
                m_eor = m_current;
                m_status.eor = dirCurrent;
            } else if (m_status.last != EuropeanNumberTerminator) {
                m_lastBeforeET = m_emptyRun ? m_eor : m_last;
            }
            break;

        // boundary neutrals should be ignored
        case BoundaryNeutral:
            if (m_eor == m_last)
                m_eor = m_current;
            break;
            // neutrals
        case BlockSeparator:
            // ### what do we do with newline and paragraph seperators that come to here?
            break;
        case SegmentSeparator:
            // ### implement rule L1
            break;
        case WhiteSpaceNeutral:
            break;
        case OtherNeutral:
            break;
        default:
            break;
        }

        if (lastLineEnded && m_eor == m_current) {
            if (!m_reachedEndOfLine) {
                m_eor = m_endOfRunAtEndOfLine;
                switch (m_status.eor) {
                case LeftToRight:
                case RightToLeft:
                case ArabicNumber:
                    m_direction = m_status.eor;
                    break;
                case EuropeanNumber:
                    m_direction = m_status.lastStrong == LeftToRight ? LeftToRight : EuropeanNumber;
                    break;
                default:
                    ASSERT_NOT_REACHED();
                }
                appendRun(m_runs);
            }
            m_current = end;
            m_status = stateAtEnd.m_status;
            m_sor = stateAtEnd.m_sor;
            m_eor = stateAtEnd.m_eor;
            m_last = stateAtEnd.m_last;
            m_reachedEndOfLine = stateAtEnd.m_reachedEndOfLine;
            m_lastBeforeET = stateAtEnd.m_lastBeforeET;
            m_emptyRun = stateAtEnd.m_emptyRun;
            m_direction = OtherNeutral;
            break;
        }

        updateStatusLastFromCurrentDirection(dirCurrent);
        m_last = m_current;

        if (m_emptyRun) {
            m_sor = m_current;
            m_emptyRun = false;
        }

        increment();
        if (!m_currentExplicitEmbeddingSequence.isEmpty()) {
            bool committed = commitExplicitEmbedding(m_runs);
            if (committed && lastLineEnded) {
                m_current = end;
                m_status = stateAtEnd.m_status;
                m_sor = stateAtEnd.m_sor;
                m_eor = stateAtEnd.m_eor;
                m_last = stateAtEnd.m_last;
                m_reachedEndOfLine = stateAtEnd.m_reachedEndOfLine;
                m_lastBeforeET = stateAtEnd.m_lastBeforeET;
                m_emptyRun = stateAtEnd.m_emptyRun;
                m_direction = OtherNeutral;
                break;
            }
        }
    }

    m_runs.setLogicallyLastRun(m_runs.lastRun());
    if (reorderRuns)
        reorderRunsFromLevels(m_runs);
    m_endOfRunAtEndOfLine = Iterator();
    m_endOfLine = Iterator();

    if (!hardLineBreak && m_runs.runCount())
        applyL1Rule(m_runs);
}

template <class Iterator, class Run>
void BidiResolver<Iterator, Run>::setMidpointStateForIsolatedRun(Run* run, const MidpointState<Iterator>& midpoint)
{
    ASSERT(!m_midpointStateForIsolatedRun.contains(run));
    m_midpointStateForIsolatedRun.add(run, midpoint);
}

template<class Iterator, class Run>
MidpointState<Iterator> BidiResolver<Iterator, Run>::midpointStateForIsolatedRun(Run* run)
{
    return m_midpointStateForIsolatedRun.take(run);
}


} // namespace blink

#endif // BidiResolver_h