textiterator.cpp

linux下开源浏览器WebKit的源码,市面上的很多商用浏览器都是移植自WebKit

                if (renderer->style()->visibility() == VISIBLE && m_offset > 0)
                    m_handledNode = handleTextNode();
            } else if (renderer && (renderer->isImage() || renderer->isWidget())) {
                if (renderer->style()->visibility() == VISIBLE && m_offset > 0)
                    m_handledNode = handleReplacedElement();
            } else
                m_handledNode = handleNonTextNode();
            if (m_positionNode)
                return;
        }

        Node* next = m_handledChildren ? 0 : m_node->lastChild();
        if (!next) {
            // Exit empty containers as we pass over them or containers
            // where [container, 0] is where we started iterating.
            if (!m_handledNode &&
                canHaveChildrenForEditing(m_node) && 
                m_node->parentNode() && 
                (!m_node->lastChild() || m_node == m_endNode && m_endOffset == 0)) {
                exitNode();
                if (m_positionNode) {
                    m_handledNode = true;
                    m_handledChildren = true;
                    return;
                }            
            }
            // Exit all other containers.
            next = m_node->previousSibling();
            while (!next) {
                if (!m_node->parentNode())
                    break;
                m_node = m_node->parentNode();
                exitNode();
                if (m_positionNode) {
                    m_handledNode = true;
                    m_handledChildren = true;
                    return;
                }
                next = m_node->previousSibling();
            }
        }
        
        m_node = next;
        m_offset = m_node ? caretMaxOffset(m_node) : 0;
        m_handledNode = false;
        m_handledChildren = false;
        
        if (m_positionNode)
            return;
    }
}

bool SimplifiedBackwardsTextIterator::handleTextNode()
{
    m_lastTextNode = m_node;

    RenderText *renderer = toRenderText(m_node->renderer());
    String str = renderer->text();

    if (!renderer->firstTextBox() && str.length() > 0)
        return true;

    m_positionEndOffset = m_offset;

    m_offset = (m_node == m_startNode) ? m_startOffset : 0;
    m_positionNode = m_node;
    m_positionStartOffset = m_offset;
    m_textLength = m_positionEndOffset - m_positionStartOffset;
    m_textCharacters = str.characters() + m_positionStartOffset;

    m_lastCharacter = str[m_positionEndOffset - 1];

    return true;
}

bool SimplifiedBackwardsTextIterator::handleReplacedElement()
{
    unsigned index = m_node->nodeIndex();
    // We want replaced elements to behave like punctuation for boundary 
    // finding, and to simply take up space for the selection preservation 
    // code in moveParagraphs, so we use a comma.  Unconditionally emit
    // here because this iterator is only used for boundary finding.
    emitCharacter(',', m_node->parentNode(), index, index + 1);
    return true;
}

bool SimplifiedBackwardsTextIterator::handleNonTextNode()
{    
    // We can use a linefeed in place of a tab because this simple iterator is only used to
    // find boundaries, not actual content.  A linefeed breaks words, sentences, and paragraphs.
    if (shouldEmitNewlineForNode(m_node) ||
        shouldEmitNewlineAfterNode(m_node) ||
        shouldEmitTabBeforeNode(m_node)) {
        unsigned index = m_node->nodeIndex();
        // The start of this emitted range is wrong, ensuring correctness would require
        // VisiblePositions and so would be slow.  previousBoundary expects this.
        emitCharacter('\n', m_node->parentNode(), index + 1, index + 1);
    }
    
    return true;
}

void SimplifiedBackwardsTextIterator::exitNode()
{
    if (shouldEmitNewlineForNode(m_node) ||
        shouldEmitNewlineBeforeNode(m_node) ||
        shouldEmitTabBeforeNode(m_node))
        // The start of this emitted range is wrong, ensuring correctness would require
        // VisiblePositions and so would be slow.  previousBoundary expects this.
        emitCharacter('\n', m_node, 0, 0);
}

void SimplifiedBackwardsTextIterator::emitCharacter(UChar c, Node *node, int startOffset, int endOffset)
{
    m_singleCharacterBuffer = c;
    m_positionNode = node;
    m_positionStartOffset = startOffset;
    m_positionEndOffset = endOffset;
    m_textCharacters = &m_singleCharacterBuffer;
    m_textLength = 1;
    m_lastCharacter = c;
}

PassRefPtr<Range> SimplifiedBackwardsTextIterator::range() const
{
    if (m_positionNode)
        return Range::create(m_positionNode->document(), m_positionNode, m_positionStartOffset, m_positionNode, m_positionEndOffset);
    
    return Range::create(m_startNode->document(), m_startNode, m_startOffset, m_startNode, m_startOffset);
}

// --------

CharacterIterator::CharacterIterator()
    : m_offset(0)
    , m_runOffset(0)
    , m_atBreak(true)
{
}

CharacterIterator::CharacterIterator(const Range *r, bool emitCharactersBetweenAllVisiblePositions, bool enterTextControls)
    : m_offset(0)
    , m_runOffset(0)
    , m_atBreak(true)
    , m_textIterator(r, emitCharactersBetweenAllVisiblePositions, enterTextControls)
{
    while (!atEnd() && m_textIterator.length() == 0)
        m_textIterator.advance();
}

PassRefPtr<Range> CharacterIterator::range() const
{
    RefPtr<Range> r = m_textIterator.range();
    if (!m_textIterator.atEnd()) {
        if (m_textIterator.length() <= 1) {
            ASSERT(m_runOffset == 0);
        } else {
            Node* n = r->startContainer();
            ASSERT(n == r->endContainer());
            int offset = r->startOffset() + m_runOffset;
            ExceptionCode ec = 0;
            r->setStart(n, offset, ec);
            r->setEnd(n, offset + 1, ec);
            ASSERT(!ec);
        }
    }
    return r.release();
}

void CharacterIterator::advance(int count)
{
    if (count <= 0) {
        ASSERT(count == 0);
        return;
    }
    
    m_atBreak = false;

    // easy if there is enough left in the current m_textIterator run
    int remaining = m_textIterator.length() - m_runOffset;
    if (count < remaining) {
        m_runOffset += count;
        m_offset += count;
        return;
    }

    // exhaust the current m_textIterator run
    count -= remaining;
    m_offset += remaining;
    
    // move to a subsequent m_textIterator run
    for (m_textIterator.advance(); !atEnd(); m_textIterator.advance()) {
        int runLength = m_textIterator.length();
        if (runLength == 0)
            m_atBreak = true;
        else {
            // see whether this is m_textIterator to use
            if (count < runLength) {
                m_runOffset = count;
                m_offset += count;
                return;
            }
            
            // exhaust this m_textIterator run
            count -= runLength;
            m_offset += runLength;
        }
    }

    // ran to the end of the m_textIterator... no more runs left
    m_atBreak = true;
    m_runOffset = 0;
}

String CharacterIterator::string(int numChars)
{
    Vector<UChar> result;
    result.reserveInitialCapacity(numChars);
    while (numChars > 0 && !atEnd()) {
        int runSize = min(numChars, length());
        result.append(characters(), runSize);
        numChars -= runSize;
        advance(runSize);
    }
    return String::adopt(result);
}

static PassRefPtr<Range> characterSubrange(CharacterIterator& it, int offset, int length)
{
    it.advance(offset);
    RefPtr<Range> start = it.range();

    if (length > 1)
        it.advance(length - 1);
    RefPtr<Range> end = it.range();

    return Range::create(start->startContainer()->document(), 
        start->startContainer(), start->startOffset(), 
        end->endContainer(), end->endOffset());
}

// --------

WordAwareIterator::WordAwareIterator()
: m_previousText(0), m_didLookAhead(false)
{
}

WordAwareIterator::WordAwareIterator(const Range *r)
: m_previousText(0), m_didLookAhead(false), m_textIterator(r)
{
    m_didLookAhead = true;  // so we consider the first chunk from the text iterator
    advance();              // get in position over the first chunk of text
}

// We're always in one of these modes:
// - The current chunk in the text iterator is our current chunk
//      (typically its a piece of whitespace, or text that ended with whitespace)
// - The previous chunk in the text iterator is our current chunk
//      (we looked ahead to the next chunk and found a word boundary)
// - We built up our own chunk of text from many chunks from the text iterator

// FIXME: Perf could be bad for huge spans next to each other that don't fall on word boundaries

void WordAwareIterator::advance()
{
    m_previousText = 0;
    m_buffer.clear();      // toss any old buffer we built up

    // If last time we did a look-ahead, start with that looked-ahead chunk now
    if (!m_didLookAhead) {
        ASSERT(!m_textIterator.atEnd());
        m_textIterator.advance();
    }
    m_didLookAhead = false;

    // Go to next non-empty chunk 
    while (!m_textIterator.atEnd() && m_textIterator.length() == 0)
        m_textIterator.advance();
    m_range = m_textIterator.range();

    if (m_textIterator.atEnd())
        return;
    
    while (1) {
        // If this chunk ends in whitespace we can just use it as our chunk.
        if (isSpaceOrNewline(m_textIterator.characters()[m_textIterator.length() - 1]))
            return;

        // If this is the first chunk that failed, save it in previousText before look ahead
        if (m_buffer.isEmpty()) {
            m_previousText = m_textIterator.characters();
            m_previousLength = m_textIterator.length();
        }

        // Look ahead to next chunk.  If it is whitespace or a break, we can use the previous stuff
        m_textIterator.advance();
        if (m_textIterator.atEnd() || m_textIterator.length() == 0 || isSpaceOrNewline(m_textIterator.characters()[0])) {
            m_didLookAhead = true;
            return;
        }

        if (m_buffer.isEmpty()) {
            // Start gobbling chunks until we get to a suitable stopping point
            m_buffer.append(m_previousText, m_previousLength);
            m_previousText = 0;
        }
        m_buffer.append(m_textIterator.characters(), m_textIterator.length());
        int exception = 0;
        m_range->setEnd(m_textIterator.range()->endContainer(), m_textIterator.range()->endOffset(), exception);
    }
}

int WordAwareIterator::length() const
{
    if (!m_buffer.isEmpty())
        return m_buffer.size();
    if (m_previousText)
        return m_previousLength;
    return m_textIterator.length();
}

const UChar* WordAwareIterator::characters() const
{
    if (!m_buffer.isEmpty())
        return m_buffer.data();
    if (m_previousText)
        return m_previousText;
    return m_textIterator.characters();
}

// --------

#if USE(ICU_UNICODE) && !UCONFIG_NO_COLLATION

static const size_t minimumSearchBufferSize = 8192;

#ifndef NDEBUG
static bool searcherInUse;
#endif

static UStringSearch* createSearcher()
{
    // Provide a non-empty pattern and non-empty text so usearch_open will not fail,
    // but it doesn't matter exactly what it is, since we don't perform any searches
    // without setting both the pattern and the text.

    // Pass empty string for the locale for now to get the Unicode Collation Algorithm,
    // rather than something locale-specific.

    UErrorCode status = U_ZERO_ERROR;
    UStringSearch* searcher = usearch_open(&newlineCharacter, 1, &newlineCharacter, 1, "", 0, &status);
    ASSERT(status == U_ZERO_ERROR);
    return searcher;
}

static UStringSearch* searcher()
{
    static UStringSearch* searcher = createSearcher();
    return searcher;
}

static inline void lockSearcher()
{
#ifndef NDEBUG
    ASSERT(!searcherInUse);
    searcherInUse = true;
#endif
}

static inline void unlockSearcher()
{
#ifndef NDEBUG
    ASSERT(searcherInUse);
    searcherInUse = false;
#endif
}

inline SearchBuffer::SearchBuffer(const String& target, bool isCaseSensitive)
    : m_target(target)
    , m_atBreak(true)
{
    ASSERT(!m_target.isEmpty());

    size_t targetLength = target.length();
    m_buffer.reserveInitialCapacity(max(targetLength * 8, minimumSearchBufferSize));
    m_overlap = m_buffer.capacity() / 4;

    // Grab the single global searcher.
    // If we ever have a reason to do more than once search buffer at once, we'll have
    // to move to multiple searchers.
    lockSearcher();

    UStringSearch* searcher = WebCore::searcher();
    UCollator* collator = usearch_getCollator(searcher);

    UCollationStrength strength = isCaseSensitive ? UCOL_TERTIARY : UCOL_PRIMARY;
    if (ucol_getStrength(collator) != strength) {
        ucol_setStrength(collator, strength);
        usearch_reset(searcher);
    }

    UErrorCode status = U_ZERO_ERROR;
    usearch_setPattern(searcher, m_target.characters(), targetLength, &status);
    ASSERT(status == U_ZERO_ERROR);
}

inline SearchBuffer::~SearchBuffer()
{
    unlockSearcher();
}

inline size_t SearchBuffer::append(const UChar* characters, size_t length)
{
    ASSERT(length);

    if (m_atBreak) {
        m_buffer.shrink(0);
        m_atBreak = false;
    } else if (m_buffer.size() == m_buffer.capacity()) {
        memcpy(m_buffer.data(), m_buffer.data() + m_buffer.size() - m_overlap, m_overlap * sizeof(UChar));
        m_buffer.shrink(m_overlap);