htmltokenizer.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

    while (!src.isEmpty()) {
        checkScriptBuffer();
        UChar ch = *src;

        if (!m_scriptCodeResync && !m_brokenComments &&
            !state.inXmp() && ch == '-' && m_scriptCodeSize >= 3 && !src.escaped() &&
            m_scriptCode[m_scriptCodeSize - 3] == '<' && m_scriptCode[m_scriptCodeSize - 2] == '!' && m_scriptCode[m_scriptCodeSize - 1] == '-' &&
            (lastDecodedEntityPosition < m_scriptCodeSize - 3)) {
            state.setInComment(true);
            state = parseComment(src, state);
            continue;
        }
        if (m_scriptCodeResync && !tquote && ch == '>') {
            src.advancePastNonNewline();
            m_scriptCodeSize = m_scriptCodeResync - 1;
            m_scriptCodeResync = 0;
            m_scriptCode[m_scriptCodeSize] = m_scriptCode[m_scriptCodeSize + 1] = 0;
            if (state.inScript())
                state = scriptHandler(state);
            else {
                state = processListing(SegmentedString(m_scriptCode, m_scriptCodeSize), state);
                processToken();
                if (state.inStyle()) { 
                    m_currentToken.tagName = styleTag.localName();
                    m_currentToken.beginTag = false;
                } else if (state.inTextArea()) { 
                    m_currentToken.tagName = textareaTag.localName();
                    m_currentToken.beginTag = false;
                } else if (state.inTitle()) { 
                    m_currentToken.tagName = titleTag.localName();
                    m_currentToken.beginTag = false;
                } else if (state.inXmp()) {
                    m_currentToken.tagName = xmpTag.localName();
                    m_currentToken.beginTag = false;
                } else if (state.inIFrame()) {
                    m_currentToken.tagName = iframeTag.localName();
                    m_currentToken.beginTag = false;
                }
                processToken();
                state.setInStyle(false);
                state.setInScript(false);
                state.setInTextArea(false);
                state.setInTitle(false);
                state.setInXmp(false);
                state.setInIFrame(false);
                tquote = NoQuote;
                m_scriptCodeSize = m_scriptCodeResync = 0;
            }
            return state;
        }
        // possible end of tagname, lets check.
        if (!m_scriptCodeResync && !state.escaped() && !src.escaped() && (ch == '>' || ch == '/' || isASCIISpace(ch)) &&
             m_scriptCodeSize >= m_searchStopperLength &&
             tagMatch(m_searchStopper, m_scriptCode + m_scriptCodeSize - m_searchStopperLength, m_searchStopperLength) &&
             (lastDecodedEntityPosition < m_scriptCodeSize - m_searchStopperLength)) {
            m_scriptCodeResync = m_scriptCodeSize-m_searchStopperLength+1;
            tquote = NoQuote;
            continue;
        }
        if (m_scriptCodeResync && !state.escaped()) {
            if (ch == '\"')
                tquote = (tquote == NoQuote) ? DoubleQuote : ((tquote == SingleQuote) ? SingleQuote : NoQuote);
            else if (ch == '\'')
                tquote = (tquote == NoQuote) ? SingleQuote : (tquote == DoubleQuote) ? DoubleQuote : NoQuote;
            else if (tquote != NoQuote && (ch == '\r' || ch == '\n'))
                tquote = NoQuote;
        }
        state.setEscaped(!state.escaped() && ch == '\\');
        if (!m_scriptCodeResync && (state.inTextArea() || state.inTitle() || state.inIFrame()) && !src.escaped() && ch == '&') {
            UChar* scriptCodeDest = m_scriptCode + m_scriptCodeSize;
            src.advancePastNonNewline();
            state = parseEntity(src, scriptCodeDest, state, m_cBufferPos, true, false);
            if (scriptCodeDest == m_scriptCode + m_scriptCodeSize)
                lastDecodedEntityPosition = m_scriptCodeSize;
            else
                m_scriptCodeSize = scriptCodeDest - m_scriptCode;
        } else {
            m_scriptCode[m_scriptCodeSize++] = ch;
            src.advance(m_lineNumber);
        }
    }

    return state;
}

HTMLTokenizer::State HTMLTokenizer::scriptHandler(State state)
{
    // We are inside a <script>
    bool doScriptExec = false;
    int startLine = m_currentScriptTagStartLineNumber + 1; // Script line numbers are 1 based, HTMLTokenzier line numbers are 0 based

    // Reset m_currentScriptTagStartLineNumber to indicate that we've finished parsing the current script element
    m_currentScriptTagStartLineNumber = 0;

    // (Bugzilla 3837) Scripts following a frameset element should not execute or, 
    // in the case of extern scripts, even load.
    bool followingFrameset = (m_doc->body() && m_doc->body()->hasTagName(framesetTag));
  
    CachedScript* cs = 0;
    // don't load external scripts for standalone documents (for now)
    if (!inViewSourceMode()) {
        if (!m_scriptTagSrcAttrValue.isEmpty() && m_doc->frame()) {
            // forget what we just got; load from src url instead
            if (!m_parser->skipMode() && !followingFrameset) {
#ifdef INSTRUMENT_LAYOUT_SCHEDULING
                if (!m_doc->ownerElement())
                    printf("Requesting script at time %d\n", m_doc->elapsedTime());
#endif
                // The parser might have been stopped by for example a window.close call in an earlier script.
                // If so, we don't want to load scripts.
                if (!m_parserStopped && (cs = m_doc->docLoader()->requestScript(m_scriptTagSrcAttrValue, m_scriptTagCharsetAttrValue)))
                    m_pendingScripts.append(cs);
                else
                    m_scriptNode = 0;
            } else
                m_scriptNode = 0;
            m_scriptTagSrcAttrValue = String();
        } else {
            // Parse m_scriptCode containing <script> info
            doScriptExec = m_scriptNode->shouldExecuteAsJavaScript();
            m_scriptNode = 0;
        }
    }

    state = processListing(SegmentedString(m_scriptCode, m_scriptCodeSize), state);
    RefPtr<Node> node = processToken();
    String scriptString = node ? node->textContent() : "";
    m_currentToken.tagName = scriptTag.localName();
    m_currentToken.beginTag = false;
    processToken();

    state.setInScript(false);
    m_scriptCodeSize = m_scriptCodeResync = 0;
    
    // FIXME: The script should be syntax highlighted.
    if (inViewSourceMode())
        return state;

    SegmentedString* savedPrependingSrc = m_currentPrependingSrc;
    SegmentedString prependingSrc;
    m_currentPrependingSrc = &prependingSrc;

    if (!m_parser->skipMode() && !followingFrameset) {
        if (cs) {
            if (savedPrependingSrc)
                savedPrependingSrc->append(m_src);
            else
                m_pendingSrc.prepend(m_src);
            setSrc(SegmentedString());

            // the ref() call below may call notifyFinished if the script is already in cache,
            // and that mucks with the state directly, so we must write it back to the object.
            m_state = state;
            bool savedRequestingScript = m_requestingScript;
            m_requestingScript = true;
            cs->addClient(this);
            m_requestingScript = savedRequestingScript;
            state = m_state;
            // will be 0 if script was already loaded and ref() executed it
            if (!m_pendingScripts.isEmpty())
                state.setLoadingExtScript(true);
        } else if (!m_fragment && doScriptExec) {
            if (!m_executingScript)
                m_pendingSrc.prepend(m_src);
            else
                prependingSrc = m_src;
            setSrc(SegmentedString());
            state = scriptExecution(ScriptSourceCode(scriptString, m_doc->frame() ? m_doc->frame()->document()->url() : KURL(), startLine), state);
        }
    }

    if (!m_executingScript && !state.loadingExtScript()) {
        m_src.append(m_pendingSrc);
        m_pendingSrc.clear();
    } else if (!prependingSrc.isEmpty()) {
        // restore first so that the write appends in the right place
        // (does not hurt to do it again below)
        m_currentPrependingSrc = savedPrependingSrc;

        // we need to do this slightly modified bit of one of the write() cases
        // because we want to prepend to m_pendingSrc rather than appending
        // if there's no previous prependingSrc
        if (!m_pendingScripts.isEmpty()) {
            if (m_currentPrependingSrc)
                m_currentPrependingSrc->append(prependingSrc);
            else
                m_pendingSrc.prepend(prependingSrc);
        } else {
            m_state = state;
            write(prependingSrc, false);
            state = m_state;
        }
    }
    
#if PRELOAD_SCANNER_ENABLED
    if (!m_pendingScripts.isEmpty() && !m_executingScript) {
        if (!m_preloadScanner)
            m_preloadScanner.set(new PreloadScanner(m_doc));
        if (!m_preloadScanner->inProgress()) {
            m_preloadScanner->begin();
            m_preloadScanner->write(m_pendingSrc);
        }
    }
#endif
    m_currentPrependingSrc = savedPrependingSrc;

    return state;
}

HTMLTokenizer::State HTMLTokenizer::scriptExecution(const ScriptSourceCode& sourceCode, State state)
{
    if (m_fragment || !m_doc->frame())
        return state;
    m_executingScript++;

    SegmentedString* savedPrependingSrc = m_currentPrependingSrc;
    SegmentedString prependingSrc;
    m_currentPrependingSrc = &prependingSrc;

#ifdef INSTRUMENT_LAYOUT_SCHEDULING
    if (!m_doc->ownerElement())
        printf("beginning script execution at %d\n", m_doc->elapsedTime());
#endif

    m_state = state;
    m_doc->frame()->loader()->executeScript(sourceCode);
    state = m_state;

    state.setAllowYield(true);

#ifdef INSTRUMENT_LAYOUT_SCHEDULING
    if (!m_doc->ownerElement())
        printf("ending script execution at %d\n", m_doc->elapsedTime());
#endif
    
    m_executingScript--;

    if (!m_executingScript && !state.loadingExtScript()) {
        m_pendingSrc.prepend(prependingSrc);        
        m_src.append(m_pendingSrc);
        m_pendingSrc.clear();
    } else if (!prependingSrc.isEmpty()) {
        // restore first so that the write appends in the right place
        // (does not hurt to do it again below)
        m_currentPrependingSrc = savedPrependingSrc;

        // we need to do this slightly modified bit of one of the write() cases
        // because we want to prepend to m_pendingSrc rather than appending
        // if there's no previous prependingSrc
        if (!m_pendingScripts.isEmpty()) {
            if (m_currentPrependingSrc)
                m_currentPrependingSrc->append(prependingSrc);
            else
                m_pendingSrc.prepend(prependingSrc);
            
#if PRELOAD_SCANNER_ENABLED
            // We are stuck waiting for another script. Lets check the source that
            // was just document.write()n for anything to load.
            PreloadScanner documentWritePreloadScanner(m_doc);
            documentWritePreloadScanner.begin();
            documentWritePreloadScanner.write(prependingSrc);
            documentWritePreloadScanner.end();
#endif
        } else {
            m_state = state;
            write(prependingSrc, false);
            state = m_state;
        }
    }

    m_currentPrependingSrc = savedPrependingSrc;

    return state;
}

HTMLTokenizer::State HTMLTokenizer::parseComment(SegmentedString& src, State state)
{
    // FIXME: Why does this code even run for comments inside <script> and <style>? This seems bogus.
    checkScriptBuffer(src.length());
    while (!src.isEmpty()) {
        UChar ch = *src;
        m_scriptCode[m_scriptCodeSize++] = ch;
        if (ch == '>') {
            bool handleBrokenComments = m_brokenComments && !(state.inScript() || state.inStyle());
            int endCharsCount = 1; // start off with one for the '>' character
            if (m_scriptCodeSize > 2 && m_scriptCode[m_scriptCodeSize-3] == '-' && m_scriptCode[m_scriptCodeSize-2] == '-') {
                endCharsCount = 3;
            } else if (m_scriptCodeSize > 3 && m_scriptCode[m_scriptCodeSize-4] == '-' && m_scriptCode[m_scriptCodeSize-3] == '-' && 
                m_scriptCode[m_scriptCodeSize-2] == '!') {
                // Other browsers will accept --!> as a close comment, even though it's
                // not technically valid.
                endCharsCount = 4;
            }
            if (handleBrokenComments || endCharsCount > 1) {
                src.advancePastNonNewline();
                if (!(state.inTitle() || state.inScript() || state.inXmp() || state.inTextArea() || state.inStyle() || state.inIFrame())) {
                    checkScriptBuffer();
                    m_scriptCode[m_scriptCodeSize] = 0;
                    m_scriptCode[m_scriptCodeSize + 1] = 0;
                    m_currentToken.tagName = commentAtom;
                    m_currentToken.beginTag = true;
                    state = processListing(SegmentedString(m_scriptCode, m_scriptCodeSize - endCharsCount), state);
                    processToken();
                    m_currentToken.tagName = commentAtom;
                    m_currentToken.beginTag = false;
                    processToken();
                    m_scriptCodeSize = 0;
                }
                state.setInComment(false);
                return state; // Finished parsing comment
            }
        }
        src.advance(m_lineNumber);
    }

    return state;
}

HTMLTokenizer::State HTMLTokenizer::parseServer(SegmentedString& src, State state)
{
    checkScriptBuffer(src.length());
    while (!src.isEmpty()) {
        UChar ch = *src;
        m_scriptCode[m_scriptCodeSize++] = ch;
        if (ch == '>' && m_scriptCodeSize > 1 && m_scriptCode[m_scriptCodeSize - 2] == '%') {
            src.advancePastNonNewline();
            state.setInServer(false);
            m_scriptCodeSize = 0;