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

                lastIsSlash = curchar == '/';

                src.advance(m_lineNumber);
            }
            break;
        case SearchValue:
            while (!src.isEmpty()) {
                UChar curchar = *src;
                if (!isASCIISpace(curchar)) {
                    if (curchar == '\'' || curchar == '\"') {
                        tquote = curchar == '\"' ? DoubleQuote : SingleQuote;
                        state.setTagState(QuotedValue);
                        if (inViewSourceMode())
                            m_currentToken.addViewSourceChar(curchar);
                        src.advancePastNonNewline();
                    } else
                        state.setTagState(Value);

                    break;
                }
                if (inViewSourceMode())
                    m_currentToken.addViewSourceChar(curchar);
                src.advance(m_lineNumber);
            }
            break;
        case QuotedValue:
            while (!src.isEmpty()) {
                checkBuffer();

                UChar curchar = *src;
                if (curchar <= '>' && !src.escaped()) {
                    if (curchar == '>' && m_attrName.isEmpty()) {
                        // Handle a case like <img '>.  Just go ahead and be willing
                        // to close the whole tag.  Don't consume the character and
                        // just go back into SearchEnd while ignoring the whole
                        // value.
                        // FIXME: Note that this is actually not a very good solution.
                        // It doesn't handle the general case of
                        // unmatched quotes among attributes that have names. -dwh
                        while (m_dest > m_buffer + 1 && (m_dest[-1] == '\n' || m_dest[-1] == '\r'))
                            m_dest--; // remove trailing newlines
                        AtomicString attributeValue(m_buffer + 1, m_dest - m_buffer - 1);
                        if (!attributeValue.contains('/'))
                            m_attrName = attributeValue; // Just make the name/value match. (FIXME: Is this some WinIE quirk?)
                        m_currentToken.addAttribute(m_attrName, attributeValue, inViewSourceMode());
                        if (inViewSourceMode())
                            m_currentToken.addViewSourceChar('x');
                        state.setTagState(SearchAttribute);
                        m_dest = m_buffer;
                        tquote = NoQuote;
                        break;
                    }
                    
                    if (curchar == '&') {
                        src.advancePastNonNewline();
                        state = parseEntity(src, m_dest, state, cBufferPos, true, true);
                        break;
                    }

                    if ((tquote == SingleQuote && curchar == '\'') || (tquote == DoubleQuote && curchar == '\"')) {
                        // some <input type=hidden> rely on trailing spaces. argh
                        while (m_dest > m_buffer + 1 && (m_dest[-1] == '\n' || m_dest[-1] == '\r'))
                            m_dest--; // remove trailing newlines
                        AtomicString attributeValue(m_buffer + 1, m_dest - m_buffer - 1);
                        if (m_attrName.isEmpty() && !attributeValue.contains('/')) {
                            m_attrName = attributeValue; // Make the name match the value. (FIXME: Is this a WinIE quirk?)
                            if (inViewSourceMode())
                                m_currentToken.addViewSourceChar('x');
                        } else if (inViewSourceMode())
                            m_currentToken.addViewSourceChar('v');
                        m_currentToken.addAttribute(m_attrName, attributeValue, inViewSourceMode());
                        m_dest = m_buffer;
                        state.setTagState(SearchAttribute);
                        tquote = NoQuote;
                        if (inViewSourceMode())
                            m_currentToken.addViewSourceChar(curchar);
                        src.advancePastNonNewline();
                        break;
                    }
                }

                *m_dest++ = curchar;
                src.advance(m_lineNumber);
            }
            break;
        case Value:
            while(!src.isEmpty()) {
                checkBuffer();
                UChar curchar = *src;
                if (curchar <= '>' && !src.escaped()) {
                    // parse Entities
                    if (curchar == '&') {
                        src.advancePastNonNewline();
                        state = parseEntity(src, m_dest, state, cBufferPos, true, true);
                        break;
                    }
                    // no quotes. Every space means end of value
                    // '/' does not delimit in IE!
                    if (isASCIISpace(curchar) || curchar == '>') {
                        AtomicString attributeValue(m_buffer + 1, m_dest - m_buffer - 1);
                        m_currentToken.addAttribute(m_attrName, attributeValue, inViewSourceMode());
                        if (inViewSourceMode())
                            m_currentToken.addViewSourceChar('v');
                        m_dest = m_buffer;
                        state.setTagState(SearchAttribute);
                        break;
                    }
                }

                *m_dest++ = curchar;
                src.advance(m_lineNumber);
            }
            break;
        case SearchEnd:
        {
            while (!src.isEmpty()) {
                UChar ch = *src;
                if (ch == '>' || ch == '<')
                    break;
                if (ch == '/')
                    m_currentToken.selfClosingTag = true;
                if (inViewSourceMode())
                    m_currentToken.addViewSourceChar(ch);
                src.advance(m_lineNumber);
            }
            if (src.isEmpty())
                break;

            searchCount = 0; // Stop looking for '<!--' sequence
            state.setTagState(NoTag);
            tquote = NoQuote;

            if (*src != '<')
                src.advance(m_lineNumber);

            if (m_currentToken.tagName == nullAtom) { //stop if tag is unknown
                m_cBufferPos = cBufferPos;
                return state;
            }

            AtomicString tagName = m_currentToken.tagName;

            // Handle <script src="foo"/> like Mozilla/Opera. We have to do this now for Dashboard
            // compatibility.
            bool isSelfClosingScript = m_currentToken.selfClosingTag && m_currentToken.beginTag && m_currentToken.tagName == scriptTag;
            bool beginTag = !m_currentToken.selfClosingTag && m_currentToken.beginTag;
            if (m_currentToken.beginTag && m_currentToken.tagName == scriptTag && !inViewSourceMode() && !m_parser->skipMode()) {
                Attribute* a = 0;
                m_scriptTagSrcAttrValue = String();
                m_scriptTagCharsetAttrValue = String();
                if (m_currentToken.attrs && !m_fragment) {
                    if (m_doc->frame() && m_doc->frame()->script()->isEnabled()) {
                        if ((a = m_currentToken.attrs->getAttributeItem(srcAttr)))
                            m_scriptTagSrcAttrValue = m_doc->completeURL(parseURL(a->value())).string();
                    }
                }
            }

            RefPtr<Node> n = processToken();
            m_cBufferPos = cBufferPos;
            if (n || inViewSourceMode()) {
                if ((tagName == preTag || tagName == listingTag) && !inViewSourceMode()) {
                    if (beginTag)
                        state.setDiscardLF(true); // Discard the first LF after we open a pre.
                } else if (tagName == scriptTag) {
                    ASSERT(!m_scriptNode);
                    m_scriptNode = static_pointer_cast<HTMLScriptElement>(n);
                    if (m_scriptNode)
                        m_scriptTagCharsetAttrValue = m_scriptNode->scriptCharset();
                    if (beginTag) {
                        m_searchStopper = scriptEnd;
                        m_searchStopperLength = 8;
                        state.setInScript(true);
                        state = parseSpecial(src, state);
                    } else if (isSelfClosingScript) { // Handle <script src="foo"/>
                        state.setInScript(true);
                        state = scriptHandler(state);
                    }
                } else if (tagName == styleTag) {
                    if (beginTag) {
                        m_searchStopper = styleEnd;
                        m_searchStopperLength = 7;
                        state.setInStyle(true);
                        state = parseSpecial(src, state);
                    }
                } else if (tagName == textareaTag) {
                    if (beginTag) {
                        m_searchStopper = textareaEnd;
                        m_searchStopperLength = 10;
                        state.setInTextArea(true);
                        state = parseSpecial(src, state);
                    }
                } else if (tagName == titleTag) {
                    if (beginTag) {
                        m_searchStopper = titleEnd;
                        m_searchStopperLength = 7;
                        State savedState = state;
                        SegmentedString savedSrc = src;
                        long savedLineno = m_lineNumber;
                        state.setInTitle(true);
                        state = parseSpecial(src, state);
                        if (state.inTitle() && src.isEmpty()) {
                            // We just ate the rest of the document as the title #text node!
                            // Reset the state then retokenize without special title handling.
                            // Let the parser clean up the missing </title> tag.
                            // FIXME: This is incorrect, because src.isEmpty() doesn't mean we're
                            // at the end of the document unless m_noMoreData is also true. We need
                            // to detect this case elsewhere, and save the state somewhere other
                            // than a local variable.
                            state = savedState;
                            src = savedSrc;
                            m_lineNumber = savedLineno;
                            m_scriptCodeSize = 0;
                        }
                    }
                } else if (tagName == xmpTag) {
                    if (beginTag) {
                        m_searchStopper = xmpEnd;
                        m_searchStopperLength = 5;
                        state.setInXmp(true);
                        state = parseSpecial(src, state);
                    }
                } else if (tagName == iframeTag) {
                    if (beginTag) {
                        m_searchStopper = iframeEnd;
                        m_searchStopperLength = 8;
                        state.setInIFrame(true);
                        state = parseSpecial(src, state);
                    }
                }
            }
            if (tagName == plaintextTag)
                state.setInPlainText(beginTag);
            return state; // Finished parsing tag!
        }
        } // end switch
    }
    m_cBufferPos = cBufferPos;
    return state;
}

inline bool HTMLTokenizer::continueProcessing(int& processedCount, double startTime, State &state)
{
    // We don't want to be checking elapsed time with every character, so we only check after we've
    // processed a certain number of characters.
    bool allowedYield = state.allowYield();
    state.setAllowYield(false);
    if (!state.loadingExtScript() && !state.forceSynchronous() && !m_executingScript && (processedCount > m_tokenizerChunkSize || allowedYield)) {
        processedCount = 0;
        if (currentTime() - startTime > m_tokenizerTimeDelay) {
            /* FIXME: We'd like to yield aggressively to give stylesheets the opportunity to
               load, but this hurts overall performance on slower machines.  For now turn this
               off.
            || (!m_doc->haveStylesheetsLoaded() && 
                (m_doc->documentElement()->id() != ID_HTML || m_doc->body()))) {*/
            // Schedule the timer to keep processing as soon as possible.
            m_timer.startOneShot(0);
#ifdef INSTRUMENT_LAYOUT_SCHEDULING
            if (currentTime() - startTime > m_tokenizerTimeDelay)
                printf("Deferring processing of data because 500ms elapsed away from event loop.\n");
#endif
            return false;
        }
    }
    
    processedCount++;
    return true;
}

void HTMLTokenizer::write(const SegmentedString& str, bool appendData)
{
    if (!m_buffer)
        return;
    
    if (m_parserStopped)
        return;

    SegmentedString source(str);
    if (m_executingScript)
        source.setExcludeLineNumbers();

    if ((m_executingScript && appendData) || !m_pendingScripts.isEmpty()) {
        // don't parse; we will do this later
        if (m_currentPrependingSrc)
            m_currentPrependingSrc->append(source);
        else {
            m_pendingSrc.append(source);
#if PRELOAD_SCANNER_ENABLED
            if (m_preloadScanner && m_preloadScanner->inProgress() && appendData)
                m_preloadScanner->write(source);
#endif
        }
        return;
    }
    
#if PRELOAD_SCANNER_ENABLED
    if (m_preloadScanner && m_preloadScanner->inProgress() && appendData)
        m_preloadScanner->end();
#endif

    if (!m_src.isEmpty())
        m_src.append(source);
    else
        setSrc(source);

    // Once a timer is set, it has control of when the tokenizer continues.
    if (m_timer.isActive())
        return;

    bool wasInWrite = m_inWrite;
    m_inWrite = true;
    
#ifdef INSTRUMENT_LAYOUT_SCHEDULING
    if (!m_doc->ownerElement())
        printf("Beginning write at time %d\n", m_doc->elapsedTime());
#endif
    
    int processedCount = 0;
    double startTime = currentTime();

    Frame* frame = m_doc->frame();

    State state = m_state;

    while (!m_src.isEmpty() && (!frame || !frame->loader()->isScheduledLocationChangePending())) {