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

                    if (inViewSourceMode())
                        processDoctypeToken();
                } else if (isWhitespace) {
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else
                    m_doctypeToken.setState(DoctypeBogus);
                break;
            }
            case DoctypePublicID: {
                if ((c == '\"' && tquote == DoubleQuote) || (c == '\'' && tquote == SingleQuote)) {
                    src.advancePastNonNewline();
                    m_doctypeToken.setState(DoctypeAfterPublicID);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else if (c == '>') {
                     // Considered bogus.  Don't process the doctype.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                    if (inViewSourceMode())
                        processDoctypeToken();
                } else {
                    m_doctypeToken.m_publicID.append(c);
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                }
                break;
            }
            case DoctypeAfterPublicID:
                if (c == '\"' || c == '\'') {
                    tquote = c == '\"' ? DoubleQuote : SingleQuote;
                    m_doctypeToken.setState(DoctypeSystemID);
                    src.advancePastNonNewline();
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else if (c == '>') {
                    // Valid doctype. Emit it now.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                    processDoctypeToken();
                } else if (isWhitespace) {
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else
                    m_doctypeToken.setState(DoctypeBogus);
                break;
            case DoctypeBeforeSystemID:
                if (c == '\"' || c == '\'') {
                    tquote = c == '\"' ? DoubleQuote : SingleQuote;
                    m_doctypeToken.setState(DoctypeSystemID);
                    src.advancePastNonNewline();
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else if (c == '>') {
                    // Considered bogus.  Don't process the doctype.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                } else if (isWhitespace) {
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else
                    m_doctypeToken.setState(DoctypeBogus);
                break;
            case DoctypeSystemID:
                if ((c == '\"' && tquote == DoubleQuote) || (c == '\'' && tquote == SingleQuote)) {
                    src.advancePastNonNewline();
                    m_doctypeToken.setState(DoctypeAfterSystemID);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else if (c == '>') {
                     // Considered bogus.  Don't process the doctype.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                    if (inViewSourceMode())
                        processDoctypeToken();
                } else {
                    m_doctypeToken.m_systemID.append(c);
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                }
                break;
            case DoctypeAfterSystemID:
                if (c == '>') {
                    // Valid doctype. Emit it now.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                    processDoctypeToken();
                } else if (isWhitespace) {
                    src.advance(m_lineNumber);
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                } else
                    m_doctypeToken.setState(DoctypeBogus);
                break;
            case DoctypeBogus:
                if (c == '>') {
                    // Done with the bogus doctype.
                    src.advancePastNonNewline();
                    state.setInDoctype(false);
                    if (inViewSourceMode())
                       processDoctypeToken();
                } else {
                    src.advance(m_lineNumber); // Just keep scanning for '>'
                    if (inViewSourceMode())
                        m_doctypeToken.m_source.append(c);
                }
                break;
            default:
                break;
        }
    }
    return state;
}

HTMLTokenizer::State HTMLTokenizer::parseTag(SegmentedString& src, State state)
{
    ASSERT(!state.hasEntityState());

    unsigned cBufferPos = m_cBufferPos;

    bool lastIsSlash = false;

    while (!src.isEmpty()) {
        checkBuffer();
        switch(state.tagState()) {
        case NoTag:
        {
            m_cBufferPos = cBufferPos;
            return state;
        }
        case TagName:
        {
            if (searchCount > 0) {
                if (*src == commentStart[searchCount]) {
                    searchCount++;
                    if (searchCount == 2)
                        m_doctypeSearchCount++; // A '!' is also part of a doctype, so we are moving through that still as well.
                    else
                        m_doctypeSearchCount = 0;
                    if (searchCount == 4) {
                        // Found '<!--' sequence
                        src.advancePastNonNewline();
                        m_dest = m_buffer; // ignore the previous part of this tag
                        state.setInComment(true);
                        state.setTagState(NoTag);

                        // Fix bug 34302 at kde.bugs.org.  Go ahead and treat
                        // <!--> as a valid comment, since both mozilla and IE on windows
                        // can handle this case.  Only do this in quirks mode. -dwh
                        if (!src.isEmpty() && *src == '>' && m_doc->inCompatMode()) {
                            state.setInComment(false);
                            src.advancePastNonNewline();
                            if (!src.isEmpty())
                                m_cBuffer[cBufferPos++] = *src;
                        } else
                          state = parseComment(src, state);

                        m_cBufferPos = cBufferPos;
                        return state; // Finished parsing tag!
                    }
                    m_cBuffer[cBufferPos++] = *src;
                    src.advancePastNonNewline();
                    break;
                } else
                    searchCount = 0; // Stop looking for '<!--' sequence
            }
            
            if (m_doctypeSearchCount > 0) {
                if (toASCIILower(*src) == doctypeStart[m_doctypeSearchCount]) {
                    m_doctypeSearchCount++;
                    m_cBuffer[cBufferPos++] = *src;
                    src.advancePastNonNewline();
                    if (m_doctypeSearchCount == 9) {
                        // Found '<!DOCTYPE' sequence
                        state.setInDoctype(true);
                        state.setTagState(NoTag);
                        m_doctypeToken.reset();
                        if (inViewSourceMode())
                            m_doctypeToken.m_source.append(m_cBuffer, cBufferPos);
                        state = parseDoctype(src, state);
                        m_cBufferPos = cBufferPos;
                        return state;
                    }
                    break;
                } else
                    m_doctypeSearchCount = 0; // Stop looking for '<!DOCTYPE' sequence
            }

            bool finish = false;
            unsigned int ll = min(src.length(), CBUFLEN - cBufferPos);
            while (ll--) {
                UChar curchar = *src;
                if (isASCIISpace(curchar) || curchar == '>' || curchar == '<') {
                    finish = true;
                    break;
                }
                
                // tolower() shows up on profiles. This is faster!
                if (curchar >= 'A' && curchar <= 'Z' && !inViewSourceMode())
                    m_cBuffer[cBufferPos++] = curchar + ('a' - 'A');
                else
                    m_cBuffer[cBufferPos++] = curchar;
                src.advancePastNonNewline();
            }

            // Disadvantage: we add the possible rest of the tag
            // as attribute names. ### judge if this causes problems
            if (finish || CBUFLEN == cBufferPos) {
                bool beginTag;
                UChar* ptr = m_cBuffer;
                unsigned int len = cBufferPos;
                m_cBuffer[cBufferPos] = '\0';
                if ((cBufferPos > 0) && (*ptr == '/')) {
                    // End Tag
                    beginTag = false;
                    ptr++;
                    len--;
                }
                else
                    // Start Tag
                    beginTag = true;

                // Ignore the / in fake xml tags like <br/>.  We trim off the "/" so that we'll get "br" as the tag name and not "br/".
                if (len > 1 && ptr[len-1] == '/' && !inViewSourceMode())
                    ptr[--len] = '\0';

                // Now that we've shaved off any invalid / that might have followed the name), make the tag.
                // FIXME: FireFox and WinIE turn !foo nodes into comments, we ignore comments. (fast/parser/tag-with-exclamation-point.html)
                if (ptr[0] != '!' || inViewSourceMode()) {
                    m_currentToken.tagName = AtomicString(ptr);
                    m_currentToken.beginTag = beginTag;
                }
                m_dest = m_buffer;
                state.setTagState(SearchAttribute);
                cBufferPos = 0;
            }
            break;
        }
        case SearchAttribute:
            while(!src.isEmpty()) {
                UChar curchar = *src;
                // In this mode just ignore any quotes we encounter and treat them like spaces.
                if (!isASCIISpace(curchar) && curchar != '\'' && curchar != '"') {
                    if (curchar == '<' || curchar == '>')
                        state.setTagState(SearchEnd);
                    else
                        state.setTagState(AttributeName);

                    cBufferPos = 0;
                    break;
                }
                if (inViewSourceMode())
                    m_currentToken.addViewSourceChar(curchar);
                src.advance(m_lineNumber);
            }
            break;
        case AttributeName:
        {
            int ll = min(src.length(), CBUFLEN - cBufferPos);
            while (ll--) {
                UChar curchar = *src;
                // If we encounter a "/" when scanning an attribute name, treat it as a delimiter.  This allows the 
                // cases like <input type=checkbox checked/> to work (and accommodates XML-style syntax as per HTML5).
                if (curchar <= '>' && (curchar >= '<' || isASCIISpace(curchar) || curchar == '/')) {
                    m_cBuffer[cBufferPos] = '\0';
                    m_attrName = AtomicString(m_cBuffer);
                    m_dest = m_buffer;
                    *m_dest++ = 0;
                    state.setTagState(SearchEqual);
                    if (inViewSourceMode())
                        m_currentToken.addViewSourceChar('a');
                    break;
                }
                
                // tolower() shows up on profiles. This is faster!
                if (curchar >= 'A' && curchar <= 'Z' && !inViewSourceMode())
                    m_cBuffer[cBufferPos++] = curchar + ('a' - 'A');
                else
                    m_cBuffer[cBufferPos++] = curchar;
                    
                src.advance(m_lineNumber);
            }
            if (cBufferPos == CBUFLEN) {
                m_cBuffer[cBufferPos] = '\0';
                m_attrName = AtomicString(m_cBuffer);
                m_dest = m_buffer;
                *m_dest++ = 0;
                state.setTagState(SearchEqual);
                if (inViewSourceMode())
                    m_currentToken.addViewSourceChar('a');
            }
            break;
        }
        case SearchEqual:
            while (!src.isEmpty()) {
                UChar curchar = *src;

                if (lastIsSlash && curchar == '>') {
                    // This is a quirk (with a long sad history).  We have to do this
                    // since widgets do <script src="foo.js"/> and expect the tag to close.
                    if (m_currentToken.tagName == scriptTag)
                        m_currentToken.selfClosingTag = true;
                    m_currentToken.brokenXMLStyle = true;
                }

                // In this mode just ignore any quotes or slashes we encounter and treat them like spaces.
                if (!isASCIISpace(curchar) && curchar != '\'' && curchar != '"' && curchar != '/') {
                    if (curchar == '=') {
                        state.setTagState(SearchValue);
                        if (inViewSourceMode())
                            m_currentToken.addViewSourceChar(curchar);
                        src.advancePastNonNewline();
                    } else {
                        m_currentToken.addAttribute(m_attrName, emptyAtom, inViewSourceMode());
                        m_dest = m_buffer;
                        state.setTagState(SearchAttribute);
                        lastIsSlash = false;
                    }
                    break;
                }
                if (inViewSourceMode())
                    m_currentToken.addViewSourceChar(curchar);