htmltokenizer.cpp

khtml在gtk上的移植版本

                    len--;
                }
                else
                    // Start Tag
                    beginTag = true;

                // Accept empty xml tags like <br/>.  We trim off the "/" so that when we call
                // getTagID, we'll look up "br" as the tag name and not "br/".
                if(len > 1 && ptr[len-1] == '/' )
                    ptr[--len] = '\0';

                // Look up the tagID for the specified tag name (now that we've shaved off any
                // invalid / that might have followed the name).
                uint tagID = getTagID(ptr, len);
                if (!tagID) {
#ifdef TOKEN_DEBUG
                    QCString tmp(ptr, len+1);
                    kdDebug( 6036 ) << "Unknown tag: \"" << tmp.data() << "\"" << endl;
#endif
                    dest = buffer;
                }
                else
                {
#ifdef TOKEN_DEBUG
                    QCString tmp(ptr, len+1);
                    kdDebug( 6036 ) << "found tag id=" << tagID << ": " << tmp.data() << endl;
#endif
                    currToken.id = beginTag ? tagID : tagID + ID_CLOSE_TAG;
                    dest = buffer;
                }
                tag = SearchAttribute;
                cBufferPos = 0;
            }
            break;
        }
        case SearchAttribute:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
                qDebug("SearchAttribute");
#endif
            bool atespace = false;
            ushort curchar;
            while(!src.isEmpty()) {
                curchar = *src;
                if(curchar > ' ') {
                    if (curchar == '<' || curchar == '>')
                        tag = SearchEnd;
                    else if(atespace && (curchar == '\'' || curchar == '"'))
                    {
                        tag = SearchValue;
                        *dest++ = 0;
                        attrName = QString::null;
                        attrNamePresent = false;
                    }
                    else
                        tag = AttributeName;

                    cBufferPos = 0;
                    break;
                }
                atespace = true;
                ++src;
            }
            break;
        }
        case AttributeName:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
                qDebug("AttributeName");
#endif
            ushort curchar;
            int ll = kMin(src.length(), CBUFLEN-cBufferPos);

            while(ll--) {
                curchar = *src;
                if(curchar <= '>') {
                    if(curchar <= ' ' || curchar == '=' || curchar == '>') {
                        unsigned int a;
                        cBuffer[cBufferPos] = '\0';
                        a = getAttrID(cBuffer, cBufferPos);
                        if (a)
                            attrNamePresent = true;
                        else {
                            attrName = QString::fromLatin1(QCString(cBuffer, cBufferPos+1).data());
                            attrNamePresent = !attrName.isEmpty();

                            // This is a deliberate quirk to match Mozilla and Opera.  We have to do this
                            // since sites that use the "standards-compliant" path sometimes send
                            // <script src="foo.js"/>.  Both Moz and Opera will honor this, despite it
                            // being bogus HTML.  They do not honor the "/" for other tags.  This behavior
                            // also deviates from WinIE, but in this case we'll just copy Moz and Opera.
                            if (currToken.id == ID_SCRIPT && curchar == '>' &&
                                attrName == "/")
                                currToken.flat = true;
                        }
                        
                        dest = buffer;
                        *dest++ = a;
#ifdef TOKEN_DEBUG
                        if (!a || (cBufferPos && *cBuffer == '!'))
                            kdDebug( 6036 ) << "Unknown attribute: *" << QCString(cBuffer, cBufferPos+1).data() << "*" << endl;
                        else
                            kdDebug( 6036 ) << "Known attribute: " << QCString(cBuffer, cBufferPos+1).data() << endl;
#endif

                        tag = SearchEqual;
                        break;
                    }
                }
                // Use tolower() instead of | 0x20 to lowercase the char because there is no 
                // performance gain in using | 0x20 since tolower() is optimized and 
                // | 0x20 turns characters such as '_' into junk.
                cBuffer[cBufferPos++] = tolower(curchar);
                ++src;
            }
            if ( cBufferPos == CBUFLEN ) {
                cBuffer[cBufferPos] = '\0';
                attrName = QString::fromLatin1(QCString(cBuffer, cBufferPos+1).data());
                attrNamePresent = !attrName.isEmpty();
                dest = buffer;
                *dest++ = 0;
                tag = SearchEqual;
            }
            break;
        }
        case SearchEqual:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
                qDebug("SearchEqual");
#endif
            ushort curchar;
            bool atespace = false;
            while(!src.isEmpty()) {
                curchar = src->unicode();
                if(curchar > ' ') {
                    if(curchar == '=') {
#ifdef TOKEN_DEBUG
                        kdDebug(6036) << "found equal" << endl;
#endif
                        tag = SearchValue;
                        ++src;
                    }
                    else if(atespace && (curchar == '\'' || curchar == '"'))
                    {
                        tag = SearchValue;
                        *dest++ = 0;
                        attrName = QString::null;
                        attrNamePresent = false;
                    }
                    else {
                        currToken.addAttribute(parser->docPtr()->document(), buffer, attrName, emptyAtom);
                        dest = buffer;
                        tag = SearchAttribute;
                    }
                    break;
                }
                atespace = true;
                ++src;
            }
            break;
        }
        case SearchValue:
        {
            ushort curchar;
            while(!src.isEmpty()) {
                curchar = src->unicode();
                if(curchar > ' ') {
                    if(( curchar == '\'' || curchar == '\"' )) {
                        tquote = curchar == '\"' ? DoubleQuote : SingleQuote;
                        tag = QuotedValue;
                        ++src;
                    } else
                        tag = Value;

                    break;
                }
                ++src;
            }
            break;
        }
        case QuotedValue:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
                qDebug("QuotedValue");
#endif
            ushort curchar;
            while(!src.isEmpty()) {
                checkBuffer();

                curchar = src->unicode();
                if (curchar == '>' && !attrNamePresent) {
                    // 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's
                    // an interim hack and doesn't handle the general case of
                    // unmatched quotes among attributes that have names. -dwh
                    while(dest > buffer+1 && (*(dest-1) == '\n' || *(dest-1) == '\r'))
                        dest--; // remove trailing newlines
                    AtomicString v(buffer+1, dest-buffer-1);
                    attrName.setUnicode(buffer+1,dest-buffer-1); 
                    currToken.addAttribute(parser->docPtr()->document(), buffer, attrName, v);
                    tag = SearchAttribute;
                    dest = buffer;
                    tquote = NoQuote;
                    break;
                }
                
                if(curchar <= '\'' && !src.escaped()) {
                    // ### attributes like '&{blaa....};' are supposed to be treated as jscript.
                    if ( curchar == '&' )
                    {
                        ++src;
                        parseEntity(src, dest, true);
                        break;
                    }
                    else if ( (tquote == SingleQuote && curchar == '\'') ||
                              (tquote == DoubleQuote && curchar == '\"') )
                    {
                        // some <input type=hidden> rely on trailing spaces. argh
                        while(dest > buffer+1 && (*(dest-1) == '\n' || *(dest-1) == '\r'))
                            dest--; // remove trailing newlines
                        AtomicString v(buffer+1, dest-buffer-1);
                        if (!attrNamePresent)
                            attrName.setUnicode(buffer+1,dest-buffer-1); 
                        currToken.addAttribute(parser->docPtr()->document(), buffer, attrName, v);

                        dest = buffer;
                        tag = SearchAttribute;
                        tquote = NoQuote;
                        ++src;
                        break;
                    }
                }
                *dest = *src;
                fixUpChar(*dest);
                ++dest;
                ++src;
            }
            break;
        }
        case Value:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
            qDebug("Value");
#endif
            ushort curchar;
            while(!src.isEmpty()) {
                checkBuffer();
                curchar = src->unicode();
                if(curchar <= '>' && !src.escaped()) {
                    // parse Entities
                    if ( curchar == '&' )
                    {
                        ++src;
                        parseEntity(src, dest, true);
                        break;
                    }
                    // no quotes. Every space means end of value
                    // '/' does not delimit in IE!
                    if ( curchar <= ' ' || curchar == '>' )
                    {
                        AtomicString v(buffer+1, dest-buffer-1);
                        currToken.addAttribute(parser->docPtr()->document(), buffer, attrName, v);
                        dest = buffer;
                        tag = SearchAttribute;
                        break;
                    }
                }

                *dest = *src;
                fixUpChar(*dest);
                ++dest;
                ++src;
            }
            break;
        }
        case SearchEnd:
        {
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 1
                qDebug("SearchEnd");
#endif
            while(!src.isEmpty()) {
                if (*src == '>' || *src == '<')
                    break;

                if (*src == '/')
                    currToken.flat = true;

                ++src;
            }
            if (src.isEmpty()) break;

            searchCount = 0; // Stop looking for '<!--' sequence
            tag = NoTag;
            tquote = NoQuote;

            if (*src != '<')
                ++src;

            if ( !currToken.id ) //stop if tag is unknown
                return;

            uint tagID = currToken.id;
#if defined(TOKEN_DEBUG) && TOKEN_DEBUG > 0
            kdDebug( 6036 ) << "appending Tag: " << tagID << endl;
#endif
            bool beginTag = !currToken.flat && (tagID < ID_CLOSE_TAG);

            if (tagID >= ID_CLOSE_TAG)
                tagID -= ID_CLOSE_TAG;
            else if (tagID == ID_SCRIPT) {
                AttributeImpl* a = 0;
                scriptSrc = QString::null;
                scriptSrcCharset = QString::null;
                if ( currToken.attrs && /* potentially have a ATTR_SRC ? */
		     parser->doc()->part() &&
                     parser->doc()->part()->jScriptEnabled() && /* jscript allowed at all? */
                     view /* are we a regular tokenizer or just for innerHTML ? */
                    ) {
                    if ( ( a = currToken.attrs->getAttributeItem( ATTR_SRC ) ) )
                        scriptSrc = parser->doc()->completeURL(parseURL( a->value() ).string() );
                    if ( ( a = currToken.attrs->getAttributeItem( ATTR_CHARSET ) ) )
                        scriptSrcCharset = a->value().string().stripWhiteSpace();
                    if ( scriptSrcCharset.isEmpty() )
                        scriptSrcCharset = parser->doc()->part()->encoding();
                    if (!(a = currToken.attrs->getAttributeItem( ATTR_LANGUAGE )))
                        a = currToken.attrs->getAttributeItem(ATTR_TYPE);
                }
                javascript = true;
                if( a ) {
                    QString lang = a->value().string();
                    lang = lang.lower();
                    if( !lang.contains("javascript") &&
                        !lang.contains("ecmascript") &&
                        !lang.contains("livescript") &&
                        !lang.contains("jscript") )
                        javascript = false;
                }
            }

            processToken();

            // we have to take care to close the pre block in
            // case we encounter an unallowed element....
            if(pre && beginTag && !DOM::checkChild(ID_PRE, tagID)) {
                kdDebug(6036) << " not allowed in <pre> " << (int)tagID << endl;
                pre = false;
            }

            switch( tagID ) {
            case ID_PRE:
                prePos = 0;
                pre = beginTag;
                break;
            case ID_SCRIPT:
                if (beginTag) {
                    searchStopper = scriptEnd;
                    searchStopperLen = 8;
                    script = true;
                    parseSpecial(src);
                }
                else if (tagID < ID_CLOSE_TAG) // Handle <script src="foo"/>
                    scriptHandler();
                break;
            case ID_STYLE:
                if (beginTag) {
                    searchStopper = styleEnd;
                    searchStopperLen = 7;
                    style = true;
                    parseSpecial(src);
                }
                break;
            case ID_TEXTAREA:
                if(beginTag) {
                    searchStopper = textareaEnd;
                    searchStopperLen = 10;
                    textarea = true;
                    parseSpecial(src);
                }
                break;
            case ID_TITLE:
                if (beginTag) {
                    searchStopper = titleEnd;
                    searchStopperLen = 7;
                    title = true;
                    parseSpecial(src);
                }
                break;
            case ID_XMP:
                if (beginTag) {
                    searchStopper = xmpEnd;
                    searchStopperLen = 5;
                    xmp = true;
                    parseSpecial(src);
                }
                break;
            case ID_SELECT:
                select = beginTag;
                break;
            case ID_PLAINTEXT:
                plaintext = beginTag;
                break;
            }
            
            if (beginTag && endTag[tagID] == FORBIDDEN)
                // Don't discard LFs since this element has no end tag.
                discard = NoneDiscard;
                
            return; // Finished parsing tag!
        }
        } // end switch
    }
    return;
}

void HTMLTokenizer::addPending()
{
    if ( select && !script )
    {
        *dest++ = ' ';
    }
    else if ( textarea || script )
    {
        switch(pending) {
        case LFPending:  *dest++ = '\n'; prePos = 0; break;
        case SpacePending: *dest++ = ' '; ++prePos; break;
        case TabPending: *dest++ = '\t'; prePos += TAB_SIZE - (prePos % TAB_SIZE); break;
        case NonePending:
            assert(0);
        }
    }
    else
    {
        int p;

        switch (pending)
        {
        case SpacePending:
            // Insert a breaking space
            *dest++ = QChar(' ');
            prePos++;
            break;

        case LFPending:
            *dest = '\n';
            dest++;
            prePos = 0;
            break;

        case TabPending:
            p = TAB_SIZE - ( prePos % TAB_SIZE );
#ifdef TOKEN_DEBUG
            qDebug("tab pending, prePos: %d, toadd: %d", prePos, p);
#endif

            for ( int x = 0; x < p; x++ )
                *dest++ = QChar(' ');
            prePos += p;
            break;

        case NonePending: