markup.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

    // Don't add namespace attribute if it is already defined for this elem.
    const AtomicString& prefix = elem->prefix();
    AtomicString attr = !prefix.isEmpty() ? "xmlns:" + prefix : "xmlns";
    return !elem->hasAttribute(attr);
}

static bool shouldAddNamespaceAttr(const Attribute* attr, HashMap<AtomicStringImpl*, AtomicStringImpl*>& namespaces)
{
    // Don't add namespace attributes twice
    DEFINE_STATIC_LOCAL(const AtomicString, xmlnsURI, ("http://www.w3.org/2000/xmlns/"));
    DEFINE_STATIC_LOCAL(const QualifiedName, xmlnsAttr, (nullAtom, "xmlns", xmlnsURI));
    if (attr->name() == xmlnsAttr) {
        namespaces.set(emptyAtom.impl(), attr->value().impl());
        return false;
    }
    
    QualifiedName xmlnsPrefixAttr("xmlns", attr->localName(), xmlnsURI);
    if (attr->name() == xmlnsPrefixAttr) {
        namespaces.set(attr->localName().impl(), attr->value().impl());
        return false;
    }
    
    return true;
}

static void appendNamespace(Vector<UChar>& result, const AtomicString& prefix, const AtomicString& ns, HashMap<AtomicStringImpl*, AtomicStringImpl*>& namespaces)
{
    if (ns.isEmpty())
        return;
        
    // Use emptyAtoms's impl() for both null and empty strings since the HashMap can't handle 0 as a key
    AtomicStringImpl* pre = prefix.isEmpty() ? emptyAtom.impl() : prefix.impl();
    AtomicStringImpl* foundNS = namespaces.get(pre);
    if (foundNS != ns.impl()) {
        namespaces.set(pre, ns.impl());
        DEFINE_STATIC_LOCAL(const String, xmlns, ("xmlns"));
        result.append(' ');
        append(result, xmlns);
        if (!prefix.isEmpty()) {
            result.append(':');
            append(result, prefix);
        }

        result.append('=');
        result.append('"');
        appendAttributeValue(result, ns, false);
        result.append('"');
    }
}

static void appendDocumentType(Vector<UChar>& result, const DocumentType* n)
{
    if (n->name().isEmpty())
        return;

    append(result, "<!DOCTYPE ");
    append(result, n->name());
    if (!n->publicId().isEmpty()) {
        append(result, " PUBLIC \"");
        append(result, n->publicId());
        append(result, "\"");
        if (!n->systemId().isEmpty()) {
            append(result, " \"");
            append(result, n->systemId());
            append(result, "\"");
        }
    } else if (!n->systemId().isEmpty()) {
        append(result, " SYSTEM \"");
        append(result, n->systemId());
        append(result, "\"");
    }
    if (!n->internalSubset().isEmpty()) {
        append(result, " [");
        append(result, n->internalSubset());
        append(result, "]");
    }
    append(result, ">");
}

static void appendStartMarkup(Vector<UChar>& result, const Node *node, const Range *range, EAnnotateForInterchange annotate, bool convertBlocksToInlines = false, HashMap<AtomicStringImpl*, AtomicStringImpl*>* namespaces = 0)
{
    bool documentIsHTML = node->document()->isHTMLDocument();
    switch (node->nodeType()) {
        case Node::TEXT_NODE: {
            if (Node* parent = node->parentNode()) {
                if (parent->hasTagName(scriptTag)
                    || parent->hasTagName(styleTag)
                    || parent->hasTagName(textareaTag)
                    || parent->hasTagName(xmpTag)) {
                    appendUCharRange(result, ucharRange(node, range));
                    break;
                }
            }
            if (!annotate) {
                appendEscapedContent(result, ucharRange(node, range), documentIsHTML);
                break;
            }
            
            bool useRenderedText = !enclosingNodeWithTag(Position(const_cast<Node*>(node), 0), selectTag);
            String markup = escapeContentText(useRenderedText ? renderedText(node, range) : stringValueForRange(node, range), false);
            if (annotate)
                markup = convertHTMLTextToInterchangeFormat(markup, static_cast<const Text*>(node));
            append(result, markup);
            break;
        }
        case Node::COMMENT_NODE:
            // FIXME: Comment content is not escaped, but XMLSerializer (and possibly other callers) should raise an exception if it includes "-->".
            append(result, "<!--");
            append(result, static_cast<const Comment*>(node)->nodeValue());
            append(result, "-->");
            break;
        case Node::DOCUMENT_NODE:
        case Node::DOCUMENT_FRAGMENT_NODE:
            break;
        case Node::DOCUMENT_TYPE_NODE:
            appendDocumentType(result, static_cast<const DocumentType*>(node));
            break;
        case Node::PROCESSING_INSTRUCTION_NODE: {
            // FIXME: PI data is not escaped, but XMLSerializer (and possibly other callers) this should raise an exception if it includes "?>".
            const ProcessingInstruction* n = static_cast<const ProcessingInstruction*>(node);
            append(result, "<?");
            append(result, n->target());
            append(result, " ");
            append(result, n->data());
            append(result, "?>");
            break;
        }
        case Node::ELEMENT_NODE: {
            result.append('<');
            const Element* el = static_cast<const Element*>(node);
            bool convert = convertBlocksToInlines & isBlock(const_cast<Node*>(node));
            append(result, el->nodeNamePreservingCase());
            NamedAttrMap *attrs = el->attributes();
            unsigned length = attrs->length();
            if (!documentIsHTML && namespaces && shouldAddNamespaceElem(el))
                appendNamespace(result, el->prefix(), el->namespaceURI(), *namespaces);

            for (unsigned int i = 0; i < length; i++) {
                Attribute *attr = attrs->attributeItem(i);
                // We'll handle the style attribute separately, below.
                if (attr->name() == styleAttr && el->isHTMLElement() && (annotate || convert))
                    continue;
                result.append(' ');

                if (documentIsHTML)
                    append(result, attr->name().localName());
                else
                    append(result, attr->name().toString());

                result.append('=');

                if (el->isURLAttribute(attr))
                    appendQuotedURLAttributeValue(result, attr->value());
                else {
                    result.append('\"');
                    appendAttributeValue(result, attr->value(), documentIsHTML);
                    result.append('\"');
                }

                if (!documentIsHTML && namespaces && shouldAddNamespaceAttr(attr, *namespaces))
                    appendNamespace(result, attr->prefix(), attr->namespaceURI(), *namespaces);
            }
            
            if (el->isHTMLElement() && (annotate || convert)) {
                Element* element = const_cast<Element*>(el);
                RefPtr<CSSMutableStyleDeclaration> style = static_cast<HTMLElement*>(element)->getInlineStyleDecl()->copy();
                if (annotate) {
                    RefPtr<CSSMutableStyleDeclaration> styleFromMatchedRules = styleFromMatchedRulesForElement(const_cast<Element*>(el));
                    // Styles from the inline style declaration, held in the variable "style", take precedence 
                    // over those from matched rules.
                    styleFromMatchedRules->merge(style.get());
                    style = styleFromMatchedRules;
                    
                    RefPtr<CSSComputedStyleDeclaration> computedStyleForElement = computedStyle(element);
                    RefPtr<CSSMutableStyleDeclaration> fromComputedStyle = CSSMutableStyleDeclaration::create();
                    
                    {
                        CSSMutableStyleDeclaration::const_iterator end = style->end();
                        for (CSSMutableStyleDeclaration::const_iterator it = style->begin(); it != end; ++it) {
                            const CSSProperty& property = *it;
                            CSSValue* value = property.value();
                            // The property value, if it's a percentage, may not reflect the actual computed value.  
                            // For example: style="height: 1%; overflow: visible;" in quirksmode
                            // FIXME: There are others like this, see <rdar://problem/5195123> Slashdot copy/paste fidelity problem
                            if (value->cssValueType() == CSSValue::CSS_PRIMITIVE_VALUE)
                                if (static_cast<CSSPrimitiveValue*>(value)->primitiveType() == CSSPrimitiveValue::CSS_PERCENTAGE)
                                    if (RefPtr<CSSValue> computedPropertyValue = computedStyleForElement->getPropertyCSSValue(property.id()))
                                        fromComputedStyle->addParsedProperty(CSSProperty(property.id(), computedPropertyValue));
                        }
                    }
                    
                    style->merge(fromComputedStyle.get());
                }
                if (convert)
                    style->setProperty(CSSPropertyDisplay, CSSValueInline, true);
                if (style->length() > 0) {
                    DEFINE_STATIC_LOCAL(const String, stylePrefix, (" style=\""));
                    append(result, stylePrefix);
                    appendAttributeValue(result, style->cssText(), documentIsHTML);
                    result.append('\"');
                }
            }
            
            if (shouldSelfClose(el)) {
                if (el->isHTMLElement())
                    result.append(' '); // XHTML 1.0 <-> HTML compatibility.
                result.append('/');
            }
            result.append('>');
            break;
        }
        case Node::CDATA_SECTION_NODE: {
            // FIXME: CDATA content is not escaped, but XMLSerializer (and possibly other callers) should raise an exception if it includes "]]>".
            const CDATASection* n = static_cast<const CDATASection*>(node);
            append(result, "<![CDATA[");
            append(result, n->data());
            append(result, "]]>");
            break;
        }
        case Node::ATTRIBUTE_NODE:
        case Node::ENTITY_NODE:
        case Node::ENTITY_REFERENCE_NODE:
        case Node::NOTATION_NODE:
        case Node::XPATH_NAMESPACE_NODE:
            ASSERT_NOT_REACHED();
            break;
    }
}

static String getStartMarkup(const Node *node, const Range *range, EAnnotateForInterchange annotate, bool convertBlocksToInlines = false, HashMap<AtomicStringImpl*, AtomicStringImpl*>* namespaces = 0)
{
    Vector<UChar> result;
    appendStartMarkup(result, node, range, annotate, convertBlocksToInlines, namespaces);
    return String::adopt(result);
}

static inline bool doesHTMLForbidEndTag(const Node *node)
{
    if (node->isHTMLElement()) {
        const HTMLElement* htmlElt = static_cast<const HTMLElement*>(node);
        return (htmlElt->endTagRequirement() == TagStatusForbidden);
    }
    return false;
}

// Rules of self-closure
// 1. No elements in HTML documents use the self-closing syntax.
// 2. Elements w/ children never self-close because they use a separate end tag.
// 3. HTML elements which do not have a "forbidden" end tag will close with a separate end tag.
// 4. Other elements self-close.
static inline bool shouldSelfClose(const Node *node)
{
    if (node->document()->isHTMLDocument())
        return false;
    if (node->hasChildNodes())
        return false;
    if (node->isHTMLElement() && !doesHTMLForbidEndTag(node))
        return false;
    return true;
}

static void appendEndMarkup(Vector<UChar>& result, const Node* node)
{
    if (!node->isElementNode() || shouldSelfClose(node) || (!node->hasChildNodes() && doesHTMLForbidEndTag(node)))
        return;

    result.append('<');
    result.append('/');
    append(result, static_cast<const Element*>(node)->nodeNamePreservingCase());
    result.append('>');
}

static String getEndMarkup(const Node *node)
{
    Vector<UChar> result;
    appendEndMarkup(result, node);
    return String::adopt(result);
}

static void appendMarkup(Vector<UChar>& result, Node* startNode, bool onlyIncludeChildren, Vector<Node*>* nodes, const HashMap<AtomicStringImpl*, AtomicStringImpl*>* namespaces = 0)
{
    HashMap<AtomicStringImpl*, AtomicStringImpl*> namespaceHash;
    if (namespaces)
        namespaceHash = *namespaces;
    
    if (!onlyIncludeChildren) {
        if (nodes)
            nodes->append(startNode);
        
        appendStartMarkup(result,startNode, 0, DoNotAnnotateForInterchange, false, &namespaceHash);
    }
    // print children
    if (!(startNode->document()->isHTMLDocument() && doesHTMLForbidEndTag(startNode)))
        for (Node* current = startNode->firstChild(); current; current = current->nextSibling())
            appendMarkup(result, current, false, nodes, &namespaceHash);
    
    // Print my ending tag
    if (!onlyIncludeChildren)
        appendEndMarkup(result, startNode);
}

static void completeURLs(Node* node, const String& baseURL)
{
    Vector<AttributeChange> changes;

    KURL parsedBaseURL(baseURL);

    Node* end = node->traverseNextSibling();
    for (Node* n = node; n != end; n = n->traverseNextNode()) {
        if (n->isElementNode()) {