Add Element-specific traversal functions

• Component: DOM
• Status: RESOLVED FIXED

[Antti Koivisto]

In many cases where we use NodeTraversal we really only want Elements. By having Element specific iteration functions we can tighten the code and make it faster too.

[Antti Koivisto]

Created attachment 178882 [details]
patch

[Ryosuke Niwa]

Comment on attachment 178882 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=178882&action=review

> Source/WebCore/ChangeLog:1529
> +        Another build fix.
> +
> +        * html/HTMLPropertiesCollection.cpp:
> +        (WebCore::nextNodeWithProperty):

Please revert these changes.

> Source/WebCore/dom/ContainerNodeAlgorithms.h:319
> +    Element* element = (shouldIncludeRoot == IncludeRoot && root->isElementNode()) ? static_cast<Element*>(root) : ElementTraversal::firstWithin(root);

I don't have toElement?

> Source/WebCore/dom/NodeTraversal.h:102
> +template <class NodeType>
> +inline Element* traverseNextElementTemplate(NodeType* current, const Node* stayWithin)
> +{
> +    Node* node = NodeTraversal::next(current, stayWithin);
> +    while (node && !node->isElementNode())
> +        node = NodeTraversal::nextSkippingChildren(node, stayWithin);
> +    return static_cast<Element*>(node);
> +}

I'm sad that we have to duplicate the template like this. Can we had a boolean template argument and share the code between two?

> Source/WebCore/dom/NodeTraversal.h:126
> +template <class NodeType>
> +inline Element* traverseNextElementSkippingChildrenTemplate(NodeType* current, const Node* stayWithin)
> +{
> +    Node* node = NodeTraversal::nextSkippingChildren(current, stayWithin);
> +    while (node && !node->isElementNode())
> +        node = NodeTraversal::nextSkippingChildren(node, stayWithin);
> +    return static_cast<Element*>(node);
> +}

Ditto.

[WebKit Review Bot]

Comment on attachment 178882 [details]
patch

Attachment 178882 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/15277319

[Antti Koivisto]

Created attachment 178892 [details]
patch2

[Antti Koivisto]

(In reply to comment #2)
> I'm sad that we have to duplicate the template like this. Can we had a boolean template argument and share the code between two?

I didn't do this as the templates get pretty messy and hard to understand if we also want optimal code.

[Antti Koivisto]

http://trac.webkit.org/changeset/137406