It seems that we do not need to heap-allocate Attribute objects. Reference counting each one seems like overkill. There must be a way to have the Vector of attributes contain the attributes themselves rather than just heap pointers. These are not variable-size objects.
But Attribute objects are exposed to JavaScript and can be removed from / added back to NamedNodeMap. How are we going to do that if they were a part of NamedNodeMap?
(In reply to comment #1) > But Attribute objects are exposed to JavaScript and can be removed from / added back to NamedNodeMap. How are we going to do that if they were a part of NamedNodeMap? Attr objects are exposed to JavaScript. Attribute objects are internal to WebKit and not exposed to JavaScript.
(In reply to comment #2) > (In reply to comment #1) > > But Attribute objects are exposed to JavaScript and can be removed from / added back to NamedNodeMap. How are we going to do that if they were a part of NamedNodeMap? > > Attr objects are exposed to JavaScript. Attribute objects are internal to WebKit and not exposed to JavaScript. Supporting Attr may still pose some challenges, but I am pretty sure that we don’t need to start doing all that heap allocating until people actually use Attr.
This has been done in http://trac.webkit.org/changeset/114870 by anttik.
By kling! *** This bug has been marked as a duplicate of bug 83440 ***
Mass moving XML DOM bugs to the "DOM" Component.