[whatwg] Use cases for Node.getElementById (was: Re: Early feedback on header association algorithm)

[Simon Pieters]

On Fri, 05 Dec 2008 19:19:04 +0100, Calogero Alex Baldacchino  
<alex.baldacchino at email.it> wrote:

> That's about the same then moving the getElementById method from the  
> Document interface to the Node interface
>  (Document inherits from Node, so the actual traversed subtree would  
> change basing on the node where the method is invocked, that is  
> 'anElement = document.getElementById("anEl")' would work as always,  
> while anElement.getElementById("anEl") would look for a descendant of  
> 'anElement' with the same id), because, essentially, IDs are a common  
> feature of all document types, despite the actual name of the attribute  
> representing an ID, so an eventual .getElementByIdInSubtree() method  
> should be defined on a somewhat DOM Core interface, and so would be out  
> of scope for HTML 5 (as I've been told .getElementById is - there is a  
> 'Web DOM Core' specification under construction).

(I'm currently the editor of that proposal, currently located at  
http://simon.html5.org/specs/web-dom-core )


> But the term 'Subtree' arises a problem with HTML 5: actually, the id  
> attribute is defined as the element unique ID in the *subtree* whithin  
> which the element is found. That is, the term subtree refers to a whole  
> document tree, but also to a disconnected subtree handled by a script  
> (and I haven't yet understood if such definition refers to a document  
> fragment containing nodes detached by any document, or a whole document  
> without a browsing context).

AIUI, it could also be a disconnected element.


> Perhaps the possible confusion arising if moving .getElementById() to  
> the Node interface might be avoided by leaving it on the document  
> interface, and overloading it with, for instance,
>
> Element   getElementById(in DOMString elementId, in Node rootElement);

You mean that it would only be possible to look for elements that are in  
the document (effectively adding it purely for the case when there are  
duplicate IDs in a document -- which is invalid)?


> so a call to document.getElementById would behave as always (or better,  
> as it will be redefined in Web DOM Core, that should be 'pick the first  
> element with a matching id'), and would coincide with a call to  
> document.getElementById("something", document); while a call to  
> document.getElementById("something", anElement) would search a matching  
> ID among the descendants of 'anElement', whether anElement be a node of  
> the current document, or a node removed by any document or created by a  
> script, or a node in another document and both the current document and  
> the current script context are enabled to access it (but a 'script  
> context' is an HTML 5 related concept, so it might be generalized as a  
> "DOM access context").

Hmm. If you allow it to work on disconnected nodes, why involve the  
'document' at all instead of using Node.getElementById? I think that's  
adding to the confusion. :-)

Anyway, apart from implementing the HTML5 table heading algorithm in  
JavaScript, what are the use cases for this? Given that there's  
Element|Document|DocumentFragment.querySelector('#id') which will return  
the first element with that ID in the given subtree, and that  
Node.getElementById seems to mostly encourage using duplicate IDs, I'm  
reluctant to add it to Web DOM Core. UAs can still have such a method  
internally if they need it.

-- 
Simon Pieters
Opera Software