[XML-SIG] DOM code now in CVS tree

[Greg Stein]

Andrew wrote:
>         * Attributes in the IDL DOM definition are implemented as
> methods, prefixed with get_.  For example, to retrieve the childNodes
> attribute, you call .get_childNodes() .
>
> 	This is not an optimal solution.  Ideally it would be as
> simple as accessing .childNodes, but this faces some problems.  First,
> the low-level tree of elements is implemented as a tree of _nodeData
> instances, with no circular references.  Element, Text, etc. instances
> are then implemented as proxies for the corresponding _nodeData
> instance.  This provides two useful features:
>
> 	   1) If there are several proxies for a given _nodeData,
> 	      changes are instantly visible to all of the proxies.
> 	   2) As Ken McLeod suggested, this should avoid circular
> 	      references, and, therefore, memory leaks.  (I haven't
> 	      actually verified that there are no leaks, though.)
>
> 	However, this also means that just setting up static
> childNodes attributes would break property 1.  So you'd have to write
> __getattr__/__setattr__ functions.  I'm worried that this would exact
> too great a performance penalty.  Thoughts?

In both cases, you're executing a method, so I don't see that there would be
much of a perf difference. Go for __getattr__.

>         * You can add extended interfaces and still remain compliant
> with the Recommendation.  Therefore, feel free to suggest other useful
> things.  I've added a few convenient aliases: the DOM spec specifies
> get_nodeName() and get_nodeValue(), so I've added get_name() and
> get_value().  We can add other alises to fix other annoyances.

You could always create subclasses of the DOM code for the Python-extended
versions. Have a "pure" version and one with the additional Python
interfaces. This goes along with the getattr from above.

Cheers,
-g

--
Greg Stein (gstein@lyra.org)