Le Vendredi 31 Janvier 2003 20:12, Stefan Seefeld a =E9crit :
> I'v the impression that Node/Attribute containers
> could/should be avoided, at least in the DOM programming
> model: Once you get hold of a node, you may iterate over
> the child nodes by means of an iterator, avoiding unnecessary
> copies (of pointers, but still...)
That make sense to me.
>
> The only context in which I can see the need for a 'NodeList'
> is when you search for matching nodes in a document (using
> xpath expressions, say), so the result will be a temporary
> list.
Agree.
>
> The SaxParser uses an 'AttributeMap', containing Attribute pointers.
> Given that an 'Attribute' is now a wrapper around a node (which is
> by design associated with the DOM programming model), I suggest
> we replace that with a simple pair of (unicode) strings. The
> current implementation will leak, as there is no way to delete
> attributes outside a document (again, by design, because Attributes
> live in the context of a document, which doesn't exist with SAX).
>
I agree that the Attribute class is no longer switable for this purpose.
As far as the pair of string, we should keep consistency between this and t=
he=20
way we do for Nodes contents : use templates... and this is true for the=20
other methods too, although I'm not sure it will make things clear/easy (bu=
t=20
this may be because it's a bit late for me :-/
> Suggestions ?
Well, nothing concrete for now, I'll think again about it tomorrow...
Best regards,
Christophe
|
