RE: NamedNodeMap

Whoops; not for *each* Node. Only one hash table per tree would be necessary.


> -----Original Message-----
> From:	Allen, Michael B (RSCH) []
> Sent:	Wednesday, July 11, 2001 4:45 PM
> To:	'Fred L. Drake, Jr.'
> Cc:
> Subject:	RE: NamedNodeMap
> Ok, I'm glad I'm not the only one that is confused by this. The only way to implement this functionality verbatium that I can see would be to have an active NodeLists table for each Node. If a Node
> is
> selected for inclusion in a NodeList a pointer to that list is saved in a list of NodeLists that reference this Node. Then, at any time, if that Node is modified, replaced, removed, etc the
> appropriate action is performed on all active NodeLists for that Node. The table could be a hashtable member of Document. Whenever a node is freed the table entry is removed (after all it's
> NodeLists
> are updated of course). I beleive it could be done in an efficient manner actually.
> I think I'll get the basics working first.
> Thanks for your help,
> Mike
> > -----Original Message-----
> > From:	Fred L. Drake, Jr. []
> > 
> > Allen, Michael B (RSCH) writes:
> >  > However, I have a new question. Imagine a NodeList is returned by
> >  > getElementsByTagName. Then a Node in the DOM tree is modified
> >  > (e.g. replaced, removed, etc). Will these changes be reflected in
> >  > the NodeList returned by getElementsByTagName? If so this requires
> > 
> >   This is the only interpretation I've ever heard, and I've heard
> > little but griping about it.  For most purposes, a fairly static
> > NodeList is sufficient; for NodeList objects used to implement the
> > childNodes attribute, there's not much of a problem, but the
> > getElementsByTagName() and getElementsByTagNameNS() return values are
> > problematic.  My own DOM implementation is non-conforming in this
> > regard, and I haven't heard any complaints from the users.
> >   I don't think the liveness requirement is described very carefully;
> > if I have a node that I remove from the tree, and then add to another
> > location, should the node be included in any NodeList objects returned
> > from getElementsByTagName()?  It's still "owned" by the document, and
> > I may well stick it back in the tree somewhere.  I'm sure this is
> > covered somewhere, but this kind of thing ends up being a sticking
> > point for me.
> > 
> > 
> >   -Fred
> > 
> > -- 
> > Fred L. Drake, Jr.  <fdrake at>
> > PythonLabs at Digital Creations
> > 

Received on Wednesday, 11 July 2001 16:48:49 UTC