Re: [xml] Does libxml2 create nodes on-the-fly?

[Andreas Pakulat <apaku gmx de>]

On 26.06.06 05:17:37, Daniel Veillard wrote:
> On Mon, Jun 26, 2006 at 10:39:05AM +0200, Andreas Pakulat wrote:
> > On 24.06.06 09:12:47, Daniel Veillard wrote:
> > > On Sat, Jun 24, 2006 at 12:26:47AM +0200, Andreas Pakulat wrote:
> > > > (though I'm not familiar with libxml2's C Interface, nor the
> > > > Python-C-Bridge-Stuff in general)...
> > > 
> > >   Should be fairly simple, make a C python binding routine based on
> > > existing ones in python/libxml.c , give it a significant name,
> > > change the beginning to expect 2 python obect references as arguments,
> > > see type.h to see how the C object is attached, get the two xmlNodePtr
> > > associated. Compare them, and return an python 0 or 1, then register
> > > the new routine at the end of libxml.c
> > >   now in libxml.py override the comparator operator to call that routine
> > > in libxml2mod, add a python test case in test/ and bingo you're done :-)
> > 
> > Hmm, this thread looks interesting now :-)
> > 
> > Thanks to William Brack I could create a proper patch that enables
> > comparison of Python objects derived from xmlCore (i.e. xmlNode and all
> > subclasses). The bugreport is 345779 and I hope it'll make it into some
> > not-so-distant future version of libxml2.
> 
>   See it wasn't that hard :-)

Yeah, but for the most part it was copy'n'paste anyway...

> William will probably commit it later if he thinks it's ready for final
> inclusion, 

Then I'd better change the test-patch to include 2 more tests now..

>   thanks, I hope doing those extra steps will help you use libxml2 in the
> long term !

Well, I think this change introduced another "problem", now I can't use
the objects as index into a python dictionary. I guess this is because I
only defined the __eq__ and __ne__ operators, but the hashing probably
needs something like greater than or less than... I'll look into that
later.

Andreas

-- 
A vivid and creative mind characterizes you.