RE: [xml] XPath / LibXML question

["Buchcik, Kasimier" <k buchcik 4commerce de>]

Hi, 

> -----Original Message-----
> From: xml-bounces gnome org [mailto:xml-bounces gnome org] On 
> Behalf Of Buchcik, Kasimier
> > -----Original Message-----
> > From: Alan_McCarthy scee net [mailto:Alan_McCarthy scee net] 
> > Sent: Thursday, June 01, 2006 5:18 PM
> > To: Buchcik, Kasimier
> > Cc: Aron Stansvik; Buchcik, Kasimier; xml gnome org
> > Subject: RE: [xml] XPath / LibXML question
> >
> > Well, that certainly returns just the one result correctly...
> >
> > I take it from your comments about it not being optimized 
> that it will
> > still find all results before returning the first, rather than just
> > terminating at the first match?
>
> Although it will return just one node, it will still try to apply the
> predicate "[ att2='2' and parent::instances/parent::item[ att1='1']]"
> on all "instance" elements it finds in the document. So 
> actually it will
> run even slower than your original expression. I'll try to optimize
> this.
>
> But the expression I proposed (even if optimized) has the 
> drawback that
> it won't be as efficient as a plain chain of node-tests in the child
> axis,
> if you *don't* know if the document will be of the expected structure.
> Hmm, don't know if I expressed this understandable enough; I 
> think this
> will become clearer if we assume that the given document has 
> a document
> element with the name "foo"; here my expression will still search the
> whole tree for "instance" elements, while your expression would stop
> already at the document element. So it all depends on how much you can
> assume about the queried tree.

[...] 

Apparently I haven't thought the issue to the end. Of course there's
also a way to restrict this to be an element inside /base/group:

"/base/group/descendant::instance[ att2='2' and
  parent::instances/parent::item[ att1='1']][1]"

Regards,

Kasimier