Re: [xml] Simple Element+Attribute Lookup with XPath - performance ?

[Martin Trappel <0xCDCDCDCD gmx at>]

-------- Original Message --------
Subject: Re: [xml] Simple Element+Attribute Lookup with XPath - 
performance ?
Date: Fri, 18 Jul 2008 08:31:52 +0200
From: 0xCDCDCDCD gmx at
To: xml gnome org
References: <487D89C5 80306 gmx at> <20080717083257 GA16927 flap>

Volker Grabsch wrote:
> On Wed, Jul 16, 2008 at 07:40:21AM +0200, Martin Trappel wrote:
> > <ROOT>
> >   <FOO Name="a">
> >     <BAR Id="2"/>
> >     <BAR Id="1"/>
> >     <BAR Id="4"/>
> >     ...
> >     <BAR Id="2345"/>
> >   </FOO>
> >   <FOO Name="b">
> >     ...
> >   </FOO>
> >   ...
> > </ROOT>
> [...]
> > Use the xpath search expression 'BAR[ id="4"]' and use the returned nodeset.
>
> While I don't know why this expression is slow, you might want to try
> the "id()" or keyref mechanisms. They are very fast because of an
> additional index structure. However, this only works properly if your
> @id values are unique in the whole document, i.e. among all <FOO>
> elements.
Unfortunately they are not.


> You could also try to use this XPath expression:
>
>     *[ id="4"]
>
> It might be faster if all sub elements of <FOO> are <BAR>s.
Will have to try that, thanks!


I should have explicitly mentioned in my first mail that interestingly
the xmlXPathEvalExpression("BAR", xpc); (in example b) is
lightning-fast, but apparently given a predicate it gets really slow.
Does anyone know what extra kind of work the XPath processor needs to do
for the simple predicate [ id="4"] other than look it up in the
attributes of the BAR elements?

cheers,
Martin