This needs just a minor variation on Muenchian grouping. Define a key on
service/@id, and run through all the service elements, outputting the
element if (a) it is the first one with that key value (standard
Muenchian test) and (b) count(key(...)) != count(/*/branch).

Michael Kay

> -----Original Message-----
> From: owner-xsl-list@xxxxxxxxxxxxxxxxxxxxxx 
> [mailto:owner-xsl-list@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of 
> Adam van den Hoven
> Sent: 25 September 2003 17:26
> To: XSL Mailing List
> Subject:  Problems selecting the right nodes. 
> 
> 
> I have a nodeset or arbitrary length that contains a set of 
> tags which 
> look something like:
> 
> <branch id="br1">
>    <service id="service1" />
>    <service id="service2" />
>    <service id="service3" />
> </branch>
> <branch id="br2">
>     <service id="service1" />
>     <service id="service4" />
> </branch>
> 
> Now what I want is a nodeset that contains (uniquely) all the 
> services 
> that do not occur in all the branches. In this case I want 
> service 2, 3 
> and 4 but not 1
> 
> My original solution was iffy. I ended up creating a tokenized string 
> ("Service2|Service3|Service4" in this case) which I then ran 
> through a 
> recursive named template which did a lot of substring-before and 
> substring-after. Since what I cared about are the ids, that was fine. 
> However, the input file is occasionally very large so this can take a 
> very long time. I want to find a better solution.
> 
> If possible, I think I'd like to avoid using extension 
> functions, unless 
> there is an obvious performance boost.
> 
> Any thoughts?
> Adam
> 
> 
>  XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
> 


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list