Re: [docbook] profiling document with link problem

[Daniel Veillard <veillard@redhat.com>]

On Tue, Jul 29, 2003 at 03:09:32PM +0200, Jirka Kosek wrote:
> Patrick Eisenacher wrote:
> 
> > Since I didn't specify any profiling options yet, the complete document 
> > should be generated to my understanding. The document is generated 
> > indeed, but the link could not be resolved and points to the document's 
> > beginning. The link is not related to any of the profiled contents.
> 
> The problem is in fact, that key() function which is used to lookup ID 
> values works over nodes from source XML document, but when you do 
> profiling the stylesheet works with temporary nodeset which was 
> converted from temporary result tree fragment and this nodeset is not 
> covered by index created for key() function.
> 
> I thought that this should work as the key() function will lookup IDs 
> for links from the full document.
> 
> As a workaround you can:
> 
> 1. Do 2-stage profiling using temporary document with profiled content
> 2. Use Saxon which works well in this situation
> 
> Maybe DV can fix this issue inside xsltproc, but question is whether 
> Saxon or xsltproc is correct in this particular case. As node-set() 
> function is not defined in XSLT standard it will be hard to judge 
> conformance if there is no standard to conform to :-(

 I don't understand fully the issue. Especially I don't understand how
the result of a node-set() could ever be the target for a key(), this
function is defined only for input documents, nothing else as far as I
can tell. And with respect to IDness I don't see either how this could
work. 
 If you can isolate a complete example independant of the DocBook
stylesheets and put it on GNOME bugzilla, I will have a look at it
    http://xmlsoft.org/XSLT/bugs.html

   thanks,

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard@redhat.com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/