RE: n squared behavior with XT for 'preceding' tests?

> hi all, we have what is probably a pretty common problem: we 
> have a series
> of elements in our input which may repeat a heading, or may 
> be different
> than the preceding instance, ...
> 
> now, this works fine (with XT) for small documents, but what 
> we find is
> that, as documents get larger, this template works more and 
> more slowly, in
> proportion to how far we've progressed through the document. 


Yes. It's actually quite tricky to optimise XSL so that an axis qualified by
a position() predicate is only evaluated until that position is reached -
there are quite a lot of special cases to be considered.

I personally think this grouping requirement is sufficiently common that it
deserves special support in the language. SAXON has a group-by attribute on
xsl:for-each which does the job quite nicely, I can't claim that it's "The
Right Way To Do It" but it's a good workaround until the standard comes up
with something.

Mike Kay


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