RE: [xsl] Mike Kay's tip 8 how to use XSLT efficiently

Subject: RE: [xsl] Mike Kay's tip 8 how to use XSLT efficiently
From: "Michael Kay" <mike@xxxxxxxxxxxx>
Date: Tue, 19 Apr 2005 09:12:03 +0100
These remarks were made a long time ago, and they were pretty off-the-cuff,
if I remember. I would still strongly advocate implementing complex
transformations as a pipeline (a sequence of simple transformations), but
more for software engineering reasons (modularity, maintainability,
reusability) than for raw performance. However, that's not contradictory:
you can't improve performance unless you can measure it and analyze it, and
that's far more likely to be feasible if you adopt a clean pipeline
architecture. A pipeline also gives you more opportunities to reconfigure
the processing, e.g. to run different parts on different machines, cache
intermediate results, do some parts eagerly and other parts lazily, etc.

Michael Kay

> -----Original Message-----
> From: Peter Gerstbach [mailto:peter@xxxxxxxxxxxx] 
> Sent: 19 April 2005 08:11
> To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> Subject: [xsl] Mike Kay's tip 8 how to use XSLT efficiently
> Hi,
> I've just read the eight tips how to use XSLT efficiently in the FAQs
> (
> But I have a question about tip number eight:
> "Split complex transformations into several stages."
> What does this mean? Is this a suggestion to apply multiple 
> transformations
> sequentielly in a pipe? What is the reason for doing this? 
> Does some small
> stylesheets perform better than one big one?
> Thanks,
> Peter

Current Thread