RE: [xsl] memory usage of xslt processing

Subject: RE: [xsl] memory usage of xslt processing
From: "Michael Kay" <mike@xxxxxxxxxxxx>
Date: Wed, 19 Apr 2006 14:44:59 +0100
> I was thinking if a tool that could analysis the XPath 
> expressions from an XSLT document and could create a kind of 
> graph nodes with the scope of the expressions could have a sens ?

The approach has often been suggested and there have been research projects
implementing such things, but I don't believe it's done routinely in any of
the widely-available XSLT processors. (I don't know about the expensive
high-end ones). One of the reasons may be that very few real stylesheets
actually meet all the required constraints; another might be that it's an
all-or-nothing optimization - one tiny change to the stylesheet and the
whole execution strategy changes.

Really you need a different processing model for serial transformation. If
you can only visit each node once then you need to be able to remember
things as you go - you need a stateful, order-sensitive language. That's why
STX was created to address this need. The only problem is, that STX
processors haven't received enough investment to make them compete
effectively even in cases where they should be able to do so.

Michael Kay
http://www.saxonica.com/ 

Current Thread