> A big thanks to James on determining the performance 
> bottleneck.  Here is a recap and resolution.
> 
SAXON will behave in much the same way: position() is far cheaper than
xsl:number. The naive algorithm for numbering n successive nodes using
xsl:number has performance proportional to n squared, and it's not at all
easy to identify the cases that are suitable for optimising.

My vote would be to throw out xsl:number and replace it with a function
position-in() that returns the position of an arbitrary node in an arbitrary
node-set. Not only would this be much easier to implement efficiently, it
would also be more flexible and more consistent with the rest of XSL.

Mike Kay


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