RE: [xsl] [xslt performance for big xml files]

["Michael Kay" <mike@xxxxxxxxxxxx>]

> > I have found Saxon to be at least 10 times faster than 
> MSXML and other 
> > competitors.
> 
> Just out of curiosity:
> 
> Are your comparisons using the same XSLT code and between an 
> XSLT 1.0 version of Saxon (which version?) and MSXML (which 
> version?), or are you comparing MSXML to a version of Saxon 
> for XSLT 2?
> 
> In the latter case you are probably not using the same code 
> in the comparison (as MSXML does not support XSLT 2) and are 
> actually comparing apples to oranges.
> 
> In the former case, could you, please, provide the exact code 
> on which the two XSLT processors were compared and the 
> versions of each processor?
> 
> 
> --
> Cheers,
> Dimitre Novatchev

Dimitre is of course quite right to point out that just because one
stylesheet runs 10 times faster under processor A compared with processor B,
it doesn't mean that every stylesheet will do so.

As regards the comparison between Saxon6 (supporting XSLT 1.0) and Saxon9
supporting 2.0, I've seen some stylesheets run 30 times the speed under
Saxon9, but for most 1.0 stylesheets, the performance of both releases is
probably fairly similar.

The conditions under which you run the code are also significant: do you
include stylesheet compilation time, document building time, serialization
time? For some use cases it's even legitimate to measure warm-up time. There
is no "correct" way to measure performance, but the metric you choose will
affect the result.

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