From: Oliver Becker <obecker@in...> - 2003-03-06 20:32:24
> Has anyone compared performance of an STX engine with a XSLT that does
> the same operations and runs on SAXON?
Though it is very true what Paul already said, I can give you the results
I measured today while developing an example for examining RDF data using
Joost as STX engine. Naturally both transformations look different. To my
surprise Joost was only around 1% slower than Saxon on a 1MB file (8.9s vs
8.8s). My experiences from other tests were that Joost is currently
unfortunately about 10% slower than XSLT engines using exactly the same
operations. But surely there are lots of line of code in Joost that can be
optimized. The main focus currently is functionality, not speed.
Now the good news: if I use a 85MB file than I cannot process this input
with Saxon. Joost can and needs 1min 40s
(AMD 500MHz, 128MB RAM, Linux, Java 1.3).
From: Christian Nentwich <c.nentwich@cs...> - 2003-03-06 20:41:14
> Though it is very true what Paul already said, I can give you the results
> I measured today while developing an example for examining RDF data using
> Joost as STX engine. Naturally both transformations look different. To my
I did some benchmarks ages ago (5 months?) and xalan and Joost were
about the same speed on an identity transformation.
My feeling is that the argument for STX should not necessarily be performance.
The other advantages are clear.
There have been some reports on the STX list that the initial
implementations are a bit slower than Xalan. But I'm sure they will =
Potentially, STX promises to be much faster than XSLT, provided that =
transformations are strictly serial in nature.
[mailto:saxon-help-admin@...] On Behalf Of Ytai =
Sent: 06 March 2003 10:10
To: Saxon Help List
Subject: [saxon] STX performance
Has anyone compared performance of an STX engine with a XSLT that does =
same operations and runs on SAXON?
I'm interested in knowing the results.
C T I Squared