XSLT with Xalan vs. STX with Joost

Question

Where can I find performance metrics (memory/time) for a non-trivial example of using XSLT (with Xalan) compared to using STX (with Joost)

Answer 1

Probably there is no universal set of benchmarks. For XSLT there is (was?) XSLTMark, but this is for comparing the XSLT engines.

There is one page with the comparison of the same transformation written in different transformation languages.

Probably the best option is to model your problem, generate test data and measure the things you are interested in.

Answer 2

I agree in that real answers are best obtained by writing your own benchmark.

For what it's worth, my recollection is that many developers had high hopes for STX to be much faster than XSLT processors; but found the actual performance of implementations to fall short on expectations. Part of the reason may be that XSLT processor implementations are ridiculously well optimized by now, and thus can handle simple transformations very efficiently, all things considered. As such, STX implementations would also need to spend time honing implementation to same degree, to produce significant speed improvements for common transformations.

Answer 3

You really should use your own benchmark to cover the things you use.

But here's one data point, ( http://www.kindle-maps.com/blog/some-performance-information-on-joost-stx.html ), a 1.3GB XML file (from OpenStreetMap data), 1,800,000ish nodes were processed with a simple STX template in 3 minutes on a low end laptop.