XML parsing performance: follow-up on 08/08 posting by Liviu Tudor

[Vadim Draluk <va...@draluk.net>]

Hi,

I was about to post my question when I discovered the
aforementioned posting. Though circumstances are
somewhat different, looks like it might be the same
problem.

I'm experiencing performance problems on the server
side, deploying a Web service under Tomcat, using
crimson parser. It works reasonably well with small
amounts of data, but some bigger messages (of mere
100K) take about an hour to parse.

In my case it is one very long String parameter that
causes the problem. All the time appears to be spent
in the content() methos of the Parser2 class. I also
suspect (not quite positive yet) that it is actually
mostly wasted in the parsedContent() method of the
InputEntity class.

In content() there's a for-loop at the very beginning.
Its first 1000 iterations take only 50ms, but then
times start growing rapidly, quickly reaching some
astonishing 10sec (!!!) per 1000 iterations.

Now comes the question: am I going to be any better
off by switching to Axis instead of Apache SOAP 2.3 ?
Or should I bite the bullet, and rewrite the code to
pass the large data CLOB as an attachment ?

Cheers

Vadim