Re: [xsl] XALAN-C++ Performance...

Subject: Re: [xsl] XALAN-C++ Performance...
From: "David N Bertoni/Cambridge/IBM" <david_n_bertoni@xxxxxxxxxx>
Date: Thu, 30 May 2002 09:00:48 -0700
> Hi All,
>
> Just want a confirmation regarding XALANC++'s  performance..
>
> 1.  I am making XERCES DOM Tree
> 2.  Using ParserLiasion
> 3.  Using XercesDOMWrapperParsedSource
> 4.  Using Compiled Stylesheet (Compiled only once... in a single run)
> 5. Using Multithreading... Different threads trasnform ...and out put the
> result
>
> Kind of same thing had been achieved using SAXON. in Java.
>
> Now transfomration time differs in Minutes....
> eg:
> 10,000 nodes transformed using SAXON in almost 5 mins.
> 10,000 nodes transformed using XALAN in almost 13 mins..
> (
>
> Step by step execution revealed that XalanTransformer.transform takes
much time.
>
> Any one can suggest idea to optimize or confirm if XALANC++ really is
that much slow.

I hesitate to answer processor-specific questions on this list, but I can't
let this one go by.

The Xalan-C++ documentation is _explicit_ as to the cost of using the
Xerces DOM for transformations:

   http://xml.apache.org/xalan-c/usagepatterns.html#dom

Not only is it going to much slower, but it's going to use a great deal
more memory.  In a multithreaded environment, things will be even slower,
as the Xerces DOM is heavily synchronized.  Use Xalan's own source tree
implementation and get rid of the Xerces DOM.

Comparing Xalan-C++ using a wrapper around another implementation's DOM and
Saxon using its native implementation is hardly a fair comparison.

If you have other questions, please post them to the Xalan-C++ user list.
It's the appropriate place, and you'll get faster and likely better
replies.

Dave



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


Current Thread