[jdom-interest] JDOM vs electric

[Dennis Sosnoski]

Hi Graham,

In the latest set of tests I've added some collections of smaller 
documents. EXML does very well on small documents (especially the SOAP 
collection, not too surprisingly!). It still performs fairly poorly on 
larger documents, especially when the elimination of whitespace is taken 
into account.

As we'd discussed previously, the differences in results look to be due 
to several factors:

1. Your test uses a single very small document, of the type that EXML 
appears optimized for
2. You test only against JDOM running Xerces, which is kind of shooting 
fish in a barrel when it comes to performance comparisons. :-) If you're 
really interested in seeing how EXML performs you should test against 
some of the other alternatives.
3. EXML discards character data sequences consisting only of whitespace, 
and thereby uses a dramatically lower number of objects in the document 
representations; while this is fine for many applications, it biases the 
tests. It's also not compliant with the XML spec, and should at least be 
pointed out to users in your documentation
4. My tests are looking for the best performance the code is able to 
give, so I take the best time of several runs, while your test uses the 
average time; using average times for benchmarking applications using 
Hot Spot JVMs is questionable, if you're going to do this at all you 
need to run an extended series of test passes and discard all but the 
last few.
5. My tests use memory settings of -Xms64M and -Xmx64M to give the JVM 
plenty of memory; this is appropriate for a server enviroment, and makes 
a big difference in how well Hot Spot handles code. Your test run uses 
the default setting.

  - Dennis

graham glass wrote:

>hi dennis,
>
>we'll run your benchmarks independently to verify them.
>
>the last time we chatted, it seemed like you were getting
>significantly different results from us, which brought
>setup and configuration issues to light.
>
>cheers,
>graham
>
>-----Original Message-----
>From: Dennis Sosnoski [mailto:dms at sosnoski.com]
>Sent: Tuesday, February 12, 2002 11:22 AM
>To: graham glass
>Cc: jdom-interest at jdom.org; Justin Wood
>Subject: Re: [jdom-interest] JDOM vs electric
>
>
>Hi Graham,
>
>When you update the page, I'd appreciate it if you'd add a link to my
>benchmarks as well (either that, or at least take out the line about
>being glad to link to other benchmarks).
>
>For more representative results you should also look at doing a couple
>of things differently in your test case. If EXML+ 4.0 still discards
>whitespace sequences (in violation of the XML spec) you should use a
>test document which does not include any extra whitespace. You should
>also use something other than Xerces 1 as the parser for the JDOM test,
>since Xerces 1 shows very poor performance on small documents (Xerces 2
>is better). Finally, you should include dom4j as a comparison, since
>dom4j is considerably more mature than JDOM and includes a number of
>performance optimizations.
>
>  - Dennis
>
>graham glass wrote:
>
>>hi there,
>>
>>one new thing worth mentioning is that EXML+ 4.0, due out in a couple
>>of weeks, includes native support for DOM. so in addition to the
>>ease-of-use,
>>you now get native DOM compatability.
>>
>>after releasing EXML+ 4.0, we'll rerun the benchmarks using xerces 2.0
>>and the latest version of DOM, and update the JDOM benchmark page.
>>
>>EXML+ 4.0 also includes transparent bidirectional serialization of Java
>>objects to/from XML, Java persistence using XML, XML document pattern
>>matching
>>and many other goodies. it includes the source code for EXML and is free
>>for most commercial uses.
>>
>>cheers,
>>graham
>>
>>http://www.themindelectric.com
>>
>>-----Original Message-----
>>From: jdom-interest-admin at jdom.org
>>[mailto:jdom-interest-admin at jdom.org]On Behalf Of Justin Wood
>>Sent: Tuesday, February 12, 2002 12:17 AM
>>To: jdom-interest at jdom.org
>>Subject: [jdom-interest] JDOM vs electric
>>
>>
>>I had a look in the mail archives from last year regarding this topic.
>>Round about March Jason and Graham Glass (from the Mind Electric) did a bit
>>of mudslinging.
>>
>>http://www.servlets.com/archive/servlet/ReadMsg?msgId=7279&listName=jdom-in
>>
>t
>
>>erest
>>
>>Graham said he would amend his page:
>>
>>http://www.themindelectric.com/products/xml/jdom.html
>>
>>Which he hasn't done.   I'm sure that JDOM has come quite a way in the last
>>year and this page is even more out of date.   Has anyone done a benchmark?
>>Are there big differences, apart from the constuction of the XML object?
>>
>Has
>
>>anyone used both and done a pro's and cons?
>>
>>Regards
>>Justin
>>
>>_______________________________________________
>>To control your jdom-interest membership:
>>http://lists.denveronline.net/mailman/options/jdom-interest/youraddr@yourho
>>
>s
>
>>t.com
>>
>>
>>_______________________________________________
>>To control your jdom-interest membership:
>>http://lists.denveronline.net/mailman/options/jdom-interest/youraddr@yourho
>>
>st.com
>
>
>
>
>