I tried a couple of archive searches to see your code, but didn't find anything - if you have a link to the particular email with the code I'd be interested to see what you'd done. I did turn off validation for the performance tests I described in the article, and used a parser that's considerably faster than the 1.4 standard (especially for smaller documents).
As for contributing to Castor, I did consider trying to make changes to Castor rather than doing a separate development to test my performance approach. The problem is that the types of things I wanted to try out couldn't really be done with incremental changes to the code. My approach would also take Castor away from its planned direction (the use of a pull parser that's central to my implementation is not compatible with JAXB, for instance). Even the binding file format would be an issue here; I initially wanted to use something closer to the Castor mapping file format, but found I couldn't easily express the relationships I wanted to convey.
With a separate open source project the Castor developers can take what they like from the code and incorporate it into Castor, or look at what I've done and find better ways to do the same thing within the existing Castor framework. I do have a lot of respect for Castor and continue to recommend it to developers in my articles and talks. Castor's mapped binding implementation was also the inspiration for much of what I'm doing in JiBX.
- Dennis
Dean Hiller wrote:
Dennis,----------------------------------------------------------- If you wish to unsubscribe from this mailing, send mail to
I actually ran perfAnal on castor and *improved their performance by 5 times for small documents* if I remember correctly. That was sufficient for our server environment. My change was never submitted as it was a hack. All I did was put in the 1.4 parser with "no configuration." Somehow, the configuration was killing them. I did post code a while ago showing what I did. I would suggest joining the castor team and improving their performance rather than going it your own. But that choice is up to you. Great thing about open source is if you don't like their performance, change their code and submit back to help their performance.
Dean
Dennis Sosnoski wrote:
Part 2 of my article comparing XML data binding frameworks for Java is now online at http://www-106.ibm.com/developerworks/library/x-databdopt2/ This part covers performance issues (Part 1 covers usage and features). Castor has some weaknesses in this area, but hopefully these can be addressed as it moves toward a 1.0 release.
I've also got some initial code and documentation available online for my JiBX XML data binding project at http://www.jibx.org This is taking an approach very similar to Castor's mapped binding from the user standpoint, although implemented very differently in order to get high performance. Most of the techniques I'm using probably won't port easily to the Castor framework (part of the reason why I'm not trying to use Castor as a basis). At least it can suggest some areas for improvement, anyway.
- Dennis
----------------------------------------------------------- If you wish to unsubscribe from this mailing, send mail to
[EMAIL PROTECTED] with a subject of:
unsubscribe castor-dev
[EMAIL PROTECTED] with a subject of:
unsubscribe castor-dev
