Monthly Archives: March 2011

JEE 6 patterns – the server side java symposium

Posted on by
1

As of Websphere 8, all major app servers support J EE 6 except one that I hadn’t heard of. Adam Bien got a number of jabs in at Rod Johnson’s comment that NetBeans isn’t prevalent.

Many patterns are gone from earlier versions of JEE including this list. Amazing how much content was discussed on one slide.

  • most interfaces – Adam mentioned another valid use of interfaces – reuse. Also noted the IXXX or XXXImpl are signs that the interface isn’t needed.
  • DAOs – should be able to state the reason if have. To avoid duplication have a service with a business name. for example OrderSearchService
  • DTOs – duplication across layers. And if you use a mapping framework like bean utils, you aren’t really decoupling. It is useful to do create a DTO when you don’t have control over the underlying object. Similar reasoning to interfaces. A good metric is whether every layer is affected for a feature request
  • ServiceLocator, Factory – use injection
  • Business Delegate – just delegates to ejb, for handling exceptions
  • Session Facade – needed to wrap CMP because didn’t work well
  • Value Object Assemblers – can just create aggregate object
  • Service Activator – asynchronous support now built in with annotation or CDI events if want ti decouple
  • Publish Subscribe with JMS – for local delivery, use CDI events instead
  • Mandatory XML configuration – Repeating the whole package name in xml so hard to refactor. Adam sounds like he favors minimal XML. I asked what belongs in XML and the answer was passwords/ips. [I don’t think he meant passwords]. He did say anything that could change often.

Other Notes

  • Keep the war as small as possible to speed up deployment. try to keep infrastructure jars out of war manage versions of jars by deploying to separate tomcat if have to use third party jars. Adam doesn’t use third party jars unless absolutely necessary
  • go for convention over configuration
  • ECB – entity (jpa) control (optional cdi bean) boundary (stateless entry point decorated by aspects), way to model a component
  • see anti if campaign website for the lack of OO in JEE

I missed a bit of this presentation even though it was really good because I was getting ready for mine.

Extreme performance in database – the server side java symposium

Posted on by
1

I wasn’t sure whether to blog about this session. Scott feels i haven’t written enough here at the Server Side Java Symposium so i decided to go for it :). Unfortunately, he didn’t say anything I could write without doing a free press release. So here’s a really short summary.

The one slide is an ad. Globalsdb.org has a free database that will be launched March 24th. Every month there will be a contest to prove what can be done on the platform.

One phrase jumped out: “what can you do with the right incentive”. And that they recognize press and money are both incentives.

Live from TSSJS – Tips for Hibernate with Patrycja

Posted on by
1

This morning I’m live blogging from TheServerSide Java Symposium 2011 at the breakout “Anti-Patterns and Best Practices for Hibernate” presented by Patrycja Wegrzynowicz.

Making Good Software
Patrycja’s primary goal is to “understand what makes good software good and bad software bad” and she is working to formalize these concepts.  She would like to develop automated code tools that can detect ahead of time whether or not code contains anti-patterns and potential performance issues.

For those not familiar with the concept, an anti-pattern is an obvious or common strategy which leads to degraded or counter-intuitive performance.

Is Java Fat
Patrycja asks the question “Is Java fat/slow?”.  She claims the language is not, in fact, slow, but that some of the ways in which we program objects is slow, which I agree with strongly.  For example, I have seen computational entensive code that improved in memory/speed by large degrees by relying on primitive variables, rather than more complex objects, in part because “new” is considered expensive in Java.

Automated Code Analysis
Requires large knowledge base.  Patrycja has been researching semantic code query system which focusses on behavior, not just structure.

Conclusion
Patrycja says that proper usage of hibernate is tricky.  Transactional issues and correctness of code under various scenarios.  Best solution may be combination of Hibernate and JDBC.

I liked that fact that she stepped back and indicate Hibernate, by itself, isn’t perfect.  I agree with her take that the best systems are often hybrid systems that take advantage of the ‘low hanging fruit’ of both worlds, and avoid the pitfalls of each.