Raible Designs

[NFJS Denver] Ramnivas Ladded and Aspect Oriented Programming

For the 2nd session of the day, I've decided to attend the session on AOP. My choices were Converting XML (Ben Galbraith), JSF (David Geary), What's new in EJB 2.1 (Monson-Haefel) and Runtime Code Generation (Glenn Vanderburg). I asked David about his JSF presentation and he said it was "boring" so I decided to skip it. This session, in the first 10 minutes, is pretty boring too. I might have to duck out of this and attend another session - but none of the other sessions really interest me. This one started this morning, and it seems to be a continuation of this morning session. It's mostly on AspectJ and he's doing the good ol' logging example. I don't really have an interest in using AspectJ since Spring's AOP is more than adequate for my needs. Oh well, I guess I'll stick around and relax.

Ramnivas is using Eclipse to compile his aspects, and shows us how if he uses "javac" to compile everything, no aspects are added. I'm guessing that there's some sort of plugin for Eclipse that allows for easy compiling of aspects. A new feature in AspectJ 1.2 is the ability to set an ASPECTPATH and then use "aj" rather than "java" to run code. This achieves the same effect as compiling the aspects in Eclipse. The "aj" script appears to just set the system classloader.

Now Ramnivas is showing us how you can use an aspect to define pointcuts on internal appserver classes. His example involves WebLogic, the J2EE Petstore and logging any EJB method calls. To use this, he has to start WebLogic with a custom script, that has a custom classloader defined. For some reason, we're now talking about log.debug() vs. if (log.isDebugEnabled()) { log.debug() }. Ramnivas thinks that most folks don't wrap their debug statements (the proper way). Who doesn't know about this?! I'd like to think that most developers are writing the 3-line version vs. the 1-line version.

An interesting quote: "So many things to show you - 3 hours is not enough." Maybe if we didn't have to wait 2 minutes for WebLogic to start each time, 3 hours would be plenty. ;-) Ramnivas just added logging to all the classes in the blueprints packages and then reloaded the page. There looks to be several thousand method calls occurring behind the scenes for a simple page. Of course, there's lots of tag libraries - but still - PetStore is a ridiculously complex application. Why did some ever impose this crap upon us as "best practices" for Java. IMO, best practices should be based around simplicity, maintainability, testability and performance. Does the PetStore even ship with unit tests? I'll bet it doesn't...

Now we're watching a pretty cool demo where Ramnivas is using aspects to enforce coding policies. In this example, he's using aspects to enforce rules in EJBs (i.e. no static variables, swing or threads). It interesting in that the aspects actually prevent compilation and display errors defined in the aspect. It's a neat idea, but I'm suspicious in that a lot of this AspectJ stuff seems to be Eclipse-specific. What I mean by this is that Eclipse seems to be required for compilation. Is it easy to aspect-enhance your classes using Ant? I would hope so.

Ramnivas seems to be quite the AspectJ expert. He mentioned that he wrote a book for Manning and has mentioned a couple of his contributions to AspectJ in this presentation. I admire authors that write and contribute to projects at the same time. It's often an indicator that the person knows what the hell they're talking about.

This afternoon's first session choices are as follows: JSF Advanced Topics (Geary), Intro to J2EE Web Services (Monson-Haefel), JSR166 (Glenn Vanderburg), Rhythm (Brian Boelsterli) and GUI Development (Ben Galbraith). Web Services and Rhythm are 3 hour presentations, so I might stay away from those - that's just too long for me. Now that I'm sitting in the 2nd half of a 3 hour presentation - it seems that these sessions are best attended in the first half. It's almost as if the presenter is trying to find things to talk about in the second half. Or maybe AOP is just a boring-ass topic. ;-)

I'd like to go to Brian's talk on Rhythm, since he's a good friend and mentor of mine. But then again, I learned Rhythm from him - and use it daily - so I don't know that I'd get anything out of it. Advanced JSF looks good, maybe I'll go to that one. For the 2nd session this afternoon, I'll probably attend Howard's Creating Powerful Web Forms with Tapestry. Maybe I should stick with JSF and Tapestry and make it a web-intensive afternoon. Both presentations are even in the same room!