What ever happened to Aspect Oriented Programming? [closed]

Question

I remember that in the late 1990s and early 2000s Aspect Oriented Programming (AOP) was supposed to be the "Next Big Thing". Nowadays I see some AOP still around, but it seems to have faded into the background.

Answer 1

There has maybe been a lot of hype in the early 2000's, and what happened is the following: there has been a lot of attempts to create aspect-oriented frameworks, and these attempts have merged into two significant projects in the Java sphere: AspectJ and Spring AOP. AspectJ is complete, complex, academic, somewhat overengineered. Spring AOP covers 80% of use cases with 20% of complexity.

If you look at Google Trends for terms "AspectJ, Spring AOP", then compare to the popularity of Java itself, you will see that the relative popularity of AspectJ is somewhat constant, but Spring AOP is raising. That means that people use AOP, but don't want the complexity of AspectJ. I think that the creators of AspectJ made a lot of tactical mistakes; AspectJ has always been a research project and has not been designed "for the masses".

In the .NET sphere, we have seen a similar interest for AOP in the early 2000's. In 2003, when I started AOP researches, there were half a dozen of AOP weavers for .NET; all were following the path of AspectJ, and all were in infant stage. None of these projects survived. Based on this analysis, I built PostSharp, who was designed to cover 80% of use cases with 20% of complexity, and yet was much more convenient to use than Spring AOP. PostSharp is now considered the leading aspect weaver for .NET. PostSharp 2.0 builds on 5 years of feedback and from industry experience of AspectJ and brings "enterprise-ready" AOP (future will judge if this claim is deserved). Besides PostSharp, other significant players are Spring Framework for .NET and Windsor Castle, two DI-centric application frameworks providing 'also' aspects (aspects are considered as dependencies injected into constructed objects). Since these technologies use runtime weaving, they have severe technical limitations, so practically they can be used only in service objects (that's what these application frameworks have been designed for). Another starting project in .NET is LinFu, which can do 'also' aspects.

To be short, the AOP landscape has undergone some consolidation last years, and probably enters the phase of productivity: customers will use it because it really saves money, not because it is cool. Even if it is very cool :).

Oh, and I forgot: most application servers have built-in support for AOP. Thinking of JBoss, WebSphere, and, to some extent, WCF.