Despite all advances in the software field, the most common bug is still the most embarrassing bug that OOP has to offer: NullPointerException.

The main weapon in fighting NPEs is known for quite some time: annotating your types with either @Nullable or @NonNull. It has also been shown how appropriate defaults enable you to omit more than 2/3 of all required annotations without loss of precision. Still, null annotations haven't yet made their way into mainstream development with Java, and so we keep developing new NPEs every day.

This talk will present an initiative to enhance the Eclipse Java Compiler with the capability to perform comprehensive analysis of null problems in your code using null annotations. Current plans state that this support shall be released as part of Juno. The new analysis together with the corresponding new quick assists will allow you to adopt null annotations at any level and gradient: from making a few API contracts explicit and checkable up-to eventually providing guarantees that a fully annotated program will never throw NPE.

Until the Juno release the new functionality is made available as a separate download that has been developed using OT/J as the implementation language. A look behind the scenes will reveal how Object Teams has once again kept the promise of extreme modularity: We need tight integration for extending a complex application (the Eclipse Java Compiler) with a complex new feature (annotation based null analysis). OT/J is a champ for integration. We didn't want to compromise the separation of concerns and again OT/J is our friend: the new feature is a cleanly separated and concise module with all the benefits for development, building and deployment.

Ecore's success stems from its power to describe deep semantic structure more concisely than Java. The downside are the tools. Certainly Ecore's structured editor is simple and effective and its graphical editor is rich and elegant but both are cumbersome compared to traditional text-based tools. The Xtext framework beckons with a solution: a textual syntax for Ecore. Going one step further, we leverage Xbase to define a concise textual notation for describing behavior and exploit it to implement constraints, derived features, operations, and data type conversion. We call this new language Xcore.

This presentation will explore Xcore's capabilities and demonstrate its powerful tools in action. They provide an experience reminiscent of JDT, as you'd expect, given their roots in Xtext. We'll also show how Xcore can be interpreted dynamically and compiled to Java statically, how it can leverage existing models currently represented as Ecore and GenModel, and how it's possible to convert between the combination of the two traditional forms and the new uniform textual representation. You'll leave this talk with an excellent understanding of some amazing new technology that will be available in the Juno release.

Leo Hsu and Regina Obe are PostGIS spatial database consultants and are showing the latest coolest stuff in PostGIS 2.0.