A Model-driven Pointcut Language for More Robust Pointcuts

Improved modularity and separation of concerns do not only intend to aid initial development, but are conceived such that developers can better manage software complexity, evolution and reuse [9]. Paradoxically, the essential techniques that AOSD proposes to improve software modularity seem to restrict the evolvability of that software. More specifically, the pointcuts that state when and where aspects need to be invoked during the execution of the base program, are fragile to evolutions of that base program. This is because these pointcut definitions typically rely heavily on the structure of the base program. This tight coupling of the pointcut definitions to the base program’s structure and behaviour can seriously hamper the evolvability of the software [12]: it implies that all pointcuts of each aspect need to be checked and possibly revised whenever the base program evolves. This problem has been coined the fragile pointcut problem [11, 5] and causes evolution problems that are specific to aspect-oriented languages. We tackle the fragile pointcut problem by replacing the intimate dependency of pointcut definitions on the base program by a more stable dependency on a conceptual model of the program. These model-based pointcuts are less likely to change upon evolution, because they are no longer defined in terms of how the program happens to be structured at a certain point in time, but rather in terms of a model of the pro-