Modularizing Invasive Aspect Languages

In domain-specific aspect languages we observe that aspects are translated to base code and subsequently require a complex integration into base code while guaranteeing the correctness of the aspect and the base code in the woven code. We call this phenomenon invasively composed aspects. Weavers for invasive aspect languages operate on the base language level and offer dedicated support for crosscutting code. Unfortunately, current implementations poorly modularize the implementation of invasive aspect languages. This hampers their (unanticipated) evolution and severely reduces the reusability of their constructs. We suggest an approach where the specification of the crosscutting behavior is expressed on a higher semantic level. To this end, we raise the abstraction level of base languages towards the specific domain of the aspect languages. As such, we enable a modular, declarative approach. We illustrate our approach with KALA, a domain-specific aspect language.