Aspect Weaving with Graph Rewriting

This paper introduces GRS-based AOP which explains a large subclass of AOP weavers as graph rewrite systems (GRS). The corresponding class of AOP problems has a formal background since it inherits all features of graph rewrite systems such as criteria for termination , connuence, and unique normal forms. In particular, it it shown that diierent kinds of rewrite rules form diierent weaver classes. At least two of them (EARS and XGRS weavers) have simple semantics since they always yield unique weaving results. Aaspect-oriented programming (AOP) allows to specify parts and behaviors of components using diierent views, or aspects. These aspects can be speci-ed independently from each other and are combined to the nal form by a weaver tool. This composition process yields software that is better readable (since speciications are modular) and better reusable (since aspects can be re-combined). However, although single instances of AOP such as Adaptive Programming Lie96] or Compositon Filters AWB + 94] rely on formal principles the technique as such does not yet have a formal background. This paper introduces GRS-based AOP which explains a large subclass of weavers as graph rewrite systems (GRS). Hence it has a formal background, inheriting all features of graph rewrite systems such as criteria for termination, connuence, and unique normal forms. In particular, it it shown that diierent kinds of rewrite rules form diierent weaver classes. Although GRS-based AOP does not cover all possible incarnations of AOP, it provides a simple formalism for aspects relying on properties that can be recognized directly from the their syntactical structure. We start with a simple example which demonstrates that graph rewrite rules can describe simple weavings. After introducing some basic terminology we present a reened classiication concering components, join points, and weavers (section 3). With GRS-based AOP, diierent kinds of rewrite rules form diier-ent weaver classes (section 3.1). Some of them have simple semantics since they always yield unique weaving results. For aspects that are encoded into aspect graphs, it is observed that GRS-based weaving must use context-sensitive graph rewriting (section 3.1). In particular, GRS-based AOP simpliies the construction of AOP weavers. Since tools exist which generate code for the graph rewriting speciications weavers can be generated automatically. We demonstrate this with an example with the graph rewrite tool OPTIMIX (section 4). A table comparing the diierent classes of GRS-based AOP concludes the paper.