GECO: A Generator Composition Approach for Aspect-Oriented DSLs

Code and model generators that are employed in model-driven engineering usually face challenges caused by complexity and tight coupling of generator implementations, particularly when multiple metamodels are involved. As a consequence maintenance, evolution and reuse of generators is expensive and error-prone. We address these challenges with a two fold approach for generator composition, called GECO, which subdivides generators in fragments and modules. (1) fragments are combined utilizing megamodel patterns. These patterns are based on the relationship between base and aspect metamodel, and define that each fragment relates only to one source and target metamodel. (2) fragments are modularized along transformation aspects, such as model navigation, and metamodel semantics. We evaluate our approach with two case studies from different domains. The obtained generators are assessed with modularity and complexity metrics, covering architecture and method level. Our results show that the generator modularity is preserved during evolution utilizing GECO.