Generator-Composition for Aspect-Oriented Domain-Specific Languages
- Zugl.: Kiel, Univ., Diss. 2016
Software systems are complex, as they must cover a diverse set of requirements describing functionality and the environment. Software engineering addresses this complexity with Model-Driven Engineering (MDE).
MDE utilizes different models and metamodels to specify views and aspects of a software system. Subsequently, these models must be transformed into code and other artifacts, which is performed by generators. Information systems and embedded systems are often used over decades.Over time, they must be modified and extended to fulfill new and changed requirements.
These alterations can be triggered by the modeling domain and by technology changes in both the platform and programming languages.
In MDE these alterations result in changes of syntax and semantics of metamodels, and subsequently of generator implementations. In MDE, generators can become complex software applications. Their complexity depends on the semantics of source and target metamodels, and the number of involved metamodels. Changes to metamodels and their semantics require generator modifications and can cause architecture and code degradation. This can result in errors in the generator, which have a negative effect on development costs and time. Furthermore, these errors can reduce quality and increase costs in projects utilizing the generator. Therefore, we propose the generator construction and evolution approach GECO, which supports decoupling of generator components and their modularization. GECO comprises three contributions: (a) a method for metamodel partitioning into views, aspects, and base models together with partitioning along semantic boundaries, (b) a generator composition approach utilizing megamodel patterns for generator fragments, which are generators depending on only one source and one target metamodel, (c) an approach to modularize fragments along metamodel semantics and fragment functionality. All three contributions together support modularization and evolvability of generators.