Towards Compositional Domain Specific Languages

The deployment of Domain Specific Languages (DSL) and in particular Domain Specific Modeling Languages (DSML) is becoming more and more prominent in various domains. In order to cope with the complexity of the realization of DSLs, common and well-established methods of software engineering such as modularization and reuse need to be adapted and applied for DSLs. This has already been noted in [2] when the emerging DSL era was still closely akin to compiler theory. As stated in this work, compositionality of DSLs can take place at several dimensions. Various contributions in this field of ongoing research reflect this and only a brief overview is given below. One form of DSL composition is the syntactic embedding such as embedding DSLs in GPLs as described in [3, 4]. In [5] a family of DSMLs are used for the generation of web information systems. There the composition is carried out via the joint usage of several languages each with their own artifacts and hence no syntactic embedding. Other contributions in the area consider the composition of the models expressed in DSLs as a constructive model transformation [1] and examine the effects of DSL composition at the infrastructure level [9] (e.g., syntax aware editors etc.). The DSL framework and workbench MontiCore [7, 8, 10, 11] was designed and realized particularly with respect to compositionality at various dimensions [6, 12]. Compositionality is of special interest if models of different modeling paradigms and hence expressed in different languages need to be combined while at the same time retaining their specific semantics. Whenever the different modeling paradigms are integrated, it can be observed that each paradigm is equipped with its own modeling language and that therefore such a paradigm integration is always also a model language composition. This holds for the composition of structural and behavioral languages as well as for the composition of languages with synchronous or real-time communication and event triggered asynchronous models, etc. In the following the dimensions of such compositions are discussed in more detail.