On the Automated Derivation of Domain-Specific UML Profiles

The model-driven engineering (MDE) of domain-specific languages (DSLs) becomes increasingly important for standardization. Both metamodels and profiles for the Unified Modeling Language (UML) are often provided for standardized DSLs. Usually, the mappings of metamodels to UML profiles are only specified in some abstract manner by such standards. Consequently, not all details required to create executable model transformations are defined. Moreover, the static semantics of metamodels and/or UML profiles are often specified only in natural language. Thus, on the one hand, ambiguities can occur, and on the other hand, the soundness of the static semantics cannot be verified and validated without a manual translation into a machine-processable language.
The main goal of this dissertation is to remedy the identified weaknesses in developing DSLs, for which a metamodel and a UML profile shall be created. To achieve this goal, we propose a holistic MDE-based approach for automatically deriving UML profiles and model transformations basedon CMOF metamodels. This approach enables an automatic transfer of DSL’s static semantics to UML profiles, so that the well-formedness of UML models with applied UML profile of a DSL can be verified automatically. In addition, the interoperability between UML and DSL models can already be validated in the development phase using the derived model transformations. Apart from new DSLs that are created from scratch, our approach also supports a migration of existing grammar-based DSLs towards metamodels, provided syntax rules exist.