Towards Systematic Management of Semantic Variability in Real-World Executable Modeling Languages

Abstract

Precise execution semantics are highly desirable for modeling languages as it enables seamless interoperability and portability of models. Yet, in practice, numerous variants of execution semantics are observed in modeling environments. We can describe modeling concepts with varying execution semantics as semantic variation points. They may be defined intentionally to cover a wide range of modeling domains or arise incidentally within the language’s ecosystem. Semantic variation points may appear in general-purpose modeling languages (such as UML) as well as in domain-specific languages. Potential causes are imprecise semantics definitions, vendor-specific extensions, or incomplete implementations, as well as inconsistent evolution of modeling tools between language versions. As a result, improved language versions are not a feasible countermeasure for this organically grown variation. Therefore, we propose to create an approach for explicitly handling semantic variation points and linking the variability model to an interpreter. To exemplify our concept, we investigate semantic variants of a domain-specific language (IEC 61499, for modeling distributed control systems). A variability model explicitly captures the semantic variants. Based on this model, we will create an adaptable model interpreter that helps identify modeling errors introduced in selected semantic variants. As models can be executed with any chosen semantic variant, our approach can support better model validation and contributes to the portability between modeling tools.