Rewriting Logic Semantics and Verification of Model Transformations

Abstract

Model transformations are used in model-driven development for mechanizing the interoperability and integration among modeling languages. Due to the graph-theoretic nature of models, the theory of graph transformation systems and its technological support provide a convenient environment for formalizing and verifying model transformations, which can then be used for defining the semantics of modelbased domain-specific languages. In this paper, we present an approach for formalizing and verifying QVT-like transformations that reuses the main concepts of graph transformation systems. Specifically, we formalize model transformations as theories in rewriting logic, so that Maude's reachability analysis and model checking features can be used for verifying them. This approach also provides a new perspective on graph transformation systems, where their formal semantics is given in rewriting logic. All the ideas presented are implemented in MOMENT2. In this way, we can define formal model transformations in the Eclipse Modeling Framework (EMF) and we can verify them in Maude. We use a model of a distributed mutual exclusion algorithm to illustrate the approach.

Contracts for Model Execution Verification

“…In all these approaches, a translational semantics is defined: a model or a model transformation is specified in another technological space in order to use third-party simulation and/or verification tools. For instance, [4] verifies invariants or temporal constraints through the Maude framework and LTL properties. [7] proposes prioritized timed Petri nets while [19] stresses the expression of transformations towards colored Petri nets.…”

What Is a Multi-modeling Language?

“…The algebraic semantics supports the notion of submodel (see [6] for details). From a graph-theoretic point of view, given…”

“…These notions are implemented in Maude and integrated within the Eclipse Modeling Framework (EMF) in the MOMENT2 tool [8,7,6]. …”

“…Furthermore, any computable function β : (M , C ) → (M , C ) can in such a case be specified by a finite set of confluent and terminating equations modulo ACU. This is exactly the approach taken in MOMENT2, where a model transformation β can be specified as a set of recursive model equations, which are automatically translated into ordinary MEL equations, as detailed in [6]. Let us consider a model transformation between the metamodel specifications for class diagrams (CD) and for relational database schemas (RDBS) in Fig.…”

“…The NAC ensures that the rule is applied only once for each attribute. MOMENT2 formalizes this model transformation as the function β cd2rdbs , which is internally defined by equations that are generated from the user defined model equations of the transformation (see [6] for further details).…”