A Framework for Model Transformations on Petri Nets in Color

Model transformations play an important role in software engineering in general and in the area of model-driven engineering in particular, representing the key mechanisms for model translations (e.g., translating an ER model into a UML class model), model augmentations (e.g., weaving aspects into a UML class model), and model alignments (e.g., mapping a content model to its GUI view), to mention just a few. Several kinds of dedicated model transformation languages have emerged during the last years, which allow specifying and executing transformations between source and target metamodels and their corre-sponding models, respectively. However, none of these languages, not even the QVT-standard proposed by the OMG, became generally accepted as a state-of-the-art technology for model transformations. This rare adoption of model transformation languages in practice is due to several reasons. First, existing model transformation languages do not provide appropriate abstraction mechanisms to deal with the complexity of structural heterogeneities of different metamodels. Second, they lack suitable reuse mechanisms in order to reduce the high and error-prone effort of specifying recurring transformations. And finally, these languages exhibit an inherent impedance mismatch between the specification and the execution of model transformations in terms of a one-to-many derivation of concepts, thus hampering both, understandability and debuggability. The aim of this project is to establish a framework called TROPIC (Transformations on Petri Nets in Color) for developing model transformations which tackles these limitations. First, TROPIC allows to specify model transformations on different abstraction levels, providing both a declarative mapping language based on UML 2 component diagrams which hides implementation details, and derived from that, an executable transformation language using Coloured Petri Nets. Second, TROPIC facilitates reusability by providing an initial library of generic transformation operators which can be bound to arbitrary metamodels and by allowing to extend this library on demand with new, user-defined, transformation operators, optionally composed out of already existing ones. Finally, TROPIC overcomes the impedance mismatch by supporting a dedicated runtime model in terms of Coloured Petri Nets, allowing for a homogeneous representation of all transformation artefacts (i.e., models, metamodels and the transformation logic itself), which fosters understandability and debuggability of model transformations. The methodology for evaluating the proposed framework builds on three major pillars. First, appropri-ate case studies for transforming heterogeneous structural as well as behavioural models will be set up and implemented with different existing model transformation languages, including TROPIC, the results being evaluated on basis of a suitable subset of the ISO 9126 software quality model. Second, the findings of these case studies will be further critically reflected by conducting an empirical study with students from our model engineering courses (around 200 master students every year). Third, dedicated workshops will be held together with internationally renowned inventors of other model transformation languages to addi-tionally review the value of our proposed framework.

Given the growing complexity of software, improvements in productivity are key in the software engineering process. A current, promising trend in this direction is called Model-Driven Engineering (MDE), targeting at specification of software on a higher level of abstraction and automating recurring implementation tasks through code generation. This leads on the one hand to shorter development processes and on the other hand to an increased quality of software, being especially of utmost importance in safety critical applications as, e.g., in aviation industry.The key mechanism in MDE are software models, which play the central role in the software engineering process. Depending on the purpose and focus, these models may be represented in different optimized formalisms. For avoiding inconsistent specifications across different formalisms, these models must be translated into one another and at the very end into code, thus, requiring so-called model transformations. For the specification of model transformations, dedicated transformation languages have been proposed. However, these transformation languages miss dedicated reuse mechanisms as well as means for debugging.In the course of the TROPIC project, for providing reuse mechanisms, recurring transformation problems have been identified, for which dedicated components have been developed. Consequently, transformation developers may use these components in the transformation development process. This leads to a more efficient and less error-prone development of model transformations. For leveraging debugging in model transformations, existing transformation languages have been translated to the visual formalism of Colored Petri Nets, which provides a formal basis and may thus be exploited for verification.For evaluating the proposed concepts, a prototypical implementation has been built. This prototype allows for the visual specification of model transformations on basis of the developed components. This specification as well as specifications in other dedicated transformation languages may be transferred into a Colored Petri Nets based view for debugging.