A dedicated approach for model composition traceability

Context: Software systems are often too complex to be expressed by a single model. Recognizing this, the Model Driven Engineering (MDE) proposes multi-modeling approaches to allow developers to describe a system from different perspectives. In this context, model composition has become important since the combination of those partial representations is inevitable. Nevertheless, no approach has been defined for keeping track of the composition effects, and this operation has been overshadowed by model transformations. Objective: This paper presents a traceability approach dedicated to the composition of models. Two aspects of quality are considered: producing relevant traces; and dealing with scalability. Method: The composition of softgoal trees has been selected to motivate the need for tracing the composition of models and to illustrate our approach. The base principle is to augment the specification of the composition with the behavior needed to generate the expected composed model accompanied with a trace model. This latter includes traces of the execution details. For that, traceability is considered as a crosscutting concern and encapsulated in an aspect. As part of the proposal, an Eclipse plug-in has been implemented as a tool support. Besides, a comparative experiment has been conducted to assess the traces relevance. We also used the regression method to validate the scalability of the tool support. Results: Our experiments show that the proposed approach allows generating relevant traces. In addition, the obtained results reveal that tracing a growing number of elements causes an acceptable increase of response time. Conclusion: This paper presents a traceability approach dedicated to the composition of models and its application to softgoal trees. The experiment results reveal that our proposal considers the composition specificities for producing valuable traceability information while supporting scalability