Automated metamodel/model co-evolution: A search-based approach

摘要

Context:Metamodels evolve over time to accommodate new features, improve existing designs, and fix errors identified in previous releases. One of the obstacles that may limit the adaptation of new metamodels by developers is the extensive manual changes that have been applied to migrate existing models. Recent studies addressed the problem of automating the metamodel/model co-evolution based on manually defined migration rules. The definition of these rules requires the list of changes at the metamodel level which are difficult to fully identify. Furthermore, different possible alternatives may be available to translate a metamodel change to a model change. Thus, it is hard to generalize these co-evolution rules.Objective:We propose an alternative automated approach for the metamodel/model co-evolution. The proposed approach refines an initial model instantiated from the previous metamodel version to make it as conformant as possible to the new metamodel version by finding the best compromise between three objectives, namely minimizing (i) the non-conformities with new metamodel version, (ii) the changes to existing models, and (iii) the textual and structural dissimilarities between the initial and revised models.Method:We formulated the metamodel/model co-evolution as a multi-objective optimization problem to handle the different conflicting objectives using the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and the Multi-Objective Particle Swarm Optimization (MOPSO).Results:We evaluated our approach on several evolution scenarios extracted from different widely used metamodels. The results confirm the effectiveness of our approach with average manual correctness, precision and recall respectively higher than 91%, 88% and 89% on the different co-evolution scenarios.Conclusion:A comparison with our previous work confirms the out-performance of our multi-objective formulation.