Automated Selection of Optimal Model Transformation Chains via Shortest-Path Algorithms

Conventional wisdom on model transformations in Model-Driven Engineering (MDE) suggests that they are crucial components in modeling environments to achieve superior automation, whether it be refactoring, simulation, or code generation. While their relevance is well-accepted, model transformations are challenging to design, implement, and verify because of the inherent complexity that they must encode. Thus, defining transformations by chaining existing ones is key to success for enhancing their reusability. This paper proposes an approach, based on well-established algorithms, to support modellers when multiple transformation chains are available to bridge a source metamodel with a target one. The all-important goal of selecting the optimal chain has been based on the quality criteria of coverage and information loss. The feasibility of the approach has been demonstrated by means of experiments operated on chains obtained from transformations borrowed from a publicly available repository.