Improved traceability for bidirectional model transformations

Conventional wisdom on bidirectionality in Model-Driven Engineering (MDE) suggests that it represents a crucial components to achieve superior model management, whether it be round-tripping, synchronisation, or consistency restoration. Despite their relevance, bidirectional transformations remain difficult to design and implement due to the complexity they must usually encode and their semantic intricacy. In particular, the invertibility of a transformation can be severely affected in case of partial or non-injective transformations. Using a proper traceability model enables transformation to be stateful and permits to deal with such cases that would be otherwise largely unfeasible. This paper proposes to leverage traceability information between source and target elements of a transformation to a first-class status in order to i) automatise its generation, ii) enable a model-based representation and iii) ease reuse and refinement in a further stage to enforce consistency restoration. The approach has been demonstrated to work in practical scenarios by means of a realisation within the JTL framework.