Maintaining consistency in networks of models: bidirectional transformations in the large

摘要

The model-driven development of systems involves multiple models, metamodels and transformations, and relationships between them. A bidirectional transformation (bx) is usually defined as a means of maintaining consistency between "two (or more)" models. This includes cases where one model may be generated from one or more others, as well as more complex ("symmetric") cases where models record partially overlapping information. In recent years, binary bx, those relating two models, have been extensively studied. Multiary bx, those relating more than two models, have received less attention. In this paper, we consider how a multiary consistency relation may be defined in terms of binary consistency relations and how consistency restoration may be carried out on a network of models and relationships between them. In particular, we consider the circumstances under which we can prove non-interference between several bidirectional transformations that impact on the same model and how the use of a more refined notion of consistency can help in cases where this is not possible. In the process, we develop an abstract theory of parts of a model that are read or modified by a bidirectional transformation. We relate the work to megamodelling and discuss further research that is needed.