Preserving constraints in horizontal model transformations

Graph rewriting is gaining credibility in the model transformation field, and tools are increasingly used to specify transformation activities. However, their use is often limited by special features of graph transformation approaches, which might not be familiar to experts in the modeling domain. On the other hand, transformations for specific domains may require special constraints to be enforced on transformation results. Preserving such constraints by manual definition of graph transformations can be a cumbersome and error-prone activity. We explore the problem of ensuring that possible violations of constraints following a transformation are repaired in a way coherent with the intended meaning of the transformation. In particular, we consider the use of transformation units within the DPO approach for intra-model transformations, where the modeling language is expressed via a type graph and graph conditions. We derive additional rules in a unit from a declarative rule expressing the principal objective of the transformation, so that the constraints set by the type graph and the graph conditions hold after the application of the unit. The approach is illustrated with reference to a diagrammatic reasoning system.