Change-Oriented Repair Propagation

Repairing software models may be a laborious task, as the number of alternatives that must be considered by engineers can be large. This is more evident in collaborative environments, where the stream of changes applied to models by different engineers is constant and unanticipated. These changes can cause multiple inconsistencies that must be fixed while preserving the changes applied. However, performing this task is not trivial, as analyzing the changes and the possible large amount of repair alternatives requires time and effort. In this work, we present an approach that aids this repair process by analyzing the stream of changes (i.e., history of changes) while exploring repair alternatives alongside their side effects. The approach generates repairs for inconsistencies identified in the model. These repairs are explored by simulating their execution while re-analyzing the model to find potential new inconsistencies created. Then, new repairs are generated to fix these new inconsistencies. This cycle repeats until the model reaches a consistent state or until repairs can no longer be generated. The approach also analyzes conflicts between repairs and changes. This analysis brings valuable information to engineers regarding how each repair alternative would impact their models and may conflict with changes as well as other repairs. We evaluated our approach in a set of 11 UML models that contain a history of changes. Our findings show how our approach can be applied in a variety of models with a different number of model elements and inconsistencies within a reasonable amount of time.