Heuristic Inference of Model Transformation Definitions from Type Mappings

Developers of model transformations are often required to produce verbose transformation definitions. This verbosity results from the need to link the elements of the target model explicitly by initializing their features, which requires developers to handwrite large amounts of repetitive transformation instructions. Whereas existing hybrid transformation languages automate the application of rules, they do not save developers from these boilerplate instructions. In this paper, we demonstrate that a large fraction of such instructions can be intelligently inferred from a set of simple mappings between the source and target types of a transformation. On this basis, we propose a heuristic inference mechanism to augment such type mappings with additional mappings for the initialization of features, thereby saving developers from the need to handwrite boilerplate instructions. We showcase the ability of our mechanism to reduce the verbosity of transformations using four case studies.