Towards Instant Automatic Model Refinement Based on OCL

Model refinement is a complex task. It is difficult for developers to refine models all by themselves. A good modeling tool should not only do routine refinement work for developers, but also guide them to make decisions during the refinement process. Such refinement work is best done by modeling tools instantly while developers refine models, to provide timely assistance. In this paper, we present a general approach for meta-model designers or experts in different fields to define and describe such automatic refinement work by rules, and these rules can instruct the modeling tool to do the refinement work instantly, whenever the developers modify the model. The automatic refinement rules in our approach are based on OCL, and their semantics and behaviors are formally defined by using Dijkstra's guarded commands, so the rules are compact, unambiguous and reliable to use. We have also implemented the editor and the interpreter for the automatic refinement rules, and integrated them into our own modeling tool to support our approach.