Section: New Results

Model Quality

Our work aims to enhance the quality of the modeling activity in the context of software engineering and language engineering. This year, this has translated in the following results:

• A benchmark that facilitates the comparison between the plethora of tools that provide some kind of quality assurance for models. Similarly to what it is done in many other domains, a common set of test benchmarks that new tools can rely on to experiment and evaluate themselves could speed up the advance in the field. Our proposal can be found [30]

• Validation of the feasibility to apply this kind of techniques in industrial settings based on two case studies [12] and [36]

• Advanced on the verification of model transformations using SMT solvers (instead of SAT or CSP-based approaches commonly used before), with some encouraging results [21] and, related to this, [13]

• A method to build models using instance-level information in terms of examples and counterexamples (gathering requirements using these instance scenarios is usually better from a stakeholder's point of view than trying to explain us general rules about the business). So far existing approaches have often focused on the generation of static models from such instance-level information but have omitted the inference of OCL business rules that could complement the static models and improve the precision of the software specification. We propose an approach to automating such inference [29] . The basic idea is based on an incorporation of the problem solving mechanism and getting user feedback: Candidates are generated by a problem solving, and irrelevant ones are eliminated using the user feedback on generated counterexamples and examples. Our approach is realized with the support tool InferOCL and has been applied on several user cases, indicating a possibility to apply this solution prototype in practice.