Towards Approximate Model Transformations

Abstract

As the size and complexity of models grow, there is a need to count on novel mechanisms and tools for transforming them. This is required, e.g., when model transformations need to provide target models without having access to the complete source models or in really short time—as it happens, e.g., with streaming models—or with very large models for which the transformation algorithms become too slow to be of practical use if the complete population of a model is investigated. In this paper we introduce Approximate Model Transformations, which aim at producing target models that are accurate enough to provide meaningful and useful results in an efficient way, but without having to be fully correct. So to speak, this kind of transformations treats accuracy for execution performance. In particular, we redefine the traditional OCL operators used to query models (e.g., allInstances, select, collect, etc.) by adopting sampling techniques and analyse the accuracy of approximate model transformations results.