Querying UML Models using OCL and Prolog: A Performance Study

The size of unified modeling language (UML) models used in practice is very large and ranges up to hundreds and thousands of classes. Querying of these models is used to support their quality assessment by information filtering and aggregating. For both, human cognition and automated analysis, there is a need for fast querying. In this context performance of model queries becomes an important issue. We investigated performance characteristics of two different querying engines: one using the object constraint language (OCL) and the other using prolog. Our comparison is based on equivalent queries in both languages. We applied the queries to 118 models of a size up to 10000 classes to analyze model load and evaluation time. Our preliminary results show that if execution time of queries is linear then prolog is faster. For one of the presented cases, the execution time in prolog was nonlinear and thus higher. Further studies should focus on a performance analysis reflecting expressiveness aspects. Our experimental material is accessible to enable future replications of this study.