Identification and Optimisation of Type-Level Model Queries

The main appeal of task-specific model management languages such as ATL, OCL, Epsilon etc. is that they offer tailored syntaxes for the tasks they target, and provide concise first-class support for recurring activities in these tasks. On the flip side, task-specific model management languages are typically interpreted and are therefore significantly slower than general purpose programming languages (which can be also used to query and modify models) such as Java. While this is not an issue for smaller models, as models grow in size, naive execution of interpreted model management programs against them can become a scalability bottleneck. In this paper, we demonstrate an architecture for optimisation of model management programs written in languages of the Epsilon platform using static analysis and program rewriting techniques. The proposed architecture facilitates optimisation of queries that target models of heterogeneous technologies in an orthogonal way. We demonstrate how the proposed architecture is used to identify and optimise typelevel queries against EMF-based models in the context of EOL programs and EVL validation constraints. We also demonstrate the performance benefits that can be delivered by this form of optimisation through a series of experiments on EMF-based models. Our experiments have shown performance improvements of up to 99.56%.