Two-Step Transformation of Model Traversal EOL Queries for Large CDO Repositories.

Recent approaches persist models in databases to overcome performance and memory limitations of XMI. Among them, Connected Data Objects CDO is a database-based model repository widely used in Model Based Engineering by academia and industry. Model traversal queries are intensively used in modelling scenarios and their performance greatly impacts tools performance and user experience. In this paper, we introduce the CDO-QT framework to transform model traversal queries from Epsilon Object Language EOL into SQL queries and execute them at CDO repositories. This way, model engineers can define queries using domain concepts at performance similar to SQL. We have evaluated CDO-QT executing a set of queries over repositories from 15﾿MB to 5﾿GB size. CDO-QT results in better performance and memory consumption with respect to other approaches Plain EMF, MDT OCL, CDO-OCL.