Towards models and metamodels co-evolution approach

Abstract

Model Driven Software Engineering has matured over the last few years and is now becoming an established technology. As a consequence dealing with evolving metamodels and the necessary co-evolution of instances of this metamodel is becoming increasingly important. Several approaches have been proposed to solve model co-evolution problem. In this paper, Firstly, existing approaches in this area are analyzed to define requirements of our approach, namely automaticity, reuse, expressiveness and intelligence. After that we present a new approach CBRMig towards a solution to metamodel and model co-evolution problem. The core of our proposal is using a case based reasoning system that automates the process of creation and selection of cases to assist with the generation of a migration algorithm used for adapting models in response to metamodel evolution.

Approaches to Co-Evolution of Metamodels and Models: A Survey

“…These learned strategies should later be applied automatically, if possible, so that the need to specify new resolution strategies is reduced over time. This proposal is called CBRMig and was published in 2013 by Anguel et al [62].…”

“…Most (14) of the 15 approaches that allow for offline change collection also enable atomic change detection ( [9], [10], [28], [32], [37], [39], [47], [53], [56], [62], [69], [70], [71], [72], see Table 2). To achieve that, most approaches (12 of 14) propose a calculation of the difference between the original and the evolved version of the metamodel, e.g.…”

“…To achieve that, most approaches (12 of 14) propose a calculation of the difference between the original and the evolved version of the metamodel, e.g. by using EMF Compare 19 (as in EMFMigrate [8], [26], [27], [28], Anguel et al [47], and CBRMig [62]). Two approaches (Williams et al [69] and MCL [30], [31], [32]) join atomic with complex change identification (as explained below).…”

“…Finally, the resolution strategy learning approach CBR-Mig [62] supports situations where a single solution is automatically applied without user interaction. However, not a predefined strategy is used, but a learned solution that is selected from a list of previously applied solutions.…”

“…Strategy Specification: Free specification of migration strategies is supported by all resolution strategy languages [23], [24], [25], [28], [29], [32], [35] as well as by individual approaches of other families: the resolution strategy generation approach of Meyers et al [40], the predefined resolution strategy approaches of Gruschko et al [14] and Wittern [61] as well as COPE/Edapt [60] and CBRMig [62].…”