On the application of the multi-evolutionary and coupling-based approach with different aspect-class integration testing strategies

Abstract

During the integration test of aspect-oriented software, it is necessary to determine an aspect-class integration and test order, associated to a minimal possible stubbing cost. To determine such optimal orders an approach based on multi-objective evolutionary algorithms was proposed. It generates a set of good orders with a balanced compromise among different measures and factors that may influence the stubbing process. However, in the literature there are different strategies proposed to aspect-class integration. For instance, the classes and aspects can be integrated in a combined strategy, or in an incremental way. The few works evaluating such strategies do not consider the multi-objective and coupling based approach. Given the importance of such approach to reduce testing efforts, in this work, we conduct an empirical study and present results from the application of the multi-objective approach with both mentioned strategies. The approach is implemented with four coupling measures and three evolutionary algorithms that are also evaluated: NSGA-II, SPEA2 and PAES. We observe that different strategies imply in different ways to explore the search space. Moreover, other results related to the practical use of both strategies are presented. © 2013 Springer-Verlag.