Generating Integration Test Orders for Aspect Oriented Software with Multi-objective Algorithms

O problema conhecido como CAITO refere-se à determinação de uma ordem para integrar e testar módulos que minimize os custos associados à construção de stubs. Algoritmos de busca, particularmente os algoritmos evolutivos, têm sido utilizados. Entretanto, o problema é bastante complexo e envolve diferentes fatores associados à construção de stubs, tais como diferentes medidas de complexidade, questões contratuais e outros. Estes fatores geralmente conflitam entre si e diferentes soluções para o problema são possíveis. Para lidar adequadamente com este problema, este trabalho explora o uso de algoritmos de otimização multiobjetivos. O artigo apresenta resultados da aplicação de dois algoritmos evolutivos NSGA-II e SPEA2 para o problema CAITO, em quatro sistemas reais implementados em AspectJ. Ambos algoritmos multiobjetivos são avaliados e comparados com o algoritmo de Tarjan e com um algoritmo genético mono-objetivo. Além disto, é apresentado um exemplo de como o testador pode utilizar as soluções obtidas de acordo com suas metas de teste. Abstract: The problem known as CAITO refers to the determination of an order to integrate and test modules that minimizes stubbing costs. Search based algorithms have been used, mainly evolutionary ones. However, the problem is very complex since it involves different factors associated to the stubbing process, such as complexity measures, contractual issues and so on. These factors are usually in conflict and different possible solutions for the problem exist. To deal properly with this problem, this study explores the use of multi-objective optimization algorithms. The paper presents results from the application of two evolutionary algorithms NSGA-II and SPEA2 to the CAITO problem in four real systems, implemented in AspectJ. Both multi-objective algorithms are evaluated and compared with the traditional Tarjan’s algorithm and with a mono-objective genetic algorithm. Moreover, it is shown how the tester can use the found solutions, according to the test goals. The problem known as CAITO refers to the determination of an order to integrate and test modules that minimizes stubbing costs. Search based algorithms have been used, mainly evolutionary ones. However, the problem is very complex since it involves different factors associated to the stubbing process, such as complexity measures, contractual issues and so on. These factors are usually in conflict and different possible solutions for the problem exist. To deal properly with this problem, this study explores the use of multi-objective optimization algorithms. The paper presents results from the application of two evolutionary algorithms NSGA-II and SPEA2 to the CAITO problem in four real systems, implemented in AspectJ. Both multi-objective algorithms are evaluated and compared with the traditional Tarjan’s algorithm and with a mono-objective genetic algorithm. Moreover, it is shown how the tester can use the found solutions, according to the test goals. ∗This is an extended version of the paper that appeared at LA-WASP 2011 (V Latin American Workshop on AspectOriented Software Development and has been recommended to RITA.) 1Computer Science Department, Federal University of Paraná (UFPR), CP:19081, CEP: 19031-970, Paraná, Brazil {wesleyk, silvia, aurora}@inf.ufpr.br 2Computer Science Department, State University of Maringá (UEM), Paraná, Brazil [email protected] Generating Integration Test Orders for Aspect-Oriented Software with Multi-objective Algorithms