Investigating the Change-Proneness of Service Patterns and Antipatterns

Like any other software systems, service-based systems (SBSs) evolve frequently to accommodate new user requirements. This evolution may degrade their design and implementation and may cause the introduction of common bad practice solutions -- antipatterns -- in opposition to patterns which are good solutions to common recurring design problems. We believe that the degradation of the design of SBSs does not only affect the clients of the SBSs but also the maintenance and evolution of the SBSs themselves. This paper presents the results of an empirical study that aimed to quantify the impact of service patterns and antipatterns on the maintenance and evolution of SBSs. We measure the maintenance effort of a service implementation in terms of the number of changes and the size of changes (i.e., Code churns) performed by developers to maintain and evolve the service, two effort metrics that have been widely used in software engineering studies. Using data collected from the evolutionary history of the SBS FraSCAti, we investigate if (1) services involved in patterns require less maintenance effort, (2) services detected as antipatterns require more maintenance effort than other services, and (3) if some particular service antipatterns are more change-prone than others. Results show that (1) services involved in patterns require less maintenance effort, but not at statistically significant level, (2) services detected as antipatterns require significantly more maintenance effort than non-antipattern services, and (3) services detected as God Component, Multi Service, and Service Chain antipatterns are more change-prone (i.e., Require more maintenance effort) than the services involved in other antipatterns. We also analysed the relation between object-oriented code smells and service patterns/antipatterns and found a significant difference in the proportion of code smells contained in the implementations of service patterns and antipatterns.