Coach: classification-based architectural patterns detection in Android apps

ABSTRACTMobile applications (apps) have become a key feature of modern life, and as such, are expected to be high-quality software products. App developers thus have to consider and apply best practices and design recommendations such as architectural patterns. One of these patterns is the popular MVC (Model View Controller) and its variants: MVP (Model View Presenter) and MVVM (Model View ViewModel). However, there is no common or established strategy that developers can follow when implementing these patterns in their apps. In this context, analyzing existing mobile apps and identifying their implementations of the MVC pattern variants becomes important to understand the current practices in mobile apps. In this paper, we present a classification-based approach to identify MVC-based patterns in Android apps. We provide a dataset of 69 Android apps manually annotated with the applied MVC-based pattern. We validate our approach on 265 open-source Android apps and show that it outperforms the only existing pattern detection tool for all the MVC-based variants.