An Approach to Test Classification in Big Android Applications

This paper intends to shed some light on how to approach challenges on automated test strategies and architecture that arise in big Android applications. The Android documentation and Google testing recommendations are overly simplistic and showcase only trivial and small examples, while the real challenges in testing big Android applications stay untouched. We discuss testing partitioning, classification of tests, and the application of the test pyramid on the Android platform. Besides a small exploratory study on the testing situation in a limited sample of open source Android applications, we show the consequences of a bad test strategy and technical debt in a big and long running Android project in the case of Catrobat. By Identifying anti patterns and smells, we establish general guidelines and patterns for testing big Android applications, as well as metrics for test code quality which have been applied and found useful in the Catrobat project.