Performance Investigation of Unit Testing in Android Programming Learning Assistance System

With increasing demands for Android application programmers, we have developed the Android Programming Learning Assistant System (APLAS) to provide a self-learning platform in Android programming. In APLAS, unit testing takes the essential role to confirm the validity of satisfying the required specifications in the answer source code from a student. However, JUnit and Robolectric, the unit testing tools adopted in APLAS, require high CPU loads and take long execution time because of the complex procedure of testing the source codes for human-interactive applications. In the previous studies, we have implemented the unit testing function in both the clientside and the server-side of the web-based online platform. In this paper, we present the performance investigations of the two unit testing tools at the validation process under various PC hardware and test case specifications. The results show that the hardware specifications, the initialization process by Gradle, and the number of test cases in a test code have significant impacts on the validation time, and the JUnit-based test code is much faster than the Robolectric-based test code.