An Empirical Study of the Performance Impacts of Android Code Smells

Abstract

Android code smells are bad implementation practices within Android applications (or apps) that may lead to poor software quality, in particular in terms of performance. Yet, performance is a main software quality concern in the development of mobile apps. Correcting Android code smells is thus an important activity to increase the performance of mobile apps and to provide the best experience to mobile end-users while considering the limited constraints of mobile devices (e.g., CPU, memory, battery). However, no empirical study has assessed the positive performance impacts of correcting mobile code smells. In this paper, we therefore conduct an empirical study focusing on the individual and combined performance impacts of three Android performance code smells (namely, Internal Getter/Setter, Member Ignoring Method, and HashMap Usage) on two open source Android apps. To perform this study, we use the Paprika toolkit to detect these three code smells in the analyzed apps, and we derive four versions of the apps by correcting each detected smell independently, and all of them. Then, we evaluate the performance of each version on a common user scenario test. In particular, we evaluate the UI and memory performance using the following metrics: frame time, number of delayed frames, memory usage, and number of garbage collection calls. Our results show that correcting these Android code smells effectively improve the UI and memory performance. In particular, we observe an improvement up to 12.4% on UI metrics when correcting Member Ignoring Method and up to 3.6% on memory-related metrics when correcting the three Android code smells. We believe that developers can benefit from these results to guide their refactoring, and thus improve the quality of their mobile apps.