Droidlens: Robust and Fine-Grained Detection for Android Code Smells

Chen-Guang Mao, Hao Wang, Gao-Jie Han, Xiaofang Zhang

Abstract: With rapid changes and rich context of user requirements, mobile applications are becoming complex software systems. Frequent iterations and mazy implementations of software functions lead Android developers to make poor design choices, called Android Code Smells. Past researches have shown that they have negative impacts on Android applications including performance, security, etc. Therefore, the automated detection of Android code smells is indispensable to help alleviate the workload of software maintainers and developers. There are already two automated detection tools, aDoctor and Paprika. However, they both have shortcomings in detecting granularity and accuracy. In this paper, we present a novel approach, called Droidlens, realizing the analysis, detection, location and refactoring of Android code smells. We also make an empirical study focusing on the performance of Droidlens, aDoctor and Paprika. The empirical result shows that Droidlens realizes the detection for 18 Android code smells. Moreover, compared to existing tools, our Droidlens can provide robust and fine-grained detection, which contributes to software refactoring and maintenance.