On the Survival of Android Code Smells in the Wild

The success of smartphones and app stores have contributed to the explosion of the number of mobile apps proposed to end-users. In this very competitive market, developers are rushed to regularly release new versions of their apps in order to retain users. Under such pressure, app developers may be tempted to adopt bad design or implementation choices, leading to the introduction of code smells. Mobile-specific code smells represent a real concern in mobile software engineering. Many studies have proposed tools to automatically detect their presence and quantify their impact on performance. However, there remains-so far-no evidence about the lifespan of these code smells in the history of mobile apps. In this paper, we present the first large-scale empirical study that investigates the survival of Android code smells. This study covers 8 types of Android code smells, 324 Android apps, 255k commits, and the history of 180k code smell instances. Our study reports that while in terms of time Android code smells can remain in the codebase for years before being removed, it only takes 34 effective commits to remove 75% of them. Also, Android code smells disappear faster in bigger projects with higher releasing trends. Finally, we observed that code smells that are detected and prioritised by linters tend to disappear before other code smells.