Repatterning: Improving the Reliability of Android Applications with an Adaptation of Refactoring

Abstract

Studies of Android applications show that NullPointerException, OutofMemoryError, and BadTokenException account for a majority of errors observed in the real world. The technical debt being born by Android developers from these systemic errors appears to be due to insufficient, or erroneous, guidance, practices, or tools. This dissertation proposes a re-engineering adaptation of refactoring, called repatterning, and pays down some of this debt. We investigated 323 Android applications for code smells, corrective patterns, or enhancement patterns related to the three exceptions. I then applied the discovered patterns to the locations suggested by the code smells in fifteen randomly selected applications. I measured the before and after reliability of the applications and observed a statistically significant improvement in reliability for two of the three exceptions. I found repatterning had a positive effect on the reliability of Android applications. This research demonstrates how refactoring can be generalized and used as a model to affect non-functional qualities other than the restructuring related attributes of maintainability and adaptability.