Tracking Down Dynamic Feature Code Changes against Python Software Evolution

Python, a typical dynamic programming language, is increasingly used in many application domains. Dynamic features in Python allow developers to change the code at runtime. Some dynamic features such as dynamic type checking play an active part in maintenance activities, thus dynamic feature code is often changed to cater to software evolution. The aim of this paper is exploring and validating the characteristics of feature changes in Python. We collected change occurrences in 85 open-source projects and discovered the relationship between feature changes and bug-fix activities. Furthermore, we went into 358 change occurrences to explore the causes and behaviors of feature changes. The results show that: (1) dynamic features are increasingly used and the code is changeable; (2) most dynamic features may behave that feature code is more likely to be changed in bug-fix activities than non-bugfix activities; (3) dynamic feature code plays both positive and negative roles in maintenance activities. Our results provide useful guidance and insights for improving automatic program repair and refactoring tools.