Masters Thesis Defense: Mohammad Sadegh Aalizadeh

Refactoring is a common maintenance practice that enables developers to improve the internal structure of a software system without altering its external behaviour. In this study we propose a novel method to automatically detect 11 motivations driving the application of EXTRACT METHOD refactoring operations. We conduct a large-scale study on 325 open-source Java repositories to automatically detect the motivations of 346K EXTRACT METHOD refactoring instances. Previous studies have been merely based on surveys, manual analysis of pull requests or commit-messages to detect the motivations of developers. In this study we develop motivation detection rules to automatically extract the developer motivations based on the context of a refactoring operation in the commit. We find that the top four motivations for EXTRACT METHOD refactoring is to introduce reusable methods, remove duplication, facilitate the implementation of new features and bug fixes, and decompose long methods to improve their readability. There is an association between the removal of duplication and the introduction of reusable methods in the refactoring instances with multiple motivations. The findings of this study provide essential feedback and insight for the research community, refactoring recommendation tool builders, and project managers, to better understand why and how developers perform EXTRACT METHOD refactorings and help them build refactoring tools tailored to the needs and practices of developers.