Understanding Widespread Changes: A Taxonomic Study

Many active research studies in software engineering, such as detection of recurring bug fixes, detection of copy-and-paste bugs, and automated program transformation tools, are motivated by the assumption that many code changes (e.g., changing an identifier name) in software systems are widespread to many locations and are similar to one another. However, there is no study so far that actually analyzes widespread changes in software systems. Understanding the nature of widespread changes could empirically support the assumption, which provides insight to improve the research studies and related tools. Our study in this paper addresses such a need. We propose a semi-automated approach that recovers code changes involving widespread changes in software systems. We further manually analyze more than nine hundred widespread changes recovered from eight software systems and categorize them into 11 families. These widespread changes and their associated families help us understand better why these widespread changes are made.