Cohesion between size of commit and type of commit

Abstract

Predicting what type of work was performed by developers in OSS has attracted several researchers. Previous studies have shown that fixing bugs are usually done by modifying very few files. In this study, we try to predict the trends and characteristics of how developers commit source code to version control systems (SVN, CVS, etc.). We want to investigate if number of files that are committed together is an indication of type of activity. In other words, is number of files in a commit related to the type of activity? Interestingly, we found that a strong correlation exits between type of activity and number of files involved in a commit. To make sure our study is not biased for a specific project, we have analyzed ten projects. Parsing messages in commits and processing the results, we found in coherent with previous studies bug fixing involves few files in a commit and commits that affect several files are usually related to adding functionality. © 2012 Springer Science+Business Media B.V.