Research Of Commit Behavior In Open Source Software Repository SVN

Open source software lead a trend of development for its efficient team work andmanagement and Subversion (SVN) have been widely used in open-source software(OSS) development. Mining the commit behavior in SVN repository is an effectiveway to analyze the features of open-source software, which can provide a new insightinto schedule planning for OSS projects based on developers’ historical commitbehavior.As we know, few of prior studies investigated the distribution of inter-committimes (known as commit intervals) and commit classification combining with release,submitter and commit size. To gain a better understanding on OSS developmentprocesses, we conducted an empirical study on four representative projects and mainlyfound two aspects as follows:(1)The distributions of commit intervals in the four projects in question roughlyfollow power laws, with commit bursts (i.e., the revisions in a SVN repository areupdated quickly over a period of time) and heavy tails. The OSS evolution maycontribute to commit bursts, while active committer’s individual behavior (such ashis/her tasks completion and illness) and long vacations are the primary factors thatresult in long inter-commit times, but new release is a weak factor.(2) Introduce TF–IDF to complete the commit classification and the distributiondiffers a lot in the four projects due to the different evolution degree. Most of thecommits tend to have only one or a few files to be modified, added or deleted.Volunteer work is mainly for error modification, which is important for qualityassurance.