Graduate School of Information Science, Nara Institute of Science and Technology
Master of Engineering
October 2012 - March 2014
Peer code review is an important quality assurance mechanism which widely used in both industrial and Open Source Software (OSS) software development projects. Effectively performing code review can increase the quality of software; but it also requires reviewers with experiences and a deep understanding of system code. However, in distributed software development projects, manually finding an appropriate reviewer is difficult and time-consuming. In order to ease developers, this thesis proposes a reviewer recommendation algorithm, called FPS algorithm, based on locations of changed files. FPS algorithm finds appropriate reviewers for a new patch by determining the similarity of its file paths with file paths of previous patches. In the experiment, FPS algorithm was evaluated using three OSS projects and compared with the existing reviewer recommendation algorithm. Moreover, this thesis studies factors that impact to the reviewer recommendation algorithms. Specifically, the impact of file path structure and time priority were investigated. The results indicate that the performance of FPS algorithm is significantly better than the existing reviewer recommendation algorithm. The results also show that file path structure has impact to the FPS algorithm while the time prioritzation cannot improve the performance.