0-1 Programming Model-Based Method for Planning Code Review Using Bug Fix History

Code review is a powerful activity for software quality improvement, and is ideal to review all source files being developed. However, such an exhaustive review would be difficult because the available time and effort are in reality limited. Thus, practitioners prioritize the source files in terms of bug-proneness by using related data such as bug fix history, and review them in decreasing order of priority-such strategy in this paper is called the "conventional method." While the conventional method is straightforward, it focuses only on the bug-proneness and cannot consider the review cost properly, so the method does not produce a cost-effective review plan. To take into account both the bug-proneness and the review cost, this paper proposes a 0-1 programming model-based method for planning code review. The proposed method formulates a review plan as a 0-1 programming problem, and the solution is the recommendation list of source files to be reviewed. Moreover, the proposed method considers the type of file-if the file is newly-developed or not. Such difference in file type may affect on how to evaluate the bug-proneness and the review strategy: newly-developed files are notable but not appeared in the bug fix history. This paper conducts a case study using popular open source software, shows that the proposed method is up to 42% more effective than the conventional method in recommending buggy files as the review targets.