基于双重维度发现代码变更模式的挖掘方法

在代码演变的历史中,识别出重复的代码变更,对工具开发者、项目开发者、代码研究者都大有裨益。本文在代码演变的历史信息中,通过抽取细粒度的变更,采用两重维度划分事务,应用频繁项集挖掘算法,提取出一些频繁的代码变更模式。并以真实项目为对象进行了实验分析,结果表明该方法能有效挖掘出项目演变历史中的频繁代码变更,检测出有意义的高层次程序转换。

Software Change Pattern Mining based on double dimensions

Identifying repetitive code changes from the history of code evolution, can benefit tool builders, program developers and code researchers. This paper employs the algorithm of change distiller to extract fine-grained changes from the code evolution history and then derives transactions using both the temporal and spacial dimentions. A frequent multiple instance itemset mining algorithm is designed and,applied to derive valuable high-level program change patterns. Experiment results on the real software projects illustrate that this approach can find the frequent code changes from the history of program evolution, revealing significant and high-level program transformations.