An Empirical Study of Code Clone Clustering Based on Clone Evolution  

作　　者：Fanlong Zhang Xiaohong Su Wen Zhao Tiantian Wang 

机构地区：[1]School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China

出　　处：《Journal of Harbin Institute of Technology(New Series)》2017年第2期10-18,共9页哈尔滨工业大学学报（英文版）

基　　金：Sponsored by the National Natural Science Foundation of China(Grant No.61173021)

摘　　要：There are lots of code clones appearing in software,which are similar code fragments with each other. In the past decades,researchers have proposed some state-of-the-art methods to detect clones. The code clones have showing some relationship with the evolution of software. In order to explore relationships between clones and their evolution,we propose a framework to cluster clones with a Fuzzy C-means clustering method.Firstly,we detect all the clones using Ni Cad,and build the clone genealogies for multiple versions software.Secondly,we extract some metrics to describe the clones and their evolution. Finally,we cluster all clone's vectors,which are generated with the different metrics for different proposes. Experimental results on six open source software packages have shown the relationships among the clone life,the number of change times,the clone pattern and et al. can help developers to understand clones.There are lots of code clones appearing in software, which are similar code fragments with each other. In the past decades, researchers have proposed some state-of-the-art methods to detect clones. The code clones have showing some relationship with the evolution of software. In order to explore relationships between clones and their evolution, we propose a framework to cluster clones with a Fuzzy C-means clustering method. Firstly, we detect all the clones using NiCad, and build the clone genealogies for multiple versions software. Secondly, we extract some metrics to describe the clones and their evolution. Finally, we cluster all clone＇s vectors, which are generated with the different metrics for different proposes. Experimental results on six open source software packages have shown the relationships among the clone life, the number of change times, the clone pattern and et al. can help developers to understand clones.

关 键 词：code clones clone clustering clone analysis clone evolution empirical study 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]