Hadoop集群下的并行克隆代码检测  

Parallel clone code detector in Hadoop cluster

作　　者：叶林[1] 姚国祥[1] 

机构地区：[1]暨南大学信息科学技术学院,广东广州510632

出　　处：《微型机与应用》2014年第2期69-71,74,共4页Microcomputer & Its Applications

基　　金：国家自然科学基金项目(61272415;61272413;61133014)

摘　　要：克隆代码会导致项目的维护困难,削弱项目的健壮性,并且克隆代码中所包含的bug会破坏整个项目。当前克隆代码检测技术或者拘泥于只能检测少数几种克隆代码,或者需要极高的检测时间。而且如果需要检测大量的源代码,一台机器的主存也许无法存储所有的信息。对克隆代码检测技术的并行运行进行了可能性研究,使用基于程序依赖图的克隆代码检测技术,这种技术不仅可以检测出语法上的克隆,也可以检测出语义上的克隆,提出了一个并行子图同构检测方法并使用MapReduce并行实现,实验结果极大地提高了该方法的运行速度。Clone codes make the project hard to maintain and weaken the robustness , and the bugs in these codes would un-dermine the whole project . The state-of-the-art clone code detectors are either not able to find code with same semantics , or computationally expensive . And if clone code detector is to be performed on plenty number of code , the main memory of one ma-chine may not able to hold all the information . In this paper we focus on the parallel of the clone code detector , we utilize the program dependence graph ( PDG )-based code clone detection method , which can not only check the code in contiguous syntax , but also the code with the same semantics . We present an approach to parallel the isomorphism matching in the PDG . By using MapReduce paradigm , we dramatically enhance the searching speed of this method .

关 键 词：克隆代码 程序依赖图 同构匹配检测 HADOOP 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP3[自动化与计算机技术—计算机科学与技术]