基于抽象语法树的数据泥团自动检测研究

摘要

Data Clumps is a common code smell, it will lead to issues such as duplicated code and increased difficulty in maintenance.As most existing code smell automatic detection tools fail to detect data clumps, and the detection type is not complete, a data clumps automatic detection based on abstract syntax tree is proposed.On the basis of existing detection tools, adding new types of data clumps to the algorithm, with some new steps as redundant data elimination and sub-data clumps extraction.Experiments are executed on 4 open source projects.Results show that the approach has high accuracy, and it is able to detect data clumps which other tools failed to detect, such as Stench Blossom, inFusion, etc.In addition, this approach has good efficiency and the execution time is directly proportionate to the size of system.%数据泥团是一种常见的代码味道,它将带来重复代码和维护难度增加等问题.针对大部分已有的代码味道自动检测工具无法检测数据泥团,且检测类型不全面等问题,提出一种基于抽象语法树的数据泥团自动检测方法.该方法在已有检测工具的基础上,增加了新的数据泥团类型,并加入了剔除冗余数据泥团和提取子数据泥团等步骤.通过对4个开源项目进行数据泥团实验,结果表明方法具有较高的精确率,与Stench Blossom、inFu-sion等工具的数据泥团自动检测功能相比,能够检测出一些其他工具无法检测的数据泥团.同时,该方法具有较好的性能,执行时间与系统规模成正比.