摘要
针对回归测试任务,提出了一种新颖的测试用例优化方法,旨在解决现有方法未同时考虑代码单元的重要性和测试用例的覆盖率,也未充分利用源代码丰富的语义和结构信息等问题。该方法采用关系图卷积网络(RGCN)以有效处理现有缺陷预测方法对代码关系学习不足的问题,并通过细粒度变更影响分析以更精确地识别受变更影响的代码单元。通过融合缺陷预测和变更影响分析信息评价代码单元的重要性,从而对测试用例进行初步排序和过滤。最后,通过代码单元和测试用例的语义相似度计算覆盖率,对测试用例作进一步排序,以实现对测试用例的多重排序。在Defect4J数据集上的实验结果表明,该方法在APFD等评价指标上均优于对比方法,验证了其在测试用例优化方面的有效性。
For the regression testing task,a novel test case optimization method is proposed,aiming to solve the problem that existing methods do not consider the importance of code units and the coverage of test cases at the same time,and do not make full use of the rich semantic and structural information of the source code.This method employs Relational Graph Convolutional Network(RGCN)to effectively deal with the problem of insufficient learning of code relationships in existing defect prediction methods,and identifies the code units affected by changes more accurately through fine-grained change impact analysis.The importance of code units is evaluated by fusing the defect prediction and change impact analysis information,so that test cases can be initially sorted and filtered.Finally,coverage is calculated by semantic similarity between code units and test cases to further rank test cases for multiple ranking of test cases.The experimental results on Defect4J dataset show that the method outperforms the comparison method in evaluation metrics such as APFD,which verifies its effectiveness in test case optimization.
作者
王浩翔
侯霞
陆佳鋆
WANG Haoxiang;HOU Xia;LU Jiayun(Computer School,Beijing Information Science&Technology University,Beijing 100101,China)
出处
《软件导刊》
2025年第8期104-111,共8页
Software Guide
基金
北京市自然科学基金青年项目(4224090)。
作者简介
王浩翔(2000-),男,北京信息科技大学计算机学院硕士研究生,研究方向为智能信息处理;通讯作者:侯霞(1976-),女,博士,北京信息科技大学计算机学院教授、硕士生导师,研究方向为智能信息处理、信息技术标准化;陆佳鋆(1999-),女,北京信息科技大学计算机学院硕士研究生,研究方向为智能信息处理。
