Research On Numerical Program Metamorphic Relations Construction And Optimization

Traditional software needs to build test oracles to verify whether outputs of the program are right or not.But when a program's input-output relation is complex,the tester is difficult or even impossible to construct the expected output.This is the Oracle problem in software testing.Metamorphic testing(MT)can effectively alleviate the Oracle problem,its verify software under testing by checking whether inputs and outputs satisfy metamorphic relation(MR).MR is the core of MT.In the process of MT,a large number of MRs will be constructed.However,if all MRs are used for testing,the efficiency of the MT will be seriously affected.Therefore,MR optimization is proposed.That is,select a small number of high-efficiency MR for testing from a large number of MR.The premise of the optimization is to measure the MR.The existing two standards,Fault Detection Ratio and the Mutation Score,are mainly used to measure the fault detection efficiency of a single MR,unable to select multiple MRs with great difference.The paper mainly studies on the MR construction method for numerical program,the optimization of MR and the related measurement of the diversity of MR.In the aspect of the construction of MR,we improved the method of constructing MR based on the equivalent class part ition,and proposed the method of constructing the metamorphic relation based on Oracle and the method of constructing the metamorphic relation based on the property of function.The experimental results show that the method can reduce the complexity of MR construction and ensuring the effectiveness of the MT.In terms of optimizing MR,two kinds of metrics are proposed in this paper for measuring the diversity of MRs.First,based on the Jaccard Distance from the perspective of test cases to measure the d iversity of MR(JDMR).Another method based on the Euclidean Distance from the perspective of mutations to measure the diversity of MR(DSMR).The experimental results show that both JDMR and DSMR can effectively measure the diversity of MRs and the JDMR is more intuitive than DSMR.Based on the JDMR,we studied the optimization strategy of MRs.The experimental results show that the selected MRs can effectively reduce the repetition rate,and JDMR also can be used in the study of composite MRs.