Research On Test Case Generation Technology Of Combinatorial Testing

Software testing running throughout the entire software development life cycle isan important method to ensure software quality. During the testing process, exhaustivetesting is impossible, but we hope to achieve the completeness of testing withoutredundancy during testing. Although equivalence partitioning, boundary value analysisand other methods manage to reduce the number of test cases, it will still generate a lotof test cases in large or complex system. Therefore, according to testing requirements, itis the key for successful testing to develop a reasonable test strategy and create anappropriate scale test case set.The combinatorial testing has the ability to ensure the rate of detecting failurestriggered by the interaction of certain factors via fewer test cases. Research shows thatthe combinatorial testing has gained significant interest in recent years both in researchand practice, but there are still some issues need to be resolved. Such as, the problem ofgenerating a minimum test sets for combinatorial testing is NP-complete, currentmethods need to be improved; the complexity of practical system causes the invalidityof the traditional method. This thesis conducts the study with the aim of improving theefficiency and applicability of the test case generation method for the combinatorialtesting.First of all, this paper introduces the study situation on combinatorial testing andanalyses、contrasts and summarizes the current method. The method based on one test ata time strategy or in parameter order strategy always generates near-minimum test caseset which can be improved. This paper proposes a test case generation strategy: one pairat a time. It creates test case set by the unit of pair. Under the guidance of the strategy, acontribution based method to generate test case for combinatorial testing is proposedhere. It firstly creates a set of ordered pair, based on which an initial test case set isgenerated, and finally reduces the number of test case set. The experimental resultshows that the method is simple and effective, and it has the characteristic of fewer testcases. It also can be extend to handle N-way testing and testing with seeds.Secondly, considering interdependence and constraints are exist among parametersof the system input model in some practical project, this paper has analyzed the impactof constrains on test case generation and proposed a method to create test case set withconstrains. It firstly extracts the implied constraints from the system constrains set, and then creates the test case set by modified horizontal expansion and vertical expansionmethod. The results show that the constraints can be effectively handled. Compared totraditional IPO method, it can reduce the steps and be more efficient.Finally, on the basis of the approach mentioned above, this dissertation designs acombinatorial testing framework and implements it by a prototyping tool. It cangenerate fewer test cases automatically and be able to handle constrains in practicalsystem.