The Reduction Theory And Application Of Mutants Based On The Semantic Correlation

Any software must be implemented extensive testing to ensure the reliability before put into use. As a fault detection test method, mutation can be used not only to generate high quality test data, but also used to evaluate error detection capability of existing test data. The first problem that needs to solve for mutation testing is using certain mutation operators to generate the corresponding mutants. However, the number of mutants generated by the conventional method is extremely large, resulting in a high cost of testing; in addition, applying conventional methods to generate test data is difficult to kill these mutants because the semantics of the original program is very close to the mutant. Therefore, how to use a targeted theory and method to generate as little as possible and better performance mutant needs to be resolved on test areas.The difficulty of killing a mutant depends on the statement correlation that between the mutant and the original program. In view of this, this paper studies the theory and method of mutation reduction based on semantics correlation. Through research, we expect to obtain semantic correlation distribution between the mutated statement and the original one for each mutant operator, and according to this to reduce mutation operators, thus greatly reduce the number of mutants, and improve the efficiency and quality of the mutation testing.Firstly, through statistics semantics correlation we obtain the distribution the five mutation operator classes. To solve this problem, first of all, giving the quantization concepts on the semantic correlation between the statement before mutation and after mutation; then, semantics correlation of mutation operators are analyzed by statistics methods and static analysis strategy; finally, applying MATLAB software to get the fitting curve and its distribution function expression of semantic correlation for these operators.Secondly, according to the semantics correlation we reduce the mutation operators, and therefore reduce the number of mutants. Firstly, using the theory of fuzzy mathematics we divide semantic correlation into four grades; then, in accordance with each correlation function expression we select the probability of each level of mutation operators, the higher the level the greater the probability of selection,and the smaller on the contrary; finally, validity of the method is evaluated through the mutant scores of the mutants before and after the reduction on the experiment.Finally, we sum up the work in this paper and give the further direction ofresearch.This paper proposes mutants reduction theory based on semantic correlation for the existed difficulties of mutation testing that reduces the number of mutants by the semantic correlation of mutant operators, and is applied to the mutation testing of practical software. The results of the experiment can reduce the number of mutants significantly, improves the efficiency of mutation testing greatly, and therefore have important theoretical and practical values.