DEVELOPMENT AND VALIDATION OF AN ADEQUACY BASED TEST DATA GENERATION ALGORITHM

Abstract

The length and the complexity of the software are rising day by day. This rising complexity has increased the demand for techniques that can generate test data effectively. Test data generation techniques selects from the input domain of the program, those input values that satisfies a pre-defined testing criteria. In this work, we propose a new test data generation algorithm. Our algorithm generates test data using adequacy based testing criteria that aims to generate an adequate test data set by using the concept of mutation analysis. In general, mutation analysis is applied after the test data is generated. But, in this work, we propose an algorithm that applies mutation analysis at the time of test data generation only, rather than applying it after the test data has been generated. Incorporation of mutation analysis at the time of test data generation leads to the generation of test data that is itself adequate and hence we need not check for its adequacy after its generation. This saves significant amount of time (required to generate adequate test cases) as compared to other techniques that applies mutation analysis after the generation of test data as the total time in these techniques is the sum of the time to generate test data and the time to apply mutation analysis. We also use genetic algorithms that explore the complete domain of the program to provide near-global optimum test data set. The use of genetic algorithms is also facilitated by the fact that many of the testing problems can be formulated as search problems. In order to analyze our algorithm, we evaluate it using fifty real time programs written in C language. The program set contains programs ranging from 35 to 250 lines of source code and includes from very basic to very complex programs. We compare our algorithm with path testing and condition testing techniques (that uses reliability based testing criteria) for these fifty programs in two categories viz. number of generated test cases and the time taken to generate test cases. The results suggest that our adequacy based algorithm is better than the reliability based path testing and condition testing techniques in both of these categories. Thus this algorithm may significantly reduce the time of test data generation.