A Value Weighting Method for Pair-wise Testing

Abstract

In this paper, we propose a weighting method for pair-wise testing. Pair-wise testing is a software testing strategy that tests every pair of parameter values at least once. Weighting allows the tester to specify desired frequency of occurrence in a test suite for each parameter value. Pair-wise testing is a widely used strategy because of its effectiveness in finding faults in software. Weighting makes this strategy more effective by allowing the tester to have more control over the resulting test suite. However, there is not much research on weighting. To our knowledge, all existing weighting methods treat weights as a second class requirement and cannot generate a test suite that sufficiently respects the given weights. The proposed method aims to overcome the problem. By taking into consideration the parameter value distribution in the current test suite and the ideal one specified by the given weights, the method generates each test case so that the resulting test suite can reflect the weights as accurately as possible. We implement the method in our testing tool and show some results to demonstrate how accurately test suites generated by this tool satisfy given weights.