Constructing Input Parameter Models from Sample Data

Black-box testing is nowadays unavoidable, especially when it comes to system or acceptance testing. The main problem in testing a function or system “completely” is the fact that testing the large number of possible input value combinations is practically unfeasible. One successful strategy to reduce the number of test cases systematically, while maintaining the goal of ensuring high-quality software, is input domain modelling. The artefact yielded by this approach is called input parameter model (IPM). The main idea of it is to partition each parameter’s domain, which form together the input domain, into equivalence classes. The elements of one equiv. class are equal in the sense that each element of them is as good as any other element for testing purposes. This will ensure that the tester only needs to test one representative of each equivalence class instead of all elements of the respective domain and thus input domain modelling reduces the number of test cases significantly. Since concepts for tool-supporting the creation of an IPM exist only barely and do not facilitate direct feedback to the user, this thesis introduces a completely new approach to the IPM creation by connecting it with sample data. This thesis points out how the tester should be assisted by discussing key properties of an IPM and problems in input domain modelling. To verify the concept, a web app named CONTROLLED was implemented with which it could be shown that through the concept substantial many violations of key properties in the created IPM could be spotted which would be otherwise not found. An IPM created with the tool will have an increased quality and may result in better test cases.