DOMAIN SPECIFIC TEST CASE GENERATION USING HIGHER ORDERED TYPED LANGUAGES FOR SPECIFICATION

Abstract

Model based testing is an approach for automatic generation of test cases on the basis of a representative model of the system. Recent studies show that model based testing has many possible advantages over manual test generation techniques including a gain in effectiveness, efficiency and reuse.

The effectiveness (ability to uncover faults in a system) of a model based testing process is determined by the correctness of the model and by the number of requirements represented in the model. In practice, test models for model based test automation techniques are usually created from requirement or design specifications of the software and hence, these techniques overtly rely on such specifications for the completeness of the test models. This may lead to failure in testing some critical requirements specific to the application domain because the user, who helps in defining the requirements, may fail to consider certain domain specific requirements. To him some may appear to be too trivial to be specified explicitly in the requirements document and the others, he may forget. Even if the requirement is complete with domain specific requirements, testers may not realize criticality of such requirements or may find them too complex to model. In all such cases, testing is incomplete and ineffective.

This dissertation describes a new model based testing technique developed to remedy such situations. The new technique is based on modeling the system under test using a strongly typed domain specific language (DSL). In the new technique, information about domain specific requirements of an application are captured automatically by exploiting properties of the DSL and are subsequently introduced in the test model. The new technique is applied to generate test cases for the applications interfacing with relational databases and the example DSL chosen for that purpose is HaskellDB. Test suites generated using the new technique are enriched with test cases addressing domain specific implicit requirements and therefore, are more effective in finding faults.

This dissertation will present details of the technique and describe an experiment and a case study to explore its effectiveness, efficiency, usability and industrial applicability.