Model-based Testing for SQL Databases

Sammanfattning: Database testing is an expensive and time consuming process, often performed with manually written test code. Automating database testing involves test case generation and test oracles which examine and verify the results. Previous work for database testing has shown a wide variety of techniques and testing tools, including specification languages and symbolic execution. In this thesis, we present a novel approach for SQL database testing. We use Modbat, an open-source tool for model-based testing which was recently used to test complex systems such as Apache Zookeeper. Our approach includes the test model, which generates test cases and builds queries in order to test the database, and the test oracle which investigates the results from the database and compares them with the expected results. We evaluate the effectiveness of our approach by using a number of metrics: execution time, test code and model size, human effort, test coverage and defect detection efficiency.