AUTOMATED SQL QUERY GENERATION - RDBMS Testing
Fadi Marouki · DiVA at Umeå University (Umeå University) · 2019
Manually writing SQL queries for Relational Database Management Systems (RDBMS) testing can be very tedious depending on the database complexity. The focus of this thesis is to develop three approaches to automatically generate SQL queries based on a given database instance. These queries can then be used to evaluate the configuration of a RDBMS. The generated queries are only partial components in RDBMS testing. However, they do reduce the amount of work required to perform such configuration assessment. The three presented approaches generate well-formed and semantically meaningful queries (i.e. queries with no logical contradictions). The first approach only consists of a context-free grammar (CFG). The second uses a CFG with an exclusion list. The third uses a CFG with a binary classification machine learning model. The results show that the binary classification algorithm approach outperforms the other two in terms of generating a higher proportion of semantically meaningful queries.