Towards a Sufficient Set of Mutation Operators for Structured Query Language (SQL)

Test suites for database applications depend on adequate test data and real-world test faults for success. An automated tool is available that quantifies test data coverage for database queries written in SQL. An automated tool is also available that mimics realworld faults by mutating SQL, however tests have revealed that these simulated faults do not completely represent real-world faults. This paper demonstrates how half of the mutation operators used by the SQL mutation tool in real-world test suites generated significantly lower detection scores than those from research test suites. Three revised mutation operators are introduced that improve detection scores and contribute toward redefining a sufficient set of mutation operators for SQL. Finally, a procedure is presented that reduces the test burden by automatically comparing SQL mutants with their original queries.