SynConSMutate: Concolic Testing of Database Applications via Synthetic Data Guided by SQL Mutants

Testing database applications using coverage analysis and mutation analysis

“…In this thesis, we leverage our basic approach [54] and develop a new approach [61] which automatically generates test cases and synthetic data if required. The generated data, in our approach, will help to improve both structural coverage and mutation score of the generated test cases.…”

“…In subsequent iterations, existing approaches generate more synthetic data for Table 5.1 and 5.2 to cover subsequent branch conditions at line10 and line12. Our work [61] focuses on generating synthetic data not only to improve structural coverage but also the mutation score, therefore generated test cases in our approach achieve high quality both in terms of coverage and mutation score. But none of these approaches (including ours) consider existing database states and the relationship among database variables, program inputs and branch conditions to generate test cases for the program.…”

“…DSE [20] or concolic testing [36] techniques can not solve branch conditions which depend on executed query result. Recent other techniques like [17,16,61] analyze previously executed path and generate synthetic data(s) to the existing database which can satisfy such branch conditions and improve the branch coverage. After executing this query against the current database table (see Table 5.1), we will get a range of values 5, 8, 9, 10 .…”

“…Otherwise, the generated synthetic data may become invalid with respect to the given database state, causing low quality test case generation for the database application. More detailed description of this module which has been leveraged from our previous work can be found in [61]. This will result only one value, i.e., packets = 10.…”

“…Since the branch condition depends on the result set returned by the queries at line2 and line8, Pex or Concolic testing techniques can not generate test cases for such branch conditions. Recent other techniques[16,17,61] analyze the queries executed in the previous execution, exploit the relationship between branch condition and query result conditions (W HERE clauses) and generate synthetic data so that the previously generated test case can satisfy such branch condition. Our approach, on the other hand, exploits therelationship among program input, query conditions, and branch conditions and checks the current database state to find any new test case which can satisfy such condition with respect to the current database state.…”