Temporal algebraic query of test sequences

Nowadays tools can generate test suites consisting of large number of test sequences. The used algorithms are typically random-based. Although more advanced variations may incorporate an advanced search algorithm to cover difficult scenarios, many decisions still have to be made randomly simply because no information is available to calculate the best decision. Because of this, many of the generated sequences may be redundant, while some others may be rare and hard to get. This paper presents a rich formalism that is based on a mix of algebraic relations and Linear Temporal Logic (LTL) to query test suites, and an efficient algorithm to execute such queries. Queries can be used as correctness specifications (oracles) to validate a test suite. They are however more general as they can be used to filter out test sequences with interesting properties, e.g. to archive them for future use. The proposed formalism is quite expressive: it can express algebraic equations with logical variables, Hoare triples, class invariants, as well as their temporal modalities. An evaluation of the query algorithm's performance is included in this paper. The whole query framework has been implemented in a testing tool for Java called T3i.