Tuesday, August 02, 2005
More on n-tuple testing
Yesterday evening I picked a journal among the too many waiting to be read :-) and found a short, interesting paper about all-pairs (more generally, n-tuple) testing:
Software Fault Interactions and Implications for Software Testing by D. Richard Kuhn, Dolores R. Wallace and Alberg M. Gallo Jr, IEEE Transactions on Software Engineering, June 2004. See also my previous post Brushing up slides on Testing for Testers for more.
The authors analyze a few real-world systems, and their results confirm the theory that testing all-pairs, and maybe all-triples, may be enough to find a large majority of errors.
However, we are still talking about a lot of test cases here! The implications on ideas like "write test cases first" and "I can safely refactor because I've created an adequate unit test suite" are not so trivial... :-)
Software Fault Interactions and Implications for Software Testing by D. Richard Kuhn, Dolores R. Wallace and Alberg M. Gallo Jr, IEEE Transactions on Software Engineering, June 2004. See also my previous post Brushing up slides on Testing for Testers for more.
The authors analyze a few real-world systems, and their results confirm the theory that testing all-pairs, and maybe all-triples, may be enough to find a large majority of errors.
However, we are still talking about a lot of test cases here! The implications on ideas like "write test cases first" and "I can safely refactor because I've created an adequate unit test suite" are not so trivial... :-)
