A Methodology for Test Selection

Software creation requires not only testing during the development cycle by the development staff, but also independent testing following the completion of the implementation. However in the latter case, the amount of testing that can be carried out is often limited by time and resources. At the very most, independent testing can be expected to provide 100% test coverage of the test requirements (or specifications) associated with the software element with the minimum of effort. This paper describes a methodology employing integer programming by which the amount of testing required to provide the maximum possible test coverage of the test requirements (for the given test set) is assured while at the same time minimizing the total number of tests to be included in a test suite. A collateral procedure provides recommendations on which tests might be eliminated if less than 100% test coverage of the test requirements is permitted. This latter procedure will be useful in determining the risk of not running the minimum set of tests for 100% test coverage. A third process selects from the test matrix the set of tests to be applied to the system following maintenance modification of any test requirements-- that is, to provide a submatrix for regression testing. The potential benefits for applying the integer programming technique in test data selection is also discussed