A Study on the Relation Between Antipatterns and the Cost of Class Unit Testing. Saban�, A., Di Penta, M., Antoniol, G., & Gu�h�neuc, Y. In Cleve, A. & Ricca, F., editors, Proceedings of the 17<sup>th</sup> European Conference on Software Maintenance and Reengineering (CSMR), pages 167–176, March, 2013. IEEE CS Press. 10 pages.
A Study on the Relation Between Antipatterns and the Cost of Class Unit Testing [pdf]Paper  abstract   bibtex   
Antipatterns are known as recurring, poor design choices; recent and past studies indicated that they negatively affect software systems in terms of understandability and maintainability, also increasing change-and defect-proneness. For this reason, refactoring actions are often suggested. In this paper, we investigate a different side-effect of antipatterns, which is their effect on testability and on testing cost in particular. We consider as (upper bound) indicator of testing cost the number of test cases that satisfy the minimal data member usage matrix (MaDUM) criterion proposed by Bashir and Goel. A study—carried out on four Java programs, Ant 1.8.3, ArgoUML 0.20, CheckStyle 4.0, and JFreeChart 1.0.13—supports the evidence that, on the one hand, antipatterns unit testing requires, on average, a number of test cases substantially higher than unit testing for non-antipattern classes. On the other hand, antipattern classes must be carefully tested because they are more defect-prone than other classes. Finally, we illustrate how specific refactoring actions—applied to classes participating in antipatterns—could reduce testing cost.

Downloads: 0