Evaluating the effects of similar-class combination on class integration test order generation

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review

View graph of relations

Related Research Unit(s)


Original languageEnglish
Article number106438
Journal / PublicationInformation and Software Technology
Online published25 Sep 2020
Publication statusOnline published - 25 Sep 2020


Context: In integration testing, the order in which classes are integrated and tested is significant for the construction of test stubs. With the existing approaches, it is usually difficult to generate the sub-optimal test orders for real applications, which have large numbers of classes. Objective: There exist moderately large numbers of classes in software systems, which is one of the main factors that complicate the generation of class integration test order (CITO). The main objectives of this study are reducing the problem space for CITO generation, and minimizing the stubbing cost of the generated test orders. Method: The approach proposed in this study is based on the hypothesis that similar-class combination can remove class dependencies and yield a smaller problem space. Identical class dependence and symmetric classes are the two main properties that are used to identify similar classes. In addition, a new cycle-breaking algorithm is introduced to minimize the stubbing cost of the generated test orders, which fully considers the two factors (number of test stubs and the corresponding stubbing complexity) that affect the overall stubbing cost. Empirical experiments are conducted on nine open-source Java programs to evaluate the performance of the proposed approach. Results: With similar-class combination, the proposed approach reduced the numbers of classes and class dependencies by over 10% and 6%, respectively, for six programs. Moreover, for four programs, the proposed approach reduced the number of cycles among class dependencies by more than 20%. The cycle-breaking algorithm achieved reduction of more than 13% in the stubbing cost, thus outperforming other competing techniques. Conclusions: The proposed method relies on the two aforementioned important properties to identify similar classes, and these properties are known to significantly improve the performance of CITO generation. The results obtained in this study confirmed the capability of the proposed approach in terms of minimizing the number of classes and class dependencies in programs. It outperformed other competing methods in minimizing the stubbing costs of the generated test orders.

Research Area(s)

  • Class integration test order, Integration testing, Similar classes, Stubbing complexity