OBJECT-ORIENTED MEASURES AS TESTABILITY INDICATORS: AN EMPIRICAL STUDY

Software measurement is one of the management strategies for developing robust and maintainable software products. The complexities inherent in software design can be controlled using software metrics. Testability is one of the sub characteristics of the maintainability attribute of software and it is the desire of quality assurance to ensure that software components are easily testable. This study empirically investigates the suitability of software metrics as indicators of software testability. The case study approach to empirical software engineering research is used in the study. Data are collected from six open-source object-oriented software products. The data consist of fifteen metric measurements of Java classes and their respective Junit test cases. Statistical analysis is conducted to show the relationship between measurements of the classes and that of their test cases. The results of the analysis showed that the complexity and coupling metrics of the Java classes are suitable indicators of the testability of the classes of object-oriented software. Also, the magnitude of the relationship is observed to be weak, which implies that the metrics do not completely measure the level of difficulty in the task of developing test cases for classes; that is, some other factors involved in test case development are not captured by software metrics.