Code Inspection Supported by Stepwise Abstraction and Visualization - An Experimental Study

Abstract

Background: In order to inspect source code effectively and efficiently, in a previous work the use of visualization for supporting the reading technique Stepwise Abstraction was proposed and implemented in the CRISTA tool. Visualization aids code comprehension, which is an essential task for a successful inspection. Goal: The objective of this paper is to evaluate the effectiveness and efficiency of using stepwise abstraction supported by visualization for defects detection, in comparison to an ad-hoc approach. Method: A controlled experiment was conducted with two groups of undergraduate students. One group inspected the Java source code of the Paint software using the approach implemented in CRISTA and the other group inspected the code using an ad-hoc approach. Results: The general performance of the subjects who used Stepwise Abstraction supported by visualization was better than that of the subjects who used the ad-hoc approach. Besides, the subjects' experience in inspection and Java did not influence the identification of defects. Conclusion: the results reveal that the use of Stepwise Abstraction and visualization promotes better performance in detecting defects than the ad-hoc approach. In future work, other approaches are being investigated as well as the support of the approaches for different types of defects.