Abstract
Context: There are many claimed advantages for the use of design patterns and their impact on software quality. However, there is no enough empirical evidence that supports these claimed benefits and some studies have found contrary results.
Objective: This empirical study aims to quantitatively measure and compare the fault density of motifs of design patterns in object-oriented systems at different levels: design level, category level, motif level, and role level.
Method: An empirical study was conducted that involved five open-source software systems. Data were analyzed using appropriate statistical test of significance differences.
Results: There is no consistent difference in fault density between classes that participate in design motifs and non-participant classes. However, classes that participate in structural design motifs tend to be less fault-dense. For creational design motifs, it was found that there is no clear tendency for the difference in fault density. For behavioral design motifs, it was found that there is no significant difference between participant classes and non-participant classes. We observed associations between five design motifs (Builder, Factory Method, Adapter, Composite and Decorator) and fault density. At the role level, we found that only one pair of roles (Adapter vs. Client) shows a significant difference in fault density.
Conclusion: There is no clear tendency for the difference in fault density between participant and non-participant classes in design motifs. However, structural design motifs have a negative association with fault density. The Builder design motif has a positive association with fault density whilst the Factory Method, Adapter, Composite, and Decorator design motifs have negative associations with fault density. Classes that participate in the Adapter role are less dense in faults than classes that participate in the Client role. (C) 2015 Elsevier B.V. All rights reserved.