Coupling measures and change ripples in C++ application software (2000)

Author(s): F.G. Wilkie and B.A. Kitchenham
Venue: Journal of Systems and Software
Date: Jun 2000

Data Collection Method: Observation


The analyzed application was written in C++. It is a commercial multimedia conferencing
system. The application comprises approximately of 114 C++ classes with 1114 member
functions spread across 25,000 lines of code. Changes to the application were carefully
recorded over a 2.5 year period from its first release. The changes came from three main
sources: bug-driven, customer-driven, and developer-driven enhancements and changes.

Wilkie and Kitchenham investigate the applicability of the CBO metric by Chidamber and
Kemerer and its ability to predict classes involved in change ripples by attempting to draw
a correlation between change ripples and coupling between objects (CBO). They found that
CBO as a coarse measurement are weak indicators of classes that are likely to exhibit coupling
problems. They found that CBO in one of the tests was able to identify the most change-
prone classes but not the classes most vulnerable to ripple effect changes.