M.P. Ware and F.G. Wilkie (UK)
Metrics, software maintenance, C++, software measurement.
This paper is concerned with an empirical validation of the use of software complexity metrics in determining the appropriate course of action when identifying overly complex components. A study is described which used complexity metrics drawn from a number of sources including the Chidamber and Kemerer (C&K) metric suite, general application specific descriptive statistics and maintenance related data. The metrics were applied to a commercial product implemented in C++. They were used in a linear study which tracked complex components across two major releases of the application. The study focused upon determining the effectiveness of redesign and the responses to the identified components when seeking to manage complexity. A discussion is presented which details the alternative courses of action available to a project manager on detection of complex classes. The conclusion to this paper includes general observations on our potential ability when seeking to control application complexity.
Important Links:
Go Back
