Fine-Grained Software Metrics in Practice

Author(s): Michael English, Jim Buckley, and Tony Cahill
Venue: Empirical Software Engineering and Measurement, First International Symposium on
Date: 2007

Type of Experiement: Case Study



A set of software metrics is proposed that are more fine grained than most metrics currently in use. These metrics are aimed at measuring the actual use of the friend construct, inheritance, and other forms of coupling. More specifically the metrics are based on the interactions for which each coupling mechanism is necessary or sufficient. The proposed metrics are then empirically evaluated using the LEDA software system. The results show that friend mechanisms are used to a limited extent to access hidden methods in a class, but used to a greater extent to access hidden attributes.

The fine grained metrics focus on measuring method-member, method-attribute, and method-method interactions. By doing it this way their results show that they are able to differentiate between the different types of coupling, which can in future work be used to find problem areas more specifically due to each of these fine-grained metrics.