A Study of the Application of Software Metrics to COBOL.

This report presents the results of a study of software metrics applied to various classes of COBOL programs. Particular attention is given to the software science metrics of Halstead, which were applied to hundreds of COBOL programs written by students at Ohio State University, as well as several production COBOL programs. The results include support of the inclusion of Data Division in the software science counting strategy, nonsupport for the use of the software science language level metric, and the identification of a weakness in the ability of the Halstead E measure to capture integration effort. Several proposed complexity metrics were compared for their ability to predict actual development effort, with none of the metrics studied behaving in an impressive manner. Some approaches for refining existing complexity metrics to overcome their apparent weaknesses are suggested.