Correlations between Internal Software Metrics and Software Dependability in a Large Population of Small C/C++ Programs

Abstract

Software metrics are often supposed to give valuable information for the development of software. In this paper we focus on several common internal metrics: Lines of Code, number of comments, Halstead Volume and McCabe's Cyclomatic Complexity. We try to find relations between these internal software metrics and metrics of software dependability: Probability of Failure on Demand and number of defects. The research is done using 59 specifications from a programming competition---The Online Judge--on the internet. Each specification provides us between 111 and 11,495programs for our analysis; the total number of programs used is 71,917. We excluded those programs that consist of a look-up table. The results for the Online Judge programs are: (1) there is a very strong correlation between Lines of Code and Hal- stead Volume; (2) there is an even stronger correlation between Lines of Code and McCabe's Cyclomatic Complexity; (3) none of the internal software metrics makes it possible to discern correct programs from incorrect ones; (4) given a specification, there is no correlation between any of the internal software metrics and the software dependability metrics.