A study on the relationship between FindBugs warnings, metrics and expert judgments

Abstract

The usage of static code analysis tools is one of the techniques to help to inspect the software quality. Many types of research have been made to evaluate such tools, but the investigate on what the analysis report further indicates and how these tools are applied in the industrial case is less concerned. This paper presents the study on three project cases in an IT Consultant company, where Findbugs, a bug pattern detection tool and SonarQube, a code quality monitoring tool are used in the development team. First, the correlation between 6 bug pattern categories and 5 source code metrics are investigated. The statistical analysis of the data extracted from tools has shown the most of the correlations are only specific to certain project. The result is partially different from expert judgments. Second, we decompose the maintainability into several characteristics, and a set of metrics measured by SonarQube are chosen to predict each of them on the basis of the practical experience. Third, the importance of metrics about test coverage and complexity is considered to be not constant among different types of projects.