Maintainability and Source Code Conventions: An Analysis of Open Source Projects

Maintainability is a desirable property of software, and a variety of metrics have been proposed for measuring it, all based on different notions of complexity. Although these metrics are useful, complexity is only one factor influencing maintainability. Practical experience in software development has led to a set of best practices and coding conventions that are believed to make source code easier to read, understand and maintain. Based on a survey of software engineers, we identify the relative importance of 71 coding conventions to maintainability. We propose a metric that offers a different perspective on maintenance, namely a “convention adherence” metric based on the number and severity of violations of these coding conventions. We examine the code repositories of four open-source Java projects to measure their adherence to coding conventions over the life of the project, based on both their self-identified conventions and those of the convention-adherence metric. Through our analysis, we discovered several interesting phenomena, including pre-release effort to bring new code in line with desirable conventions, effective usage of automated code convention checkers as part of the build process to improve adherence, variations in adherence over the software lifecycle, and a class of conventions consistently ignored in open source projects.