On the maintainability of CRAN packages

Abstract

When writing software, developers are confronted with a trade-off between depending on existing components and reimplementing similar functionality in their own code. Errors may be inadvertently introduced because of dependencies to unreliable components, and it may take longer time to fix these errors. We study such issues in the context of the CRAN archive, a long-lived software ecosystem consisting of over 5000 R packages being actively maintained by over 2500 maintainers, with different flavors of each package depending on the development status and target operating system. Based on an analysis of package dependencies and package status, we present preliminary results on the sources of errors in these packages per flavor, and the time that is needed to fix these errors.