Possible Directions for Improving Dependency Versioning in R

Abstract

One of the most powerful features of R is its infrastructure for contributed code. The built-in package manager and complementary repositories provide a great system for development and exchange of code, and have played an important role in the growth of the platform towards the de-facto standard in statistical computing that it is today. However, the number of packages on CRAN and other repositories has increased beyond what might have been foreseen, and is revealing some limitations of the current design. One such problem is the general lack of dependency versioning in the infrastructure. This paper explores this problem in greater detail, and suggests approaches taken by other open source communities that might work for R as well. Three use cases are defined that exemplify the issue, and illustrate how improving this aspect of package management could increase reliability while supporting further growth of the R community.

An introduction to Docker for reproducible research

“…These challenges can be significantly reduced because Docker defines the software environment to a particular operating system and suite of libraries, such as the Ubuntu or Debian distribution. Such distributions use a staged release model with stable, testing and unstable phases subjected to extensive testing to catch such potential problems [20], while also providing regular security updates to software within each stage. Nonetheless, this cannot completely avoid the challenge of code-rot, particularly when it is necessary to install software that is not (yet) available for a given distribution.…”

Enhancing Reproducibility and Collaboration via Management of R Package Cohorts

“…Updating package versions within a library mid-project can be costly, potentially involving the regeneration of results or the modification of analysis or package code (Ooms 2013). Freezing package versions over the course of long-running projects, however, introduces a risk of generating incorrect results which would have been correct if generated using up-to-date software.…”

“…Collaborators often need to synchronize package versions to guarantee comparability of their results (Gentleman and Temple Lang 2004) (RStudio Inc 2014) (FitzJohn, Pennell, Zanne, and Cornwell 2014). A package author might switch between contexts when maintaining multiple software branches, or when alternating between development and analysis work (Ooms 2013) (Gentleman, Carey, Dudoit, Ellis, Gautier, Gentry, Huber, Irizarry, Rossini, Smyth, and others 2003). Large collaborative or enterprise organizations might formally support this synchronization by automating the testing and publication of a canonical cohort of package versions (Revolution Analytics 2014).…”

maintaineR: A Web-Based Dashboard for Maintainers of CRAN Packages

“…In addition, limitations of R's dependency versioning system have been reported and possible directions for improvement (such as staged package distributions and versioned package management) have been proposed [2]. Another problem is a lack of coordination between maintainers of dependent packages.…”