How renv restores packages from r-universe for reproducibility or production

[This article was first published on rOpenSci - open tools for open science, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

This post is part of a series of technotes about r-universe, a new umbrella project by rOpenSci under which we experiment with various ideas for improving publication and discovery of research software in R. As the project evolves, we will post updates to document features and technical details. For more information, visit the r-universe project page.

What is renv

RStudio’s renv package is a powerful dependency management toolkit for R. It allows you to create a lockfile that records the exact versions of R packages used in a given project, and provides tooling to install exactly those same versions on another machine, or at a later point in time. This is very useful to create an isolated project environment for reproducibility or production purposes.

The latest version of renv (0.15.0) now supports restoring packages that were installed from r-universe. In this post we explain how this works, and why it has to work this way.

Why r-universe is not an archive

The letter A in CRAN stands for ‘archive’: the source code for all versions of all packages that were ever released on CRAN are stored indefinitely on an ever-growing ftp server. The CRAN homepage for each package has a link to “Old sources” which lists all the source packages (with timestamps) in the archive, some dating back over 20 years. This is a fantastic service, but quite expensive. For this reason CRAN has to limit the size of source packages and release frequency per package.

A fundamental difference between r-universe and CRAN is that r-universe has rolling development model. The system automatically tracks the upstream git repositories and continuously builds and deploys from the latest commits, sometimes many times per day per package. We do not archive all of these versions in r-universe, but instead rely on the upstream git repositories to provide historical sources.

Because packages published in r-universe are by definition built from a public git server, R and renv can treat such packages the same way as packages installed directly from git using e.g. install_github(). Effectively, r-universe serves a gateway to content on git, and if we know the upstream git-url and commit-id, it is unimportant how the package was installed exactly.

How it works

For every R package on r-universe, the build system adds some extra fields to the DESCRIPTION file with details on where the package was installed from. This includes the cranlike repository on r-universe, and the upstream Git url, branch and commit that the package was built from.

For example, have a look at the DESCRIPTION file from the gert package:

# Download gert source package and extract the DESCRIPTION file
gert <- download.packages('gert', '.', type = 'source', repos = '')
untar(gert[,2], 'gert/DESCRIPTION', tar = 'internal')

The following fields were added by r-universe:

## ...
## [30] "Repository:"
## [31] "RemoteUrl:"
## [32] "RemoteRef: HEAD"
## [33] "RemoteSha: b9dcd4c00449702cf70e782dc28eabf80c8d5971"

We can also retrieve the same data though the r-universe APIs, for example:

pkg <- jsonlite::fromJSON('')
## [1] "b9dcd4c00449702cf70e782dc28eabf80c8d5971"

Now this gives us an unambiguous reference to the exact source code, in this case:

This provides renv and similar tools with the required information for installing precisely this same version on another machine. There is no way to cheat: any change in source code will result in a different sha, even if the version number in the description file stays the same.

How renv restores from r-universe

flow chart from upstream git server to r

Depending on where a package originates from, renv uses different methods for finding an older source. For CRAN packages, it uses the CRAN archive, and for packages installed with install_github() it will checkout the old version from git based on the commit hash.

For r-universe it uses a hybrid approach: it first checks if the required version is available from the cranlike repository that was mentioned in the package Repository field, i.e. for the gert example above. The benefit of using the cranlike repository is that binary packages are available for Windows and Mac, which will ease the installation a bit.

If the package is no longer available on r-universe, renv will clone and checkout the source code from the RemoteUrl and RemoteSha values, in the same way as for packages installed with install_github().

To leave a comment for the author, please follow the link and comment on their blog: rOpenSci - open tools for open science. offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)