The ‘pkgcheck’ computational environment

Results of running pkgcheck package on your local computer may differ from results generated by rOpenSci’s automated checking system. The pkgcheck-action GitHub action provides a standard computational system for running pkgcheck, and can be used to trigger checks on every push to GitHub. Some packages may nevertheless depend on other system libraries, or packages from other programming languages, that are not included in the standard environment used by the pkgcheck-action. This vignette describes the computational environment for pkgcheck, including advice on how it may be modified to ensure your package passes rOpenSci’s automated checks.

The pkgcheck Docker container

Both pkgcheck-action and rOpenSci’s own checking system use the Docker container provided in the pkgcheck Dockerfile. This Docker environment builds on the GitHub “runner” images, with the first couple of hundred lines installing all system libraries needed to replicate that environment.

That Docker environment also includes installations of:

  • NixOS
  • The Julia language, along with the JULL installer to enable easy package installation.
  • A selection of pre-installed R packages, notably including several very large yet commonly installed packages.
  • The reticulate package, with an associated virtual python environment, and any python packages needed in rOpenSci’s suite of packages.
  • quarto to enable packages to use quarto vignettes, or documents in any other format provided by quarto.

How to get your package to pass rOpenSci’s ‘pkgcheck’ system

If you are preparing a package for submission, and it is failing pkgcheck-action because of missing software, whether system dependencies not otherwise covered in sysreqsdb, missing packages from either Julia or python, or some other system requirement, you can first modify your local pkgcheck-action workflow file to install these requirements as additional “steps”. Once you’ve got the pkgcheck-action passing on your local repository, you may also submit a pull request to update our own Dockerfile with whatever additional system dependencies your package may require.