Minimal Docker images for R

What is R?

R is a free software environment for statistical computing and graphics. It compiles and runs on a wide variety of UNIX platforms, Windows and MacOS. See more at https://www.r-project.org/

Goals and features

The main goal of these images is to keep them minimal, so they can be used as part of a bigger (web) application, or as a base image. Currently the (R 4.1.0) r-minimal image is less than 21MB compressed, and 35.4MB uncompressed.

All images use Alpine Linux.

The images include the installr tools that can install R packages from CRAN or GitHub:

❯ installr -h
Usage: ./installr [ -c | -d ] [ -a pkgs ] [ -t pkgs ] [ -r ] [ -p ] REMOTES ...

Options:
  -c    install C and C++ compilers and keep them
  -d    install C and C++ compilers, temporarily
  -a    install Alpine packages and keep them
  -t    install Alpine packages, temporarily
  -p    do not remove pak after the installation (ignored if -r is given).

REMOTES may be:
  * package names from CRAN/Bioconductor, e.g.    ggplot2
  * slugs of GitHub repos, e.g.                   tidyverse/ggplot2
  * GitHub branch, tag or commit, e.g             tidyverse/[email protected]
  * URLs to package .tar.gz files, e.g.           url::https://x.com/pkg.tar.gz
  * path to a local directory, e.g.               local::.

Recent r-minimal containers use pak (https://github.com/r-lib/pak) for R packages installation. If you have problems with pak, or need to install a package from a source that pak does not support, but the remotes package does, then install the remotes package first.

Limitations

To keep the images minimal, they do not include a number of parts and features that most users would prefer to have for interactive R development:

• Recommended R packages are not installed.
• Documentation is not included.
• No X11 support.
• No Java support.
• No OpenMP support.
• No JPEG, PNG or TIFF support.
• No Cairo support.
• No Tcl/Tk support.
• No translations, only English.
• The image does not have C, C++ or Fortran compilers.
• Limited time zone data: GMT, UTC and America/New_York, see below if you need better time zone data.

Usage

Get the image from Docker Hub:

docker pull docker.io/rhub/r-minimal:latest

or from GitHub Packages:

docker pull docker.pkg.github.com/r-hub/r-minimal/r-minimal:latest

Platforms

All images are available on linux/amd64 and linux/arm64 platforms.

Supported R versions

Currently we support the last patch version of the last five minor R versions. The latest tag always uses the last R release.

Dockerfile examples

One of our main goals is to be able to use rhub/r-minimal as a base image, and easily add R packages from CRAN or GitHub to it, to create a new image. Run installr from a Dockerfile to add R packages to the r-minimal image:

FROM rhub/r-minimal
RUN installr praise
CMD [ "R", "--slave", "-e", "cat(praise::praise())" ]

Package with compiled code:

FROM rhub/r-minimal
RUN installr -d glue

After the package(s) have been installed, installr removed the compilers, as these are typically not needed on the final image. If you want to keep them use installr -c instead of installr -d.

Package with system requirements:

FROM rhub/r-minimal
RUN installr -d -t linux-headers pingr
CMD [ "R", "-q", "-e", "pingr::is_online() || stop('offline')" ]

Similarly to compilers, system packages are removed after the R packages have been installed. If you want to keep (some of) them, use installr -a instead of installr -t. (You can also mix the two.)

Popular packages:

Hints on installing some popular R packages:

See also the Dockerfiles in the examples directory.

Note that package and system dependencies change over time, so if any of these commands do not work any more, please let us know.

See the [examples/rmarkdown/Dockerfile] for installing pandoc.

Time zones

The image uses R’s internal time zone database, but most time zones are removed from, to save space. The only supported ones are GMT, UTC and America/New_York. If you need more time zones, then install Alpine’s time zone package and point R to it:

apk add --no-cache tzdata
export TZDIR=/usr/share/zoneinfo

See also the discussion at #24

Known failures and workarounds

• The ps package needs the linux-headers Alpine package at compile time. Many tidyverse packages depend on ps, so they’ll need it as well:

  installr -d -t linux-headers ps
  

• The arrow package are hard to install, because Alpine Linux does not have the required libraries. For the details, please see: #7

• The V8 package does not compile on Alpine Linux by default. Luckily, you can now download a static binary of V8:

  DOWNLOAD_STATIC_LIBV8=1 installr -d -t curl-dev V8
  

• The prophet package depens on V8, through rstan, and you can use the same trick to install it:

  DOWNLOAD_STATIC_LIBV8=1 installr -d -t "linux-headers gfortran curl-dev" prophet
  

• The readxl package does not compile on Alpine Linux currently. You can install this branch from GitHub:

  installr -d gaborcsardi/readxl@fix/alpine-linux
  

• The tidyverse package depends on readxl, so you’ll need to do the same:

  installr -d -t "curl-dev libxml2-dev linux-headers gfortran" \
      -a "libcurl libxml2" gaborcsardi/readxl@fix/alpine-linux tidyverse
  

• To install the magick package, you need both the imagemagick and imagemagick-dev Alpine packages, both at install time and run time:

  installr -d -a "imagemagick imagemagick-dev" -t "curl-dev" magick
  

License

See https://www.r-project.org/Licenses/ for the R licenses

These Dockerfiles are licensed under the MIT License.

© R Consortium