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 4.4.1) r-minimal image is 23.74MB compressed, and 45.88MB 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 ] [ -e ] [ -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).
  -e    use renv to restore the renv.lock file if present.

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/ggplot2@v1.0.0
  * 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 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 OpenMP support. (But you can configure it for a package, see examples/data.table.)
• 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 ghcr.io/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.

image	R version	tags	note
R devel	4.5.0-devel	devel, 4.5.0, 4.5, 4.5.0-devel, 4.5-devel, 2024-11-10	Built daily
R next	4.4.2-Patched	next, patched, 4.4.2-patched, 4.4-patched	Built daily
R release	4.4.2	4.4.2, 4.4, release, latest
R 4.3.x	4.3.3	4.3.3, 4.3
R 4.2.x	4.2.3	4.2.3, 4.2
R 4.1.x	4.1.3	4.1.3, 4.1
R 4.0.x	4.0.5	4.0.5, 4.0
R 3.6.x	3.6.3	3.6.3, 3.6

Daily R-devel builds

We tag our daily R-devel builds with the build date. These images can be useful for tracking down bugs and regressions in R. E.g.:

docker run -ti ghcr.io/r-hub/r-minimal/r-minimal:2022-11-25

The tags start on 2021-07-25. linux/arm64 images are available from 2022-01-13.

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.)

Using with renv:

To use renv to restore the renv.lock file, use the -e option:

FROM rhub/r-minimal
COPY .Rprofile .Rprofile
COPY renv renv
COPY renv.lock .
RUN installr -d -e

If you copy the entire folder with renv, including the activate.R and .Rprofile, renv will bootstrap itself with the same version as the lock and restore the packages with the proper versions. All the necessary compilers and libraries needed at runtime need to be installed with the -a and -t options. Please refer to examples/renv-shiny for an example that install shiny and rmarkdown in a container.

Popular packages:

Hints on installing some popular R packages:

package	installr command	~ image size (uncompressed)
data.table	See examples/data.table for OpenMP support	26.2 MB (50.0 MB)
dplyr	installr -d dplyr	31.9 MB (59.7 MB)
ggplot2	installr -d -t gfortran ggplot2	56.1 MB (93.5 MB)
h2o	See examples/h2o.	354.0 MB (511.0 MB)
knitr	installr -d knitr	25.3 MB (48.3 MB)
shiny	See examples/shiny.	49.8 MB (103.5 MB)
sf	See examples/sf.	79.7 MB (194.8 MB)
plumber	See examples/plumber.	56.3 MB (127.7 MB)
rmarkdown	installr -d rmarkdown	58.1 MB (153.5 MB) (with pandoc)
tidyverse	See examples/tidyverse.	113.7 MB (208.6 MB)
rstan	See examples/rstan.	92.0 MB (298.5 MB)
xgboost	installr -d -t "gfortran libexecinfo-dev" -a libexecinfo xgboost	35.5 MB (71.1 MB)

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 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 needs a Makevars file to add a link flag. See the example Dockerfile in the examples/arrow directory.

• The V8 packagees do not compile on aarch64 machines by default. On x86_64 it installs fine:

  installr -d -t curl-dev V8
  

  This means that other packages that need V8 (e.g. rstan and prophet) do not work on aarch64, either.

• 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
  

• pingr 2.0.4 does not build on Alpine, use the dev version:

  installr -d -t linux-headers r-lib/pingr
  

• Rcpp 1.0.13 does not compile with R 4.4.2. Use the dev version of Rcpp until a new version is released on CRAN. This affects all packages that depend on Rcpp, naturally. E.g. shiny, V8, rstan, odbc, golem, prophet, pagedown, plumber, sf, etc. See the updated [examples].

  installr -d Rcppcore/Rcpp
  

License

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

These Dockerfiles are licensed under the MIT License.

© R Consortium