docker-packrat-example

docker-packrat-example copied to clipboard

Installing R packages to a Docker image with packrat

This is a minimal example that installs a single package each from CRAN, bioconductor, and github to a Docker image using packrat.

For more info, see my blogpost.

Setup

Keep Dockerfile and install_packages.R in the root of the project directory (you can clone this repo as an example project).

First step: make packrat/packrat.lock file

Launch rocker/rstudio (or any other rocker image of your choice) with tag set to same version we will use in the Dockerfile, and this directory mounted.

Note that the rocker images come with various packages pre-installed, so if you want to track those packages properly you would have to install them like the other example packages in install_packages.R.

docker run -it -e DISABLE_AUTH=true -v /path/to/project/:/home/rstudio/project rocker/rstudio:3.5.3 bash

Inside the docker container, run the script to install R packages with packrat. Exit when done.

cd home/rstudio/project
Rscript install_packages.R

This will install current versions of all packages in install_packages.R, but the main reason is to write packrat/packrat.lock.

Second step: actually build the image

Now we can use packrat.lock to restore (i.e., install) packages when we build the image.

docker build . -t mycontainer

We should now be able to run the container with the packages we installed.

docker run -it mycontainer R

Inside the container running R, check that the packages are installed:

library(minimal)
packageVersion("minimal")

Third step: rinse, repeat

Edit the packages in install_packages.R as needed to add new packages or update old ones, and repeat Steps 1 and 2.