R and Docker

Warwick R User Group

James Tripp

Senior Research Software Engineer, IT Services (University of Warwick)

Who am I?

James Tripp

Research Software Engineer

james.tripp@warwick.ac.uk

Researcher in Cognitive Science
Senior Academic Technologist @ Interdisciplinary Centre
Senior Research Software Engineer @ IT Services

Plan

Why?
Docker?
Next Steps

Why?

Developers

Software developers sometimes wrote code which only ran on their machine
That could be a problem

Developers

Developers could build and update images
Images are stored in a registry
Production servers download images
Copies of these images called containers are used in production

Researchers

How does this help research?

Aids computational reproducibility (Boettiger 2015)

Allows other to run code without having to install lots of dependencies (Eglen et al. 2017)

Language agnostic and used widely (Pittard and Li 2020)

Provides the basis for technologies like the research compendium (Marwick, Boettiger, and Mullen 2018), binder (“Get Started with Binder Binder 0.1b Documentation,” n.d.) and dev containers (Smith 2022)

Docker

What?

R images on DockerHub

The R project has an official r-base image

Rocker Project has lots of images:

-   rocker/r-ver - alternative to r-base
-   rocker/rstudio - R and RStudio
-   rocker/tidyverse- R and tidyverse packages
-   rocker/shiny - Shiny server built in
-   many others

Base image

To use the official image

docker run --rm -ti r-base

The r-base image downloaded
Container created from image
Terminal enters container

The command line options are to remove the container on exit (-rm) and to create a command line interface with interactivity (-ti)

RStudio image

docker run --rm -ti -e DISABLE_AUTH="true" -p 8787:8787 rocker/rstudio

Downloads image, creates and terminal enters container
The container provides access to RStudio via localhost:8787
You now have an isolated RStudio container running

Command line options are remove the container on exit (-rm), create a command line interface (-ti), set environment variables (-e) and set the port (-p) so that connecting via port 8787 will connect to port 8787 in the container.

Custom images

Docker images are created from dockerfiles. A sample dockerfile is below¹.

FROM r-base
COPY . /usr/local/src/myscripts
WORKDIR /usr/local/src/myscripts
CMD ["Rscript", "myscript.R"]

To build the image, go to the folder containing the dockerfile and run.

docker build -t jamestripp/myimage .

Finally, to create a container and jump into it.

docker run --rm -ti jamestripp/myimage

You can publish images to the Docker Hub (or another registry) for others to use (see instructions here and here)

Local files

Copying files when the image is created. Baked into the image.

FROM r-base
COPY . /usr/local/src/myscripts
WORKDIR /usr/local/src/myscripts
CMD ["Rscript", "myscript.R"]

Volumes are folders on your local file system which are accessible to a container

# MacOS and Linux
docker run --rm -ti -e DISABLE_AUTH="true" -p 8787:8787 -v  $(pwd):/home/rstudio/data rocker/rstudio
# Windows
docker run --rm -ti -e DISABLE_AUTH="true" -p 8787:8787 -v  absolute_path:/home/rstudio/data rocker/rstudio
#May work on Windows
docker run --rm -ti -e DISABLE_AUTH="true" -p 8787:8787 -v  %cd%:/home/rstudio/data rocker/rstudio

Where absolute_path is the full Windows path

Example

Fortune

myscript.R

library(fortunes)

print(fortune())
print(fortune())

dockerfile

FROM r-base
COPY . /usr/local/src/myscripts
WORKDIR /usr/local/src/myscripts
RUN ["install2.r", "fortune"]
CMD ["Rscript", "myscript.R"]

Build and run it

docker build -t jamestripp/fortune .
docker run --rm -ti jamestripp/fortune

Fortune

Other examples

RMarkdown

Shiny apps

Deploying models

To consider

Changes in the base image can influence the images build upon them
Ensure you use trustworthy images such as Rocker (Boettiger and Eddelbuettel 2017)

Smaller images with very little software may be preferable as these are easier to maintain (Gruening et al. 2019a)

Containers are, perhaps, only part of a reproducible workflow(Mölder et al. 2021; Gruening et al. 2019b)
Do we need to replicate the entire computing environment? Perhaps automagic or renv, which captures the R version and libraries, is all we need… (see the previous WRUG presentation by Carlos Camara Menoyo)

Next steps

Learning materials

Look through the R Blogger posts on Docker

Work through the UserR2022 workshop by rsangole (repo, blog post)

Read through the docker section in the DevOps for Data Science book

Papers

Some R packages are mentioned in Nüst, Eddelbuettel, et al. (2020). I was not able to get these packages working
- Stevedore - Sends commands to Docker. Required reticulate and the Python docker module. Recieved the error ‘Did not find required python module ’docker’’. Issue reported and not addressed in the past year…
- Dockyard - Aims to help you create and run a container. Last commit was 3 years ago and the example code on the github page does not work…
- Dockermachine - Last updated 5 years ago.

Nüst, Sochat, et al. (2020) outlines 10 rules for creating dockerfiles, talks about good practice and sign posts other tools

Peikert and Brandmaier (2021) offer a workflow for rendering markdown documents which includes a make file for dependencies
Not an exhaustive list

Tools

Binder

Run a docker container on a remote server. Place a button on your public github repo README.md.

ShinyProxy

Spins up shiny containers with one container per user (see the R Bloggers post)

Questions?

References

Boettiger, Carl. 2015. “An Introduction to Docker for Reproducible Research.” ACM SIGOPS Operating Systems Review 49 (1): 7179. https://doi.org/10.1145/2723872.2723882.

Boettiger, Carl, and Dirk Eddelbuettel. 2017. “An Introduction to Rocker: Docker Containers for R.” The R Journal 9 (2): 527–36. https://journal.r-project.org/archive/2017/RJ-2017-065/index.html.

Eglen, Stephen J., Ben Marwick, Yaroslav O. Halchenko, Michael Hanke, Shoaib Sufi, Padraig Gleeson, R. Angus Silver, et al. 2017. “Toward Standard Practices for Sharing Computer Code and Programs in Neuroscience.” Nature Neuroscience 20 (6): 770–73. https://doi.org/10.1038/nn.4550.

“Get Started with Binder Binder 0.1b Documentation.” n.d. https://mybinder.readthedocs.io/en/latest/introduction.html.

Gruening, Bjorn, Olivier Sallou, Pablo Moreno, Felipe da Veiga Leprevost, Hervé Ménager, Dan Søndergaard, Hannes Röst, et al. 2019a. “Recommendations for the Packaging and Containerizing of Bioinformatics Software.” https://doi.org/10.12688/f1000research.15140.2.

———, et al. 2019b. “Recommendations for the Packaging and Containerizing of Bioinformatics Software.” https://doi.org/10.12688/f1000research.15140.2.

Marwick, Ben, Carl Boettiger, and Lincoln Mullen. 2018. “Packaging Data Analytical Work Reproducibly Using r (and Friends).” The American Statistician 72 (1): 80–88. https://doi.org/10.1080/00031305.2017.1375986.

Mölder, Felix, Kim Philipp Jablonski, Brice Letcher, Michael B. Hall, Christopher H. Tomkins-Tinch, Vanessa Sochat, Jan Forster, et al. 2021. “Sustainable Data Analysis with Snakemake.” F1000Research 10 (April): 33. https://doi.org/10.12688/f1000research.29032.2.

Nüst, Daniel, Dirk Eddelbuettel, Dom Bennett, Robrecht Cannoodt, Dav Clark, Gergely Daróczi, Mark Edmondson, et al. 2020. “The Rockerverse: Packages and Applications for Containerisation with R.” The R Journal 12 (1): 437–61. https://journal.r-project.org/archive/2020/RJ-2020-007/index.html.

Nüst, Daniel, Vanessa Sochat, Ben Marwick, Stephen J. Eglen, Tim Head, Tony Hirst, and Benjamin D. Evans. 2020. “Ten Simple Rules for Writing Dockerfiles for Reproducible Data Science.” PLOS Computational Biology 16 (11): e1008316. https://doi.org/10.1371/journal.pcbi.1008316.

Peikert, Aaron, and Andreas M. Brandmaier. 2021. “A Reproducible Data Analysis Workflow With R Markdown, Git, Make, and Docker.” Quantitative and Computational Methods in Behavioral Sciences, May, 1–27. https://doi.org/10.5964/qcmb.3763.

Pittard, W. Stephen, and Shuzhao Li. 2020. “The Essential Toolbox of Data Science: Python, R, Git, and Docker.” In, edited by Shuzhao Li, 265–311. Methods in Molecular Biology. New York, NY: Springer US. https://doi.org/10.1007/978-1-0716-0239-3_15.

Smith, David. 2022. “Easy r Tutorials with Dev Containers | r-Bloggers.” https://www.r-bloggers.com/2022/08/easy-r-tutorials-with-dev-containers/.