In meterds/mdsrocker: Docker images for CI/CD and dockerized AWS Solutions at METER Data Science

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

# libraries
library(tibble)
library(dplyr)
library(purrr)
devtools::load_all(".")

# dockerhub account
account = desc::desc_get_field("Config/Dockerhub/Account")

mdsrocker

badge_name = "Docker Image CI"
src = glue::glue("https://github.com/{account}/mdsrocker/actions/workflows/publish-docker-images.yml/badge.svg")
href = glue::glue("https://github.com/{account}/mdsrocker/actions/workflows/publish-docker-images.yml")
glue("[![{badge_name}]({src})]({href})")

The goal of mdsrocker is to provide a framework for building a stack of version-stable docker images based on rocker/r-ver, an ubuntu-based docker image containing version-stable base R & a set of src build tools.

We follow the version tag approach of the rocker-project, using the image tag for specifying which version of R is desired (e.g. r paste("rocker/r-ver:", as.character(getRversion()))). Omit the tag or specify :latest to always receive the latest versions.

Image stack

stack = mdsrocker_dockerfiles |> 
  dplyr::select(image, description)

stack = stack |> 
  dplyr::mutate(
    image = kableExtra::cell_spec(
      image, 
      "html",
      link = glue::glue("https://hub.docker.com/r/{account}/{image}")
    )
  )

kableExtra::kable(stack, escape = FALSE)

Architecture

This building system is designed as an R package.

Data

All relevant information is stored in two data.frames (as package data):

• mdsrocker_installation: three different software bundles for the various scopes of the docker images, each containing predefined sets of software packages (system libraries, extra software packages, Python packages, R packages) to install.

kableExtra::kable(mdsrocker_installation)

• mdsrocker_dockerfiles: four dockerfile specifications, each containing the relevant information for generating the dockerfiles.

mdsrocker_dockerfiles |> 
  dplyr::select(-description) |> 
  kableExtra::kable()

Scripts & Files

In this building system,

• standalone shell scripts containing all the install logic are located under the scripts/ directory,
• dockerfiles used in the Github Actions workflow to build the images are located under dockerfiles/, and
• two github actions workflows are stored under .github/workflows/:
• update-building-system.yml: re-builds all scripts if any changes were done in the repository (e.g. adding an R version to the DESCRIPTION file or adding some R packages to the software bundles),
• publish-docker-images.yml: actually builds and pushes the defined image stack.

Updating

All scripts are automatically updated in the Github Actions Workflow update-building-system.yml (by running execute.R) when commited and pushed.

For adding changes (e.g. additional R packages), do:

1. git pull the latest changes in the remote repository,
2. run:
3. mdseasy::easy_version(element = "dev", check_renv = FALSE, check_gitlab_ci = FALSE, push = FALSE)
4. update and run the code generating the package data in data-raw/datasets.R,
5. run:
6. devtools::document()
7. devtools::check()
8. update NEWS.md
9. run
10. gert::git_commit_all("YOUR COMMIT MESSAGE")
11. mdseasy::easy_version(check_renv = FALSE, check_gitlab_ci = FALSE)

Execution

The Github Actions workflow publish-docker-images.yml is:

• scheduled to run every Sunday,
• triggered by successful runs of the update-building-system.yml workflow on main branch,
• or can be started manually on Github.

meterds/mdsrocker documentation built on Dec. 2, 2023, 11:56 p.m.