This packages hubValidations in a docker container so they can be run without requiring R to be installed on a machine.

The main motivation for this image is to simplify the validation process for our workflows.

Images on GitHub need to be run at the job level. You can use the container key for this. For this container, the working directory is /project, so you need to link your github workspace the working directory using the volumes keyword.

jobs:
  validate-submission:
    runs-on: ubuntu-latest
    env:
      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
    container:
      image: ghcr.io/hubverse-org/test-docker-hubvalidations:main
      ports:
        - 80
      volumes:
        - ${{ github.workspace }}:/project

After that, you only need 2 steps: checkout and validate:

    steps:
      - name: Checkout Pull Request
        uses: actions/checkout@v4
      - name: Run validations
        run: |
          # R code you want to run
        shell: Rscript {0}

To run locally, you can use this incantation with any command or script to run

docker run \
--platform=linux/amd64 \
--rm \
-it \
-v "/path/to/hub":"/project" \
ghcr.io/hubverse-org/test-docker-hubvalidations:main \
# script to run

validate.R . <org/repo> <PR>

This script will run hubValidations::validate_pr() with the local pull request and it will also print any error attributes to the console on failure when they exist so that it's easier to determine exactly what caused the error.

The docker container will be built automatically on GitHub. It can also be built locally to an image called hubvalidaitons with:

docker build --platform=linux/amd64 -t hubvalidations .