A GitHub Action to run DataLinter—a contextual data and code linter for machine-learning and statistical-modelling workflows—inside a reproducible Docker container. Results are automatically posted as a comment on pull requests.

• What is DataLinter?
• Features
• Quick Start
• Inputs
• Outputs
• Workflow Examples
• How It Works
• DataLinter Configuration
• Troubleshooting
• Contributing
• License

DataLinter detects potential issues in data files and associated code (R, Python, etc.) used in ML/statistical pipelines.

• Reproducible execution via a pre-built Docker image (ghcr.io/zgornel/datalinter-compiled:latest).
• Automatic mounting of data, code, and configuration directories.
• Full output capture (stdout + stderr) wrapped in a Markdown code block.
• Automatic PR commenting (edits previous comment if it exists).
• Composite action—works on any runner that supports Docker.
• Flexible logging levels.

Add the following workflow to .github/workflows/datalinter.yml:

name: DataLinter

env:
  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: true   # Required by the embedded PR comment action on newer runners
  PROJECT_DIR_NAME: "target-project"

on:
  pull_request:
    branches: [ main, master ]
  push:
    branches: [ main, master ]

jobs:
  datalinter:
    name: Run DataLinter
    runs-on: ubuntu-latest
    permissions:
      pull-requests: write   # Needed for automatic PR comments
      # issues: write        # Optional; not currently used

    steps:
      - name: Checkout repository
        uses: actions/checkout@v6
        with:
          path: ${{ env.PROJECT_DIR_NAME }}

      - name: Run DataLinter
        id: datalinter
        uses: OxoaResearch/DataLinter-Action@v1
        with:
          target-project-checkout-dir: ${{ env.PROJECT_DIR_NAME }}
          data-path: "data/my_data.csv"
          code-path: "code/my_analysis.r"
          config-path: "config/datalinter.toml"
          log-level: "warning"

• Pull-request, recommended for reviews – see Quick Start above.
• Push-only i.e. logs only, no comment, use

on:
  push:
    branches: [ main ]
# ... rest of job identical, but remove `pull-requests: write` if desired

datalinter-github-action has the following inputs:

• target-project-checkout-dir (required), the name of the directory where the repository will be checked out
• data-path (required), path to the data relative to target-project-checkout-dir
• code-path (required), path to code, relative to target-project-checkout-dir
• config-path (required), path to the datalinter .toml file configuration, relative to target-project-checkout-dir
• log-level (optional, default is 'debug'), logging level; can be 'debug', 'info', 'warning' and 'error'.

Complete DataLinter output (stdout + stderr) formatted as a Markdown code block. You can reference it with ${{ steps.<id>.outputs.result }} if you need further processing.

1. The action splits the supplied paths into directory and filename components.

2. It pulls ghcr.io/zgornel/datalinter-compiled:latest

3. Three Docker volumes are mounted:

   • Code directory → /tmp
   • Data directory → /_data
   • Config directory → /_config

4. datalinter binary is executed inside the container.

5. All output is captured and stored in the result output.

6. On pull_request events the output is automatically posted (or edited) as a PR comment using thollander/actions-comment-pull-request@v3.

Create a datalinter.toml file in your repository. Full documentation and example configurations are available in the DataLinter repository.

• “No such file or directory” → Verify the three paths exist relative to the checkout directory.
• Docker pull fails → GitHub runners have Docker; rate-limiting on GHCR is rare but can be mitigated with a personal access token if needed.
• PR comment not appearing → Ensure the job has pull-requests: write permission and the workflow is triggered on a PR (not a push from a fork without approval).
• Output too verbose → Set log-level: 'warning' or 'error'.
• Spaces/special characters in paths → The action quotes paths internally; it should handle them.

Please file an issue to report a bug or request a feature.

This project is licensed under the MIT License. See LICENSE for details.