JabRef GitHub Action Demo

Demonstration for JabRef's GitHub action.

name: Check

on:
  pull_request:

jobs:
  test-action:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout
        uses: actions/checkout@v7

      - name: Run jabref-action
        uses: jabref/jabref-action@main
        with:
          bibfile: Chocolate.bib
          output-format: github-actions

With output-format: github-actions, the consistency findings are emitted as GitHub Actions workflow commands, so they appear as annotations on the pull request and the commit. The raw output looks like:

::error file=Chocolate.bib,line=33,col=1,title=Richard_2017:groups::unknown field for entry type Article
::error file=Chocolate.bib,line=42,col=1,title=Parker_2006:groups::unknown field for entry type Article
...

The default output-format is errorformat. Other supported values are csv, github-actions and txt. The txt format is actually Markdown — it produces a table, so it renders nicely in the GitHub Actions job summary.

Checking a whole library collection

The check-collection.yml workflow runs the action against every .bib and .bb file in the JabRef/bibtex-library-collection repository, which is included here as a git submodule.

The submodule tracks the collection's flattened mirror branch (shallow clone), so all bibliography files are available without nested submodules:

[submodule "bibtex-library-collection"]
	path = bibtex-library-collection
	url = https://github.com/JabRef/bibtex-library-collection.git
	branch = mirror
	shallow = true

The workflow uses two jobs:

1. discover — lists every .bib and .bb file in the submodule and emits them as a JSON array used to build the matrix.
2. check — consumes the matrix and runs jabref-action once per file. It uses max-parallel: 1 so the files are checked sequentially, and fail-fast: false so one file with inconsistencies does not cancel the rest. It sets output-format: txt, so each file's findings render as a Markdown table in the job summary.

Note

A GitHub Actions matrix is limited to 256 jobs. If the collection ever contains more than 256 .bib/.bb files, the matrix has to be split (e.g. batched into multiple jobs).