Inhibits commits with bad messages from getting merged

platisd, updated 🕥 2022-01-22 11:27:03

Bad Commit Message Blocker Build Status

A simple script, doing some natural language processing, to inhibit git commits with bad messages from getting merged.



A Python3 script, easy to integrate with various CI machinery (e.g. see GitHub Actions example) that is meant to keep bad commit messages out of a project. It verifies whether the seven rules of a great Git commit message by Chris Beams, are being followed:

  1. Separate subject from body with a blank line
  2. Limit the subject line to 50 characters
  3. Capitalize the subject line
  4. Do not end the subject line with a period
  5. Use the imperative mood in the subject line
  6. Wrap the body at 72 characters
  7. Use the body to explain what and why vs. how


The more a project grows in contributors and size, the more important the quality of its commit messages becomes.

It is ultimately hard to quantify what makes a good commit message. However, we can easily determine automatically whether it adheres to most of the git commit message best practices. Having a script to automatically verify that, aside of other rules that you might already enforce (e.g. checking whether a task or a requirement is referenced) can save a lot of time in the long run as well as even avoid conflicts between the reviewer(s) and the reviewee.

After adopting this script, what you get is a set of best practices, agreed upon by the project and automatically applied. This way, everyone has to follow them or CI will not allow that commit in. Simple as that! :innocent:


Most of the rules are trivial to implement in code, except two of them, no. 5 and no. 7. Specifically, checking whether the commit subject begins with imperative mood is tricky due to limitations of the Natural Language Processing libraries being utilized. You can read more about the constraints here. Essentially, it boils down to the lack of many imperative sentences existing in the datasets used when training the relevant statistical models. Subsequently, the enforcement of this rule might produce some false positive and false negative errors.

On the other hand, verifying whether the commit body actually explains what and why instead of how is not (?) possible, due to the subjective nature of the problem. All in all, in most cases, this is all the reviewers would have to do themselves manually, to ensure the quality of a commit message.

Get started

You need to have Python3 installed and follow the steps bellow: * Install TextBlob * pip3 install --user textblob * Install NLTK corpora * python3 -m textblob.download_corpora * Run the script to verify a commit message * python3 --message "Add a really cool feature" * To define your own maximum character limits, call the script with the appropriate arguments: * --subject-limit (defaults to 50) to set the subject line limit. E.g.: * python3 --subject-limit 80 --message "Add a really cool feature" * --body-limit (defaults to 72) to set the body line limit. E.g.: * python3 --body-limit 120 --message "Add a really cool feature"

GitHub Action

Now you can use this script as part of your GitHub Actions CI pipeline.

An example configuration can be seen below:

```yaml name: Commit messages

on: [pull_request]

jobs: check-commit-message: if: github.event.pull_request.user.type != 'Bot' # a number of GitHub Apps that can send PRs don't have configurable commits runs-on: ubuntu-20.04 steps: - name: Verify commit messages follow best practices in CI uses: platisd/[email protected] with: github_token: ${{ secrets.GITHUB_TOKEN }} # Optionally set the subject character limit (default 50) subject_limit: 60 # Optionally set the body character limit (default 72) body_limit: 100 # Optionally set the remote branch name to merge (default master) remote_branch: dev ```


add example yml for use in github and gitlab repos

opened on 2022-06-23 14:10:10 by funkyfisch

It would be nice to have example config files for using this script in:

  • github actions (currently part of the README)
  • gitlab ci
  • circleci

Not sure if its better to have them as part of the repository as files (it might be a pollution) or to have them as part of the README

Retrieve proper commitish for comparison from forked repos in PRs

opened on 2022-01-22 03:15:52 by webknjaz

The current action implementation assumes that commits are always coming from the upstream repository but in reality, they don't, in the case of forks.


This needs to be fixed here:

Action items: * [ ] Check the GitHub Actions docs to figure out the variable that exposes this information * [ ] Use that to send a PR

Convert the docker-based GHA into a composite action

opened on 2021-12-13 20:50:14 by webknjaz

I think it would be reasonable to refactor the GHA to use a better action type. Docker is limited to Ubuntu workers and has a certain amount of runtime overhead that I believe can be addressed with Composite GitHub Actions. I've promptly checked the current setup and I don't believe there's anything that is tightly coupled with the docker runtime and can't be substituted.



opened on 2021-09-16 22:18:28 by platisd None

Adding file checker

opened on 2019-11-12 13:27:57 by edmundwatson

Adding file checker so it can be used with a git commit hook. Also adding example commit-msg to be added to .git/hooks


Version 1.0.1 2021-09-16 22:48:24

  • Fix Action not working with private repositories

Version 1.0.0 2020-11-08 17:08:52

Dimitris Platis

Software Engineer & Maker. Currently with a special interest in the fields of Embedded Systems, Internet of things and Robotics.

GitHub Repository

git nlp nltk textblob commit-message best-practices travis-ci continuous-integration code-review