Conventional Commits about conventional-commits

[thclark]

@cortadocodes I've assigned this to you, we'll come to it fairly soon I think, as the more I think about it, the more it's really going to make a difference to workflow.

[thclark]

@cortadocodes if you're building a small equivalent of the conform tool, plus a pre-commit hook, perhaps we'll group them in this repo and rename the repo (eg pre-commit-conventional-commit), rather than having the tool in a separate repo, and the hook in this one.

Probably scratch that last, I didn't realise how small the checker would be.

[cortadocodes]

I've had a look at the conform tool written in go which gives some useful features for this:

• Enforces Conventional Commits
• Enforces header of given length in the imperative with an ending that isn't invalid (e.g. a full stop)
• Allows enforcing (or not) of the presence of a body
• Spell checks the commit message in the given language
• Allows a discrete set of custom commit codes (e.g. the ones I've already been using)

For me currently, it works if a pre-built release is downloaded and manually added as a hook in the .git directory, but it doesn't seem to work properly with

• go get
• go build
• The provided GitHub action
• The provided pre-commit hook

So I think it may be better to write our own, but I'm wondering which of the features above are must haves and which are could haves @thclark?

[cortadocodes]

It seems to me like these are the must-haves:

• Enforces Conventional Commits
• Enforces header of given length
• Enforces header with an ending that isn't invalid
• Allows a discrete set of custom commit codes

[thclark]

So I think it may be better to write our own,

Our own what? Conform tool? Or GHA and pre-commit hook to run the existing conform tool? The latter, yes - the former would very much raise an eyebrow at how far deep we're getting.

It seems to me like these are the must-haves:

• Enforces Conventional Commits
• Enforces header of given length
• Enforces header with an ending that isn't invalid
• Allows a discrete set of custom commit codes

I agree that skipping the spell checker is sensible. It's mildly annoying when there are typos, it's a crashing marshmallowman on your dev flow when the spell checker doesn't have your word or the thing you deliberately want to type.

It seems like you've intentionally missed the imperative tense constraint on the header (e.g: enforcing the form between "fixes the issue" (passive) "fix the issue" (imperative)). Why is that? Anything to add to the pros and cons below?

Pros:

• Automated releases will read strangely if the tenses are mixed
  Cons:
• it may cause too much friction as we adopt the tool (and have our developers adopt it)
• Similar logic might apply as for the spellchecker (too constraining when you really are wanting to type a specific message).

[cortadocodes]

I meant writing our own conform tool, pre-commit package, and GHA. I've started doing it and it's turnint out to be quite simple - I'm nearly done!

Agree about spell checker - that's why I left it out.

With the imperative, I think all commits should be in the imperative but I'm aware that not everybody wants to do that. I also didn't know how to easily check if a sentence is in the imperative without NLP, although it looks like conform just checks if the first word in the header after the commit code is in a list of known imperative verbs.

I do think it's part of good practice to make commits imperative, so I could/should probably easily add what conform has.

[thclark]

@cortadocodes if you're building a small equivalent of the conform tool, plus a pre-commit hook, perhaps we'll group them in this repo and rename the repo (eg pre-commit-conventional-commit), rather than having the tool in a separate repo, and the hook in this one.

[cortadocodes]

This is now released in https://github.com/octue/octue-sdk-python/releases/tag/0.1.20

The only difference from the issue description is that we agreed to only check if the setup.py version is semantically correct rather than actually updating it - this allows us complete control over the version number of the next release in case we need to override what git-mkver says.

[cortadocodes]

This has now also been rolled out to amy, twined, and planex-site.