Commitlint: feat: update conventional commit linter

[bcoe]

Hey @marionebl :wave:

We've been making some slight tweaks to the conventional commit spec, as we march towards a v1.0.0 of the spec. Mainly introducing ! a shorthand for breaking changes.

I've also finally taken the time to write a preset based directly on conventionalcommits.org.

It would be slick if we could update commitlint to match conventional-changelog-conventionalcommits and conventionalcommits.org, or its conventional commits linter.

[marionebl]

Hey @bcoe, thanks for reaching out! I'm rolling some scenarios around, input and thoughts on the following would be awesome:

1. Do we update commitlint-config-conventional to match the new spec or create commitlint-config-conventionalcommits. I am leaning towards option 1.

2. Do you see cases for teams to limit / validate things around the !? Still wrapping my head around the implications.

[bcoe]

re: 1, I agree let's just update the existing commitlint-config-conventional.

re: 2, I think, with how I'm currently picturing ! will work, we just need to allow this as a valid character in the type/scope prefix; there shouldn't be any validation needed beyond this.

[byCedric]

Allowing it in the types/scopes will work just fine I think. We might add rules like prefer-exclamation, to warn/error for the "old" BREAKING: ... notation right?

[byCedric]

I'll close issue #632 in favour of this one!

[byCedric]

Ok, so I double-checked the current conventional config. I don't think a lot has to be changed to match the spec. The only "biggest" (probably breaking) change is to use the conventional-changelog-conventionalcommits parser options. Besides that, I found some other parts that probably require some discussion.

Are there more types allowed than currently listed?
In the specification, I found that rule 2 and rule 3 are more "user-guidelines" (which still should be followed). But rule 10 kind of "opens" a possibility of more types, which are not listed in the current config. Should we "remove" it, or let this be as is? In the writer, provided by Benjamin, I see the same types are used as headers in the output changelog. But again, I'm not sure if forcing this set actually "matches the spec".

Is forcing types/scopes/subjects to a specific-case a problem?
Rule 11 basically says "everything, with the exception of BREAKING CHANGE, should be considered case-insensitive. I think this means that writing a type like Feat should be handled like feat. My personal preference is lower case enforcement, but I have no clue if that violates the spec.

Is forcing to forbid full stops in subjects a problem?
Again, I couldn't find this within the spec. An example at rule 5 actually uses this full stop. In HTML, it's rendered as listed below, which kind of includes the . within the <em />.

<em>fix: array parsing issue when multiple spaces were contained in string.</em>

[byCedric]

I'll make some time available this weekend to finish the new configuration and support for !. 😄

[bcoe]

awesome thank you @byCedric 👍

[byCedric]

It took some effort, and a lot of sweat/tears, but it's there. The "next" (now 8.3.3) commitlint version now supports the new conventionalcommits parser, and it's included by default when using @commitlint/config-conventional@next (now 8.3.3). Try it out, let us know what you think or encounter 😄🚀

$ npm i --save @commitlint/cli@next @commitlint/config-conventional@next

I'll try to push for the full spec update because some other developers are reporting conflicting types or casing issues. But that might take a bit longer and probably requires a major bump. Rest assured, we are on it!