feat: Validate multi-line messages

[grahamhar]

The documented conventional commit specification states that on multi-line commit message there must be a blank line between the title and the body.

I was unsure on updating the error output or not. If it is updated I guess the error out put should guide towards which element failed the title line or the multi line issue. Let me know your thoughts.

fixes #67

[thekaveman]

Thanks for working on this @grahamhar.

I have a few comments on the implementation having thought about this a little more:

1. The is_conventional() function should do this validation (rather than having to call a separate function), since as you note, it is part of the spec
2. To that end, I think we can probably do this by modifying the existing regex for r_subject(), perhaps splitting that into r_subject() and r_body()
3. Finally (and this is a little trickier) -- making the blank line a requirement is technically a breaking change (while acknowledging again that yes, the spec does call for a blank line); so I'd like to start with this feature hidden behind our --strict flag (see feat: introduce "strict" mode #61) so we can release a 2.x update. In a future 3.x update, we can make it the default.

[grahamhar]

Thanks for working on this @grahamhar.

I have a few comments on the implementation having thought about this a little more:

1. The is_conventional() function should do this validation (rather than having to call a separate function), since as you note, it is part of the spec
2. To that end, I think we can probably do this by modifying the existing regex for r_subject(), perhaps splitting that into r_subject() and r_body()
3. Finally (and this is a little trickier) -- making the blank line a requirement is technically a breaking change (while acknowledging again that yes, the spec does call for a blank line); so I'd like to start with this feature hidden behind our --strict flag (see feat: introduce "strict" mode #61) so we can release a 2.x update. In a future 3.x update, we can make it the default.

@thekaveman I think I have managed to implement this in the way you are asking, let me know if you have any other suggestions.

I do have one thought on how we should handle the following example (A subject followed by a blank line but no body) as I couldn't find an answer in the spec on this

feat: A valid

Also I am unsure if the use of \n in the regex is sufficient, How does git running on Windows based systems behave? The editor used might also use \r\n.

[thekaveman]

@grahamhar:

I do have one thought on how we should handle the following example (A subject followed by a blank line but no body)

I'm OK to ignore this case -- you're right, I can't see anything in the spec on this either, and normally git will just delete those extra blank lines from the commit message in post.

Also I am unsure if the use of \n in the regex is sufficient, How does git running on Windows based systems behave?

Good point. I think a regex like \r?\n will work here, as that will catch both Windows and Unix-style line endings. Seems like Mac OS X and later also uses \n so that should cover our bases. https://stackoverflow.com/a/39022365/453168

[grahamhar]

@thekaveman I think this should be good now.

[thekaveman]

Thanks for your work on this and going back and forth with me on some reviews @grahamhar!

[thekaveman]

Thanks again @grahamhar!

I have released v3.0.0 which includes the change to enforce the blank line, but drops the effect of the --strict flag -- the blank line is always required now.