the ascii version is nicer / expected for automatic parsing

You’re entitled to your opinion. My opinion is that the Unicode version is nicer, and automatic parsers should conform to the commit messages we actually have. (I’m willing to be overruled, but I think it’s important to note that our opinions differ.)

#155309 (comment)

One issue with switching update.nix to use ASCII is that update.nix is also used by repos which do not allow the ASCII variant, whereas Nixpkgs seems to have been tolerating the Unicode variant so far.

What repos are those

automatic parsers should conform to the commit messages we actually have

As long as it's clear that we use that format, that's right. Currently, anyone who writes some one-off script without specifically knowing about this difference (e.g. me) will use the existing docs and expect ->. They'll also tell people they don't comply with CONTRIBUTING if they notice a different format.

To be clear, I'm not opposed to the unicode symbol being used. I'm proposing a bug fix where we document one format, but do something different.

But the current state is broken. If we explicitly wants to use unicode, then we should make a different bug fix which:

• updates CONTRIBUTING.md
• updates the docs
• (optionally) changes the defaults in other scripts, because having inconsistent approaches means the parsing has to deal with (->|→)

Also, from the other comment:

it is as simple as pressing Compose followed by those two characters.

That really depends on the system / layout / os, etc. I don't have the compose key, so when discussing this, I need to keep copy-pasting the symbol.

My opinion is that the Unicode version is nicer, and automatic parsers should conform to the commit messages we actually have.

but it isn't and this is the quickest fix to get it working and things inline with contributing. If ofborg and so on support the unicode character, we can revisit this.

Also side note: I don't even know how to type the unicode version on my keyboard.

What repos are those

I use update.nix in my personal repos and do not want ugly “pretend arrows” there.

CONTRIBUTING.md is not a formal specification and commit messages are not written in a formal language. The only requirement is that the commit messages roughly match the described shape and Unicode arrows do just fine in that regard. That is why they are generally accepted and any tool that wants to attempt to parse historic commit messages will quickly discover that it needs to support them.

That is why they are generally accepted

To my knowledge they are not accepted.

any tool that wants to attempt to parse historic commit messages

This was mostly about current new commit messages as eg ofborg is not parsing historic data.

That is why they are generally accepted and any tool that wants to attempt to parse historic commit messages will quickly discover that it needs to support them.

It is generally accepted that some share of commit messages do not follow CONTRIBUTING.md precisely. Getting to zero is hard, but if a safe change in a script can reduce the amount of deviations, that's a nice thing.

A lot of those -> commits are coming from the update bots without human involvement and could be tweaked either way. +1 for Unicode support; Unicode lets you mean what you say (that is in this case a rightwards arrow, not dash greater than).

I don't even know how to type the unicode version on my keyboard.

If you commit via the CLI git commit will likely bring you into a Vim-compatible editor to format your commit message. The diagraph is ->, or with default bindings <C-k>->, or Ctrl+k then - then >.

If using the Compose key, you can type Compose - >.

If you use Kitty, you use Crtl+Shift+U to bring up the Unicode picker & type (or partially type) Rightwards arrow.

You can also make your own macro, or add it to a keyboard layer if it is used enough (I added it to my phone keyboard).

TIL

The diagraph is ->, or with default bindings <C-k>->, or Ctrl+k then - then >.

You can do 99% of your work with 1% of vim's features.