maintainers/scripts/update.nix: Remove unicode from message

[viraptor]

Remove unicode from message and comply with CONTRIBUTING.md:

Format the commit messages in the following way:

(pkg-name | nixos/<module>): (from -> to | init at version | refactor | etc)

It's not used frequently and the ascii version is nicer (because it's type-able regardless of layout) / expected (because it's documented) for automatic parsing.

git log --grep '->' --oneline --since '2022-01-01' | wc -l
   67791
git log --grep '→' --oneline --since '2022-01-01' | wc -l
    2914

Description of changes

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 23.05 Release Notes (or backporting 22.11 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

[andersk]

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.)

[Artturin]

-> is the dominant one and used in CONTRIBUTING.md

It's also easy to type and minor aesthetics don't matter as we're not in a gallery

[Artturin]

#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

[viraptor]

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.

[SuperSandro2000]

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.

[jtojnar]

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.

[SuperSandro2000]

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.

[7c6f434c]

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.

[toastal]

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.

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 >.

[SuperSandro2000]

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.