Add homebrew formula tap

[sudosubin]

I wanted to use check-jsonschema as a command line tool, but there was no brew formula package.

Contributing Homebrew/homebrew-core upstream has failed due the GitHub repository wasn't notable enough. (Homebrew/homebrew-core#105012)

So I have made a homebrew tap repository (sudosubin/homebrew-cask-jsonschema), but if it exists in an upstream organization(python-jsonschema), it might be better managed.

# before (sudosubin/homebrew-check-jsonschema)
brew tap sudosubin/check-jsonschema

# after (python-jsonschema/homebrew-check-jsonschema)
brew tap python-jsonschema/check-jsonschema

How do others think?

[sirosen]

I have a few thoughts on this. I'm just going to share what I'm thinking -- there's no implied resolution or plan here.

First, is installation with pipx install check-jsonschema not sufficient? What is the value-add of being in homebrew vs being able to brew install pipx; pipx install check-jsonschema?

check-jsonschema might gain enough traction to go into homebrew-core at some point.

• This project is fairly new, it only has about a year of history. It seems to gain a star every 1-2 weeks, and I hope to see the trend accelerate, but I also do not actively promote it.
• It is slowly/steadily going to become the official CLI for jsonschema, and I believe that Julian wants to drop the CLI from jsonschema at some point.

I've used homebrew but never maintained anything homebrew related.
I took a look at your tap repo, to get minimally familiar, and it seems like it lists pinned versions of dependent python packages. Is that a strict requirement for homebrew? It seems like you'd have to replicate the extras (json5, pyjson5, and toml), and I don't know how that works.

I have a personal project ( pyschemata ) in which I'm toying with an alternative way of vendoring the schemastore schemas needed here for pre-commit hooks. That could become an auto-releasing calver-versioned package, and would help isolate check-jsonschema from upstream changes. But changing dependencies rapidly would need to percolate down to supported install methods. I'm a little worried about how this could impact homebrew support.

This project is still marked beta and is doing v0.x releases. I might be more interested in revisiting once the interface stabilizes enough that I call it 1.0 and mark it production. (At which point, I intend to maintain semver-like release behavior, within reason.)

[sirosen]

Oh, also, thanks very much for your interest in this project and for taking the initiative to put together a tap repo!
I think it's super cool that you're interested in this. I'll continue to think about it and may have more conclusive thoughts after my workday.

[sudosubin]

Thanks for the comment :)

First, many users and projects use Brewfile to bootstrap most dependencies. ex) rails/Brewfile. So I believe it will be easier to use check-jsonschema when it's available via brew.

Second, brew formula does not support optional dependencies/resources. The formula file should include all the necessary dependencies. (AFAIK)

[sirosen]

After giving this some more thought and returning to this issue, I think the appropriate action here is to close as a wontfix.

Although I'm happy to offer advice or guidance to anyone looking to repackage and redistribute this code, I'm not prepared to take on the burden of maintaining a non-python distribution format. Plus, when I weigh the concerns I mentioned above about the tap formula format being somewhat-incompatible with the design of this project, I find I'm not fully convinced that this is a good idea.
There are already some problems introduced by supporting both pypi packaging and pre-commit usage, so expanding the packaging surface area further seems unwise.

Given that pipx is redistributed as a formula, I would tend to recommend using that to bootstrap usage. That will allow for things like pipx inject check-jsonschema tomli to get TOML support.