Replace black formatter with ruff

[Chubercik]

This PR replaces black static checking suite with ruff.

It's a pretty big change, and I understand it warrants discussion, which I welcome and look forward to.

The basic premise of this change is that the ruff linter and formatter is slowly creeping its way to becoming the new standard - more importantly, it promises orders of magnitude better performance than previously available solutions.

If this is a welcome change we might want to also update Godot's code style guidelines doc entry.

---

From ruff's README file:

• 10-100x faster than existing linters (like Flake8) and formatters (like Black)
• Installable via pip
• pyproject.toml support
• Python 3.12 compatibility
• Drop-in parity with Flake8, isort, and Black
• Built-in caching, to avoid re-analyzing unchanged files
• Fix support, for automatic error correction (e.g., automatically remove unused imports)
• Over 800 built-in rules, with native re-implementations of popular Flake8 plugins, like flake8-bugbear
• First-party editor integrations for VS Code and more
• Monorepo-friendly, with hierarchical and cascading configuration

---

Local usage ought to be similar to that of black:

pip3 install ruff --user

ruff -l 120 <path/to/file(s)>

[Chubercik]

Messed up SCons pretty badly, will tend to that tomorrow.. 😪

[AThousandShips]

You used a merge commit to update your branch, please use rebase in the future instead, see here

[Repiteo]

Holy hell, ruff is making an insanely good first impression! A linter/formatter that has the capability of recognizing SConstruct in editor without major tweaking is amazing. VSCode can even recognize a .ruff.toml file natively, so local formatting can actually occur right out of the box. When you have the chance, be sure to add a .ruff.toml to the repo root with this content:

extend-exclude = ["thirdparty"]
extend-include = ["SConstruct", "SCSub"]
line-length = 120
target-version = "py38"

[Repiteo]

After some adjustments to ensure this is synced with pyproject.toml & suppressing certain rules in SConstruct/SCsub files, everything looks good to go!

We can revisit those suppressed rules once this is established as a baseline, probably around the same point we decide which extra rules we should enable.

[Riteo]

Looks quite nice!

[Chubercik]

@Repiteo's changes/additions lgtm!

[akien-mga]

That's technically correct, but a bit sad to lose the alignment.

[Chubercik]

Nothing we can do I'm afraid, seems to be a shortcoming of idiomatic Python.

[akien-mga]

This could use a multiline f-string, but I'm not sure this would make things much better for 3 lines (the big drawback of multiline strings is losing the indentation, which also looks bad). So I guess it's fine as is.

[Chubercik]

One other way of solving this would be to drop f-strings in this instance altogether:

if godot_platform != "windows":
    configurations += [
        '<ProjectConfiguration Include="editor|' + proj_plat + '">',
        "  <Configuration>editor</Configuration>",
        "  <Platform>" + proj_plat + "</Platform>",
        "</ProjectConfiguration>",
    ]

Though that would probably raise a warning about not using f-strings, so there's no winning here.

[akien-mga]

Looks good to me overall.

[Calinou]

Tested locally, it works as expected.

I suggest adding this in pyproject.toml (after the [tool.ruff] section) to enable isort functionality in ruff check --fix (note that this does not affect ruff format):

[tool.ruff.lint]
extend-select = [
	"I", # isort
]

This way, contributors won't need to install isort separately anymore.

[qarmin]

I think that also some other usable rule groups could be enabled here in this or other MR - https://docs.astral.sh/ruff/rules/ like:

• https://docs.astral.sh/ruff/rules/#flake8-bugbear-b
• https://docs.astral.sh/ruff/rules/#flake8-simplify-sim
• https://docs.astral.sh/ruff/rules/#perflint-perf
• some lints from https://docs.astral.sh/ruff/rules/#ruff-specific-rules-ruf

[akien-mga]

Let's assess those in a follow-up PR, I think this one is big enough for now :)

[akien-mga]

Is this done by isort? What's the rule applied here?

It makes sense to me to separate stdlib imports from our local ones, but I'd then expect core_builders and methods to be adjacent, not separated.

Or is it because core_builders is actually local (relative path) and methods was added to sys.path?

[akien-mga]

Thanks!

[bruvzg]

This is causing issues on macOS, since the downloaded ruff binary is unsignedunnotarized and can't run (it's fixable by removing the quarantine flag or re-signing it manually, but pretty annoying).