Could ruff also check files according to file's shebang?

[praiskup]

$ ruff  check .
warning: No Python files found under the given path(s)
All checks passed!
$ head -1 vcs-diff-lint
#! /usr/bin/python3
$ ruff check vcs-diff-lint
All checks passed!

Even though the file is not *.py, it still is a Python file and it is worth checking.

[dhruvmanila]

Related #2192

tl;dr It's recommended to use the include / extend-include config options to explicitly list down the files that you want Ruff to check for. Adding support for it might come with a performance cost although not sure.

[dhruvmanila]

Does using include / extend-include solve your use case? If so, we can close this issue.

[praiskup]

The differential analysis tool we develop uses Ruff as one of the backends, and while we appreciate Ruff's overall performance - analyzing the shebang is worth it in our case.

The goal is to make future project enablements (= "start using vcs-diff-lint + ruff" in more projects) as trivial as possible. Explicitly asking for include/extend-include per/project raises the initial enablement barrier, and incomplete analysis is also not nice...

That said, having something like analyze-shebangs = true opt-in would be completely fine for our use case; we can afford to spend some time on this.

[praiskup]

fedora-copr/vcs-diff-lint#25 seems to help a bit here; vcs-diff-lint knows the set of changed Python, including those that do not end with *.py. So we can indeed use extend-include to at least cover those changed files, but we are still not checking unchanged non-*.py files.

[MichaReiser]

Detecting all Python files based on shebang has the downside that Ruff needs to read at least the first 80 or so bytes to determine whether the file should be included in the analysis. This is significantly more costly than just looking at a file extension, so I would not want this behavior to be the default. I'm open to adding such a feature if we can come up with an interface that doesn't penalty the default use case where this isn't needed.

[praiskup]

Detecting all Python files based on shebang has the downside that Ruff needs to read at least the first 80 or so

Could we, e.g., only include files with no "suffixes"? That would be the most common case, I assume.

Otherwise, opt-in is completely fine!

[pmp-p]

Also some html files - though valid python - have a non standard shebang (like it used to be on MS-DOS) that Ruff cannot yet support/handle ( unlike black -x / python -x , fails with SyntaxError: Expected a statement )

eg source : https://github.com/pygame-web/showroom/blob/main/pygame-scripts/org.pygame.touchpong.html

for run with python-wasm or python3 -x : https://pygame-web.github.io/showroom/pygame-scripts/org.pygame.touchpong.html

ref: psf/black#3214

[praiskup]

Also some html files - though valid python - have a non standard shebang

Indeed, this should be only an additional/optional analysis, and I don't think we need to have 100% success rate (nb. the basic .py suffix rule isn't 100% correct either).