Make pyastgrep usable as a pre-commit hook

[ashrub-holvi]

Hi,

I played a bit for make pyastgrep usable as pre-commit hook, not really sure how practical is to use it this ways, but I guess it's cheap to have it and give people chance to try, at least for simple cases it should be good.

[holvi-noor]

Looks good to me, lets try 👍

[spookylukey]

Thanks so much for this, it's an interesting approach.

I've documented the patterns I use for this kind of thing here: https://lukeplant.me.uk/blog/posts/pyastgrep-and-custom-linting/

There are some significant issues with pyastgrep for linting, especially that non-matches can start failing silently very easily when you change Python version, due to changes in Python's AST, or just because you didn't check the XPath expression properly - the XPath is tricky enough to need its own tests. That's why I prefer the method in the post that uses explicit "can I find the needle" checks, with example of what exactly the needle looks like.

Regarding the example in the PR:

            name: No usage of legacy_function_name
            args: ['.//FunctionDef[@name="legacy_function_name"]']

This actually checks for a definition def legacy_function_name(), not usages of it. Usages are much harder to check exhaustively, and for this kind of thing I use Semgrep. So maybe the example serves as an illustration of why other tools are better for linting?

I'd be interested in your feedback!

[ashrub-holvi]

https://lukeplant.me.uk/blog/posts/pyastgrep-and-custom-linting/

Thanks!

This actually checks for a definition def legacy_function_name(), not usages of it.

yes, not carefully enough prepared example

I'd be interested in your feedback!

I think this

There are some significant issues with pyastgrep for linting, especially that non-matches can start failing silently very easily when you change Python version, due to changes in Python's AST, or just because you didn't check the XPath expression properly

is a quite good explanation, so closing it, thank you!