Detect inefficient math mixing float and int

[cspotcode]

Is it feasible for pyright to detect when code does something like my_float * 2 and suggest changing 2 to 2.0 to take advantage of python 3.11's "Specializing Adapting Interpreter"?

Multiplying a float by an int isn't a type error, so pyright shouldn't really flag this; it's out of scope. Nevertheless, for a linter to flag this kind of thing, it needs to have typechecking machinery, so I feel like this can only be accomplished by pyright or a pyright plugin.

I'm curious if anyone else has thoughts on this. Maybe a fork of pyright is the best approach, re-using as much pyright as possible, adding a diagnostic check for when int and float types are mixed in math operators.

[erictraut]

I consider this out of scope for a static type checker. This is more of a linter feature, so you might want to reach out to the maintainers of pylint or other popular Python linters with this request.

[cspotcode]

That makes sense and I agree. Do you know if any linters do their own type analysis sufficient to know the static types of both sides of a binary operator? I feel like probably only pyright and mypy have the necessary analysis to know.

[r3m0t]

Hi cspotcode, pylint/astroid has inference and a good tutorial on writing checkers. However it predates type annotations so I'm not sure whether it respects these. https://pylint.pycqa.org/en/latest/development_guide/contributor_guide/index.html / https://pylint.pycqa.org/projects/astroid/en/latest/inference.html