Summary

numpy's not-a-number (np.nan) will never be matched in a match-case statement. But neither ruff nor mypy complain.

The following prints out Found a number: nan, because np.nan does not match with np.nan.

import numpy as np


def my_func(number: float) -> None:
    match number:
        case np.nan:
            print("Found np.nan")
        case float(number):
            print(f"Found a number: {number}")


my_func(np.nan)

I believe using np.nan in a match-case statement should be considered a mistake and it would be cool to have a rule as part of the numpy-specific-rules