Numpy typing is a little weird

[JonoLF]

Describe the Bug

Some of the numpy types being inferred are technically correct, but not as expected.

import numpy as np
from numpy.typing import NDArray

def calculate() -> NDArray:
     return np.zeros(1000)

In a minimal example as above, I get the type hint that zeros is an attribute:

When I navigate to zeros, it takes me to the all list that exports it, but not to the function:

It also doesn't understand the NDArray return type at all:

Neovim version:

NVIM v0.12.2
Build type: Release
LuaJIT 2.1.1774638290

Pyrefly version: 1.0.0
Numpy version: 2.4.2
Python version: 3.14

Please let me know what other information is needed, not sure what logs would be helpful, but happy to provide whatever requested

Sandbox Link

No response

(Only applicable for extension issues) IDE Information

No response

[jorenham]

The problem you have with np.zeros was an issue in the numpy stubs. This was recently fixed, so this will no longer be a problem in the upcoming NumPy 2.5 release.

numpy.typing.NDArray requires you to pass the scalar type, so you'd have to write NDArray[np.float64] for it to be valid. But I agree that it's a bit strange that this causes Pyrefly to infer it as Unknown instead of NDArray[Unknown].