Not supported: np.float32() can accept a list and produce an ndarray

[mikolajz]

The following code works but doesn't pass type checking:

x = None  # type: Optional[np.ndarray]
x = np.float32([1.0, 0.0, 0.0])
assert isinstance(x, np.ndarray)

But I don't know if this can be expressed in mypy - when I tried adding @overload def __new__(cls, x: List[float]) -> ndarray: ..., mypy now accepted the arguments but ignored the return type and assumed this produces an np.float32.

[person142]

After discussion on the mailing list:

https://mail.python.org/pipermail/numpy-discussion/2020-April/080566.html

I think we've decided that we want to discourage such uses and instead encourage people to write

np.array([1.0, 0.0, 0.0], dtype=np.float32)

I've opened #61 to clarify that intent.

[mikolajz]

Makes sense.