Pytest is broken on my computer #166
Comments
|
I just started getting lots of errors like |
|
@PhilipVinc I'm experiencing the same issue. Haven't yet figured out what's wrong. |
|
I think a doctest is the culprit: $ pytest docs/comparison.md
<the above error> |
|
For some reason, the below refuses to run as a doctest and now throws an error: from plum import dispatch, parametric
from typing import Any, Optional, Tuple, Union
import numpy as np
class NDArrayMeta(type):
def __instancecheck__(self, x):
if self.concrete:
shape, dtype = self.type_parameter
else:
shape, dtype = None, None
return (
isinstance(x, np.ndarray)
and (shape is None or x.shape == shape)
and (dtype is None or x.dtype == dtype)
)
@parametric
class NDArray(np.ndarray, metaclass=NDArrayMeta):
@classmethod
@dispatch
def __init_type_parameter__(
cls,
shape: Optional[Tuple[int, ...]],
dtype: Optional[Any],
):
"""Validate the type parameter."""
return shape, dtype
@classmethod
@dispatch
def __le_type_parameter__(
cls,
left: Tuple[Optional[Tuple[int, ...]], Optional[Any]],
right: Tuple[Optional[Tuple[int, ...]], Optional[Any]],
):
"""Define an order on type parameters. That is, check whether
`left <= right` or not."""
shape_left, dtype_left = left
shape_right, dtype_right = right
return (
(shape_right is None or shape_left == shape_right)
and (dtype_right is None or dtype_left == dtype_right)
)
@dispatch
def f(x: np.ndarray):
print("Any NP array!")
@dispatch
def f(x: NDArray[(2, 2), None]):
print("A 2x2 array!")
@dispatch
def f(x: NDArray[None, int]):
print("An int array!")
@dispatch
def f(x: NDArray[(2, 2), int]):
print("A 2x2 int array!")
f(np.ones((3, 3))) # Any NP array!
f(np.ones((3, 3), int)) # An int array!
f(np.ones((2, 2))) # A 2x2 array!
f(np.ones((2, 2), int)) # A 2x2 int array! |
|
Upgrading |
|
(Numpy 2.0 was released today and I guess that might cause some breakage) |
|
yep |
|
@nstarman Are you still experiencing issues? |
|
Yes. I just opened GalacticDynamics/galax#357, which works on my un-upgraded plum/numpy/etc. |
|
I went lower down the stack. I'm seeing the same thing in GalacticDynamics/quaxed#56 |
|
@nstarman I can confirm that I’m also seeing double registrations in some packages of mine. This seems to be a separate issue from the current thread, and revealed by the recently added warning. Do the registrations only pollute your output or actually break something? I will keep investigating on my end. |
|
Let's move this do a different thread, but the issue seems to be that plum is trying to register the same function twice! E.g. I'm seeing this and both |
Since some recent changes in the testing infrastructure, I keep getting those errors
Does anyone know why?
The text was updated successfully, but these errors were encountered: