BUG: RuntimeWarning emitted sporadically during __setitem__ on a masked array #23000
Description
Describe the issue:
Over in Dask's CI we are encountering a RuntimeWarning randomly in CI when running this test. Specifically in this line where we have a __setitem__ call on a masked array.
We elevate warnings from numpy to errors in the Dask test suite, so we get the following traceback (here's an example CI build where we see this):
_______________ test_setitem_extended_API_2d_mask[index1-value1] _______________
[gw0] darwin -- Python 3.11.0 /Users/runner/miniconda3/envs/test-environment/bin/python3.11
index = (slice(1, 5, 2), [7, 5])
value = masked_array(
data=[[--, --],
[--, --]],
mask=[[ True, True],
[ True, True]],
fill_value=1e+20,
dtype=float64)
@pytest.mark.parametrize(
"index, value",
[
[(1, slice(1, 7, 2)), np.ma.masked],
[(slice(1, 5, 2), [7, 5]), np.ma.masked_all((2, 2))],
],
)
def test_setitem_extended_API_2d_mask(index, value):
x = np.ma.arange(60).reshape((6, 10))
dx = da.from_array(x.data, chunks=(2, 3))
dx[index] = value
> x[index] = value
dask/array/tests/test_array_core.py:4167:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
self = masked_array(
data=[[ 0, 1,
2, 3... 56, 57,
58, 59]],
mask=False,
fill_value=999999)
indx = (slice(1, 5, 2), [7, 5])
value = masked_array(
data=[[--, --],
[--, --]],
mask=[[ True, True],
[ True, True]],
fill_value=1e+20,
dtype=float64)
def __setitem__(self, indx, value):
"""
x.__setitem__(i, y) <==> x[i]=y
Set item described by index. If value is masked, masks those
locations.
"""
if self is masked:
raise MaskError('Cannot alter the masked element.')
_data = self._data
_mask = self._mask
if isinstance(indx, str):
_data[indx] = value
if _mask is nomask:
self._mask = _mask = make_mask_none(self.shape, self.dtype)
_mask[indx] = getmask(value)
return
_dtype = _data.dtype
if value is masked:
# The mask wasn't set: create a full version.
if _mask is nomask:
_mask = self._mask = make_mask_none(self.shape, _dtype)
# Now, set the mask to its value.
if _dtype.names is not None:
_mask[indx] = tuple([True] * len(_dtype.names))
else:
_mask[indx] = True
return
# Get the _data part of the new value
dval = getattr(value, '_data', value)
# Get the _mask part of the new value
mval = getmask(value)
if _dtype.names is not None and mval is nomask:
mval = tuple([False] * len(_dtype.names))
if _mask is nomask:
# Set the data, then the mask
> _data[indx] = dval
E RuntimeWarning: invalid value encountered in cast
../../../miniconda3/envs/test-environment/lib/python3.11/site-packages/numpy/ma/core.py:3371: RuntimeWarningUnfortunately, we've not been able to reproduce the issue locally (have only seen it in CI). Though I thought it was still worth opening an issue in case NumPy devs have any insight into what might be happening.
I initially thought this might have been a mac-specific issue (possibly similar to conda-forge/numpy-feedstock#229) but e've also observed this failure on Windows (see this CI build).
Reproduce the code example:
# This is the relevant snippet extracted from the `dask/dask` test
import numpy as np
x = np.ma.arange(60).reshape((6, 10))
index = (slice(1, 5, 2), [7, 5])
value = np.ma.masked_all((2, 2))
x[index] = valueError message:
No response
Runtime information:
In [1]: import sys, numpy; print(numpy.__version__); print(sys.version)
1.24.1
3.11.0 | packaged by conda-forge | (main, Oct 25 2022, 06:24:51) [Clang 14.0.4 ]
In [2]: import numpy
In [3]: print(numpy.show_runtime())
[{'simd_extensions': {'baseline': ['SSE', 'SSE2', 'SSE3'],
'found': ['SSSE3',
'SSE41',
'POPCNT',
'SSE42',
'AVX',
'F16C',
'FMA3',
'AVX2',
'AVX512F',
'AVX512CD',
'AVX512_SKX',
'AVX512_CLX',
'AVX512_CNL',
'AVX512_ICL'],
'not_found': ['AVX512_KNL']}},
{'architecture': 'Haswell',
'filepath': '/Users/james/mambaforge/envs/bad-env/lib/libopenblasp-r0.3.21.dylib',
'internal_api': 'openblas',
'num_threads': 8,
'prefix': 'libopenblas',
'threading_layer': 'openmp',
'user_api': 'blas',
'version': '0.3.21'},
{'filepath': '/Users/james/mambaforge/envs/bad-env/lib/libomp.dylib',
'internal_api': 'openmp',
'num_threads': 8,
'prefix': 'libomp',
'user_api': 'openmp',
'version': None}]
NoneNote this is from my local machine where I'm not able to reproduce the issue.
From this dask/dask CI build, where the RuntimeWarning is emitted, I'm able to see numpy=1.24.1=py311h62c7003_0 from conda-forge is being used.
Context for the issue:
No response