def setup_class(self):
self.a = np.array(['2020-12-31', '2021-01-01', '2021-01-02'], dtype='M')
self.mask_a = np.array([False, True, False])
self.ma = Masked(self.a, mask=self.mask_a)
self.b = np.array([['2021-01-07'], ['2021-01-31']], dtype='M')
self.mask_b = np.array([[False], [True]])
self.mb = Masked(self.b, mask=self.mask_b)
def setup_class(self):
self.a = np.array([15, 255, 0], dtype='u1')
self.mask_a = np.array([False, True, False])
self.ma = Masked(self.a, mask=self.mask_a)
self.b = np.unpackbits(self.a).reshape(6, 4)
self.mask_b = np.array([False]*15 + [True, True] + [False]*7).reshape(6, 4)
self.mb = Masked(self.b, mask=self.mask_b)
def setup_class(self):
super().setup_class()
self.qpv = self.pv << self.pv_unit
self.pv_mask = np.array([(True, False),
(False, False),
(False, True)], [('p', bool), ('v', bool)])
self.mpv = Masked(self.qpv, mask=self.pv_mask)
def test_all_explicit(self):
a1 = np.array([[1., 2.],
[3., 4.]])
a2 = np.array([[1., 0.],
[3., 4.]])
if self._data_cls is not np.ndarray:
a1 = self._data_cls(a1, self.a.unit)
a2 = self._data_cls(a2, self.a.unit)
ma1 = Masked(a1, mask=[[False, False],
[True, True]])
ma2 = Masked(a2, mask=[[False, True],
[False, True]])
ma1_eq_ma2 = ma1 == ma2
assert_array_equal(ma1_eq_ma2.unmasked, np.array([[True, False],
[True, True]]))
assert_array_equal(ma1_eq_ma2.mask, np.array([[False, True],
[True, True]]))
assert ma1_eq_ma2.all()
assert not (ma1 != ma2).all()
ma_eq1 = ma1_eq_ma2.all(1)
assert_array_equal(ma_eq1.mask, np.array([False, True]))
assert bool(ma_eq1[0]) is True
assert bool(ma_eq1[1]) is False
ma_eq0 = ma1_eq_ma2.all(0)
assert_array_equal(ma_eq0.mask, np.array([False, True]))
assert bool(ma_eq1[0]) is True
assert bool(ma_eq1[1]) is False
def setup_class(self):
self.a = np.arange(1., 7.).reshape(2, 3)
self.mask_a = np.array([[True, False, False], [False, True, False]])
self.ma = Masked(self.a, mask=self.mask_a)
self.b = np.array([2.5, 10., 3.])
self.mask_b = np.array([False, True, False])
self.mb = Masked(self.b, mask=self.mask_b)
def test_nonzero_0d(self):
res1 = Masked(1, mask=False).nonzero()
assert len(res1) == 1
assert_array_equal(res1[0], np.ones(()).nonzero()[0])
res2 = Masked(1, mask=True).nonzero()
assert len(res2) == 1
assert_array_equal(res2[0], np.zeros(()).nonzero()[0])
def choose(a, choices, out=None, mode='raise'):
"""Construct an array from an index array and a set of arrays to choose from.
Like `numpy.choose`. Masked indices in ``a`` will lead to masked output
values and underlying data values are ignored if out of bounds (for
``mode='raise'``). Any values masked in ``choices`` will be propagated
if chosen.
"""
from astropy.utils.masked import Masked
a_data, a_mask = Masked._get_data_and_mask(a)
if a_mask is not None and mode == 'raise':
# Avoid raising on masked indices.
a_data = a.filled(fill_value=0)
kwargs = {'mode': mode}
if out is not None:
if not isinstance(out, Masked):
raise NotImplementedError
kwargs['out'] = out.unmasked
data, masks = _get_data_and_masks(*choices)
data_chosen = np.choose(a_data, data, **kwargs)
if out is not None:
kwargs['out'] = out.mask
mask_chosen = np.choose(a_data, masks, **kwargs)
if a_mask is not None:
mask_chosen |= a_mask
return Masked(data_chosen, mask_chosen) if out is None else out
def setup_class(self):
self.a = np.array([[-np.inf, +np.inf, np.nan, 3., 4.]] * 2)
self.mask_a = np.array([[False] * 5, [True] * 4 + [False]])
self.ma = Masked(self.a, mask=self.mask_a)
self.b = np.array([[3.0001], [3.9999]])
self.mask_b = np.array([[True], [False]])
self.mb = Masked(self.b, mask=self.mask_b)
def setup_class(self):
super().setup_class()
self.ma = Masked(self.a, mask=self.mask_a)
self.mb = Masked(self.b, mask=self.mask_b)
self.mc = Masked(self.c, mask=self.mask_c)
self.msa = Masked(self.sa, mask=self.mask_sa)
self.msb = Masked(self.sb, mask=self.mask_sb)
def test_can_initialize_with_masked_values(self):
mcls = Masked(self.MyArray)
mms = mcls(Masked(np.asarray(self.a), mask=self.m))
assert isinstance(mms, Masked)
assert isinstance(mms, self.MyArray)
assert_array_equal(mms.unmasked, self.a)
assert_array_equal(mms.mask, self.m)
def setup_class(self):
self.a = np.array([[np.nan, np.nan, 3.], [4., 5., 6.]])
self.mask_a = np.array([[True, False, False], [False, True, False]])
self.b = np.arange(1, 7).reshape(2, 3)
self.mask_b = self.mask_a
self.ma = Masked(self.a, mask=self.mask_a)
self.mb = Masked(self.b, mask=self.mask_b)
def setup_class(self):
self.ra = np.array([3., 5., 0.]) << u.hourangle
self.dec = np.array([4., 12., 1.]) << u.deg
self.sc = SkyCoord(self.ra, self.dec)
self.mask = np.array([False, False, True])
self.mra = Masked(self.ra, self.mask)
self.mdec = Masked(self.dec, self.mask)
self.msc = SkyCoord(self.mra, self.mdec)
def test_full_like(self, value):
o = np.full_like(self.ma, value)
if value is np.ma.masked:
expected = Masked(o.unmasked, True)
else:
expected = Masked(np.empty_like(self.a))
expected[...] = value
assert_array_equal(o.unmasked, expected.unmasked)
assert_array_equal(o.mask, expected.mask)
def check(self, func, *args, **kwargs):
ma_list = kwargs.pop('ma_list', [self.ma, self.ma])
a_list = [Masked(ma).unmasked for ma in ma_list]
m_list = [Masked(ma).mask for ma in ma_list]
o = func(ma_list, *args, **kwargs)
expected = func(a_list, *args, **kwargs)
expected_mask = func(m_list, *args, **kwargs)
assert_array_equal(o.unmasked, expected)
assert_array_equal(o.mask, expected_mask)
def test_block(self):
self.check(np.block)
out = np.block([[0., Masked(1., True)],
[Masked(1, False), Masked(2, False)]])
expected = np.array([[0, 1.], [1, 2]])
expected_mask = np.array([[False, True], [False, False]])
assert_array_equal(out.unmasked, expected)
assert_array_equal(out.mask, expected_mask)
def test_equality_strings(self, op):
m1 = Masked(np.array(['a', 'b', 'c']), mask=[True, False, False])
m2 = Masked(np.array(['a', 'b', 'd']), mask=[False, False, False])
result = op(m1, m2)
assert_array_equal(result.unmasked, op(m1.unmasked, m2.unmasked))
assert_array_equal(result.mask, m1.mask | m2.mask)
result2 = op(m1, m2.unmasked)
assert_masked_equal(result2, result)
def test_indirect_creation(self):
assert self.MyArray not in Masked._masked_classes
mms = Masked(self.a, mask=self.m)
assert isinstance(mms, Masked)
assert isinstance(mms, self.MyArray)
assert_array_equal(mms.unmasked, self.a)
assert_array_equal(mms.mask, self.m)
assert self.MyArray in Masked._masked_classes
assert Masked(self.MyArray) is type(mms)
def test_viewing(self):
mms = Masked(self.a, mask=self.m)
mms2 = mms.view()
assert type(mms2) is mms.__class__
assert_masked_equal(mms2, mms)
ma = mms.view(np.ndarray)
assert type(ma) is MaskedNDArray
assert_array_equal(ma.unmasked, self.a.view(np.ndarray))
assert_array_equal(ma.mask, self.m)
def setup_class(self):
self.x = np.array([3., 5., 0.]) << u.m
self.y = np.array([4., 12., 1.]) << u.m
self.z = np.array([0., 0., 1.]) << u.m
self.c = r.CartesianRepresentation(self.x, self.y, self.z)
self.mask = np.array([False, False, True])
self.mx = Masked(self.x, self.mask)
self.my = Masked(self.y, self.mask)
self.mz = Masked(self.z, self.mask)
self.mc = r.CartesianRepresentation(self.mx, self.my, self.mz)
def test_choose_masked(self):
ma = Masked(np.array([-1, 1]), mask=[True, False]).reshape((2, 1))
out = ma.choose((self.ma, self.mb))
expected = np.choose(ma.filled(0), (self.a, self.b))
expected_mask = np.choose(ma.filled(0), (self.mask_a, self.mask_b)) | ma.mask
assert_array_equal(out.unmasked, expected)
assert_array_equal(out.mask, expected_mask)
with pytest.raises(ValueError):
ma.unmasked.choose((self.ma, self.mb))
def test_interp(self):
xp = np.arange(5.)
fp = np.array([1., 5., 6., 19., 20.])
mask_fp = np.array([False, False, False, True, False])
mfp = Masked(fp, mask=mask_fp)
x = np.array([1.5, 17.])
mask_x = np.array([False, True])
mx = Masked(x, mask=mask_x)
out = np.interp(mx, xp, mfp)
expected = np.interp(x, xp[mask_fp], fp[mask_fp])
assert_array_equal(out.unmasked, expected)
assert_array_equal(out.mask, mask_x)
def where(condition, *args):
from astropy.utils.masked import Masked
if not args:
return condition.nonzero(), None, None
condition, c_mask = Masked._get_data_and_mask(condition)
data, masks = _get_data_and_masks(*args)
unmasked = np.where(condition, *data)
mask = np.where(condition, *masks)
if c_mask is not None:
mask |= c_mask
return Masked(unmasked, mask=mask)
def test_masked_list_subclass(self):
class MyList(list):
pass
ml = MyList(range(3))
mml = Masked(ml, mask=[False, True, False])
assert isinstance(mml, Masked)
assert isinstance(mml, MyList)
assert isinstance(mml.unmasked, MyList)
assert mml.unmasked == [0, 1, 2]
assert_array_equal(mml.mask, np.array([False, True, False]))
assert Masked(MyList) is type(mml)
def test_meshgrid(self):
a = np.arange(1., 4.)
mask_a = np.array([True, False, False])
ma = Masked(a, mask=mask_a)
b = np.array([2.5, 10., 3., 4.])
mask_b = np.array([False, True, False, True])
mb = Masked(b, mask=mask_b)
oa, ob = np.meshgrid(ma, mb)
xa, xb = np.broadcast_arrays(a, b[:, np.newaxis])
ma, mb = np.broadcast_arrays(mask_a, mask_b[:, np.newaxis])
for o, x, m in ((oa, xa, ma), (ob, xb, mb)):
assert_array_equal(o.unmasked, x)
assert_array_equal(o.mask, m)
def nanfunc(a, *args, **kwargs):
from astropy.utils.masked import Masked
a, mask = Masked._get_data_and_mask(a)
if issubclass(a.dtype.type, np.inexact):
nans = np.isnan(a)
mask = nans if mask is None else (nans | mask)
if mask is not None:
a = Masked(a, mask)
if fill_value is not None:
a = a.filled(fill_value)
return np_func(a, *args, **kwargs)
def test_insert(self):
obj = (1, 1)
values = Masked([50., 25.], mask=[True, False])
out = np.insert(self.ma.flatten(), obj, values)
expected = np.insert(self.a.flatten(), obj, [50., 25.])
expected_mask = np.insert(self.mask_a.flatten(), obj, [True, False])
assert_array_equal(out.unmasked, expected)
assert_array_equal(out.mask, expected_mask)
with pytest.raises(TypeError):
np.insert(self.a.flatten(), obj, values)
with pytest.raises(TypeError):
np.insert(self.ma.flatten(), Masked(obj), values)
def test_whole_mask_setting_simple(self):
ma = Masked(self.a)
assert ma.mask.shape == ma.shape
assert not ma.mask.any()
ma.mask = True
assert ma.mask.shape == ma.shape
assert ma.mask.all()
ma.mask = [[True], [False]]
assert ma.mask.shape == ma.shape
assert_array_equal(ma.mask, np.array([[True] * 3, [False] * 3]))
ma.mask = self.mask_a
assert ma.mask.shape == ma.shape
assert_array_equal(ma.mask, self.mask_a)
assert ma.mask is not self.mask_a
assert np.may_share_memory(ma.mask, self.mask_a)
def test_regression_12978(self):
"""Regression tests for https://github.com/astropy/astropy/pull/12978"""
# This case produced incorrect results
mask = [False, True, False]
x = np.array([1, 2, 3])
xp = Masked(np.array([1, 2, 3]), mask=mask)
fp = Masked(np.array([1, 2, 3]), mask=mask)
result = np.interp(x, xp, fp)
assert_array_equal(result, x)
# This case raised a ValueError
xp = np.array([1, 3])
fp = Masked(np.array([1, 3]))
result = np.interp(x, xp, fp)
assert_array_equal(result, x)
class TestStringFunctions:
# More elaborate tests done in test_masked.py
@classmethod
def setup_class(self):
self.ma = Masked(np.arange(3), mask=[True, False, False])
def test_array2string(self):
out0 = np.array2string(self.ma)
assert out0 == '[— 1 2]'
# Arguments are interpreted as usual.
out1 = np.array2string(self.ma, separator=', ')
assert out1 == '[—, 1, 2]'
# If we do pass in a formatter, though, it should be used.
out2 = np.array2string(self.ma, separator=', ', formatter={'all': hex})
assert out2 == '[———, 0x1, 0x2]'
# Also as positional argument (no, nobody will do this!)
out3 = np.array2string(self.ma, None, None, None, ', ', '',
np._NoValue, {'int': hex})
assert out3 == out2
# But not if the formatter is not relevant for us.
out4 = np.array2string(self.ma,
separator=', ',
formatter={'float': hex})
assert out4 == out1
def test_array_repr(self):
out = np.array_repr(self.ma)
assert out == 'MaskedNDArray([—, 1, 2])'
ma2 = self.ma.astype('f4')
out2 = np.array_repr(ma2)
assert out2 == 'MaskedNDArray([——, 1., 2.], dtype=float32)'
def test_array_str(self):
out = np.array_str(self.ma)
assert out == '[— 1 2]'
def setup_class(self):
self.a = np.arange(54.).reshape(3, 3, 6)
self.mask_a = np.zeros(self.a.shape, dtype=bool)
self.mask_a[1, 1, 1] = True
self.mask_a[0, 1, 4] = True
self.mask_a[1, 2, 5] = True
self.ma = Masked(self.a, mask=self.mask_a)