def jarnoTransf(x_p, y_p, x, y, ra, dec, log):
"""
Source: https://elonen.iki.fi/code/misc-notes/affine-fit/
Slightly modified to work better with (ra, dec) and simplified using
np.linalg.solve() instead of the original Gauss-Jordan method.
"""
from_pt = np.array([x, y]).T
to_pt = np.array([ra, dec]).T
trn = Affine_Fit(from_pt, to_pt)
print("\nTransformation is:")
print(trn.To_Str())
print("\nTransformation is:", file=log)
print(trn.To_Str(), file=log)
ra_coeffs, dec_coeffs = trn.coeffs()
# Transform cross-matched (x, y), for plotting
ra_x, dec_y = trn.Transform(np.array([x, y]))
# Fitting error as mean of distances.
err = np.sqrt((ra - ra_x)**2 + (dec - dec_y)**2).mean()
print("Fitting error = {:.9f}\n".format(err))
print("Fitting error = {:.9f}".format(err), file=log)
ra_p, dec_p = trn.Transform(np.array([x_p, y_p]))
ra_p = MaskedColumn(ra_p, name='ra')
dec_p = MaskedColumn(dec_p, name='dec')
return ra_p, dec_p, ra_x, dec_y, ra_coeffs, dec_coeffs, err
def merge_tables() -> Table:
"""Merges the tables."""
data = join(load_gaia_tyc(), load_tyc_spec(), keys=['TYC'], join_type='left')
data = join(
data, load_ascc(),
keys=['TYC'],
join_type='left',
table_names=('gaia', 'ascc'),
metadata_conflicts='silent',
)
data['SpType'] = MaskedColumn(data['SpType_gaia'].filled(data['SpType_ascc'].filled('')))
data['SpType'].mask = data['SpType'] == ''
data.remove_columns(['SpType_gaia', 'SpType_ascc'])
data = join(data, load_tyc_hd(), keys=['TYC'], join_type='left', metadata_conflicts='silent')
data = join(
data,
load_tyc_teff(),
keys=['TYC'],
join_type='left',
table_names=('gaia', 'tycteff'),
)
data['teff_val'] = MaskedColumn(
data['teff_val_gaia'].filled(data['teff_val_tycteff'].filled(np.nan)),
)
data['teff_val'].mask = np.isnan(data['teff_val'])
data.remove_columns(['teff_val_tycteff', 'teff_val_gaia'])
data = join(data, load_sao(), keys=['TYC'], join_type='left')
return data
def px2Eq(x_p, y_p, cr_m_data):
"""
Transform pixels to (ra, dec) using the correlated astrometry.net file.
"""
ra, dec = cr_m_data['field_ra'], cr_m_data['field_dec']
x, y = cr_m_data['field_x'], cr_m_data['field_y']
def func(X, a, b, c, d, e, f):
x, y = X
return a + b * x + c * y + d * x * y + e * x**2 + f * y**2
# Find transformation from (x,y) to (ra,dec)
p0 = (100., 0., 0., 0., 0., 0.)
x2ra = curve_fit(func, (x, y), ra, p0)[0]
y2de = curve_fit(func, (x, y), dec, p0)[0]
print(("ra = {:.7f} + {:.7f} * x + {:.7f} * y + {:.7f} * xy +"
" {:.7f} * x^2 + {:.7f} * y^2 +").format(*x2ra))
print(("dec = {:.7f} + {:.7f} * x + {:.7f} * y + {:.7f} * xy + "
"{:.7f} * x^2 + {:.7f} * y^2 +").format(*y2de))
# Transform pixel (x,y) coordinates into the (ra,dec) system
ra_p = func((x_p, y_p), *x2ra)
dec_p = func((x_p, y_p), *y2de)
ra_p = MaskedColumn(ra_p, name='ra')
dec_p = MaskedColumn(dec_p, name='dec')
return ra_p, dec_p
def nudgeTransf(x_p, y_p, cr_m_data):
"""
This works but gives *very* poor results (July 2019)
"""
ra, dec = cr_m_data['field_ra'], cr_m_data['field_dec']
x, y = cr_m_data['field_x'], cr_m_data['field_y']
import nudged
dom = list(map(list, zip(x, y)))
ran = list(map(list, zip(-ra, dec)))
trans = nudged.estimate(dom, ran)
cc = list(map(list, list(np.array([x_p, y_p]).T)))
ra_p, dec_p = np.array(trans.transform(cc)).T
ra_p = -ra_p
print(trans.get_matrix())
print(trans.get_rotation())
print(trans.get_scale())
print(trans.get_translation())
# plt.subplot(221);plt.title("dom");plt.scatter(x, y)
# plt.subplot(222);plt.title("ran");plt.scatter(ra, dec)
# plt.subplot(223);plt.title("(x, y)");plt.scatter(x_p, y_p)
# plt.subplot(224);plt.title("(x_t, y_t)");plt.scatter(ra_p, dec_p)
# plt.show()
ra_p = MaskedColumn(ra_p, name='ra')
dec_p = MaskedColumn(dec_p, name='dec')
return ra_p, dec_p
def test_coalesce_basic():
col1 = MaskedColumn(data=[1, 0], mask=[False, True])
col2 = MaskedColumn(data=[0, 2], mask=[True, False])
for result in [coalesce((col1, col2)), coalesce((col2, col1))]:
assert result[0] == 1
assert result[1] == 2
assert result.mask.any() == False
def _map_ids(tbl: Table) -> dict[int, int]:
gaia_ids = np.zeros(len(tbl), dtype=np.int64)
cel_ids = np.zeros(len(tbl), dtype=np.uint32)
id_idx = {}
for i, ids in enumerate(tbl['ids']):
for ident in ids.split('|'):
if ident.startswith('HIP'):
hip = int(ident[3:].strip())
assert hip not in id_idx
id_idx[hip] = i
cel_ids[i] = hip
elif ident.startswith('TYC'):
tycs = [int(t) for t in ident[3:].strip().split('-')]
tyc = tycs[0] + tycs[1] * 10000 + tycs[2] * 1000000000
assert tyc not in id_idx
id_idx[tyc] = i
if cel_ids[i] == 0:
cel_ids[i] = tyc
elif ident.startswith('Gaia DR2'):
gaia_ids[i] = int(ident[8:].strip())
tbl.add_columns([
MaskedColumn(data=gaia_ids, name='gaia'),
MaskedColumn(data=cel_ids, name='hip'),
])
return id_idx
def _fromat_column_entry(self, t):
for n in self._numeric_fields:
if n in t.colnames:
try:
if None in t[n].data:
t[n] = MaskedColumn(t[n].data,
name=n,
dtype=np.float,
mask=t[n].data == None)
else:
t[n] = MaskedColumn(t[n].data, name=n, dtype=np.float)
t[n].format = '%e'
except:
for ID, v in enumerate(t[n].data):
try:
c = ast.literal_eval(t[n].data[ID])
if type(c) == int:
t[n].data[ID] = '%d' % c
else:
t[n].data[ID] = '%e' % c
except:
pass
def test_mask_copy():
"""Test that the mask is copied when copying a table (issue #7362)."""
c = MaskedColumn([1, 2], mask=[False, True])
c2 = MaskedColumn(c, copy=True)
c2.mask[0] = True
assert np.all(c.mask == [False, True])
assert np.all(c2.mask == [True, True])
def setup_method(self, method):
self.a = MaskedColumn(name='a', data=[1, 2, 3], fill_value=1)
self.b = MaskedColumn(name='b', data=[4, 5, 6], mask=True)
self.c = MaskedColumn(name='c', data=[7, 8, 9], mask=False)
self.d_mask = np.array([False, True, False])
self.d = MaskedColumn(name='d', data=[7, 8, 7], mask=self.d_mask)
self.t = Table([self.a, self.b], masked=True)
self.ca = Column(name='ca', data=[1, 2, 3])
def merge_all() -> Table:
"""Merges the HIP and TYC data."""
hip_data = process_hip()
# extract the non-Gaia sources to make the merging easier
non_gaia = hip_data[hip_data['source_id'].mask]
# merge object data for objects in both catalogues
hip_data = join(hip_data[np.logical_not(hip_data['source_id'].mask)],
process_tyc(),
keys=['source_id'],
table_names=['hip', 'tyc'],
join_type='outer')
# Mask blank spectral type and component identifiers
for str_col in (c for c in hip_data.colnames if hip_data[c].dtype.kind == 'U'):
hip_data[str_col].mask = np.logical_or(hip_data[str_col].mask, hip_data[str_col] == '')
prefer_tyc = {'HD', 'SAO', 'Comp'}
for base_col in (c[:-4] for c in hip_data.colnames if c.endswith('_hip')):
hip_col = base_col + '_hip'
tyc_col = base_col + '_tyc'
hip_data.rename_column(hip_col, base_col)
if isinstance(hip_data[base_col], MaskedColumn):
mask = np.logical_and(hip_data[base_col].mask, hip_data[tyc_col].mask)
if base_col in prefer_tyc:
base_data = hip_data[tyc_col].filled(hip_data[base_col])
else:
base_data = hip_data[base_col].filled(hip_data[tyc_col])
hip_data[base_col] = MaskedColumn(base_data, mask=mask)
hip_data.remove_column(tyc_col)
hip_data['HIP'] = hip_data['HIP'].filled(hip_data['TYC'])
hip_data.remove_columns('TYC')
# Add the non-Gaia stars back into the dataset
hip_data = vstack([hip_data, non_gaia], join_type='outer', metadata_conflicts='silent')
# Merge SAO, preferring the values from the SAO catalogue
sao = load_sao()
sao = sao[np.isin(sao['HD'], hip_data[np.logical_not(hip_data['HD'].mask)]['HD'])]
hip_data['SAO'].mask = np.logical_or(hip_data['SAO'].mask,
np.isin(hip_data['SAO'], sao['SAO']))
hd_sao = join(hip_data[np.logical_not(hip_data['HD'].mask)],
sao,
keys=['HD'],
table_names=['xref', 'sao'],
join_type='left')
hd_sao.rename_column('SAO_xref', 'SAO')
hd_sao['SAO'] = MaskedColumn(hd_sao['SAO_sao'].filled(hd_sao['SAO']),
mask=np.logical_and(hd_sao['SAO'].mask,
hd_sao['SAO_sao'].mask))
hd_sao.remove_column('SAO_sao')
return vstack([hip_data[hip_data['HD'].mask], hd_sao], join_type='exact')
def test_add_masked_row_to_masked_table_mapping4(self):
# When adding values to a masked table, if the mask is specified as a
# dict, then keys in values should match keys in mask
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
with pytest.raises(ValueError) as exc:
t.add_row({'b': 5}, mask={'a': True})
assert exc.value.args[0] == 'keys in mask should match keys in vals'
def test_add_masked_row_to_masked_table_mismatch(self):
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
with pytest.raises(TypeError) as exc:
t.add_row([2, 5], mask={'a': 1, 'b': 0})
assert exc.value.args[0] == "Mismatch between type of vals and mask"
with pytest.raises(TypeError) as exc:
t.add_row({'b': 5, 'a': 2}, mask=[1, 0])
assert exc.value.args[0] == "Mismatch between type of vals and mask"
def setup_method(self, method):
self.c1 = MaskedColumn(name='col1',
data=[1, 2, 3],
mask=[False, False, False])
self.c2 = MaskedColumn(name='col2',
data=[4, 5, 6],
mask=[True, False, False])
self.c3 = MaskedColumn(name='col3',
data=[7, 8, 9],
mask=[False, True, False])
def setup_method(self, method):
self.a = MaskedColumn(name='a', data=[1, 2, 3], fill_value=1)
self.b = MaskedColumn(name='b', data=[4, 5, 6], mask=True)
self.c = MaskedColumn(name='c', data=[7, 8, 9], mask=False)
self.d_mask = np.array([False, True, False])
self.d = MaskedColumn(name='d', data=[7, 8, 7], mask=self.d_mask)
self.t = Table([self.a, self.b], masked=True)
self.ca = Column(name='ca', data=[1, 2, 3])
self.sc = MaskedColumn(name='sc', data=[(1, 1.), (2, 2.), (3, 3.)],
dtype='i8,f8', fill_value=(0, -1.))
def test_remove_masked_column(self):
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1, 2, 3], mask=[0, 1, 0]))
t['a'].fill_value = 42
t.add_column(MaskedColumn(name='b', data=[4, 5, 6], mask=[1, 0, 1]))
t.remove_column('b')
assert t.masked
assert np.all(t['a'] == np.array([1, 2, 3]))
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert t['a'].fill_value == 42
assert t.colnames == ['a']
def test_add_masked_row_to_masked_table_mapping1(self):
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
t.add_row({'b': 5, 'a': 2}, mask={'a': 1, 'b': 0})
t.add_row({'a': 3, 'b': 6}, mask={'b': 1, 'a': 0})
assert t.masked
assert np.all(np.array(t['a']) == np.array([1, 2, 3]))
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert np.all(np.array(t['b']) == np.array([4, 5, 6]))
assert np.all(t['b'].mask == np.array([1, 0, 1], bool))
def test_add_masked_column_to_masked_table(self):
t = Table(masked=True)
assert t.masked
t.add_column(MaskedColumn(name='a', data=[1, 2, 3], mask=[0, 1, 0]))
assert t.masked
t.add_column(MaskedColumn(name='b', data=[4, 5, 6], mask=[1, 0, 1]))
assert t.masked
assert np.all(t['a'] == np.array([1, 2, 3]))
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert np.all(t['b'] == np.array([4, 5, 6]))
assert np.all(t['b'].mask == np.array([1, 0, 1], bool))
def test_add_masked_row_to_masked_table_iterable(self):
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
t.add_row([2, 5], mask=[1, 0])
t.add_row([3, 6], mask=[0, 1])
assert t.masked
assert np.all(np.array(t['a']) == np.array([1, 2, 3]))
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert np.all(np.array(t['b']) == np.array([4, 5, 6]))
assert np.all(t['b'].mask == np.array([1, 0, 1], bool))
def test_add_masked_row_to_masked_table_mapping2(self):
# When adding values to a masked table, if the mask is specified as a
# dict, then values not specified will have mask values set to True
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
t.add_row({'b': 5}, mask={'b': 0})
t.add_row({'a': 3}, mask={'a': 0})
assert t.masked
assert t['a'][0] == 1 and t['a'][2] == 3
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert t['b'][1] == 5
assert np.all(t['b'].mask == np.array([1, 0, 1], bool))
def test_add_masked_row_to_masked_table_mapping3(self):
# When adding values to a masked table, if mask is not passed to
# add_row, then the mask should be set to False if values are present
# and True if not.
t = Table(masked=True)
t.add_column(MaskedColumn(name='a', data=[1], mask=[0]))
t.add_column(MaskedColumn(name='b', data=[4], mask=[1]))
t.add_row({'b': 5})
t.add_row({'a': 3})
assert t.masked
assert t['a'][0] == 1 and t['a'][2] == 3
assert np.all(t['a'].mask == np.array([0, 1, 0], bool))
assert t['b'][1] == 5
assert np.all(t['b'].mask == np.array([1, 0, 1], bool))
def test_masked_column_with_unit_in_qtable():
"""Test that adding a MaskedColumn with a unit to QTable issues warning"""
t = QTable()
t['a'] = MaskedColumn([1, 2])
assert isinstance(t['a'], MaskedColumn)
t['b'] = MaskedColumn([1, 2], unit=u.m)
assert isinstance(t['b'], u.Quantity)
with pytest.warns(UserWarning,
match="dropping mask in Quantity column 'c'") as w:
t['c'] = MaskedColumn([1, 2], unit=u.m, mask=[True, False])
assert len(w) == 1
assert isinstance(t['b'], u.Quantity)
def test_masked_column_with_unit_in_qtable():
"""Test that adding a MaskedColumn with a unit to QTable creates a MaskedQuantity."""
MaskedQuantity = Masked(u.Quantity)
t = QTable()
t['a'] = MaskedColumn([1, 2])
assert isinstance(t['a'], MaskedColumn)
t['b'] = MaskedColumn([1, 2], unit=u.m)
assert isinstance(t['b'], MaskedQuantity)
assert np.all(t['b'].mask == False) # noqa
t['c'] = MaskedColumn([1, 2], unit=u.m, mask=[True, False])
assert isinstance(t['c'], MaskedQuantity)
assert np.all(t['c'].mask == [True, False])
def test_coercing_fill_value_type():
"""
Test that masked column fill_value is coerced into the correct column type.
"""
# This is the original example posted on the astropy@scipy mailing list
t = Table({'a': ['1']}, masked=True)
t['a'].set_fill_value('0')
t2 = Table(t, names=['a'], dtype=[np.int32])
assert isinstance(t2['a'].fill_value, np.int32)
# Unit test the same thing.
c = MaskedColumn(['1'])
c.set_fill_value('0')
c2 = MaskedColumn(c, dtype=np.int32)
assert isinstance(c2.fill_value, np.int32)
def test_parsing_columns( columns_file ):
"""
Test parsing of COLUMN elements
<COLUMN dmtype ref >
<OPTIONMAPPING>
<OPTION>"string"</OPTION>
<ENUMLITERAL>"vodml-id"</ENUMLITERAL> or <SEMANTICCONCEPT>
</OPTIONMAPPING>
</COLUMN>
The parser pushes the Table rows into value arrays
"""
luminosity = columns_file.find_instances( LuminosityMeasurement )[0]
# COLUMN with dmtype, value (ivoa:string)
assert len(luminosity.description) == 2
assert luminosity.description[0] == "some descriptive text"
assert luminosity.description[1] == "more descriptive text"
# COLUMN with dmtype, value, unit (ivoa:RealQuantity)
# - uses astropy Quantity
# o luminosity.value = Quantity with value array length 2 + unit
# o luminosity.value[0] = Quantity with first value + unit
assert len(luminosity.value) == 2
expected_lum = numpy.array([15.718, 14.847], dtype='float32') * u.Unit('mag')
numpy.testing.assert_array_equal( expected_lum, luminosity.value )
# COLUMN with OPTIONMAPPING
# MCD NOTE: TODO - enumeration type == 'string'.. OPTIONMAPPING not parsed/interpreted
assert len(luminosity.type) == 2
assert luminosity.type[0] == "magnitude"
assert luminosity.type[1] == "magnitude"
#assert (luminosity.optionmapping) == 2
# Check multiplicity handling
# o values dimension 1 == 'row'
elem = columns_file.find_instances( MultiObj )[0]
assert len(elem.a) == 2
assert isinstance(elem.a, MaskedColumn)
assert elem.a.shape == (2,)
numpy.testing.assert_array_equal( elem.a, MaskedColumn([100.0, 100.1], dtype='float32'))
# MCD NOTE: TODO - check this, I do not think 'b' should be a List of MaskedColumn
assert len(elem.b) == 1 # b is a list of length=1
assert isinstance(elem.b[0], MaskedColumn) # of MaskedColumn
assert elem.b[0].shape == (2,2) # 2 rows x 2 values each
numpy.testing.assert_array_equal( elem.b[0], MaskedColumn([[200.0, 201.0],[200.1,201.1]], dtype='float32'))
def test_set_get_fill_value_for_str_column(self):
c = MaskedColumn(name='c', data=['xxxx', 'yyyy'], mask=[True, False])
# assert np.all(c.filled() == ['N/A', 'yyyy'])
c.fill_value = 'ABCDEF'
assert c.fill_value == 'ABCD' # string truncated to dtype length
assert np.all(c.filled() == ['ABCD', 'yyyy'])
assert np.all(c.filled('XY') == ['XY', 'yyyy'])
def test_astropy(self):
"""test astropy table
+ test null_value
+ test masked column
"""
tablemaker = CDSTablesMaker()
ntab = Table([(1, 2, None, 4, 999), (4.0, 1115.0, np.NaN, None, 999),
(1.1, 2., 999, 12.3, 12.3),
(-1.001, 2., 0, -99.12, np.NaN)],
names=['a', 'b', 'col3', 'col4'])
ntab["col4"] = MaskedColumn(ntab["col4"],
mask=[(val > 0) for val in ntab["col4"]])
tablename = ReadMeCase.__result_file("astropy")
table = tablemaker.addTable(ntab,
name=tablename,
description="test astropy table",
nullvalue=999)
col = table.get_column("col4")
col.set_null_value(-1.001)
tablemaker.writeCDSTables()
self.assertTrue(self.__test_file(tablename, 3), "astropy file")
readme = ReadMeCase.__readme_file("astropy")
with open(readme, "w") as fd:
tablemaker.makeReadMe(out=fd)
self.assertTrue(self.__test_file(readme, 5), "Readme for astropy")
sys.stderr.write("generate {0}/{1} {0}/{2}\n".format(
RESULT_DIR, tablename, readme))
def _gemini_json_to_table(json):
"""
takes a JSON object as returned from the Gemini archive webserver and turns it into an `~astropy.table.Table`
Parameters
----------
json : dict
A JSON object from the Gemini archive webserver
Returns
-------
response : `~astropy.table.Table`
"""
data_table = Table(masked=True)
for key in __keys__:
col_data = np.array([obj.get(key, None) for obj in json])
atype = str
col_mask = np.equal(col_data, None)
data_table.add_column(MaskedColumn(col_data.astype(atype), name=key, mask=col_mask))
return data_table
def setup_method(self, method):
mask = [True, False, False]
self.meta = {'a': 1, 'b': [2, 3]}
self.a = MaskedColumn(name='a',
data=[1, 2, 3],
fill_value=10,
mask=mask,
meta={'a': 1})
self.b = MaskedColumn(name='b',
data=[4.0, 5.0, 6.0],
fill_value=10.0,
mask=mask)
self.c = MaskedColumn(name='c',
data=['7', '8', '9'],
fill_value='1',
mask=mask)
def merge_gaia(tbl: Table) -> Table:
"""Merges Gaia data for non-Celestia stars."""
with gzip.open(GAIA_PATH, 'rb') as f:
gaia = votable.parse_single_table(f).to_table()
bp_rp = gaia['bp_rp'].filled(0)
bp_rp2 = bp_rp * bp_rp
gaia.add_column(
MaskedColumn(data=gaia['phot_g_mean_mag'].filled(np.nan) + 0.01760 +
bp_rp * 0.006860 + bp_rp2 * 0.1732,
name='flux',
mask=gaia['phot_g_mean_mag'].mask))
gaia.remove_columns(['phot_g_mean_mag', 'bp_rp'])
gaia.rename_column('source_id', 'gaia')
gaia.rename_column('r_est', 'dist')
has_gaia = tbl[np.logical_not(tbl['gaia'].mask)]
merged = join(has_gaia,
gaia,
keys=['gaia'],
join_type='left',
table_names=['cel', 'gaia'])
merged['ra'] = np.where(merged['ra_gaia'].mask, merged['ra_cel'],
merged['ra_gaia'])
merged['dec'] = np.where(merged['dec_gaia'].mask, merged['dec_cel'],
merged['dec_gaia'])
merged.add_columns([
MaskedColumn(data=np.where(merged['dist_gaia'].mask,
merged['dist_cel'], merged['dist_gaia']),
name='dist',
mask=np.logical_and(merged['dist_gaia'].mask,
merged['dist_cel'].mask)),
MaskedColumn(data=np.where(merged['flux_cel'].mask,
merged['flux_gaia'], merged['flux_cel']),
name='flux',
mask=np.logical_and(merged['flux_cel'].mask,
merged['flux_gaia'].mask))
])
merged.remove_columns([
'ra_cel', 'ra_gaia', 'dec_cel', 'dec_gaia', 'dist_cel', 'dist_gaia',
'flux_cel', 'flux_gaia'
])
return vstack([tbl[tbl['gaia'].mask], merged], join_type='exact')
def test_set_get_fill_value_for_structured_column(self):
assert self.sc.fill_value == np.array((0, -1.), self.sc.dtype)
sc = self.sc.copy()
assert sc.fill_value.item() == (0, -1.)
sc.fill_value = (-1, np.inf)
assert sc.fill_value == np.array((-1, np.inf), self.sc.dtype)
sc2 = MaskedColumn(sc, fill_value=(-2, -np.inf))
assert sc2.fill_value == np.array((-2, -np.inf), sc2.dtype)