def test_index_mismatch(self):
"""Indexes can be different as long as they're the same length.
The returned Series will have an index matching the first column's."""
df = pd.DataFrame(data=self.reference_data)
# Same-length columns with mismatched indexes compare just fine.
a = df.loc[:, 'A']
a.index = test_index = ['v', 'w', 'x', 'y', 'z']
result = coalesce(a, df['C'])
self.assertTrue((result.index == test_index).all())
self.assertTrue((result.index != df['C'].index).all())
self.assertTrue(
(result.values == pd.Series(['c', 'aa', 'ccc', 'cccc',
'ccccc']).values).all())
# Columns must be the same length, however.
too_short = pd.Series(['foo', 'bar'])
too_long = pd.Series(['foo', 'bar', 'baz', 'qux', 'quux', 'corge'])
with self.assertRaises(Exception):
result = coalesce(a, too_short)
with self.assertRaises(Exception):
result = coalesce(a, too_long)
def test_consider_null_cross_type(self):
"""Test the optional keyword argument test_consider_null."""
df = pd.DataFrame(data=self.reference_data)
# Don't allow cross-type comparison with mixed types...
# ...for now.
with self.assertRaises(TypeError):
coalesce(df['two'], df['C'], consider_null=['cccc'])
def test_consider_null(self):
"""Test the optional keyword argument test_consider_null."""
df = pd.DataFrame(data=self.reference_data)
# Maybe zero is a bad number. Consider it null.
result = coalesce(df['two'], df['three'], consider_null=[0])
self.assertTrue((result == pd.Series([2, 2, 2.2, 3, 3])).all())
# consider_nulls takes multiple values.
result = coalesce(df['D'], df['C'], consider_null=['d', ''])
self.assertTrue(
(result == pd.Series(['c', 'cc', 'ccc', 'cccc', 'ccccc'])).all())
def test_numeric_columns(self):
"""Test the basic case with numbers."""
df = pd.DataFrame(data=self.reference_data)
# Two columns
result = coalesce(df['one'], df['two'])
result = result.fillna('nan')
self.assertTrue((result == pd.Series([1., 2., 2.2, 'nan', 0.])).all())
# Three columns
result = coalesce(df['one'], df['two'], df['three'])
self.assertTrue((result == pd.Series([1., 2., 2.2, 3., 0.])).all())
def test_value_preservation(self):
"""Make sure valid values aren't overwritten by nulls."""
df = pd.DataFrame(data=self.reference_data)
result = coalesce(df['C'], df['A'])
self.assertTrue((result == df['C']).all())
def test_cross_type_comparison(self):
"""Cross type comparison is allowed in the standard use case."""
df = pd.DataFrame(data=self.reference_data)
result = coalesce(df['A'], df['one'], df['E'])
result = result.fillna('nan')
self.assertTrue((result == pd.Series([1, 'aa', 'nan', 'nan',
7])).all())
def test_one_column(self):
"""Test that using one column is a no-op, returning no changes."""
df = pd.DataFrame(data=self.reference_data)
for c in df.columns:
col = df.loc[:, c]
result = coalesce(col)
self.assertTrue((result.fillna('nan') == col.fillna('nan')).all())
self.assertTrue((result.index == col.index).all())
def test_string_columns(self):
"""Test the basic case with strings."""
df = pd.DataFrame(data=self.reference_data)
# Two columns
result = coalesce(df['A'], df['C'])
self.assertTrue(
(result == pd.Series(['c', 'aa', 'ccc', 'cccc', 'ccccc'])).all())
# Three columns
result = coalesce(df['A'], df['D'], df['C'])
self.assertTrue(
(result == pd.Series(['d', 'aa', 'ccc', 'cccc', 'ccccc'])).all())
# None is equivalent to NaN
result = coalesce(df['B'], df['C'])
self.assertTrue(
(result == pd.Series(['b', 'bb', 'ccc', 'cccc', 'bbbbb'])).all())