def test_with_dict_2_values(self, todf_cls, df):
pipe = Pipeline([
('union', DictFeatureUnion([
('b', ItemSelector('b')),
('a', ItemSelector('a')),
])),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
assert result.equals(df)
def test_with_dict_column_name_raises(self, todf_cls, df):
pipe = Pipeline([
('union', DictFeatureUnion([
('b', ItemSelector('b')),
('a', ItemSelector('a')),
])),
('to_df', todf_cls(columns=['c', 'd'])),
])
with pytest.raises(ValueError) as exc:
pipe.fit_transform(df)
expected = ("ToDataFrame with explicit column names cannot "
"transform a dictionary because the dictionary's "
"keys already determine the column names.")
assert str(exc.value) == expected
def test_with_df(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector(['a', 'b'])),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
assert result.equals(df)
def test_with_series(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector('a')),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
expected = df[['a']]
assert result.equals(expected)
def test_with_series_column_name(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector('a')),
('to_df', todf_cls(columns=['c'])),
])
result = pipe.fit_transform(df)
expected = df.rename(columns={'a': 'c'})[['c']]
assert result.equals(expected)
def test_with_dict_1_value(self, todf_cls, df):
pipe = Pipeline([
('union', DictFeatureUnion([
('a', ItemSelector('a')),
])),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
expected = df[['a']]
assert result.equals(expected)
def test_with_2d_array(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector(['a', 'b'])),
('values', FunctionTransformer(
lambda x: x.values, validate=False)),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
expected = df.rename(columns={'a': 0, 'b': 1})
assert result.equals(expected)
def test_with_list(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector('a')),
('values', FunctionTransformer(
lambda x: x.values.tolist(), validate=False)),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
expected = df.rename(columns={'a': 0})[[0]]
assert result.equals(expected)
def test_with_series_wrong_number_of_column_names_raises(
self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector('a')),
('to_df', todf_cls(columns=['c', 'd'])),
])
with pytest.raises(ValueError) as exc:
pipe.fit_transform(df)
expected = ("ToDataFrame with more than one column name cannot "
"transform a Series object.")
assert str(exc.value) == expected
def test_with_df_column_name_raises(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector(['a', 'b'])),
('to_df', todf_cls(columns=['c', 'd'])),
])
with pytest.raises(ValueError) as exc:
pipe.fit_transform(df)
expected = ("ToDataFrame with explicit column names cannot "
"transform a DataFrame because the DataFrame's "
"columns already determine the column names.")
assert str(exc.value) == expected
def test_with_list_wrong_number_of_column_names_raises(self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector('a')),
('values', FunctionTransformer(
lambda x: x.values.tolist(), validate=False)),
('to_df', todf_cls(columns=['c', 'd'])),
])
with pytest.raises(ValueError) as exc:
pipe.fit_transform(df)
expected = ("ToDataFrame with more than one column name cannot "
"transform a list.")
assert str(exc.value) == expected
def test_with_array_wrong_number_of_column_names_raises(
self, todf_cls, df):
pipe = Pipeline([
('select', ItemSelector(['a', 'b'])),
('values', FunctionTransformer(
lambda x: x.values, validate=False)),
('to_df', todf_cls(columns=['c', 'd', 'e'])),
])
with pytest.raises(ValueError) as exc:
pipe.fit_transform(df)
expected = ("ToDataFrame was given data with 2 columns but "
"was initialized with 3 column names.")
assert str(exc.value) == expected
def test_with_dict_columns_sorted(self, todf_cls):
import string
az = string.ascii_lowercase
df = pd.DataFrame({c: np.arange(3) for c in az})
pipe = Pipeline([
('union', DictFeatureUnion([
(c, ItemSelector(c)) for c in az[::-1]
])),
('to_df', todf_cls()),
])
result = pipe.fit_transform(df)
cols = result.columns.tolist()
assert cols == sorted(cols)