def test_inverse_transform_cols_map_str_and_tuples(self):
"""Test 1:1 and n:1 in the same map. """
class C(r.RecipipeTransformer):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def get_column_mapping(self):
return {"c1": "c1", ("c1", "t1"): "c1t1"}
def _transform(self, df):
df = df[["c1", "t1"]]
df.columns = ["c1", "c1t1"]
return df
def _inverse_transform(self, df):
df = df[["c1", "c1t1"]]
df.columns = ["c1", "t1"]
return df
t = C("c1", "t1")
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
df = t.inverse_transform(df)
out_cols = ["c1", "t1", "c2"]
self.assertListEqual(list(df.columns), out_cols)
def test_fit_cols_all(self):
"""When not cols are specified we need to fit all of them. """
t = RecipipeTransformerMock()
t.fit(create_df_3dtypes())
self.assertListEqual(t.cols, ["c1", "c2", "t1"])
self.assertEqual(t.n_fit, 1)
def _test_fit_check_column_mapping(self):
t = r.RecipipeTransformer()
t.get_column_mapping = MagicMock(return_value={
"c1": ["c1", "c2"],
"c2": []
})
with self.assertRaises(ValueError):
t.fit(create_df_3dtypes())
def test_transform_all_columns(self):
"""Transform a df and return the same columns. """
t = RecipipeTransformerMock()
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
self.assertListEqual(list(df.columns), ["c1", "c2", "t1"])
def test_cols_taken_from_col_map(self):
"""If no cols are given, the col_map should be used to obtain them. """
class C(r.RecipipeTransformer):
def get_column_mapping(self):
return {"c1": ["hi", "bye"]}
t = C()
t.fit(create_df_3dtypes())
self.assertListEqual(t.cols, ["c1"])
def test_transform_some_columns(self):
class C(r.RecipipeTransformer):
def _transform(self, df):
return df[self.cols]
t = C("c1", "c2")
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
self.assertListEqual(list(df.columns), ["c1", "c2", "t1"])
def test_transform_keep_original_false_and_format(self):
class C(r.RecipipeTransformer):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def _transform(self, df):
df = df[self.cols]
df.columns = [self.col_format.format(i) for i in df.columns]
return df
t = C("c1", "c2", keep_original=False, col_format="{}_out")
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
out_cols = ["c1_out", "c2_out", "t1"]
self.assertListEqual(list(df.columns), out_cols)
def test_transform_cols_map_tuples(self):
class C(r.RecipipeTransformer):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def get_column_mapping(self):
return {"c1": ("c1_1", "c1_2"), ("c1", "t1"): "c1t1"}
def _transform(self, df):
df = df[["c1", "c1", "t1"]]
df.columns = ["c1_1", "c1_2", "c1t1"]
return df
t = C("c1", "t1")
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
out_cols = ["c1_1", "c1_2", "c1t1", "c2"]
self.assertListEqual(list(df.columns), out_cols)
def test_inverse_transform_keep_original_without_original(self):
class C(r.RecipipeTransformer):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def _transform(self, df):
df = df[self.cols]
df.columns = ["c1_out", "c2_out"]
return df
def _inverse_transform(self, df):
df = df[["c1_out", "c2_out"]]
df.columns = ["c1", "c2"]
return df
t = C("c*", keep_original=True, col_format="{}_out")
df = create_df_3dtypes()
t.fit(df)
df = t.transform(df)
df = df.drop(["c1", "c2"], axis=1)
df = t.inverse_transform(df)
out_cols = ["c1", "c2", "t1"]
self.assertListEqual(list(df.columns), out_cols)
def test_fit_columns_no_match(self):
with self.assertRaises(ValueError):
cols = fit_columns(create_df_3dtypes(), ["r*"])
def test_fit_columns_cols_dtype(self):
cols = fit_columns(create_df_3dtypes(), ["c*"], int)
self.assertListEqual(cols, ["c1"])
def test_fit_columns_cols_empty(self):
cols = fit_columns(create_df_3dtypes(), [], [])
self.assertListEqual(cols, ["c1", "c2", "t1"])
def test_fit_exclude(self):
t = RecipipeTransformerMock(exclude=["c1", ["c*"]])
t.fit(create_df_3dtypes())
self.assertListEqual(t.cols, ["t1"])
self.assertEqual(t.n_fit, 1)
def test_fit_columns_no_match_no_error(self):
cols = fit_columns(create_df_3dtypes(), ["r*"], raise_error=False)
self.assertListEqual(cols, [])
def test_fit_cols_and_dtype_exclude(self):
t = RecipipeTransformerMock("c*", dtype=dict(exclude=int))
t.fit(create_df_3dtypes())
self.assertListEqual(t.cols, ["c2"])
self.assertEqual(t.n_fit, 1)
def test_fit_columns_duplicates_drop(self):
cols = fit_columns(create_df_3dtypes(), cols=["c*", "c1"])
self.assertListEqual(cols, ["c1", "c2"])
def test_fit_columns_no_dtype_in_df(self):
cols = fit_columns(create_df_3dtypes(), dtype=float)
self.assertListEqual(cols, [])
def test_fit_columns_no_cols_dtype(self):
cols = fit_columns(create_df_3dtypes())
self.assertListEqual(cols, ["c1", "c2", "t1"])
def test_fit_cols_keep_original_collision(self):
"""Keep original only works when no name collisions exist. """
t = RecipipeTransformerMock(keep_original=True)
with self.assertRaises(ValueError):
t.fit(create_df_3dtypes())
def test_fit_columns_cols(self):
cols = fit_columns(create_df_3dtypes(), ["c*"])
self.assertListEqual(cols, ["c1", "c2"])
