def test_coldrop_non_str_lbl():
"""Testing the ColDrop pipeline stage."""
df = _test_df2()
assert 2 in df.columns
stage = ColDrop(2)
res_df = stage.apply(df)
assert 2 not in res_df.columns
assert 'num1' in res_df.columns
assert False in res_df.columns
def test_coldrop_lambda():
"""Testing the ColDrop pipeline stage."""
df = _test_df()
assert 'num1' in df.columns
assert 'num2' in df.columns
stage = ColDrop(lambda col: 'num' in col.name)
res_df = stage.apply(df)
assert 'num1' not in res_df.columns
assert 'num2' not in res_df.columns
assert 'char' in res_df.columns
def test_coldrop_multi_col():
"""Testing the ColDrop pipeline stage."""
df = _test_df()
assert 'num1' in df.columns
assert 'num2' in df.columns
stage = ColDrop(['num1', 'num2'])
res_df = stage.apply(df)
assert 'num1' not in res_df.columns
assert 'num2' not in res_df.columns
assert 'char' in res_df.columns
def test_fitonly_w_coldrop():
df = _test_df()
stage = FitOnly(ColDrop('num1'))
res = stage(df)
assert 'num1' not in res.columns
assert 'num2' in res.columns
assert 'char' in res.columns
df = _test_df()
res = stage(df)
assert 'num1' in res.columns
assert 'num2' in res.columns
assert 'char' in res.columns
df = _test_df()
res = stage.fit_transform(df)
assert 'num1' not in res.columns
assert 'num2' in res.columns
assert 'char' in res.columns
df = _test_df()
res = stage(df, verbose=True)
assert 'num1' in res.columns
assert 'num2' in res.columns
assert 'char' in res.columns
def test_coldrop_one_col():
"""Testing the ColDrop pipeline stage."""
df = _test_df()
assert 'num1' in df.columns
stage = ColDrop('num1')
res_df = stage.apply(df)
assert 'num1' not in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
# make sure fit is null operation for unfittable stages
res_df = stage.fit(df)
assert 'num1' in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
res_df = stage.fit(df, verbose=True)
assert 'num1' in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
# make sure transform and fit_transform are equivalent to apply
# for unfittable stages
res_df = stage.transform(df)
assert 'num1' not in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
res_df = stage.transform(df, verbose=True)
assert 'num1' not in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
res_df = stage.fit_transform(df)
assert 'num1' not in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
res_df = stage.fit_transform(df, verbose=True)
assert 'num1' not in res_df.columns
assert 'num2' in res_df.columns
assert 'char' in res_df.columns
def test_pipeline_error(time):
"""Test exceptions at pipeline level"""
# test fit
df = _test_df()
func = lambda df: df['num1'] == df['num3']
pipeline = PdPipeline([ColByFrameFunc("Equality", func), ColDrop("B")])
with pytest.raises(PipelineApplicationError):
pipeline.fit(df, verbose=True, time=time)
# test transform
df = _test_df()
with pytest.raises(PipelineApplicationError):
pipeline.transform(df, verbose=True, time=time)
# test fit_transform
df = _test_df()
with pytest.raises(PipelineApplicationError):
pipeline.fit_transform(df, verbose=True, time=time)
def test_coldrop_missing_col():
"""Testing the ColDrop pipeline stage."""
df = _test_df()
assert 'num1' in df.columns
stage = ColDrop('num3')
with pytest.raises(FailedPreconditionError):
res_df = stage.apply(df)
res_df = stage.apply(df, exraise=False)
assert res_df.equals(df)
# make sure fit is null operation for unfittable stages
with pytest.raises(FailedPreconditionError):
res_df = stage.fit(df)
with pytest.raises(FailedPreconditionError):
res_df = stage.fit(df, verbose=True)
res_df = stage.fit(df, exraise=False)
assert res_df.equals(df)
# make sure transform and fit_transform are equivalent to apply
# for unfittable stages
with pytest.raises(FailedPreconditionError):
res_df = stage.transform(df)
with pytest.raises(FailedPreconditionError):
res_df = stage.transform(df, verbose=True)
res_df = stage.transform(df, exraise=False)
assert res_df.equals(df)
with pytest.raises(FailedPreconditionError):
res_df = stage.fit_transform(df)
with pytest.raises(FailedPreconditionError):
res_df = stage.fit_transform(df, verbose=True)
res_df = stage.fit_transform(df, exraise=False)
assert res_df.equals(df)
stage = ColDrop('num3', errors='ignore')
res_df = stage.apply(df)
assert res_df.equals(df)
# make sure fit is null operation for unfittable stages
res_df = stage.fit(df)
assert res_df.equals(df)
# make sure transform and fit_transform are equivalent to apply
# for unfittable stages
res_df = stage.fit_transform(df)
assert res_df.equals(df)
res_df = stage.transform(df)
assert res_df.equals(df)
def test_fitonly_w_coldrop_missing_col():
df = _test_df()
stage = FitOnly(ColDrop('num3'))
with pytest.raises(FailedPreconditionError):
stage.apply(df)