def test_labelencoder_unfitted():
""" Try calling `.transform()` without fitting first
"""
df = cudf.Series(np.random.choice(10, (10, )))
le = LabelEncoder()
assert not le._fitted
with pytest.raises(NotFittedError):
le.transform(df)
def test_labelencoder_unseen():
""" Try encoding a value that was not present during fitting
"""
df = cudf.Series(np.random.choice(10, (10, )))
le = LabelEncoder().fit(df)
assert le._fitted
with pytest.raises(KeyError):
le.transform(cudf.Series([-1]))
def test_unfitted_inverse_transform():
""" Try calling `.inverse_transform()` without fitting first
"""
df = cudf.Series(np.random.choice(10, (10, )))
le = LabelEncoder()
assert (not le._fitted)
with pytest.raises(RuntimeError):
le.transform(df)
def test_labelencoder_transform(length, cardinality):
""" Try fitting and then encoding a small subset of the df
"""
df = cudf.Series(np.random.choice(cardinality, (length, )))
le = LabelEncoder().fit(df)
assert le._fitted
subset = df.iloc[0:df.shape[0] // 2]
encoded = le.transform(subset)
subset_arr = _df_to_similarity_mat(subset)
encoded_arr = _df_to_similarity_mat(encoded)
assert ((encoded_arr == encoded_arr.T) == (
subset_arr == subset_arr.T)).all()