def test_index_with_float_array_raises(self, multi_indexing):
X = np.zeros(10)
i = np.arange(3, 0.5)
with pytest.raises(IndexError) as exc:
multi_indexing(X, i)
assert exc.value.args[0] == (
"arrays used as indices must be of integer (or boolean) type")
def test_index_with_float_array_raises(self, multi_indexing):
X = np.zeros(10)
i = np.arange(3, 0.5)
with pytest.raises(IndexError) as exc:
multi_indexing(X, i)
assert exc.value.args[0] == (
"arrays used as indices must be of integer (or boolean) type")
def test_index_with_float_array_raises(self, multi_indexing):
# sklearn < 0.22 raises IndexError with msg0
# sklearn >= 0.22 raises ValueError with msg1
X = np.zeros(10)
i = np.arange(3, 0.5)
with pytest.raises((IndexError, ValueError)) as exc:
multi_indexing(X, i)
msg0 = "arrays used as indices must be of integer (or boolean) type"
msg1 = ("No valid specification of the columns. Only a scalar, list or "
"slice of all integers or all strings, or boolean mask is allowed")
result = exc.value.args[0]
assert result in (msg0, msg1)
def test_boolean_index_2d_with_torch_tensor(self, multi_indexing):
X = torch.LongTensor(np.arange(9).reshape(3, 3))
i = np.eye(3).astype(bool)
res = multi_indexing(X, i)
expected = torch.LongTensor([0, 4, 8])
assert all(res == expected)
def test_boolean_index_2d_with_torch_tensor(self, multi_indexing):
X = torch.LongTensor(np.arange(9).reshape(3, 3))
i = np.eye(3).astype(bool)
res = multi_indexing(X, i)
expected = torch.LongTensor([0, 4, 8])
assert all(res == expected)
def test_list_of_dataframe_and_series_slice(self, multi_indexing, pd):
data = [
pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0]),
pd.Series(data=[0, 1, 2], index=[2, 1, 0]),
]
result = multi_indexing(data, np.s_[:2])
assert result[0].equals(
pd.DataFrame({'a': [0, 1], 'b': [3, 4]}, index=[2, 1]))
assert result[1].equals(pd.Series(data=[0, 1], index=[2, 1]))
def test_list_of_dataframe_and_series_slice(self, multi_indexing, pd):
data = [
pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0]),
pd.Series(data=[0, 1, 2], index=[2, 1, 0]),
]
result = multi_indexing(data, np.s_[:2])
assert result[0].equals(
pd.DataFrame({'a': [0, 1], 'b': [3, 4]}, index=[2, 1]))
assert result[1].equals(pd.Series(data=[0, 1], index=[2, 1]))
def test_dict_of_torch_tensors(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result.keys() == expected.keys()
for k in result:
try:
val = result[k].long().numpy()
except AttributeError:
val = result[k]
assert np.allclose(val, expected[k])
def test_list_of_dataframe_and_series(self, multi_indexing, pd):
data = [
pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0]),
pd.Series(data=[0, 1, 2], index=[2, 1, 0]),
]
result = multi_indexing(data, 0)
assert result[0].equals(
pd.Series(data=[0, 3], index=['a', 'b'], name=2))
assert result[1] == 0
def test_list_of_dataframe_and_series(self, multi_indexing, pd):
data = [
pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0]),
pd.Series(data=[0, 1, 2], index=[2, 1, 0]),
]
result = multi_indexing(data, 0)
assert result[0].equals(
pd.Series(data=[0, 3], index=['a', 'b'], name=2))
assert result[1] == 0
def test_dict_of_torch_tensors(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result.keys() == expected.keys()
for k in result:
try:
val = result[k].long().numpy()
except AttributeError:
val = result[k]
assert np.allclose(val, expected[k])
def test_mixed_data(self, multi_indexing):
data = [
[1, 2, 3],
np.arange(3),
torch.arange(3, 6),
{'a': [4, 5, 6], 'b': [7, 8, 9]},
]
result = multi_indexing(data, 0)
expected = [1, 0, 3, {'a': 4, 'b': 7}]
assert result == expected
def test_mixed_data(self, multi_indexing):
data = [
[1, 2, 3],
np.arange(3),
torch.arange(3, 6),
{'a': [4, 5, 6], 'b': [7, 8, 9]},
]
result = multi_indexing(data, 0)
expected = [1, 0, 3, {'a': 4, 'b': 7}]
assert result == expected
def test_mixed_data_slice(self, multi_indexing):
data = [
[1, 2, 3],
np.arange(3),
torch.arange(3, 6),
{'a': [4, 5, 6], 'b': [7, 8, 9]},
]
result = multi_indexing(data, np.s_[:2])
assert result[0] == [1, 2]
assert np.allclose(result[1], np.arange(2))
assert np.allclose(result[2].long().numpy(), np.arange(3, 5))
assert result[3] == {'a': [4, 5], 'b': [7, 8]}
def test_mixed_data_slice(self, multi_indexing):
data = [
[1, 2, 3],
np.arange(3),
torch.arange(3, 6),
{'a': [4, 5, 6], 'b': [7, 8, 9]},
]
result = multi_indexing(data, np.s_[:2])
assert result[0] == [1, 2]
assert np.allclose(result[1], np.arange(2))
assert np.allclose(result[2].long().numpy(), np.arange(3, 5))
assert result[3] == {'a': [4, 5], 'b': [7, 8]}
def test_pandas_dataframe_slice(self, multi_indexing, pd):
import pandas as pd
df = pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0])
result = multi_indexing(df, np.s_[:2])
expected = pd.DataFrame({'a': [0, 1], 'b': [3, 4]}, index=[2, 1])
assert result.equals(expected)
def test_pandas_dataframe(self, multi_indexing, pd):
df = pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0])
result = multi_indexing(df, 0)
# Note: taking one row of a DataFrame returns a Series
expected = pd.Series(data=[0, 3], index=['a', 'b'], name=2)
assert result.equals(expected)
def test_pandas_series(self, multi_indexing, pd):
series = pd.Series(data=[0, 1, 2], index=[2, 1, 0])
result = multi_indexing(series, 0)
assert result == 0
def test_index_torch_tensor_with_numpy_int_array(self, multi_indexing):
X = torch.zeros((1000, 10))
i = np.arange(100)
result = multi_indexing(X, i)
assert (result == X[:100]).all()
def test_pandas_dataframe_slice(self, multi_indexing, pd):
import pandas as pd
df = pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0])
result = multi_indexing(df, np.s_[:2])
expected = pd.DataFrame({'a': [0, 1], 'b': [3, 4]}, index=[2, 1])
assert result.equals(expected)
def test_list(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert np.allclose(result, expected)
def test_pandas_dataframe(self, multi_indexing, pd):
df = pd.DataFrame({'a': [0, 1, 2], 'b': [3, 4, 5]}, index=[2, 1, 0])
result = multi_indexing(df, 0)
# Note: taking one row of a DataFrame returns a Series
expected = pd.Series(data=[0, 3], index=['a', 'b'], name=2)
assert result.equals(expected)
def test_index_torch_tensor_with_numpy_int_array(self, multi_indexing):
X = torch.zeros((1000, 10))
i = np.arange(100)
result = multi_indexing(X, i)
assert (result == X[:100]).all()
def test_dict_of_lists(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result == expected
def test_boolean_index_2d(self, multi_indexing):
X = np.arange(9).reshape(3, 3)
i = np.eye(3).astype(bool)
result = multi_indexing(X, i)
expected = np.asarray([0, 4, 8])
assert np.allclose(result, expected)
def test_dict_of_arrays(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result.keys() == expected.keys()
for k in result:
assert np.allclose(result[k], expected[k])
def test_index_torch_tensor_with_numpy_bool_array(self, multi_indexing):
X = torch.zeros((1000, 10))
i = np.asarray([True] * 100 + [False] * 900)
result = multi_indexing(X, i)
assert (result == X[:100]).all()
def test_pandas_series(self, multi_indexing, pd):
series = pd.Series(data=[0, 1, 2], index=[2, 1, 0])
result = multi_indexing(series, 0)
assert result == 0
def test_pandas_series_slice(self, multi_indexing, pd):
series = pd.Series(data=[0, 1, 2], index=[2, 1, 0])
result = multi_indexing(series, np.s_[:2])
expected = pd.Series(data=[0, 1], index=[2, 1])
assert result.equals(expected)
def test_dict_of_lists(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result == expected
def test_dict_of_arrays(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert result.keys() == expected.keys()
for k in result:
assert np.allclose(result[k], expected[k])
def test_pandas_series_slice(self, multi_indexing, pd):
series = pd.Series(data=[0, 1, 2], index=[2, 1, 0])
result = multi_indexing(series, np.s_[:2])
expected = pd.Series(data=[0, 1], index=[2, 1])
assert result.equals(expected)
def test_index_torch_tensor_with_numpy_bool_array(self, multi_indexing):
X = torch.zeros((1000, 10))
i = np.asarray([True] * 100 + [False] * 900)
result = multi_indexing(X, i)
assert (result == X[:100]).all()
def test_list(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i)
assert np.allclose(result, expected)
def test_boolean_index_2d(self, multi_indexing):
X = np.arange(9).reshape(3, 3)
i = np.eye(3).astype(bool)
result = multi_indexing(X, i)
expected = np.asarray([0, 4, 8])
assert np.allclose(result, expected)
def test_torch_tensor(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i).long().numpy()
assert np.allclose(result, expected)
def test_sparse_csr_matrix(self, multi_indexing, data, i, expected):
data = sparse.csr_matrix(data)
result = multi_indexing(data, i).toarray()
assert np.allclose(result, expected)
def test_torch_tensor(self, multi_indexing, data, i, expected):
result = multi_indexing(data, i).long().numpy()
assert np.allclose(result, expected)
