def test_data_types(
metric: MetricTypes, dtype: str, expected: str, device: DeviceTypes
) -> None:
skip_gpu_tests_if_no_gpu(device)
x = get_vectors(dtype=dtype)
distance_matrix = pairwise_distance(x, metric=metric, device=device).compute()
assert distance_matrix.dtype.name == expected
def test_distance_euclidean(
size: typing.Tuple[int, int], chunk: typing.Tuple[int, int]
) -> None:
x = get_vectors(size=size, chunk=chunk)
distance_matrix = pairwise_distance(x, metric="euclidean")
expected_matrix = squareform(pdist(x))
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_distance_correlation(size: typing.Tuple[int, int],
chunk: typing.Tuple[int, int]) -> None:
x = get_vectors(size=size, chunk=chunk)
distance_matrix = pairwise_distance(x, metric="correlation")
distance_array = pdist(x, metric="correlation")
expected_matrix = squareform(distance_array)
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_pairwise_split_every(
size: typing.Tuple[int, int],
chunk: typing.Tuple[int, int],
split_every: int,
metric: MetricTypes,
) -> None:
x = get_vectors(size=size, chunk=chunk)
distance_matrix = pairwise_distance(x, metric=metric, split_every=split_every)
expected_matrix = squareform(pdist(x, metric=metric))
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_pairwise(
size: typing.Tuple[int, int],
chunk: typing.Tuple[int, int],
device: DeviceTypes,
metric: MetricTypes,
) -> None:
skip_gpu_tests_if_no_gpu(device)
x = get_vectors(size=size, chunk=chunk)
distance_matrix = pairwise_distance(x, metric=metric, device=device)
distance_array = pdist(x, metric=metric)
expected_matrix = squareform(distance_array)
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_missing_values(
metric: str,
metric_func: typing.Callable[[ArrayLike, ArrayLike], np.float64],
dtype: str,
) -> None:
x = get_vectors(array_type="np", dtype=dtype)
ri_times = np.random.randint(5, 20)
m, n = x.shape
for i in range(ri_times):
if dtype == "f8":
x[np.random.randint(0, m)][np.random.randint(0, m)] = np.nan
x[np.random.randint(0, m)][np.random.randint(0, m)] = np.random.randint(
-100, -1
)
distance_matrix = pairwise_distance(x, metric=metric)
expected_matrix = create_distance_matrix(x, metric_func)
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_wrong_dimension_array() -> None:
with pytest.raises(ValueError):
pairwise_distance(da.arange(6).reshape(1, 2, 3))
with pytest.raises(ValueError):
pairwise_distance(da.arange(10))
def test_invalid_device() -> None:
x = get_vectors(array_type="np")
with pytest.raises(ValueError):
pairwise_distance(x, device="invalid-device") # type: ignore[arg-type]
def test_undefined_metric() -> None:
x = get_vectors(array_type="np")
with pytest.raises(NotImplementedError):
pairwise_distance(x, metric="not-implemented-metric") # type: ignore[arg-type]
def test_distance_ndarray() -> None:
x = get_vectors(array_type="np")
distance_matrix = pairwise_distance(x, metric="euclidean")
expected_matrix = squareform(pdist(x))
np.testing.assert_almost_equal(distance_matrix, expected_matrix)
def test_data_types(dtype, expected):
x = get_vectors(dtype=dtype)
distance_matrix = pairwise_distance(x)
assert distance_matrix.dtype.name == expected
def test_data_types(metric: MetricTypes, dtype: str, expected: str) -> None:
x = get_vectors(dtype=dtype)
distance_matrix = pairwise_distance(x, metric=metric).compute()
assert distance_matrix.dtype.name == expected
