class TestUnit(TestCase):
@given(get_array_1d2d()) # type: ignore
def test_mloc(self, array: np.ndarray) -> None:
x = util.mloc(array)
self.assertTrue(isinstance(x, int))
@given(get_array_1d2d()) # type: ignore
def test_shape_filter(self, shape: np.ndarray) -> None:
self.assertTrue(len(util.shape_filter(shape)), 2)
@given(get_dtype_pairs()) # type: ignore
def test_resolve_dtype(self, dtype_pair: tp.Tuple[np.dtype,
np.dtype]) -> None:
x = util.resolve_dtype(*dtype_pair)
self.assertTrue(isinstance(x, np.dtype))
@given(get_dtypes(min_size=1)) # type: ignore
def test_resolve_dtype_iter(self, dtypes: tp.Iterable[np.dtype]) -> None:
x = util.resolve_dtype_iter(dtypes)
self.assertTrue(isinstance(x, np.dtype))
@given(get_labels(min_size=1)) # type: ignore
def test_resolve_type_iter(self, objects: tp.Iterable[object]) -> None:
known_types = set(
(None, type(None), bool, str, object, int, float, complex,
datetime.date, datetime.datetime, fractions.Fraction))
resolved, has_tuple, values_post = util.resolve_type_iter(objects)
self.assertTrue(resolved in known_types)
@given(get_arrays_2d_aligned_columns()) # type: ignore
def test_concat_resolved_axis_0(self, arrays: tp.List[np.ndarray]) -> None:
array = util.concat_resolved(arrays, axis=0)
self.assertEqual(array.ndim, 2)
self.assertEqual(array.dtype,
util.resolve_dtype_iter((x.dtype for x in arrays)))
@given(get_arrays_2d_aligned_rows()) # type: ignore
def test_concat_resolved_axis_1(self, arrays: tp.List[np.ndarray]) -> None:
array = util.concat_resolved(arrays, axis=1)
self.assertEqual(array.ndim, 2)
self.assertEqual(array.dtype,
util.resolve_dtype_iter((x.dtype for x in arrays)))
@given(get_dtype(), get_shape_1d2d(), get_value()) # type: ignore
def test_full_or_fill(self, dtype: np.dtype,
shape: tp.Union[tp.Tuple[int], tp.Tuple[int, int]],
value: object) -> None:
array = util.full_for_fill(dtype, shape, fill_value=value)
self.assertTrue(array.shape == shape)
if isinstance(value, (float, complex)) and np.isnan(value):
pass
else:
self.assertTrue(value in array)
@given(get_dtype()) # type: ignore
def test_dtype_to_na(self, dtype: util.DtypeSpecifier) -> None:
post = util.dtype_to_na(dtype)
self.assertTrue(post in {0, False, None, '', np.nan, util.NAT})
@given(get_array_1d2d(dtype_group=DTGroup.NUMERIC)) # type: ignore
def test_ufunc_axis_skipna(self, array: np.ndarray) -> None:
has_na = util.isna_array(array).any()
for nt in UFUNC_AXIS_SKIPNA.values():
ufunc = nt.ufunc
ufunc_skipna = nt.ufunc_skipna
# dtypes = nt.dtypes
# composable = nt.composable
# doc = nt.doc_header
# size_one_unity = nt.size_one_unity
with np.errstate(over='ignore', under='ignore', divide='ignore'):
post = util.ufunc_axis_skipna(array=array,
skipna=True,
axis=0,
ufunc=ufunc,
ufunc_skipna=ufunc_skipna)
if array.ndim == 2:
self.assertTrue(post.ndim == 1)
@given(get_array_1d2d()) # type: ignore
def test_ufunc_unique(self, array: np.ndarray) -> None:
post = util.ufunc_unique(array, axis=0)
self.assertTrue(len(post) <= array.shape[0])
@given(get_array_1d(min_size=1), st.integers()) # type: ignore
def test_roll_1d(self, array: np.ndarray, shift: int) -> None:
post = util.roll_1d(array, shift)
self.assertEqual(len(post), len(array))
self.assertEqualWithNaN(array[-(shift % len(array))], post[0])
@given(get_array_2d(min_rows=1, min_columns=1),
st.integers()) # type: ignore
def test_roll_2d(self, array: np.ndarray, shift: int) -> None:
for axis in (0, 1):
post = util.roll_2d(array, shift=shift, axis=axis)
self.assertEqual(post.shape, array.shape)
start = -(shift % array.shape[axis])
if axis == 0:
a = array[start]
b = post[0]
else:
a = array[:, start]
b = post[:, 0]
self.assertAlmostEqualValues(a, b)
@given(get_array_1d(dtype_group=DTGroup.OBJECT)) # type: ignore
def test_iterable_to_array_a(self, array: np.ndarray) -> None:
values = array.tolist()
post, _ = util.iterable_to_array(values)
self.assertAlmostEqualValues(post, values)
# explicitly giving object dtype
post, _ = util.iterable_to_array(values, dtype=util.DTYPE_OBJECT)
self.assertAlmostEqualValues(post, values)
@given(get_labels()) # type: ignore
def test_iterable_to_array_b(self, labels: tp.Iterable[tp.Any]) -> None:
post, _ = util.iterable_to_array(labels)
self.assertAlmostEqualValues(post, labels)
self.assertTrue(isinstance(post, np.ndarray))
@given(st.slices(10)) # type: ignore #pylint: disable=E1120
def test_slice_to_ascending_slice(self, key: slice) -> None:
post_key = util.slice_to_ascending_slice(key, size=10)
self.assertEqual(set(range(*key.indices(10))),
set(range(*post_key.indices(10))))
# to_datetime64
# to_timedelta64
# key_to_datetime_key
@given(get_array_1d2d()) # type: ignore
def test_array_to_groups_and_locations(self, array: np.ndarray) -> None:
groups, locations = util.array_to_groups_and_locations(array, 0)
if len(array) > 0:
self.assertTrue(len(groups) >= 1)
# always 1dm locations
self.assertTrue(locations.ndim == 1)
self.assertTrue(len(np.unique(locations)) == len(groups))
@given(get_array_1d2d()) # type: ignore
def test_isna_array(self, array: np.ndarray) -> None:
post = util.isna_array(array)
self.assertTrue(post.dtype == bool)
values = np.ravel(array)
count_na = sum(util.isna_element(x) for x in values)
self.assertTrue(np.ravel(post).sum() == count_na)
@given(get_array_1d(dtype_group=DTGroup.BOOL)) # type: ignore
def test_binary_transition(self, array: np.ndarray) -> None:
post = util.binary_transition(array)
# could be 32 via result of np.nonzero
self.assertTrue(post.dtype in (np.int32, np.int64))
# if no True in original array, result will be empty
if array.sum() == 0:
self.assertTrue(len(post) == 0)
# if all True, result is empty
elif array.sum() == len(array):
self.assertTrue(len(post) == 0)
else:
# the post selection shold always be indices that are false
self.assertTrue(array[post].sum() == 0)
@given(get_array_1d2d()) # type: ignore
def test_array_to_duplicated(self, array: np.ndarray) -> None:
if array.ndim == 2:
for axis in (0, 1):
post = util.array_to_duplicated(array, axis=axis)
if axis == 0:
unique_count = len(set(tuple(x) for x in array))
else:
unique_count = len(
set(tuple(array[:, i]) for i in range(array.shape[1])))
if unique_count < array.shape[axis]:
self.assertTrue(post.sum() > 0)
else:
post = util.array_to_duplicated(array)
# if not all value are unique, we must have some duplicated
if len(set(array)) < len(array):
self.assertTrue(post.sum() > 0)
self.assertTrue(post.dtype == bool)
@given(get_array_1d2d()) # type: ignore
def test_array_shift(self, array: np.ndarray) -> None:
for shift in (-1, 1):
for wrap in (True, False):
tests = []
post1 = util.array_shift(array=array,
shift=shift,
axis=0,
wrap=wrap)
tests.append(post1)
if array.ndim == 2:
post2 = util.array_shift(array=array,
shift=shift,
axis=1,
wrap=wrap)
tests.append(post2)
for post in tests:
self.assertTrue(array.shape == post.shape)
# type is only always maintained if we are wrapping
if wrap:
self.assertTrue(array.dtype == post.dtype)
@given(st.lists(get_array_1d(), min_size=2, max_size=2)) # type: ignore
def test_union1d(self, arrays: tp.Sequence[np.ndarray]) -> None:
post = util.union1d(arrays[0], arrays[1], assume_unique=False)
self.assertTrue(post.ndim == 1)
# nan values in complex numbers make direct comparison tricky
self.assertTrue(len(post) == len(set(arrays[0]) | set(arrays[1])))
# complex results are tricky to compare after forming sets
if (post.dtype.kind not in ('O', 'M', 'm', 'c', 'f')
and not np.isnan(post).any()):
self.assertTrue(set(post) == (set(arrays[0]) | set(arrays[1])))
@given(st.lists(get_array_1d(), min_size=2, max_size=2)) # type: ignore
def test_intersect1d(self, arrays: tp.Sequence[np.ndarray]) -> None:
post = util.intersect1d(arrays[0], arrays[1], assume_unique=False)
self.assertTrue(post.ndim == 1)
# nan values in complex numbers make direct comparison tricky
self.assertTrue(len(post) == len(set(arrays[0]) & set(arrays[1])))
if (post.dtype.kind not in ('O', 'M', 'm', 'c', 'f')
and not np.isnan(post).any()):
self.assertTrue(set(post) == (set(arrays[0]) & set(arrays[1])))
@given(get_arrays_2d_aligned_columns(min_size=2,
max_size=2)) # type: ignore
def test_union2d(self, arrays: tp.Sequence[np.ndarray]) -> None:
post = util.union2d(arrays[0], arrays[1], assume_unique=False)
if post.dtype == object:
self.assertTrue(post.ndim == 1)
else:
self.assertTrue(post.ndim == 2)
self.assertTrue(
len(post) == len(
set(util.array2d_to_tuples(arrays[0]))
| set(util.array2d_to_tuples(arrays[1]))))
@given(get_arrays_2d_aligned_columns(min_size=2,
max_size=2)) # type: ignore
def test_intersect2d(self, arrays: tp.Sequence[np.ndarray]) -> None:
post = util.intersect2d(arrays[0], arrays[1], assume_unique=False)
if post.dtype == object:
self.assertTrue(post.ndim == 1)
else:
self.assertTrue(post.ndim == 2)
self.assertTrue(
len(post) == len(
set(util.array2d_to_tuples(arrays[0]))
& set(util.array2d_to_tuples(arrays[1]))))
@given(get_arrays_2d_aligned_columns()) # type: ignore
def test_array_set_ufunc_many(self,
arrays: tp.Sequence[np.ndarray]) -> None:
for union in (True, False):
post = util.ufunc_set_iter(arrays, union=union)
if post.dtype == object:
# returned object arrays might be 2D or 1D of tuples
self.assertTrue(post.ndim in (1, 2))
else:
self.assertTrue(post.ndim == 2)
class TestUnit(TestCase):
@given(sfst.get_array_1d(dtype_group=sfst.DTGroup.NUMERIC, max_size=100))
def test_rank_1d_ordinal(self, value: np.ndarray) -> None:
a1 = rankdata(value, method='ordinal')
a2 = rank_1d(value, method='ordinal', start=1)
self.assertEqual(a1.tolist(), a2.tolist())
if len(value):
a3 = rank_1d(value, method='ordinal')
self.assertEqual(a3.min(), 0)
a4 = rank_1d(value, method='ordinal', ascending=False)
self.assertEqual(a4.min(), 0)
@given(sfst.get_array_1d(dtype_group=sfst.DTGroup.NUMERIC, max_size=100))
def test_rank_1d_dense(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
a1 = rankdata(value, method='dense')
a2 = rank_1d(value, method='dense', start=1)
self.assertEqual(a1.tolist(), a2.tolist())
if len(value):
a3 = rank_1d(value, method='dense')
self.assertEqual(a3.min(), 0)
a4 = rank_1d(value, method='dense', ascending=False)
self.assertEqual(a4.min(), 0)
@given(sfst.get_array_1d(dtype_group=sfst.DTGroup.NUMERIC, max_size=100))
def test_rank_1d_min(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
a1 = rankdata(value, method='min')
a2 = rank_1d(value, method='min', start=1)
self.assertEqual(a1.tolist(), a2.tolist())
if len(value):
a3 = rank_1d(value, method='min')
self.assertEqual(a3.min(), 0)
a4 = rank_1d(value, method='min', ascending=False)
self.assertEqual(a4.min(), 0)
@given(sfst.get_array_1d(dtype_group=sfst.DTGroup.NUMERIC, max_size=100))
def test_rank_1d_max(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
a1 = rankdata(value, method='max')
a2 = rank_1d(value, method='max', start=1)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_1d(dtype_group=sfst.DTGroup.NUMERIC, max_size=100))
def test_rank_1d_average(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
a1 = rankdata(value, method='average')
a2 = rank_1d(value, method='mean', start=1)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_2d(dtype_group=sfst.DTGroup.NUMERIC, max_rows=20, max_columns=20))
def test_rank_2d_ordinal(self, value: np.ndarray) -> None:
for axis in (0, 1):
a1 = rankdata(value, method='ordinal', axis=axis)
a2 = rank_2d(value, method='ordinal', start=1, axis=axis)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_2d(dtype_group=sfst.DTGroup.NUMERIC, max_rows=20, max_columns=20))
def test_rank_2d_dense(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
for axis in (0, 1):
a1 = rankdata(value, method='dense', axis=axis)
a2 = rank_2d(value, method='dense', start=1, axis=axis)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_2d(dtype_group=sfst.DTGroup.NUMERIC, max_rows=20, max_columns=20))
def test_rank_2d_min(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
for axis in (0, 1):
a1 = rankdata(value, method='min', axis=axis)
a2 = rank_2d(value, method='min', start=1, axis=axis)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_2d(dtype_group=sfst.DTGroup.NUMERIC, max_rows=20, max_columns=20))
def test_rank_2d_max(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
for axis in (0, 1):
a1 = rankdata(value, method='max', axis=axis)
a2 = rank_2d(value, method='max', start=1, axis=axis)
self.assertEqual(a1.tolist(), a2.tolist())
@given(sfst.get_array_2d(dtype_group=sfst.DTGroup.NUMERIC, max_rows=20, max_columns=20))
def test_rank_2d_average(self, value: np.ndarray) -> None:
# cannot compare values with NaN as scipy uses quicksort
if np.isnan(value).any():
return
for axis in (0, 1):
a1 = rankdata(value, method='average', axis=axis)
a2 = rank_2d(value, method='mean', start=1, axis=axis)
self.assertEqual(a1.tolist(), a2.tolist())
class TestUnit(TestCase):
@given(get_array_1d2d()) # type: ignore
def test_mloc(self, array: np.ndarray) -> None:
x = util.mloc(array)
self.assertTrue(isinstance(x, int))
@given(get_dtype_pairs()) # type: ignore
def test_resolve_dtype(self, dtype_pair: tp.Tuple[np.dtype, np.dtype]) -> None:
x = util.resolve_dtype(*dtype_pair)
self.assertTrue(isinstance(x, np.dtype))
@given(get_dtypes(min_size=1)) # type: ignore
def test_resolve_dtype_iter(self, dtypes: tp.Iterable[np.dtype]) -> None:
x = util.resolve_dtype_iter(dtypes)
self.assertTrue(isinstance(x, np.dtype))
@given(get_labels(min_size=1)) # type: ignore
def test_resolve_type_iter(self, objects: tp.Iterable[object]) -> None:
known_types = set((
None,
type(None),
bool,
str,
object,
int,
float,
complex,
datetime.date,
datetime.datetime,
fractions.Fraction
))
resolved, has_tuple, values_post = util.resolve_type_iter(objects)
self.assertTrue(resolved in known_types)
@given(get_arrays_2d_aligned_columns()) # type: ignore
def test_concat_resolved_axis_0(self, arrays: tp.List[np.ndarray]) -> None:
array = util.concat_resolved(arrays, axis=0)
self.assertEqual(array.ndim, 2)
self.assertEqual(array.dtype, util.resolve_dtype_iter((x.dtype for x in arrays)))
@given(get_arrays_2d_aligned_rows()) # type: ignore
def test_concat_resolved_axis_1(self, arrays: tp.List[np.ndarray]) -> None:
array = util.concat_resolved(arrays, axis=1)
self.assertEqual(array.ndim, 2)
self.assertEqual(array.dtype, util.resolve_dtype_iter((x.dtype for x in arrays)))
@given(get_dtype(), get_shape_1d2d(), get_value()) # type: ignore
def test_full_or_fill(self,
dtype: np.dtype,
shape: tp.Union[tp.Tuple[int], tp.Tuple[int, int]],
value: object) -> None:
array = util.full_for_fill(dtype, shape, fill_value=value)
self.assertTrue(array.shape == shape)
if isinstance(value, (float, complex)) and np.isnan(value):
pass
else:
self.assertTrue(value in array)
@given(get_dtype()) # type: ignore
def test_dtype_to_na(self, dtype: util.DtypeSpecifier) -> None:
post = util.dtype_to_na(dtype)
self.assertTrue(post in {0, False, None, '', np.nan, util.NAT})
@given(get_array_1d(min_size=1, dtype_group=DTGroup.NUMERIC)) # type: ignore
def test_ufunc_skipna_1d(self, array: np.ndarray) -> None:
has_na = util.isna_array(array).any()
for ufunc, ufunc_skipna, dtype in UFUNC_AXIS_SKIPNA.values():
with np.errstate(over='ignore', under='ignore'):
v1 = ufunc_skipna(array)
# this should return a single value
self.assertFalse(isinstance(v1, np.ndarray))
if has_na:
v2 = ufunc(array)
self.assertFalse(isinstance(v2, np.ndarray))
@given(get_array_1d2d()) # type: ignore
def test_ufunc_unique(self, array: np.ndarray) -> None:
post = util.ufunc_unique(array, axis=0)
self.assertTrue(len(post) <= array.shape[0])
@given(get_array_1d(min_size=1), st.integers()) # type: ignore
def test_roll_1d(self, array: np.ndarray, shift: int) -> None:
post = util.roll_1d(array, shift)
self.assertEqual(len(post), len(array))
self.assertEqualWithNaN(array[-(shift % len(array))], post[0])
@given(get_array_2d(min_rows=1, min_columns=1), st.integers()) # type: ignore
def test_roll_2d(self, array: np.ndarray, shift: int) -> None:
for axis in (0, 1):
post = util.roll_2d(array, shift=shift, axis=axis)
self.assertEqual(post.shape, array.shape)
start = -(shift % array.shape[axis])
if axis == 0:
a = array[start]
b = post[0]
else:
a = array[:, start]
b = post[:, 0]
self.assertAlmostEqualValues(a, b)
@given(get_array_1d(dtype_group=DTGroup.OBJECT)) # type: ignore
def test_iterable_to_array_a(self, array: np.ndarray) -> None:
values = array.tolist()
post, _ = util.iterable_to_array(values)
self.assertAlmostEqualValues(post, values)
# explicitly giving object dtype
post, _ = util.iterable_to_array(values, dtype=util.DTYPE_OBJECT)
self.assertAlmostEqualValues(post, values)
@given(get_labels()) # type: ignore
def test_iterable_to_array_b(self, labels: tp.Iterable[tp.Any]) -> None:
post, _ = util.iterable_to_array(labels)
self.assertAlmostEqualValues(post, labels)
self.assertTrue(isinstance(post, np.ndarray))
@given(st.slices(10)) # type: ignore
def test_slice_to_ascending_slice(self, key: slice) -> None:
post_key = util.slice_to_ascending_slice(key, size=10)
self.assertEqual(
set(range(*key.indices(10))),
set(range(*post_key.indices(10)))
)
# to_datetime64
# to_timedelta64
# key_to_datetime_key
@given(get_array_1d2d()) # type: ignore
def test_array_to_groups_and_locations(self, array: np.ndarray) -> None:
groups, locations = util.array_to_groups_and_locations(array, 0)
if len(array) > 0:
self.assertTrue(len(groups) >= 1)
# always 1dm locations
self.assertTrue(locations.ndim == 1)
self.assertTrue(len(np.unique(locations)) == len(groups))
@given(get_array_1d2d()) # type: ignore
def test_isna_array(self, array: np.ndarray) -> None:
post = util.isna_array(array)
self.assertTrue(post.dtype == bool)
values = np.ravel(array)
count_na = sum(util.isna_element(x) for x in values)
self.assertTrue(np.ravel(post).sum() == count_na)
@given(get_array_1d(dtype_group=DTGroup.BOOL)) # type: ignore
def test_binary_transition(self, array: np.ndarray) -> None:
post = util.binary_transition(array)
# could be 32 via result of np.nonzero
self.assertTrue(post.dtype in (np.int32, np.int64))
# if no True in original array, result will be empty
if array.sum() == 0:
self.assertTrue(len(post) == 0)
# if all True, result is empty
elif array.sum() == len(array):
self.assertTrue(len(post) == 0)
else:
# the post selection shold always be indices that are false
self.assertTrue(array[post].sum() == 0)