class TestUnit(TestCase):
@given(sfst.get_frame_or_frame_go())
def test_basic_attributes(self, f1: Frame) -> None:
self.assertEqual(len(f1.dtypes), f1.shape[1])
# self.assertEqual(f1.shape, f1.shape)
self.assertEqual(f1.ndim, 2)
# self.assertEqual(f1.unified, len(f1.mloc) <= 1)
if f1.shape[0] > 0 and f1.shape[1] > 0:
self.assertTrue(f1.size > 0)
self.assertTrue(f1.nbytes > 0)
else:
self.assertTrue(f1.size == 0)
self.assertTrue(f1.nbytes == 0)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC))
def test_unary_operators_numeric(self, f1: Frame) -> None:
for op in UFUNC_UNARY_OPERATORS:
if op == '__invert__': # invalid on non Boolean
continue
func = getattr(operator, op)
values = f1.values
# must coerce all blocks to same type to compare to what NP does
a = func(f1.astype(values.dtype)).values
b = func(values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BOOL))
def test_unary_operators_boolean(self, f1: Frame) -> None:
for op in UFUNC_UNARY_OPERATORS:
if op != '__invert__': # valid on Boolean
continue
func = getattr(operator, op)
a = func(f1).values
b = func(f1.values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC))
def test_binary_operators_numeric(self, f1: Frame) -> None:
for op in UFUNC_BINARY_OPERATORS:
if op in {
'__matmul__',
'__pow__',
'__lshift__',
'__rshift__',
'__and__',
'__xor__',
'__or__',
'__mod__',
'__floordiv__',
}:
continue
func = getattr(operator, op)
values = f1.values
# must coerce all blocks to same type to compare to what NP does
f2 = f1.astype(values.dtype)
a = func(f2, f2).values
b = func(values, values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BOOL))
def test_binary_operators_boolean(self, f1: Frame) -> None:
for op in UFUNC_BINARY_OPERATORS:
if op not in {
'__and__',
'__xor__',
'__or__',
}:
continue
func = getattr(operator, op)
a = func(f1, f1).values
values = f1.values
b = func(values, values)
self.assertAlmostEqualArray(a, b)
@given(
sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC,
index_dtype_group=sfst.DTGroup.STRING,
min_rows=3,
max_rows=3,
min_columns=3,
max_columns=3))
def test_binary_operators_matmul(
self,
f1: Frame,
) -> None:
f2 = f1.relabel(columns=f1.index)
f3 = f2 @ f1
self.assertAlmostEqualArray(f3.values, f2.values @ f1.values)
# from hypothesis import reproduce_failure
# NOTE: was able to improve many of these, but continued to get compliated type cases, and complications
@given(
sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC_REAL,
min_rows=1,
min_columns=1))
def test_ufunc_axis(self, f1: Frame) -> None:
for attr, attrs in UFUNC_AXIS_SKIPNA.items():
if attr in ('std', 'var'):
continue
for axis in (0, 1):
values = f1.values
# must coerce all blocks to same type to compare to what NP does
# f2 = f1.astype(values.dtype)
a = getattr(f1, attr)(axis=axis).values # call the method
b = attrs.ufunc_skipna(values, axis=axis)
# if a.dtype != b.dtype:
# continue
# try:
# self.assertAlmostEqualArray(a, b)
# except:
# import ipdb; ipdb.set_trace()
# raise
@given(sfst.get_frame())
def test_frame_isin(self, f1: Frame) -> None:
value = f1.iloc[0, 0]
if (not isna_element(value) and not isinstance(value, np.datetime64)
and not isinstance(value, np.timedelta64)):
self.assertTrue(f1.isin((value, )).iloc[0, 0])
# # TODO: intger tests with pow, mod
#---------------------------------------------------------------------------
@given(sfst.get_frame_go(), sfst.get_label())
def test_frame_go_setitem(self, f1: Frame, label: tp.Hashable) -> None:
shape = f1.shape
f1['foo'] = label # type: ignore
self.assertEqual(shape[1] + 1, f1.shape[1])
@given(sfst.get_arrays_2d_aligned_rows(min_size=2, max_size=2))
def test_frame_go_extend(self, arrays: tp.Sequence[np.ndarray]) -> None:
f1 = FrameGO(arrays[0], columns=self.get_letters(arrays[0].shape[1]))
shape = f1.shape
f2 = Frame(arrays[1])
f1.extend(f2)
self.assertEqual(f1.shape[1], shape[1] + f2.shape[1])
@given(sfst.get_arrays_2d_aligned_rows(min_size=3))
def test_frame_go_extend_items(self,
arrays: tp.Sequence[np.ndarray]) -> None:
frame_array = arrays[0]
# just take first columm form 2d arrays
series_arrays = [a[:, 0] for a in arrays[1:]]
f1 = FrameGO(frame_array)
shape = f1.shape
letters = self.get_letters(len(series_arrays))
def items() -> tp.Iterator[tp.Tuple[tp.Hashable, Series]]:
for idx, label in enumerate(letters):
s = Series(series_arrays[idx], index=f1.index)
yield label, s
f1.extend_items(items())
self.assertEqual(f1.shape[1], shape[1] + len(series_arrays))
#---------------------------------------------------------------------------
# exporters
@given(sfst.get_frame_or_frame_go())
def test_frame_to_pairs(self, f1: Frame) -> None:
for i in range(0, 1):
post = f1.to_pairs(i)
if i == 1:
self.assertEqual(len(post), f1.shape[1]) # type: ignore
else:
self.assertEqual(len(post[0][1]), f1.shape[0]) # type: ignore
self.assertTrue(isinstance(post, tuple))
@given(
sfst.get_frame_or_frame_go(
dtype_group=sfst.DTGroup.BASIC,
index_dtype_group=sfst.DTGroup.BASIC,
))
def test_frame_to_pandas(self, f1: Frame) -> None:
post = f1.to_pandas()
self.assertTrue(post.shape == f1.shape)
if not f1.isna().any().any():
self.assertTrue((post.values == f1.values).all())
@given(
sfst.get_frame_or_frame_go(
dtype_group=sfst.DTGroup.BASIC,
index_dtype_group=sfst.DTGroup.BASIC,
))
def test_frame_to_parquet(self, f1: Frame) -> None:
import pyarrow
with temp_file('.parquet') as fp:
try:
f1.to_parquet(fp)
self.assertTrue(os.stat(fp).st_size > 0)
except pyarrow.lib.ArrowNotImplementedError:
# could be Byte-swapped arrays not supported
pass
@given(
sfst.get_frame_or_frame_go(
dtype_group=sfst.DTGroup.CORE,
index_dtype_group=sfst.DTGroup.CORE,
))
def test_frame_to_msgpack(self, f1: Frame) -> None:
msg = f1.to_msgpack()
f2 = Frame.from_msgpack(msg)
assert f1.equals(f2,
compare_name=True,
compare_dtype=True,
compare_class=True)
f2 = Frame.from_msgpack(f1.to_msgpack())
assert f1.equals(f2,
compare_name=True,
compare_dtype=True,
compare_class=True)
@given(
sfst.get_frame_or_frame_go(
dtype_group=sfst.DTGroup.BASIC,
index_dtype_group=sfst.DTGroup.BASIC,
))
def test_frame_to_xarray(self, f1: Frame) -> None:
xa = f1.to_xarray()
self.assertTrue(tuple(xa.keys()) == tuple(f1.columns))
@given(
sfst.get_frame(
dtype_group=sfst.DTGroup.BASIC,
index_dtype_group=sfst.DTGroup.BASIC,
))
def test_frame_to_frame_go(self, f1: Frame) -> None:
f2 = f1.to_frame_go()
f2['__new__'] = 10
self.assertTrue(len(f2.columns) == len(f1.columns) + 1)
@skip_win # type: ignore # get UnicodeEncodeError: 'charmap' codec can't encode character '\u0162' in position 0: character maps to <undefined>
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BASIC, ))
def test_frame_to_csv(self, f1: Frame) -> None:
with temp_file('.csv') as fp:
f1.to_csv(fp)
self.assertTrue(os.stat(fp).st_size > 0)
# not yet validating result, as edge cases with unusual unicode and non-unique indices are a problem
# f2 = Frame.from_csv(fp,
# index_depth=f1.index.depth,
# columns_depth=f1.columns.depth)
@skip_win # type: ignore # UnicodeEncodeError
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BASIC, ))
def test_frame_to_tsv(self, f1: Frame) -> None:
with temp_file('.txt') as fp:
f1.to_tsv(fp)
self.assertTrue(os.stat(fp).st_size > 0)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BASIC, ))
def test_frame_to_xlsx(self, f1: Frame) -> None:
with temp_file('.xlsx') as fp:
f1.to_xlsx(fp)
self.assertTrue(os.stat(fp).st_size > 0)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BASIC, ))
def test_frame_to_sqlite(self, f1: Frame) -> None:
with temp_file('.sqlite') as fp:
try:
f1.to_sqlite(fp)
self.assertTrue(os.stat(fp).st_size > 0)
except (sqlite3.IntegrityError, sqlite3.OperationalError,
OverflowError):
# some indices, after translation, are not unique
# SQLite is no case sensitive, and does not support unicide
# OverflowError: Python int too large to convert to SQLite INTEGER
pass
@given(
sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BASIC,
columns_dtype_group=sfst.DTGroup.STRING,
index_dtype_group=sfst.DTGroup.STRING))
def test_frame_to_hdf5(self, f1: Frame) -> None:
f1 = f1.rename('f1')
with temp_file('.hdf5') as fp:
try:
f1.to_hdf5(fp)
self.assertTrue(os.stat(fp).st_size > 0)
except ValueError:
# will happen for empty strings and unicde that cannot be handled by HDF5
pass
@given(sfst.get_frame_or_frame_go())
def test_frame_to_html(self, f1: Frame) -> None:
post = f1.to_html()
self.assertTrue(len(post) > 0)
@skip_win # type: ignore # UnicodeEncodeError
@given(sfst.get_frame_or_frame_go())
def test_frame_to_html_datatables(self, f1: Frame) -> None:
post = f1.to_html_datatables(show=False)
self.assertTrue(len(post) > 0)
@given(sfst.get_frame_or_frame_go())
def test_frame_to_rst(self, f1: Frame) -> None:
post = f1.to_rst()
self.assertTrue(len(post) > 0)
@given(sfst.get_frame_or_frame_go())
def test_frame_to_markdown(self, f1: Frame) -> None:
post = f1.to_markdown()
self.assertTrue(len(post) > 0)
@given(sfst.get_frame_or_frame_go())
def test_frame_to_latex(self, f1: Frame) -> None:
post = f1.to_latex()
self.assertTrue(len(post) > 0)
@given(sfst.get_frame_or_frame_go())
def test_frame_blocks_dont_have_reference_cycles(self, f1: Frame) -> None:
self.assertEqual([f1], gc.get_referrers(f1._blocks))
class TestUnit(TestCase):
@given(sfst.get_frame_or_frame_go()) # type: ignore
def test_basic_attributes(self, f1: Frame) -> None:
self.assertEqual(len(f1.dtypes), f1.shape[1])
# self.assertEqual(f1.shape, f1.shape)
self.assertEqual(f1.ndim, 2)
# self.assertEqual(f1.unified, len(f1.mloc) <= 1)
if f1.shape[0] > 0 and f1.shape[1] > 0:
self.assertTrue(f1.size > 0)
self.assertTrue(f1.nbytes > 0)
else:
self.assertTrue(f1.size == 0)
self.assertTrue(f1.nbytes == 0)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC)
) # type: ignore
def test_unary_operators_numeric(self, f1: Frame) -> None:
for op in _UFUNC_UNARY_OPERATORS:
if op == '__invert__': # invalid on non Boolean
continue
func = getattr(operator, op)
values = f1.values
# must coerce all blocks to same type to compare to what NP does
a = func(f1.astype(values.dtype)).values
b = func(values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BOOL)
) # type: ignore
def test_unary_operators_boolean(self, f1: Frame) -> None:
for op in _UFUNC_UNARY_OPERATORS:
if op != '__invert__': # valid on Boolean
continue
func = getattr(operator, op)
a = func(f1).values
b = func(f1.values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.NUMERIC)
) # type: ignore
def test_binary_operators_numeric(self, f1: Frame) -> None:
for op in _UFUNC_BINARY_OPERATORS:
if op in {
'__matmul__',
'__pow__',
'__lshift__',
'__rshift__',
'__and__',
'__xor__',
'__or__',
'__mod__',
}:
continue
func = getattr(operator, op)
values = f1.values
# must coerce all blocks to same type to compare to what NP does
f2 = f1.astype(values.dtype)
a = func(f2, f2).values
b = func(values, values)
self.assertAlmostEqualArray(a, b)
@given(sfst.get_frame_or_frame_go(dtype_group=sfst.DTGroup.BOOL)
) # type: ignore
def test_binary_operators_boolean(self, f1: Frame) -> None:
for op in _UFUNC_BINARY_OPERATORS:
if op not in {
'__and__',
'__xor__',
'__or__',
}:
continue
func = getattr(operator, op)
a = func(f1, f1).values
values = f1.values
b = func(values, values)
self.assertAlmostEqualArray(a, b)
@given(
sfst.get_frame_or_frame_go(
dtype_group=sfst.DTGroup.NUMERIC, # type: ignore
min_rows=3,
max_rows=3,
min_columns=3,
max_columns=3))
def test_binary_operators_matmul(
self,
f1: Frame,
) -> None:
f2 = f1.relabel(columns=f1.index)
f3 = f2 @ f1
self.assertAlmostEqualArray(f3.values, f2.values @ f1.values)
# from hypothesis import reproduce_failure
# NOTE: was able to improve many of these, but continued to get compliated type cases, and complications
@given(
sfst.get_frame_or_frame_go( # type: ignore
dtype_group=sfst.DTGroup.NUMERIC_REAL,
min_rows=1,
min_columns=1))
def test_ufunc_axis(self, f1: Frame) -> None:
for attr, attrs in UFUNC_AXIS_SKIPNA.items():
if attr in ('std', 'var'):
continue
for axis in (0, 1):
values = f1.values
# must coerce all blocks to same type to compare to what NP does
# f2 = f1.astype(values.dtype)
a = getattr(f1, attr)(axis=axis).values # call the method
b = attrs.ufunc_skipna(values, axis=axis)
# if a.dtype != b.dtype:
# continue
# try:
# self.assertAlmostEqualArray(a, b)
# except:
# import ipdb; ipdb.set_trace()
# raise
# # TODO: intger tests with pow, mod
#---------------------------------------------------------------------------
@given(sfst.get_frame_go(), sfst.get_label()) # type: ignore
def test_frame_go_setitem(self, f1: Frame, label: tp.Hashable) -> None:
shape = f1.shape
f1['foo'] = label # type: ignore
self.assertEqual(shape[1] + 1, f1.shape[1])
@given(sfst.get_arrays_2d_aligned_rows(min_size=2,
max_size=2)) # type: ignore
def test_frame_go_extend(self, arrays: tp.Sequence[np.ndarray]) -> None:
f1 = FrameGO(arrays[0], columns=self.get_letters(arrays[0].shape[1]))
shape = f1.shape
f2 = Frame(arrays[1])
f1.extend(f2)
self.assertEqual(f1.shape[1], shape[1] + f2.shape[1])
@given(sfst.get_arrays_2d_aligned_rows(min_size=3)) # type: ignore
def test_frame_go_extend_items(self,
arrays: tp.Sequence[np.ndarray]) -> None:
frame_array = arrays[0]
# just take first columm form 2d arrays
series_arrays = [a[:, 0] for a in arrays[1:]]
f1 = FrameGO(frame_array)
shape = f1.shape
letters = self.get_letters(len(series_arrays))
def items() -> tp.Iterator[tp.Tuple[tp.Hashable, Series]]:
for idx, label in enumerate(letters):
s = Series(series_arrays[idx], index=f1.index)
yield label, s
f1.extend_items(items())
self.assertEqual(f1.shape[1], shape[1] + len(series_arrays))
#---------------------------------------------------------------------------
# exporters
@given(sfst.get_frame_or_frame_go()) # type: ignore
def test_frame_to_pairs(self, f1: Frame) -> None:
for i in range(0, 1):
post = f1.to_pairs(i)
if i == 1:
self.assertEqual(len(post), f1.shape[1]) # type: ignore
else:
self.assertEqual(len(post[0][1]), f1.shape[0]) # type: ignore
self.assertTrue(isinstance(post, tuple))
@given(
sfst.get_frame_or_frame_go( # type: ignore
dtype_group=sfst.DTGroup.BASIC,
index_dtype_group=sfst.DTGroup.BASIC,
))
def test_frame_to_pandas(self, f1: Frame) -> None:
post = f1.to_pandas()
self.assertTrue(post.shape == f1.shape)
if not f1.isna().any().any(): # type: ignore
self.assertTrue((post.values == f1.values).all())