def test_cast_dtypes_inplace(self):
_cast_dtypes(self.dft)
dtypes = set(self.dft.dtypes)
self.assertEqual(
dtypes, set([np.dtype("int"),
np.dtype("float"),
np.dtype("O")]))
self.assertEqual(self.dft["F"][0], "false")
def test_schema(self):
df = pd.DataFrame({
"a": [1, 2],
"b": [None, 3],
"c": pd.array([4, np.nan], dtype="Int64")
})
_cast_dtypes(df)
names, types = _get_schema(df)
self.assertListEqual(names, ["a", "b", "c"])
self.assertListEqual(types, ["bigint", "double", "bigint"])
def test_query_builder(self):
df = pd.DataFrame({
"a": [1, 2],
"b": [None, 3],
"c": pd.array([4, np.nan], dtype="Int64")
})
_cast_dtypes(df)
q = self.writer._build_query(
"foo", "bar", list(df.itertuples(index=False, name=None)),
df.columns)
# column 'b' is handled as float64 because of null
q_expected = "INSERT INTO foo.bar (a, b, c) VALUES (1, null, 4), (2, 3.0, null)"
self.assertEqual(q, q_expected)
def test_cast_dtypes(self):
dft = _cast_dtypes(self.dft, inplace=False)
dtypes = set(dft.dtypes)
self.assertEqual(
dtypes,
set([
np.dtype("int64"),
np.dtype("float"),
np.dtype("O"),
pd.Int64Dtype()
]),
)
self.assertEqual(dft["F"][0], "false")
self.assertTrue(isinstance(dft["H"][1], str))
self.assertEqual(dft["H"][1], "[1, 2, 3]")
self.assertIsNone(dft["H"][2])
self.assertIsNone(dft["I"][2])
self.assertIsNone(dft["J"][2])
self.assertIsNone(dft["K"][1])
self.assertIsNone(dft["L"][1])
self.assertIsNone(dft["M"][1])
self.assertTrue(np.isnan(dft["N"][1]))
# Nullable int will be float dtype by pandas default
self.assertTrue(isinstance(dft["N"][0], float))
# _cast_dtypes keeps np.nan/pd.NA when None in Int64 column given
# This is for consistency of _get_schema
self.assertTrue(pd.isna(dft["O"][2]))
def test_cast_dtypes_keep_list(self):
_cast_dtypes(self.dft, keep_list=True)
dtypes = set(self.dft.dtypes)
self.assertEqual(
dtypes, set([np.dtype("int"),
np.dtype("float"),
np.dtype("O")]))
self.assertTrue(
self.dft["H"].apply(lambda x: isinstance(x, list)).all())
self.assertTrue(
self.dft["I"].apply(lambda x: isinstance(x, list)).all())
self.assertTrue(
self.dft["J"].apply(lambda x: isinstance(x, list)).all())
self.assertTrue(isinstance(self.dft["H"].iloc[0][2], int))
# numpy.ndarray containing numpy.nan will be converted as float type
self.assertTrue(isinstance(self.dft["I"].iloc[0][2], float))
self.assertTrue(isinstance(self.dft["I"].iloc[1][2], int))
self.assertTrue(self.dft["I"].iloc[0][1] is None)
def test_cast_dtypes(self):
dft = _cast_dtypes(self.dft, inplace=False)
dtypes = set(dft.dtypes)
self.assertEqual(
dtypes, set([np.dtype("int"),
np.dtype("float"),
np.dtype("O")]))
self.assertEqual(dft["F"][0], "false")
self.assertTrue(isinstance(dft["H"][1], str))
self.assertEqual(dft["H"][1], "[1, 2, 3]")
def test_cast_dtypes_nullable(self):
dft = pd.DataFrame({
"P":
pd.Series([True, False, None], dtype="boolean"),
"Q":
pd.Series(["foo", "bar", None], dtype="string"),
})
dft = _cast_dtypes(dft, inplace=False)
dtypes = set(dft.dtypes)
self.assertEqual(dtypes, set([np.dtype("O")]))
self.assertIsNone(dft["P"][2])
self.assertIsNone(dft["Q"][2])
def test_cast_dtypes_keep_list(self):
_cast_dtypes(self.dft, keep_list=True)
dtypes = set(self.dft.dtypes)
self.assertEqual(
dtypes,
set([
np.dtype("int64"),
np.dtype("float"),
np.dtype("O"),
pd.Int64Dtype()
]),
)
self.assertTrue(self.dft["H"].apply(_isinstance_or_null,
args=(list, )).all())
self.assertTrue(self.dft["I"].apply(_isinstance_or_null,
args=(list, )).all())
self.assertTrue(self.dft["J"].apply(_isinstance_or_null,
args=(list, )).all())
self.assertTrue(isinstance(self.dft["H"].iloc[0][2], int))
# numpy.ndarray containing numpy.nan will be converted as float type
self.assertTrue(isinstance(self.dft["I"].iloc[0][2], float))
self.assertTrue(isinstance(self.dft["I"].iloc[1][2], int))
self.assertIsNone(self.dft["I"].iloc[0][1])
