assignment with datetime64[ns, UTC] raises TypeError

[Lmmejia11]

Code Sample

# Your code here

import datetime, pandas as pd

data = [[datetime.datetime(2020, 2, 28, 13, 51, 27, tzinfo=datetime.timezone.utc)],
 [datetime.datetime(2020, 2, 28, 13, 51, 27, tzinfo=datetime.timezone.utc)],
 [datetime.datetime(2020, 2, 28, 13, 51, 27, tzinfo=datetime.timezone.utc)],
 [datetime.datetime(2020, 2, 28, 13, 51, 27, tzinfo=datetime.timezone.utc)]]
df = pd.DataFrame(data, columns=['cd'])     # dtype: datetime64[ns, UTC]
df2 = pd.DataFrame(index=df.index)

df2.loc[:,'cd2'] = df['cd']                  # no error
df2.loc[df.index,'cd3'] = df['cd']     # error
df2.loc[df.index,'cd2'] = df['cd']     # no error

Problem description

If you try to assign datetime values (with zone and indexes) to a column, it will raise TypeError: data type not understood.
No errors raise with index ':', or when the column already has the correct type. Note that this only happens if the datetime has zone information. With tzinfo=None, no errors occur.

Output of pd.show_versions()

I noticed this bug in version 1.0.1
No errors occur in 0.25.3

INSTALLED VERSIONS

commit : None
pandas : 1.0.1
numpy : 1.16.4
pytz : 2019.1
dateutil : 2.8.0
pip : 20.0.1
setuptools : 45.1.0
Cython : None
pytest : None
hypothesis : None
sphinx : 2.4.2
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : 2.8.2 (dt dec pq3 ext lo64)
jinja2 : 2.10.1
IPython : 7.8.0
pandas_datareader: None
bs4 : 4.8.2
bottleneck : None
fastparquet : None
gcsfs : 0.6.0
lxml.etree : None
matplotlib : 3.1.0
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 0.16.0
pytables : None
pytest : None
pyxlsb : None
s3fs : None
scipy : 1.2.0
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None
numba : None

[TomAugspurger]

Thanks for the report. Things look weird around

pandas/pandas/core/internals/blocks.py

Lines 842 to 857 in f25ed6f

	# current dtype cannot store value, coerce to common dtype
	find_dtype = False
	if hasattr(value, "dtype"):
	dtype = value.dtype
	find_dtype = True
	elif lib.is_scalar(value) and not isna(value):
	dtype, _ = infer_dtype_from_scalar(value, pandas_dtype=True)
	find_dtype = True
	if find_dtype:
	dtype = find_common_type([values.dtype, dtype])
	if not is_dtype_equal(self.dtype, dtype):
	b = self.astype(dtype)
	return b.setitem(indexer, value)

,

(Pdb) pp self
DatetimeBlock: 4 dtype: datetime64[ns]
(Pdb) pp value
<DatetimeArray>
['2020-02-28 13:51:27+00:00', '2020-02-28 13:51:27+00:00',
 '2020-02-28 13:51:27+00:00', '2020-02-28 13:51:27+00:00']
Length: 4, dtype: datetime64[ns, UTC]
(Pdb) pp dtype
<class 'object'>
(Pdb) pp self.values
array(['NaT', 'NaT', 'NaT', 'NaT'], dtype='datetime64[ns]')

I see a couple options, both of which are likely valuable.

1. Ensure we create a datetime64[ns, <tz>] array of all-NA values, where the tz matches the target dtype
2. Ensure that Block.setitem correctly handles datetime64[ns] -> datetime64[ns, tz].

Are you interested in working on this?

[jorisvandenbossche]

(tagged this as a regression)

[h-vishal]

In 0.25 the DatetimeArray value to be assigned was coerced to an array of Timestamp values when it was assigned to an object dtype, I think that that part of the code can be reintroduced.

    def _try_coerce_args(self, other):
        """ provide coercion to our input arguments """

        if isinstance(other, ABCDatetimeIndex):
            # May get a DatetimeIndex here. Unbox it.
            other = other.array

        if isinstance(other, DatetimeArray):
            # hit in pandas/tests/indexing/test_coercion.py
            # ::TestWhereCoercion::test_where_series_datetime64[datetime64tz]
            # when falling back to ObjectBlock.where
            other = other.astype(object)

        return other

I can work on this.