BUG: Construction of Series with NAs and interval-of-datetime dtype raises error

[tlaytongoogle]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Code Sample, a copy-pastable example

import pandas as pd

print(pd.Series(data=[None], dtype='interval[datetime64[ns]]'))

Problem description

Instantiating a Series with NA values in data fails if dtype is specified as interval[datetime64[ns]], raising an error:

Traceback (most recent call last):
File "main.py", line 3, in
print(pd.Series(
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/series.py", line 364, in init
data = sanitize_array(data, index, dtype, copy, raise_cast_failure=True)
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/construction.py", line 480, in sanitize_array
subarr = _try_cast(data, dtype, copy, raise_cast_failure)
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/construction.py", line 579, in _try_cast
subarr = array_type(arr, dtype=dtype, copy=copy)
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/arrays/interval.py", line 280, in _from_sequence
return cls(scalars, dtype=dtype, copy=copy)
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/arrays/interval.py", line 201, in new
return cls._simple_new(
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/arrays/interval.py", line 227, in _simple_new
left = left.astype(dtype.subtype)
File "/opt/virtualenvs/python3/lib/python3.8/site-packages/pandas/core/indexes/numeric.py", line 355, in astype
raise TypeError(
TypeError: Cannot convert Float64Index to dtype datetime64[ns]; integer values are required for conversion

Expected Output

0    NaN
dtype: interval

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 2cb9652
python : 3.8.10.final.0
python-bits : 64
OS : Linux
OS-release : 5.4.0-1043-gcp
Version : #46-Ubuntu SMP Mon Apr 19 19:17:04 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : en_US.UTF-8
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.2.4
numpy : 1.20.3
pytz : 2021.1
dateutil : 2.8.1
pip : 19.3.1
setuptools : 56.0.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : 3.0.1
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.4.2
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : 1.4.15
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[mzeitlin11]

Thanks for reporting this @tlaytongoogle! Could you please add an informative title to the issue?

[tlaytongoogle]

Title added.

[weikhor]

Hi, I try to work on this problem. I am not sure whether this is considered bug. I notice that in this file pandas\tests\indexes\interval\test_astype.py, the function

index = IntervalIndex.from_arrays(
                  [-1.5, np.nan, 0.0, 0.0, 1.5], [-0.5, np.nan, 1.0, 1.0, 3.0], closed="both"
             )

@pytest.mark.parametrize("subtype", ["datetime64[ns]", "timedelta64[ns]"])
    def test_subtype_datetimelike(self, index, subtype):
        dtype = IntervalDtype(subtype, "right")
        msg = "Cannot convert .* to .*; subtypes are incompatible"
        with pytest.raises(TypeError, match=msg):
            index.astype(dtype)

The series [-1.5, np.nan, 0.0, 0.0, 1.5] in IntervalIndex index contains nan values. Based on test case, the expected result is TypeError: Cannot convert Float64Index to dtype datetime64[ns]; integer values are required for conversion which is same result from issue reported here.

Thank

[tlaytongoogle]

@weikhor Consider that, in contrast to that test, it is entirely possible to convert the values from a Float64Index (including NaN values) into datetime64[ns] values via something like:

DatetimeIndex(index.left, dtype="datetime64[ns]")  # index.left is a Float64Index

The key point here is that passing values into a constructor is not quite equivalent to casting with astype. Thus, I don't think that this test is in-and-of-itself evidence that the reported issue constitutes intended behavior.