BUG: Inconsistent initialization of `timedelta64` between Series vs numpy & pd.TimeDelta

[galipremsagar]

• 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.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

The creation of a pandas Series with any dtype other than timedelta64[ns] is discarding values and setting the entire series to 0's. Whereas numpy.array and pd.TimeDelta are correctly instantiating and preserving the passed values which can be retrieved by a type-cast too.

Code Sample, a copy-pastable example

>>> import pandas as pd
>>> pd.__version__
'1.1.0'
>>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[ns]')
0   0 days 00:00:00.001000
1   0 days 00:00:00.000200
2   0 days 00:00:00.003000
dtype: timedelta64[ns]
>>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[s]')
0   0 days
1   0 days
2   0 days
dtype: timedelta64[ns]
>>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[s]').astype('int')
0    0
1    0
2    0
dtype: int64
>>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[ns]').astype('int')
0    1000000
1     200000
2    3000000
dtype: int64

>>> import numpy as np
>>> np.array([1000000, 200000, 3000000], dtype='timedelta64[ns]')
array([1000000,  200000, 3000000], dtype='timedelta64[ns]')
>>> np.array([1000000, 200000, 3000000], dtype='timedelta64[s]')
array([1000000,  200000, 3000000], dtype='timedelta64[s]')
>>> np.array([1000000, 200000, 3000000], dtype='timedelta64[s]').astype('int')
array([1000000,  200000, 3000000])

>>> pd.Timedelta(10)
Timedelta('0 days 00:00:00.000000010')
>>> pd.Timedelta(10, unit='s')
Timedelta('0 days 00:00:10')
>>> pd.Timedelta(10, unit='ns')
Timedelta('0 days 00:00:00.000000010')

Problem description

The expected behaviour is Series initialization must be in-line with numpy array initialization in this case.

Expected Output

Output of pd.show_versions()

INSTALLED VERSIONS

commit : d9fff27
python : 3.7.3.final.0
python-bits : 64
OS : Darwin
OS-release : 19.6.0
Version : Darwin Kernel Version 19.6.0: Sun Jul 5 00:43:10 PDT 2020; root:xnu-6153.141.1~9/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : None
LOCALE : en_US.UTF-8
pandas : 1.1.0
numpy : 1.19.0
pytz : 2020.1
dateutil : 2.8.1
pip : 20.1.1
setuptools : 49.1.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[simonjayhawkins]

Thanks @galipremsagar for the report.

Further investigation and PRs welcome.

[simonjayhawkins]

I'll mark as a regression as this raised Exception in 0.25.3

>>> pd.__version__
'0.25.3'
>>>
>>> pd.Series([1000000, 200000, 3000000], dtype="timedelta64[s]")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\simon\Anaconda3\lib\site-packages\pandas\core\series.py", line 314, in __init__
    data = sanitize_array(data, index, dtype, copy, raise_cast_failure=True)
  File "C:\Users\simon\Anaconda3\lib\site-packages\pandas\core\internals\construction.py", line 679, in sanitize_array
    subarr = _try_cast(data, dtype, copy, raise_cast_failure)
  File "C:\Users\simon\Anaconda3\lib\site-packages\pandas\core\internals\construction.py", line 784, in _try_cast
    subarr = maybe_cast_to_datetime(arr, dtype)
  File "C:\Users\simon\Anaconda3\lib\site-packages\pandas\core\dtypes\cast.py", line 1035, in maybe_cast_to_datetime
    "dtype [{dtype}]".format(dtype=dtype)
TypeError: cannot convert timedeltalike to dtype [timedelta64[s]]
>>>

[galipremsagar]

So is the expected behavior to raise an exception or initialization similar to numpy?

[mroeschke]

I think the recent non-nano support probably fixed the Series initialization case with timedelta64[s] case. Could use a unit test

In [1]: >>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[ns]')
Out[1]:
0   0 days 00:00:00.001000
1   0 days 00:00:00.000200
2   0 days 00:00:00.003000
dtype: timedelta64[ns]

In [2]: >>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[s]')
Out[2]:
0   0 days 00:00:00.001000
1   0 days 00:00:00.000200
2   0 days 00:00:00.003000
dtype: timedelta64[ns]

In [3]: >>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[s]').astype('int')
Out[3]:
0    1000000
1     200000
2    3000000
dtype: int64

In [4]: >>> pd.Series([1000000, 200000, 3000000], dtype='timedelta64[ns]').astype('int')
Out[4]:
0    1000000
1     200000
2    3000000
dtype: int64