BUG: Date offset on numpy datetime64 does not work for week and month offsets

[yohplala]

• 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

import numpy as np
import pandas as pd

# test data: a numpy Datetime64
ts_np = np.Datetime64('2021-01-01T08:00:00.00')

# 1 hour offset: works as expected (1 day offset also works as expected)
do=pd.tseries.frequencies.to_offset('1h')
ts_np + do

Result

Timestamp('2021-01-01 09:00:00')

# 1 week offset returns an error
do=pd.tseries.frequencies.to_offset('1W')
ts_np + do

Result

ts_np+do
Traceback (most recent call last):

  File "<ipython-input-11-e20c02eb9ea8>", line 1, in <module>
    ts_np+do

UFuncTypeError: ufunc 'add' cannot use operands with types dtype('<M8[ms]') and dtype('O')

# 1 month begin offset returns an error
do=pd.tseries.frequencies.to_offset('1MS')

Result

ts_np+do
Traceback (most recent call last):

  File "<ipython-input-13-e20c02eb9ea8>", line 1, in <module>
    ts_np+do

UFuncTypeError: ufunc 'add' cannot use operands with types dtype('<M8[ms]') and dtype('O')

Problem description

Applying week or month offset on numpy Datetime64 does not work.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 2cb9652
python : 3.8.5.final.0
python-bits : 64
OS : Linux
OS-release : 5.8.0-53-generic
Version : #60~20.04.1-Ubuntu SMP Thu May 6 09:52:46 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : fr_FR.UTF-8
LOCALE : fr_FR.UTF-8

pandas : 1.2.4
numpy : 1.20.1
pytz : 2021.1
dateutil : 2.8.1
pip : 21.1.1
setuptools : 52.0.0.post20210125
Cython : 0.29.23
pytest : 6.2.3
hypothesis : None
sphinx : 4.0.1
blosc : None
feather : None
xlsxwriter : 1.3.8
lxml.etree : 4.6.3
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : 3.0.0
IPython : 7.22.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : 1.3.2
fsspec : 0.9.0
fastparquet : 0.6.0
gcsfs : None
matplotlib : 3.3.4
numexpr : 2.7.3
odfpy : None
openpyxl : 3.0.7
pandas_gbq : None
pyarrow : 2.0.0
pyxlsb : None
s3fs : None
scipy : 1.6.2
sqlalchemy : 1.4.15
tables : 3.6.1
tabulate : None
xarray : None
xlrd : 2.0.1
xlwt : 1.3.0
numba : 0.51.2

[ghost]

A possible workaround to this can be to use the timedelta64 function of numpy:

import numpy as np
import pandas as pd

ts_np = np.Datetime64('2021-01-01T08:00:00.00')
do = np.timedelta64(1, 'W')
ts_np + do

Output:

2021-01-08T08:00:00.000

[jbrockmendel]

This should be fixed by defining __array_priority__ on BaseOffset. PR would be welcome.

[jbrockmendel]

#44532 fixed this behavior but only implemented tests for ops with td64. A PR adding tests from the OP would close this.

[dhruvsamdani]

When performing this code, it does not give the correct result.

import pandas as pd
import numpy as np

ts_np = np.Datetime64('2021-01-01T08:00:00.00')

do=pd.tseries.frequencies.to_offset('1W')
ts_np + do

Result:

2021-01-03 08:00:00

This is only two days and not the full week. Is this how it is supposed to work?

[jbrockmendel]

do here is a Week offset with weekday=6, so ts_np + do is giving you the next date with res.weekday() ==6. I think you want either to_offset("7D") or pd.offsets.Week(n=1, weekday=None)

[ChrisAlbertsen]

take