BUG: Regression adding Timedelta to DatetimeIndex, get ValueError: Inferred frequency None from passed values does not conform to passed frequency

[nathanielatom]

• 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
pd.date_range(start=pd.Timestamp('2019-03-26 00:00:00-0400', tz='Canada/Eastern'), 
              end=pd.Timestamp('2020-10-17 00:00:00-0400', tz='Canada/Eastern'), 
              freq='D') + pd.Timedelta('1D')

Traceback

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
~/.miniconda/lib/python3.7/site-packages/pandas/core/arrays/datetimelike.py in _validate_frequency(cls, index, freq, **kwargs)
   1156             if not np.array_equal(index.asi8, on_freq.asi8):
-> 1157                 raise ValueError
   1158         except ValueError as e:

ValueError: 

The above exception was the direct cause of the following exception:

ValueError                                Traceback (most recent call last)
<ipython-input-2-d8f454fa3ee3> in <module>
----> 1 pd.date_range(start=pd.Timestamp('2019-03-26 00:00:00-0400', tz='Canada/Eastern'), end=pd.Timestamp('2020-10-17 00:00:00-0400', tz='Canada/Eastern'), freq='D') + pd.Timedelta('1D')

~/.miniconda/lib/python3.7/site-packages/pandas/core/indexes/extension.py in method(self, other)
    145             return NotImplemented
    146         meth = getattr(self._data, opname)
--> 147         result = meth(_maybe_unwrap_index(other))
    148         return _wrap_arithmetic_op(self, other, result)
    149 

~/.miniconda/lib/python3.7/site-packages/pandas/core/ops/common.py in new_method(self, other)
     63         other = item_from_zerodim(other)
     64 
---> 65         return method(self, other)
     66 
     67     return new_method

~/.miniconda/lib/python3.7/site-packages/pandas/core/arrays/datetimelike.py in __add__(self, other)
   1399             result = self._add_nat()
   1400         elif isinstance(other, (Tick, timedelta, np.timedelta64)):
-> 1401             result = self._add_timedeltalike_scalar(other)
   1402         elif isinstance(other, BaseOffset):
   1403             # specifically _not_ a Tick

~/.miniconda/lib/python3.7/site-packages/pandas/core/arrays/datetimelike.py in _add_timedeltalike_scalar(self, other)
   1254             new_freq = self.freq
   1255 
-> 1256         return type(self)(new_values, dtype=self.dtype, freq=new_freq)
   1257 
   1258     def _add_timedelta_arraylike(self, other):

~/.miniconda/lib/python3.7/site-packages/pandas/core/arrays/datetimes.py in __init__(self, values, dtype, freq, copy)
    282 
    283         if inferred_freq is None and freq is not None:
--> 284             type(self)._validate_frequency(self, freq)
    285 
    286     @classmethod

~/.miniconda/lib/python3.7/site-packages/pandas/core/arrays/datetimelike.py in _validate_frequency(cls, index, freq, **kwargs)
   1169                 f"Inferred frequency {inferred} from passed values "
   1170                 f"does not conform to passed frequency {freq.freqstr}"
-> 1171             ) from e
   1172 
   1173     # monotonicity/uniqueness properties are called via frequencies.infer_freq,

ValueError: Inferred frequency None from passed values does not conform to passed frequency D

Problem description

Unexpected ValueError when adding a Timedelta to a timezone-aware DatetimeIndex. It appears using timezone-aware Timestamp objects is necessary to reproduce - simple date strings won't.

Expected Output

The output from the same code sample in v0.25.3 is:

DatetimeIndex(['2019-03-27 00:00:00-04:00', '2019-03-28 00:00:00-04:00',
               '2019-03-29 00:00:00-04:00', '2019-03-30 00:00:00-04:00',
               '2019-03-31 00:00:00-04:00', '2019-04-01 00:00:00-04:00',
               '2019-04-02 00:00:00-04:00', '2019-04-03 00:00:00-04:00',
               '2019-04-04 00:00:00-04:00', '2019-04-05 00:00:00-04:00',
               ...
               '2020-10-09 00:00:00-04:00', '2020-10-10 00:00:00-04:00',
               '2020-10-11 00:00:00-04:00', '2020-10-12 00:00:00-04:00',
               '2020-10-13 00:00:00-04:00', '2020-10-14 00:00:00-04:00',
               '2020-10-15 00:00:00-04:00', '2020-10-16 00:00:00-04:00',
               '2020-10-17 00:00:00-04:00', '2020-10-18 00:00:00-04:00'],
              dtype='datetime64[ns, Canada/Eastern]', length=572, freq=None)

don't know about later versions, before v1.1.3.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : db08276
python : 3.7.6.final.0
python-bits : 64
OS : Darwin
OS-release : 16.7.0
Version : Darwin Kernel Version 16.7.0: Thu Dec 20 21:53:35 PST 2018; root:xnu-3789.73.31~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_CA.UTF-8
LOCALE : en_CA.UTF-8

pandas : 1.1.3
numpy : 1.19.2
pytz : 2020.1
dateutil : 2.8.1
pip : 20.2.3
setuptools : 50.3.0
Cython : 0.29.20
pytest : 6.1.1
hypothesis : None
sphinx : 1.8.5
blosc : 1.9.2
feather : None
xlsxwriter : 1.2.1
lxml.etree : 4.5.2
html5lib : 0.9999999
pymysql : 0.9.3
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.16.1
pandas_datareader: 0.9.0
bs4 : 4.9.3
bottleneck : 1.3.2
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.2.1
numexpr : 2.7.1
odfpy : None
openpyxl : 2.6.3
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : None
scipy : 1.5.2
sqlalchemy : 1.3.11
tables : 3.6.1
tabulate : None
xarray : 0.16.1
xlrd : 1.2.0
xlwt : 1.3.0
numba : 0.51.2

[jreback]

try on master pls

[nathanielatom]

Yep, same behaviour as v1.1.3 on master; edited.

[dsaxton]

Regression is due to #32799

cc @jbrockmendel

[jbrockmendel]

cc @mroeschke this looks like another case where we need freq="D" to mean DayDST or something

[jbrockmendel]

OK, so there are at least three things wrong here:

1. In this context, freq="D" needs to mean something like DayDST
2. dti.infer_freq() is incorrectly returning "D" instead of None
3. _add_timedeltalike_scalar should use type(self)._simple_new instead of type(self) (once the other two problems are resolved)

[simonjayhawkins]

moved off 1.1.4 milestone (scheduled for release tomorrow) as no PRs to fix in the pipeline

[nathanielatom]

Thanks for the heads up