BUG: round() on tz-aware DateTimeIndex (spanning DST switch) throws AmbiguousTimeError as if tz-unaware

[bluppfisk]

• 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.

Code Sample, a copy-pastable example

import pandas as pd
dt = pd.Timestamp("2020-10-25 02:00:00").tz_localize("Europe/Brussels", ambiguous=False).round('min')

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pandas/_libs/tslibs/timestamps.pyx", line 489, in pandas._libs.tslibs.timestamps.Timestamp.round
  File "pandas/_libs/tslibs/timestamps.pyx", line 443, in pandas._libs.tslibs.timestamps.Timestamp._round
  File "pandas/_libs/tslibs/timestamps.pyx", line 826, in pandas._libs.tslibs.timestamps.Timestamp.tz_localize
  File "pandas/_libs/tslibs/tzconversion.pyx", line 232, in pandas._libs.tslibs.tzconversion.tz_localize_to_utc
pytz.exceptions.AmbiguousTimeError: Cannot infer dst time from 2020-10-25 02:00:00, try using the 'ambiguous' argument

import pandas as pd
dt = pd.Timestamp("2020-10-25 02:00:00").tz_localize("Europe/Brussels", ambiguous=True).round('min')

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pandas/_libs/tslibs/timestamps.pyx", line 489, in pandas._libs.tslibs.timestamps.Timestamp.round
  File "pandas/_libs/tslibs/timestamps.pyx", line 443, in pandas._libs.tslibs.timestamps.Timestamp._round
  File "pandas/_libs/tslibs/timestamps.pyx", line 826, in pandas._libs.tslibs.timestamps.Timestamp.tz_localize
  File "pandas/_libs/tslibs/tzconversion.pyx", line 232, in pandas._libs.tslibs.tzconversion.tz_localize_to_utc
pytz.exceptions.AmbiguousTimeError: Cannot infer dst time from 2020-10-25 02:00:00, try using the 'ambiguous' argument

Problem description

A localized Timestamp does not give an error when rounding.
A Series of a localized DateTimeIndex shouldn't either.

[this should explain why the current behaviour is a problem and why the expected output is a better solution]

Expected Output

Output is expected to be identical as for

pandas.Timestamp("2020-10-25 02:00:00+01:00").round('min')

respectively

pandas.Timestamp("2020-10-25 02:00:0+02:00").round('min')

i.e. without AmbiguousTimeError

Output of pd.show_versions()

[paste the output of pd.show_versions() here leaving a blank line after the details tag]

INSTALLED VERSIONS

commit : db08276
python : 3.8.5.final.0
python-bits : 64
OS : Linux
OS-release : 5.9.1-050901-generic
Version : #202010170731 SMP Sat Oct 17 07:42:59 UTC 2020
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.1.3
numpy : 1.19.3
pytz : 2020.1
dateutil : 2.8.1
pip : 20.0.2
setuptools : 44.0.0

rest is not installed

[bluppfisk]

Note that a workaround exists by converting to UTC first:

>>> pd.Timestamp("2020-10-25 02:00:00").tz_localize("Europe/Brussels", ambiguous=True).astimezone("UTC").round('min').astimezone("Europe/Brussels")

[jbrockmendel]

@mroeschke enough of these are building up that I think we can make the Day -> DayDST switch and call it a bugfix instead of an API change. what do you think?

[mroeschke]

100% in support for 'D' and offsets.Day to mean calendar day xref #22864.

I think it would affect several of operations though (IIRC resample has some sticky logic around 'D'), so I agree it's a "bug fix" in operations like this but an "API change" in others.

[tnielens]

Same issue with ceil and floor:

import pandas as pd
pd.Timestamp('2021-10-31 01:30:12+02', tzinfo=dateutil.tz.gettz('Europe/Brussels')).ceil('H')

AmbiguousTimeError                        Traceback (most recent call last)
/tmp/ipykernel_67812/2238318785.py in <module>
----> 1 pd.Timestamp('2021-10-31 01:30:12+02', tzinfo=dateutil.tz.gettz('Europe/Brussels')).ceil('H')
pandas/_libs/tslibs/timestamps.pyx in pandas._libs.tslibs.timestamps.Timestamp.ceil()
pandas/_libs/tslibs/timestamps.pyx in pandas._libs.tslibs.timestamps.Timestamp._round()
pandas/_libs/tslibs/timestamps.pyx in pandas._libs.tslibs.timestamps.Timestamp.tz_localize()
pandas/_libs/tslibs/tzconversion.pyx in pandas._libs.tslibs.tzconversion.tz_localize_to_utc_single()
pandas/_libs/tslibs/tzconversion.pyx in pandas._libs.tslibs.tzconversion.tz_localize_to_utc()
AmbiguousTimeError: Cannot infer dst time from 2021-10-31 02:00:00, try using the 'ambiguous' argument

[jreback]

see #44287

this is as expected

[mroeschke]

Yes, and the documentation will be updated in #44357 (available in version 1.4) to reflect this. Closing.

[sasaki000]

A similar behavior has occurred in Pandas: [1.4.1].
Should I create another record?

import platform
import pandas as pd
print(f'- Python:[{platform.python_version()}], Pandas:[{pd.__version__}]') # <-- This AmbiguousTimeError occurred in Python:[3.9.1], Pandas:[1.4.1]

datetime_index = pd.DatetimeIndex([pd.Timestamp('2022-02-07',tz='America/New_York')], freq='D')
# datetime_index.freq = None # <--- You can avoid errors by executing this line.
print('Input:', datetime_index)

datetime_index = datetime_index.shift(periods=-92, freq='D')
print('Output:', datetime_index)

• Python:[3.9.1], Pandas:[1.4.1]
  Input: DatetimeIndex(['2022-02-07 00:00:00-05:00'], dtype='datetime64[ns, America/New_York]', freq='D')

---

AmbiguousTimeError Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_4168/2621838263.py in
8 print('Input:', datetime_index)
9
---> 10 datetime_index = datetime_index.shift(periods=-92, freq='D')
11 print('Output:', datetime_index)

~\AppData\Local\Programs\Python\Python39\lib\site-packages\pandas\core\indexes\datetimelike.py in shift(self, periods, freq)
354 arr = self._data.view()
355 arr._freq = self.freq
--> 356 result = arr._time_shift(periods, freq=freq)
357 return type(self)._simple_new(result, name=self.name)
358

~\AppData\Local\Programs\Python\Python39\lib\site-packages\pandas\core\arrays\datetimelike.py in _time_shift(self, periods, freq)
1247 # appropriate timezone from start and end, so tz does not need
1248 # to be passed explicitly.
-> 1249 return self._generate_range(start=start, end=end, periods=None, freq=self.freq)
1250
1251 @unpack_zerodim_and_defer("add")

~\AppData\Local\Programs\Python\Python39\lib\site-packages\pandas\core\arrays\datetimes.py in _generate_range(cls, start, end, periods, freq, tz, normalize, ambiguous, nonexistent, inclusive)
453
454 if tz is not None and index.tz is None:
--> 455 arr = tzconversion.tz_localize_to_utc(
456 index.asi8, tz, ambiguous=ambiguous, nonexistent=nonexistent
457 )

~\AppData\Local\Programs\Python\Python39\lib\site-packages\pandas_libs\tslibs\tzconversion.pyx in pandas._libs.tslibs.tzconversion.tz_localize_to_utc()

AmbiguousTimeError: Cannot infer dst time from 2021-11-07 01:00:00, try using the 'ambiguous' argument