DatetimeArray._from_sequence with mixture of tz-naive and tz-aware data.

[TomAugspurger]

This should probably raise

In [1]: import pandas as pd; import numpy as np

In [2]: arr = np.array([pd.Timestamp('2000'), pd.Timestamp('2000', tz='CET')])

In [3]: pd.arrays.DatetimeArray._from_sequence(arr)
Out[3]:
<DatetimeArrayMixin>
['2000-01-01 01:00:00+01:00', '2000-01-01 00:00:00+01:00']
Length: 2, dtype: datetime64[ns, CET]

For reference, a mix of different timezones does raise

In [4]: mix = np.array([pd.Timestamp('2000', tz="US/Central"), pd.Timestamp('2000', tz='CET')])

In [5]: pd.arrays.DatetimeArray._from_sequence(mix)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
~/sandbox/pandas/pandas/core/arrays/datetimes.py in objects_to_datetime64ns(data, dayfirst, yearfirst, utc, errors, require_iso8601, allow_object)
   1753         try:
-> 1754             values, tz_parsed = conversion.datetime_to_datetime64(data)
   1755             # If tzaware, these values represent unix timestamps, so we

~/sandbox/pandas/pandas/_libs/tslibs/conversion.pyx in pandas._libs.tslibs.conversion.datetime_to_datetime64()
    178                     if not tz_compare(val.tzinfo, inferred_tz):
--> 179                         raise ValueError('Array must be all same time zone')
    180                 else:

ValueError: Array must be all same time zone

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
<ipython-input-5-b2c990c41b55> in <module>
----> 1 pd.arrays.DatetimeArray._from_sequence(mix)

~/sandbox/pandas/pandas/core/arrays/datetimes.py in _from_sequence(cls, data, dtype, copy, tz, freq, dayfirst, yearfirst, ambiguous)
    326         subarr, tz, inferred_freq = sequence_to_dt64ns(
    327             data, dtype=dtype, copy=copy, tz=tz,
--> 328             dayfirst=dayfirst, yearfirst=yearfirst, ambiguous=ambiguous)
    329
    330         freq, freq_infer = dtl.validate_inferred_freq(freq, inferred_freq,

~/sandbox/pandas/pandas/core/arrays/datetimes.py in sequence_to_dt64ns(data, dtype, copy, tz, dayfirst, yearfirst, ambiguous)
   1660             #  or M8[ns] to denote wall times
   1661             data, inferred_tz = objects_to_datetime64ns(
-> 1662                 data, dayfirst=dayfirst, yearfirst=yearfirst)
   1663             tz = maybe_infer_tz(tz, inferred_tz)
   1664

~/sandbox/pandas/pandas/core/arrays/datetimes.py in objects_to_datetime64ns(data, dayfirst, yearfirst, utc, errors, require_iso8601, allow_object)
   1757             return values.view('i8'), tz_parsed
   1758         except (ValueError, TypeError):
-> 1759             raise e
   1760
   1761     if tz_parsed is not None:

~/sandbox/pandas/pandas/core/arrays/datetimes.py in objects_to_datetime64ns(data, dayfirst, yearfirst, utc, errors, require_iso8601, allow_object)
   1748             dayfirst=dayfirst,
   1749             yearfirst=yearfirst,
-> 1750             require_iso8601=require_iso8601
   1751         )
   1752     except ValueError as e:

~/sandbox/pandas/pandas/_libs/tslib.pyx in pandas._libs.tslib.array_to_datetime()
    458 @cython.wraparound(False)
    459 @cython.boundscheck(False)
--> 460 cpdef array_to_datetime(ndarray[object] values, str errors='raise',
    461                         bint dayfirst=False, bint yearfirst=False,
    462                         object utc=None, bint require_iso8601=False):

~/sandbox/pandas/pandas/_libs/tslib.pyx in pandas._libs.tslib.array_to_datetime()
    535                             iresult[i] = _ts.value
    536                         else:
--> 537                             raise ValueError('Tz-aware datetime.datetime '
    538                                              'cannot be converted to '
    539                                              'datetime64 unless utc=True')

ValueError: Tz-aware datetime.datetime cannot be converted to datetime64 unless utc=True

[jreback]

I @jbrockmendel already noted this, its in to_datetime, but really because these hit different paths in array_to_datetime

In [1]: arr = np.array([pd.Timestamp('2000'), pd.Timestamp('2000', tz='CET')])
In [2]: pd.to_datetime(arr)
Out[2]: DatetimeIndex(['2000-01-01 01:00:00+01:00', '2000-01-01 00:00:00+01:00'], dtype='datetime64[ns, CET]', freq=None)

In [3]: mix = np.array([pd.Timestamp('2000', tz="US/Central"), pd.Timestamp('2000', tz='CET')])

In [4]: pd.to_datetime(mix)
ValueError: Tz-aware datetime.datetime cannot be converted to datetime64 unless utc=True

[jbrockmendel]

Yah, #24006 would handle these correctly, needs to come off the back-burner.

IIRC there was a "M8ns/i8 are treated inconsistently" issue in there too.

[TomAugspurger]

This isn't a blocker for 0.24.0 I think.