datetime like series .dt properties fail after ts.dt.tz_localize(tz1).dt.tz_convert(tz2) sequence

[witosx]

Second tz_convert fails with AmbiguousTimeError ('UTC' in the example, but other timezones fail too).

EDIT: Actually it's not only second tz_convert that fails, it looks like most (if not all) of the .dt accessed properties are affected

t = pd.Series(pd.date_range('2015-01-01', '2016-01-01', freq='T'))
t = t.dt.tz_localize('UTC').dt.tz_convert('America/Chicago')
t.head(10)

0   2014-12-31 18:00:00-06:00
1   2014-12-31 18:01:00-06:00
2   2014-12-31 18:02:00-06:00
3   2014-12-31 18:03:00-06:00
4   2014-12-31 18:04:00-06:00
5   2014-12-31 18:05:00-06:00
6   2014-12-31 18:06:00-06:00
7   2014-12-31 18:07:00-06:00
8   2014-12-31 18:08:00-06:00
9   2014-12-31 18:09:00-06:00
dtype: datetime64[ns, America/Chicago]

t.dt.tz_convert('UTC')

---------------------------------------------------------------------------

AmbiguousTimeError                        Traceback (most recent call last)

/usr/lib/python3.5/site-packages/pandas/core/series.py in _make_dt_accessor(self)
   2639         try:
-> 2640             return maybe_to_datetimelike(self)
   2641         except Exception:


/usr/lib/python3.5/site-packages/pandas/tseries/common.py in maybe_to_datetimelike(data, copy)
     48     if is_datetime64_dtype(data.dtype) or is_datetime64tz_dtype(data.dtype):
---> 49         return DatetimeProperties(DatetimeIndex(data, copy=copy, freq='infer'), index, name=data.name)
     50     elif is_timedelta64_dtype(data.dtype):


/usr/lib/python3.5/site-packages/pandas/util/decorators.py in wrapper(*args, **kwargs)
     88                     kwargs[new_arg_name] = new_arg_value
---> 89             return func(*args, **kwargs)
     90         return wrapper


/usr/lib/python3.5/site-packages/pandas/tseries/index.py in __new__(cls, data, freq, start, end, periods, copy, name, tz, verify_integrity, normalize, closed, ambiguous, dtype, **kwargs)
    343                     subarr = tslib.tz_localize_to_utc(ints, tz,
--> 344                                                       ambiguous=ambiguous)
    345 


pandas/tslib.pyx in pandas.tslib.tz_localize_to_utc (pandas/tslib.c:64516)()


AmbiguousTimeError: Cannot infer dst time from Timestamp('2015-11-01 01:00:00'), try using the 'ambiguous' argument


During handling of the above exception, another exception occurred:


AttributeError                            Traceback (most recent call last)

<ipython-input-2-eb75730cd6fc> in <module>()
----> 1 t.dt.tz_convert('UTC')


/usr/lib/python3.5/site-packages/pandas/core/generic.py in __getattr__(self, name)
   2239                 or name in self._metadata
   2240                 or name in self._accessors):
-> 2241             return object.__getattribute__(self, name)
   2242         else:
   2243             if name in self._info_axis:


/usr/lib/python3.5/site-packages/pandas/core/base.py in __get__(self, instance, owner)
    186             # this ensures that Series.str.<method> is well defined
    187             return self.accessor_cls
--> 188         return self.construct_accessor(instance)
    189 
    190     def __set__(self, instance, value):


/usr/lib/python3.5/site-packages/pandas/core/series.py in _make_dt_accessor(self)
   2640             return maybe_to_datetimelike(self)
   2641         except Exception:
-> 2642             raise AttributeError("Can only use .dt accessor with datetimelike "
   2643                                  "values")
   2644 


AttributeError: Can only use .dt accessor with datetimelike values

[jorisvandenbossche]

This seems to come down to the failing of DatetimeIndex(t) (which is used within the dt accessor):

In [37]: t.head()
Out[37]:
0   2014-12-31 18:00:00-06:00
1   2014-12-31 18:01:00-06:00
2   2014-12-31 18:02:00-06:00
3   2014-12-31 18:03:00-06:00
4   2014-12-31 18:04:00-06:00
dtype: datetime64[ns, America/Chicago]

In [38]: pd.DatetimeIndex(t)
---------------------------------------------------------------------------
AmbiguousTimeError                        Traceback (most recent call last)
<ipython-input-38-3e6430dea298> in <module>()
----> 1 pd.DatetimeIndex(t)

c:\users\vdbosscj\scipy\pandas-joris\pandas\util\decorators.pyc in wrapper(*args
, **kwargs)
     87                 else:
     88                     kwargs[new_arg_name] = new_arg_value
---> 89             return func(*args, **kwargs)
     90         return wrapper
     91     return _deprecate_kwarg

c:\users\vdbosscj\scipy\pandas-joris\pandas\tseries\index.pyc in __new__(cls, da
ta, freq, start, end, periods, copy, name, tz, verify_integrity, normalize, clos
ed, ambiguous, dtype, **kwargs)
    342                     ints = subarr.view('i8')
    343                     subarr = tslib.tz_localize_to_utc(ints, tz,
--> 344                                                       ambiguous=ambiguou
s)
    345
    346                 subarr = subarr.view(_NS_DTYPE)

tslib.pyx in pandas.tslib.tz_localize_to_utc (pandas\tslib.c:55271)()

AmbiguousTimeError: Cannot infer dst time from Timestamp('2015-11-01 01:00:00'),
 try using the 'ambiguous' argument

When providing tz data to DatetimeIndex, it first makes it naive and then converts back to the tz, which fails without supplying ambiguous='infer'.

[jreback]

@witosx give #11301 a try. This was passing thru the ambiguous kw incorrectly.