API: tz_convert within DatetimeIndex constructor

[jbrockmendel]

At the moment the following raises:

>>> dti = pd.date_range('2016-01-01', periods=3, tz='US/Central')
>>> pd.DatetimeIndex(dti, tz='Asia/Tokyo')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python2.7/site-packages/pandas/core/indexes/datetimes.py", line 413, in __new__
    raise TypeError(msg.format(data.tz, tz))
TypeError: data is already tz-aware US/Central, unable to set specified tz: Asia/Tokyo

It isn't clear to me that raising is the right thing to do; shouldn't this just be equivalent to dti.tz_convert('Asia/Tokyo')? Or is this ambiguous for some reason?

[mroeschke]

This works for Timestamp albeit I am not really a fan of tz= meaning localizing and converting. But if this is properly documented, we might as well follow Timestamp's behavior unless I am missing something

In [1]: pd.Timestamp(pd.Timestamp('2016-01-01', tz='US/Central'), tz='Asia/Tokyo')
Out[1]: Timestamp('2016-01-01 15:00:00+0900', tz='Asia/Tokyo')

[jreback]

I would rather leave this and be very explicit
doing anything non explicit with tz localize vs conversion has bitten lots in the past

[mroeschke]

For consistency sake then, we should depreciate the Timestamp behavior then.

[jreback]

yep that sounds right