DataFrame.reset_index() loses timezone

[jim22k]

When timezone-aware datetime.datetime objects are used as the index of a DataFrame via set_index(), undoing the operation with reset_index() loses the timezone.

After the round-trip, the datetime.datetime column becomes a datetime64[ns] column, which can be considered a feature, not a bug. The times are correctly converted to UTC (which is good if you're going to lose the timezone), but the loss of timezone information seems like a bug.

This has been tested against version 0.9.1 on Python 2.7 (Windows 32-bit)

Code to recreate

import datetime, dateutil, pandas as pd
tzoffset = dateutil.tz.tzoffset(None, -5*60*60)
ts = datetime.datetime(2012,11,15,10,38,4,tzinfo=tzoffset)
print ts # this has a timezone
df = pd.DataFrame([[ts, 15]], columns=['ts', 'val'])
dfi = df.set_index('ts')
print dfi.index # this has a timezone
df2 = dfi.reset_index() # <-- This is where the timezone gets lost
print df2['ts'][0] # this doesn't have a timezone
print df2['ts'][0].to_datetime() # neither does this