to_datetime does not handle timezones correctly for timestamps

[CodeSturgeon]

It seems that the to_datetime function always assumes that a timestamp is in local time and converts it to UTC. This does not seem to be the case with converting from strings.

This code:

from datetime import datetime
import time
import pandas as pd
now = datetime.now()
ts = time.mktime(now.timetuple())
print ts
print now.strftime('%Y-%m-%d %H:%M')
print pd.to_datetime(ts, unit='s')
print pd.to_datetime(ts, unit='s', utc=True)
print pd.to_datetime(now.strftime('%Y-%m-%d %H:%M'))

Give this result (pandas 0.15.1):

1421967906.0
2015-01-22 18:05
2015-01-22 23:05:06
2015-01-22 23:05:06+00:00
2015-01-22 18:05:00

Note that when to_datetime is converting a timestamp five hours are added no matter the utc= arg value.

[jreback]

thanks for reporting
this is a dupe of #6415 pull requests are welcome