Creating tz aware pandas timestamp objects from an integer

Question

I have a integer that is the number of microseconds after the unix epoch. (in GMT)

How can I convert 1349863207154117 using astype to a pandas.Timestamp("2012-10-10T06:00:07.154117", tz=¨UTC¨)? The documentation on astype is not very thorough. I have tried the following.

x = 1349863207154117
dt64 = np.int64(x).astype("M8[us]")
print dt64

returns:

np.datetime64("2012-10-10T06:00:07.154117-0400")

2. if I only want seconds, this works:

time = pd.Timestamp(datetime.datetime.fromtimestamp(int(x / 1e6)), tz=¨UTC¨)

Answer 1

• Pandas: Epoch timestamps covers this.

In [2]: pd.to_datetime(1349863207154117,unit='us')
Out[2]: Timestamp('2012-10-10 10:00:07.154117')

• If you want this in a local timezone

In [6]: pd.to_datetime(1349863207154117,unit='us').tz_localize('US/Eastern')
Out[6]: Timestamp('2012-10-10 10:00:07.154117-0400', tz='US/Eastern')

• If your time is in UTC, but you want it in another tz.

In [9]: pd.to_datetime(1349863207154117,unit='us').tz_localize('UTC').tz_convert('US/Eastern')
Out[9]: Timestamp('2012-10-10 06:00:07.154117-0400', tz='US/Eastern')

• Or this

In [10]: pd.to_datetime(1349863207154117,unit='us',utc=True).tz_convert('US/Eastern')
Out[10]: Timestamp('2012-10-10 06:00:07.154117-0400', tz='US/Eastern')