Unix timestamp converting by python pandas

Question

I want to convert a time(String) to unix timestamp, then convert it back to readable datetime, however, there are 7 hours different. What is the problem in my python code?

import datetime

#Convert a Datetime String to unix timestamp   
def get_Timestamp_from_String(datetimeStr):    

    timestamp = int(datetime.datetime.strptime(datetimeStr, '%Y-%m-%d %H:%M:%S').timestamp())*1000

    return timestamp


# Convert unix timestamp to DateTime    
def get_Datetime_from_Timestamp(timestamp):

    dt = pd.to_datetime(timestamp, unit='ms')

    return dt

Test the datetime: "2016-04-01 01:10:04", however, if I convert it back, then I got the result: '2016-04-01 08:10:04', there are 7 hours different

timestamp = get_Timestamp_from_String("2016-04-01 01:10:04")
res = get_Datetime_from_Timestamp(timestamp)
print(res)

>>>Timestamp('2016-04-01 08:10:04')

What is the problem in my code?

Answer 1

I think you need to_datetime twice, for ms divide by 10**6 :

datetimeStr = '2016-04-01 01:10:04'
timestamp = pd.to_datetime(datetimeStr).value //10**6
print (timestamp)
1459473004000

dt = pd.to_datetime(timestamp, unit='ms')
print (dt)
2016-04-01 01:10:04

For ns , native format of timestamps in numpy dividing is not necessary:

datetimeStr = '2016-04-01 01:10:04'
timestamp = pd.to_datetime(datetimeStr).value
print (timestamp)
1459473004000000000

dt = pd.to_datetime(timestamp, unit='ns')
print (dt)
2016-04-01 01:10:04

But if have timezone information:

datetimeStr = '2016-04-01 01:10:04-07:00'
timestamp = pd.to_datetime(datetimeStr).value //10**6
print (timestamp)
1459498204000

dt = pd.to_datetime(timestamp, unit='ms')
print (dt)
2016-04-01 08:10:04