Pandas corrupting datetime object

Question

I am trying to create timezone aware date column in a pandas DataFrame. When I run the code below, the resulting pandas column does not have the same datetime as the one I inputted. What am I doing wrong here?

I am using python 3.6.2 and pandas 0.20.3

from datetime import datetime
import pandas as pd
import pytz 
date_string = "12/14/2016 12:00"
timezone = pytz.timezone("US/Pacific")
input_datetime =  datetime.strptime(date_string, "%m/%d/%Y %H:%M").replace(tzinfo=timezone)
df = pd.DataFrame({"datetime":[input_datetime]})

If I run that code, df['datetime'][0].minute returns 53 while input_datetime.minute returns 0.

When I don't replace the tzinfo I do not have a problem.

Post your actual code, not a screenshot. Also take a look at [this link](https://stackoverflow.com/questions/20109391/how-to-make-good-reproducible-pandas-examples). — pault, Dec 14 '17 at 21:25
What do you mean by changing? Is it because you are calling minutes of `x` which is `10/5/16 0:00`? — Joseph K., Dec 14 '17 at 21:30
@JosephK. `datetime.datetime(2016, 12, 14, 12, 0, tzinfo=)` becomes `Timestamp('2016-12-14 11:53:00-0800', tz='US/Pacific')` after I create the pandas DataFrame — jackalack, Dec 14 '17 at 23:49

score 1 · Answer 1 · edited Jan 29 '20 at 14:23

1

If you first convert your input_datetime you can call the minutes (or years etc) of your dataframe with .dt.minute

input_datetime = pd.to_datetime(datetime.strptime(date_string, 
                                                  "%m/%d/%Y %H:%M")).replace(tzinfo=timezone)


df = pd.DataFrame({"datetime":[input_datetime]})
df['datetime'].dt.minute

edited Jan 29 '20 at 14:23

ohe

3,461
3
26
50

answered Jan 29 '20 at 14:03

Chiel

1,865
1
11
24

score 0 · Answer 2 · answered Jan 29 '20 at 14:24

You can use pandas .dt and tz_localize:

from datetime import datetime
import pandas as pd
date_string = "12/14/2016 12:00"
df = pd.DataFrame({'datetime':[datetime.strptime(date_string, "%m/%d/%Y %H:%M")]})
df['datetime'].dt.tz_localize('US/Pacific')

Output:

0   2016-12-14 12:00:00-08:00
Name: datetime, dtype: datetime64[ns, US/Pacific]

Pandas corrupting datetime object

2 Answers2