TypeError: unsupported operand type(s) for -: 'datetime.time' and 'datetime.time&#039-CodePudding

I am trying to subtract values of two datetime objects like the following:

df["Time Taken"] = (pd.to_datetime(df['end_time']).dt.tz_convert('Asia/Kolkata').dt.time) - \
                           (pd.to_datetime(df['start_time']).dt.tz_convert('Asia/Kolkata').dt.time)

in a Dataframe like the following:

     id                    serial reference_number                         date                warehouse          owner                   start_time                     end_time
0  352                 655556555             5002  2022-05-15T13:33:00.208423Z  Delhivery Goa Warehouse  Delhivery_Goa  2022-05-31T11:26:10.977240Z  2022-05-31T11:59:55.421159Z
1  362  E2806995000040043136895A             5007  2022-05-18T10:37:28.871625Z  Delhivery Goa Warehouse  Delhivery_Goa  2022-05-31T11:26:10.977240Z  2022-05-31T11:59:55.421159Z
2  363  E28069950000500431368519             5008  2022-05-18T10:44:32.398842Z  Delhivery Goa Warehouse  Delhivery_Goa  2022-05-31T11:26:10.977240Z  2022-05-31T11:59:55.421159Z
3  364  E28069950000400431368536             5008  2022-05-18T10:44:32.398842Z  Delhivery Goa Warehouse  Delhivery_Goa  2022-05-31T11:26:10.977240Z  2022-05-31T11:59:55.421159Z
4  370  E28069950000500431368549             5010  2022-05-18T12:30:34.599759Z  Delhivery Goa Warehouse  Delhivery_Goa  2022-05-31T11:26:10.977240Z  2022-05-31T11:59:55.421159Z

I tried timedelta as well but the I got the following error:

only leading negative signs are allowed

How do I do it?

CodePudding user response：

First subtract columns converted to datetimes by Series.sub, for seconds use Series.dt.total_seconds::

df["Time Taken"] = (pd.to_datetime(df['end_time']).sub(pd.to_datetime(df['start_time']))
                      .dt.total_seconds())

For format HH:MM:SS from timedeltas use:

def f(x):
    ts = x.total_seconds()
    hours, remainder = divmod(ts, 3600)
    minutes, seconds = divmod(remainder, 60)
    return ('{}:{:02d}:{:02d}').format(int(hours), int(minutes), int(seconds)) 

df["Time Taken"] = (pd.to_datetime(df['end_time']).sub(pd.to_datetime(df['start_time']))
                      .apply(f))

CodePudding user response：

You're making this way more complicated than it needs to be:

Just

df["Time Taken"] = (
    pd.to_datetime(df['end_time'])
    - pd.to_datetime(df['start_time'])
)

This should give you a column of timedeltas which you can format as you wish.

Steps you can skip:

Converting the time zone. The times in the database already specify specific points in time, so the difference can be calculated. Converting from UTC to local doesn't change the result.
Extracting the time from the date with .dt.time. This is what causes the error because times don't support subtraction. Subtraction isn't implemented for these because the concept of a difference between two times isn't well-defined. Like is (1am - 4am) -3 hours or 21 hours or even -2 or -4 hours because a switch between daylight savings and normal time occured?