how to convert a datetime.time() object to datetime.datetime object pandas

Question

import pandas as pd
import datetime
df = pd.DataFrame([{'st':datetime.datetime.strptime('21:00:00','%H:%M:%S').time(),'et':datetime.datetime.strptime('22:00:00','%H:%M:%S').time()}, {'st':datetime.datetime.strptime('1:00:00','%H:%M:%S').time(),'et':datetime.datetime.strptime('3:00:00','%H:%M:%S').time()}])


Out[183]: df
         et        st
0  22:00:00  21:00:00
1  03:00:00  01:00:00

I would like to be able to convert the above dataframe with new fields having datetime.datetime objects with two other extra columns such as here having any dummy date in it and using the time from their respective rows:

      et        st      sdate_time                edate_time
0  22:00:00  21:00:00   2018-01-01 21:00:00      2018-01-01 22:00:00  
1  03:00:00  01:00:00   2018-01-01 1:00:00       2018-01-01 3:00:00

The approach I have tried is using apply method

df['et'].apply(lambda et: pd.datetime.combine(datetime.datetime.strptime('2018-01-01', '%Y-%m-%d').date(),et))

but turns out that the dataframe could be really huge and I would like to vectorize the above operation without the apply method .

Does `df.apply(pd.to_datetime, format='%H:%M:%S')` work for you? Where the date is 1900-01-01? — Scott Boston, Oct 23 '18 at 18:32

score 4 · Accepted Answer · answered Oct 23 '18 at 18:38

4

Try this

date = str(datetime.datetime.strptime('2018-01-01', '%Y-%m-%d').date())

df['edate_time'] = pd.to_datetime(date + " " + df.et.astype(str))

       et          st            edate_time
0   22:00:00    21:00:00    2018-01-01 22:00:00
1   03:00:00    01:00:00    2018-01-01 03:00:00

answered Oct 23 '18 at 18:38

ResidentSleeper

2,385
2
10
20

1

This anwer relies on a "fake" date – Tobbey Sep 25 '19 at 07:17
@Tobbey What do you mean by a "fake date" ? – ResidentSleeper Sep 25 '19 at 07:24
An arbitrary date '2018-01-01' with an arbitrary format needs to be provided. – Tobbey Sep 25 '19 at 08:08
@Tobbey Do you have a better idea for an object, which *only* holds time information? You need to define a fixed data. – gies0r May 14 '20 at 23:28

score 3 · Answer 2 · answered Oct 23 '18 at 18:40

Try:

df.assign(sdate_time=pd.to_datetime(df['et'], format='%H:%M:%S'), 
          edate_time=pd.to_datetime(df['st'], format='%H:%M:%S'))

Output:

         et        st          sdate_time          edate_time
0  22:00:00  21:00:00 1900-01-01 22:00:00 1900-01-01 21:00:00
1  03:00:00  01:00:00 1900-01-01 03:00:00 1900-01-01 01:00:00

score 2 · Answer 3 · answered Oct 26 '20 at 12:44

Is there no better way of doing this than to assign a random date?

I have matplotlib version '3.3.1' and still face this issue of not being able to plot a datetime.time object for which I do not need a date but only the 24 hours of a day.

I've seen workarounds for the label, using import matplotlib.dates as mdates but still not the best way.

how to convert a datetime.time() object to datetime.datetime object pandas

3 Answers3

Linked