Difference between two timestamps rounded up to hours and adding column to dataframe

Question

I have a dataframe that looks like this, the last two columns I added to help me understand the time difference. I would like the time difference between the timestamp and forecastRan to be rounded up to the next hour and add it as a column to my dataframe. However, it just gives me the hours (not rounded up).

            timeStamp      forecastRan  time_diff_in_hours  time_diff_seconds   

12 2016-11-23 23:00:00   2016-11-23 12:18:00    10.0              38520.0     
13 2016-11-24 00:00:00   2016-11-23 12:18:00    11.0              42120.0    
14 2016-11-24 01:00:00   2016-11-23 12:18:00    12.0              45720.0 

These are ways I've gotten the difference

df['time_diff_in_hours'] = (df['timeStamp'] - df['forecastRan']).astype('timedelta64[h]')

df['time_diff']= (df['timeStamp'] - df['forecastRan']).dt.total_seconds()

However, the first one just gives me the hours. With the second one I tried dividing by 3600 to get the hours and then using math.ceil to round up, like this:

import datetime
import math

df['time_diff']= math.ceil(((df['timeStamp'] - df['forecastRan']).dt.total_seconds()/3600))

I get: TypeError: cannot convert the series to <class 'float'>. I think the problem happened when I introudced math.ceil, because when I just used (df['timeStamp'] - df['forecastModelRun']).dt.total_seconds()/3600, I didn't get any errors. Not sure where exactly to round up and how to.

Answer 1

I believe need np.ceil:

df['time_diff']= np.ceil(((df['timeStamp'] - df['forecastRan']).dt.total_seconds()/3600))
print (df)
            timeStamp         forecastRan  time_diff_in_hours  \
0 2016-11-23 23:00:00 2016-11-23 12:18:00                10.0   
1 2016-11-24 00:00:00 2016-11-23 12:18:00                11.0   
2 2016-11-24 01:00:00 2016-11-23 12:18:00                12.0   

   time_diff_seconds  time_diff  
0            38520.0       11.0  
1            42120.0       12.0  
2            45720.0       13.0  

Or dt.ceil:

df['time_diff']= (df['timeStamp'] - df['forecastRan']).dt.ceil('h')
print (df)
            timeStamp         forecastRan  time_diff_in_hours  \
0 2016-11-23 23:00:00 2016-11-23 12:18:00                10.0   
1 2016-11-24 00:00:00 2016-11-23 12:18:00                11.0   
2 2016-11-24 01:00:00 2016-11-23 12:18:00                12.0   

   time_diff_seconds time_diff  
0            38520.0  11:00:00  
1            42120.0  12:00:00  
2            45720.0  13:00:00 

---

df['time_diff' ]= (df['timeStamp'] - df['forecastRan']).dt.ceil('h').astype('timedelta64[h]')
print (df)
            timeStamp         forecastRan  time_diff_in_hours  \
0 2016-11-23 23:00:00 2016-11-23 12:18:00                10.0   
1 2016-11-24 00:00:00 2016-11-23 12:18:00                11.0   
2 2016-11-24 01:00:00 2016-11-23 12:18:00                12.0   

   time_diff_seconds  time_diff  
0            38520.0       11.0  
1            42120.0       12.0  
2            45720.0       13.0