Pandas: Return Hour from Datetime Column Directly

Question

Assume I have a DataFrame sales of timestamp values:

timestamp               sales_office 2014-01-01 09:01:00     Cincinnati 2014-01-01 09:11:00     San Francisco 2014-01-01 15:22:00     Chicago 2014-01-01 19:01:00     Chicago 

I would like to create a new column time_hour. I can create it by writing a short function as so and using apply() to apply it iteratively:

def hr_func(ts):     return ts.hour  sales['time_hour'] = sales['timestamp'].apply(hr_func) 

I would then see this result:

timestamp               sales_office         time_hour 2014-01-01 09:01:00     Cincinnati           9 2014-01-01 09:11:00     San Francisco        9 2014-01-01 15:22:00     Chicago              15 2014-01-01 19:01:00     Chicago              19 

What I'd like to achieve is some shorter transformation like this (which I know is erroneous but gets at the spirit):

sales['time_hour'] = sales['timestamp'].hour 

Obviously the column is of type Series and as such doesn't have those attributes, but it seems there's a simpler way to make use of matrix operations.

Is there a more-direct approach?

Answer 1

Assuming timestamp is the index of the data frame, you can just do the following:

hours = sales.index.hour 

If you want to add that to your sales data frame, just do:

import pandas as pd pd.concat([sales, pd.DataFrame(hours, index=sales.index)], axis = 1) 

Edit: If you have several columns of datetime objects, it's the same process. If you have a column ['date'] in your data frame, and assuming that 'date' has datetime values, you can access the hour from the 'date' as:

hours = sales['date'].hour 

Edit2: If you want to adjust a column in your data frame you have to include dt:

sales['datehour'] = sales['date'].dt.hour