Pandas resample/groupby day of week and year

Question

I am trying to create a report that is grouped by day of week for each year.

I have a df that looks like this:

                        s1      s2   srd
dt          
2004-02-04 11:21:00 2365.79 2372.37 -7.0
2004-02-05 10:15:00 2365.79 2368.03 -2.0
2004-02-17 06:43:00 2421.05 2425.26 -4.0
2004-02-17 12:43:00 2418.42 2420.53 -2.0
2004-02-17 12:44:00 2420.39 2420.53 -0.0

The dt index is in datetime format.

What I am looking for is a dataframe that looks like this (I only need srd column and function to group can be anything, like sum, count, etc.):

                  srd
dayOfWeek  year         
Mon        2004    10
           2005    11
           2006    8
           2007    120
Tues       2004    105
           2005    105

I have tried dayOfWeekDf = df.resample('B') , but I get a dataframe that looks like it is split by week number.

I also tried df.groupby([df.index.weekday, df.index.year])['srd'].transform('sum') , but it does not even group for some reason, as I get the following (Feb 17th appears 3 times).

                    srd
dt  
2004-02-04 11:21:00 81.0
2004-02-05 10:15:00 203.0
2004-02-17 06:43:00 37.0
2004-02-17 12:43:00 37.0
2004-02-17 12:44:00 37.0

Answer 1

If you want the dayOfWeek and year names in the index, you can assign them:

>>> df.assign(year=df.index.year, dayOfWeek = df.index.weekday_name).groupby(['dayOfWeek','year']).srd.sum()
dayOfWeek  year
Thursday   2004   -2.0
Tuesday    2004   -6.0
Wednesday  2004   -7.0
Name: srd, dtype: float64

Otherwise, you can use the way you were doing, but omit the transform :

>>> df.groupby([df.index.weekday_name, df.index.year])['srd'].sum()
dt         dt  
Thursday   2004   -2.0
Tuesday    2004   -6.0
Wednesday  2004   -7.0
Name: srd, dtype: float64