Pandas group by multiple custom aggregate function on multiple columns

Question

Given data:

grp	data1	data2	data3
a	2	1	2
a	4	6	3
b	3	2	1
b	7	3	5

Expected output:

grp	sum(data1)	sum(data2)/sum(data1)	sum(data3)/sum(data1)
a	6	1.166666667	0.83
a	10	0.5	0.6

Assume custom aggregation can be dependent on multiple columns and not always a simple division operation. I know using SQL query it's possible, but I am interested in an answer with apply and aggregate function if possible.

Answer 1

You can use groupby + assign here to generate required aggregations. You can apply whatever aggregate function is needed.

g = df.groupby('grp')
#                                         for custom agg func use .agg(custom_agg_func)
#                                                          ^^^^^
g[['data1']].agg('sum').assign(sum2 = lambda df: g['data2'].sum()/df['data1'],
                               sum3 = lambda df: g['data3'].sum()/df['data1'])
#                ^^^^^^
#    you can use custom agg func of your choice

     data1      sum2      sum3
grp                           
a        6  1.166667  0.833333
b       10  0.500000  0.600000