Subtract sum of two rows from another row based on condition in Pandas

Question

I have a dataframe with three columns- Col1 , Col2 and Col3

Col1	Col2	Col3
x	a	10
x	b	12
x	c	25
y	a	13
y	b	14
y	c	37

I want to update the value of Col3 for each c from Col2 for each unique Col1 values by the below logic:

c-(a+b)

The desired output:

Col1	Col2	Col3
x	a	10
x	b	12
x	c	3
y	a	13
y	b	14
y	c	10

I would like to know the appropriate operation that I need to do to achieve the desired solution. Thanks

Answer 1

We can select the subset of rows where the Col2 values is either a or b , then group these rows by Col1 and transform using sum to calculate the transformed sum a + b per group, finally subtract the transformed sum from Col3 where the corresponding Col2 is c

m = df['Col2'].isin(['a', 'b'])
s = df['Col3'].where(m).groupby(df['Col1']).transform('sum')
df.loc[df['Col2'].eq('c'), 'Col3'] -= s

---

  Col1 Col2  Col3
0    x    a  10.0
1    x    b  12.0
2    x    c   3.0
3    y    a  13.0
4    y    b  14.0
5    y    c  10.0

Answer 2

Might be verbose, but a possible way if each group has only one a , b and c .

def subtract(group):
    result = group.loc[df['Col2'] == 'c', 'Col3'].values - (group.loc[df['Col2'] == 'a', 'Col3'].values + group.loc[df['Col2'] == 'b', 'Col3'].values)
    group.loc[df['Col2'] == 'c', 'Col3'] = result
    return group

df_ = df.groupby('Col1').apply(subtract)

print(df_)

  Col1 Col2  Col3
0    x    a    10
1    x    b    12
2    x    c     3
3    y    a    13
4    y    b    14
5    y    c    10

Answer 3

Well, If you wanna perform this kind of operation. Why not restructure the data?. You can use pivot_table ->

df1 = df.pivot_table(index = 'Col1',columns='Col2', values='Col3')
print(df1)

Output-

Col2   a   b   c
Col1            
x     10  12  25
y     13  14  37

Then you can perform the operations that you want easily and after that, if you want data back in its original format you can always use the melt function.

df1['c'] -= df1['a']+ df1['b']
df1.melt()