I have 2 dataframes looking like this :
In each dataframe there is pattern of 1-2 in the values column. (the values are not significant to my problem, just to demonstrate the pattern)
df1 = {'idx': [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
'values': [20, 1000, 10001, 21, 1000, 1002, 22, 1003, 1007,23]}
df2 = {'idx': [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
'values': [1000, 21, 22, 1000, 22, 23, 1000, 20, 21, 1000]}
I need to swap rows between the two dataframes so that the outcome would be :
df_expected1 = {'idx': [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
'values': [20, 21, 22, 21, 22, 23, 22, 20, 21,23]}
df_expected2 = {'idx': [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
'values': [1000, 1000, 10001, 1000, 1000, 1002, 1000, 1003, 1007, 1000]}
Knowing that swap-needed indices are from 3n-2, you can make mask then use numpy.where:
m = df1["idx"].add(2).mod(3).eq(0)
s1 = np.where(m, df1["values"], df2["values"])
s2 = np.where(~m, df1["values"], df2["values"])
df1["values"] = s1
df2["values"] = s2
Output:
idx values
0 1 20
1 2 21
2 3 22
3 4 21
4 5 22
5 6 23
6 7 22
7 8 20
8 9 21
9 10 23
idx values
0 1 1000
1 2 1000
2 3 10001
3 4 1000
4 5 1000
5 6 1002
6 7 1000
7 8 1003
8 9 1007
9 10 1000