Complex pandas isin function

Question

I have a dataframe:

In [47]: df
Out[47]: 
    uid   a   b
0   111   1   2
1   111   2   3
2   111   4   5
3   111   6   7
4   111   5   8
5   222   0   9
6   222  11  12
7   222  13  11
8   222   2   1
9   333  14  16
10  333   3   2
11  333  16  19
12  333   3   4
13  444  21  20
14  444   9   5
15  444  20  24
16  444   5   6

I want to check if values in b are present in a,vice-versa only if the uid is same. I used isin :

df[(df.b.isin(df.a))|(df.a.isin(df.b))]

but this goes through all values and does not give me the desired output.

Desired output:

 Out[49]: 
    uid   a   b
0   111   1   2
1   111   2   3
2   111   4   5
4   111   5   8
6   222  11  12
7   222  13  11
9   333  14  16
11  333  16  19
13  444  21  20
14  444   9   5
15  444  20  24
16  444   5   6

Answer 1

I think you need groupby and apply boolean indexing :

print (df.groupby('uid').apply(lambda x: x[(x.b.isin(x.a))|(x.a.isin(x.b))]))
        uid   a   b
uid                
111 0   111   1   2
    1   111   2   3
    2   111   4   5
    4   111   5   8
222 6   222  11  12
    7   222  13  11
333 9   333  14  16
    11  333  16  19
444 13  444  21  20
    14  444   9   5
    15  444  20  24
    16  444   5   6

print (df.groupby('uid')
         .apply(lambda x: x[(x.b.isin(x.a))|(x.a.isin(x.b))])
         .reset_index(drop=True))
    uid   a   b
0   111   1   2
1   111   2   3
2   111   4   5
3   111   5   8
4   222  11  12
5   222  13  11
6   333  14  16
7   333  16  19
8   444  21  20
9   444   9   5
10  444  20  24
11  444   5   6