Merging pandas dataframes on 2 columns but in either order

Question

The problem:

I've got a situation with 2 dataframes:

test1 = pd.DataFrame({'id_A':['Ben', 'Julie', 'Jack', 'Jack'],
                  'id_B':['Julie', 'Ben', 'Nina', 'Julie']})

test2 = pd.DataFrame({'id_a':['Ben', 'Ben', 'Ben', 'Julie', 'Julie', 'Nina'],
                      'id_b':['Julie', 'Nina', 'Jack', 'Nina', 'Jack', 'Jack'],
                      'value':[1,1,0,0,1,0]})

>>> test1
    id_A   id_B
0    Ben  Julie
1  Julie    Ben
2   Jack   Nina
3   Jack  Julie

>>> test2
    id_a   id_b  value
0    Ben  Julie      1
1    Ben   Nina      1
2    Ben   Jack      0
3  Julie   Nina      0
4  Julie   Jack      1
5   Nina   Jack      0

What I'd like to do is merge test2 with test1 where id_A == id_a and id_B == id_b OR where id_A == id_b and id_B == id_a, resulting in the following dataframe:

>>> final_df
    id_A   id_B  value
0    Ben  Julie      1
1  Julie    Ben      1
2   Jack   Nina      0
3   Jack  Julie      1

Current Solution:

My solution works but seems messy, and I'd like to see if I'm overlooking some more clever way to do things. It involves concatenating test2 with itself, but reversing the 2 columns of interest (id_a becomes id_b and vice-versa), and then merging from there.

test3 = pd.concat([test2, test2.rename(columns = {'id_a':'id_b', 'id_b':'id_a'})])

final_df = (test1.merge(test3, left_on = ['id_A', 'id_B'],
                        right_on = ['id_a', 'id_b'])
            .drop(['id_a', 'id_b'], axis=1))

Question:

Does anyone know a neater way to do this? I feel like I'm probably overlooking some amazing pandorable way of doing things.

Thanks for your help!

Answer 1

With frozenset

test1.assign(
    value=test1.apply(frozenset, 1).map({frozenset(a): b for *a, b in test2.values}))

    id_A   id_B  value
0    Ben  Julie      1
1  Julie    Ben      1
2   Jack   Nina      0
3   Jack  Julie      1

---

Less cute, more robust. Remove what you need to afterwards.

t1 = test1.assign(ref=list(map(frozenset, zip(test1.id_A, test1.id_B))))
t2 = test2.assign(ref=list(map(frozenset, zip(test2.id_a, test2.id_b))))

t1.merge(t2, on='ref')

    id_A   id_B            ref   id_a   id_b  value
0    Ben  Julie   (Julie, Ben)    Ben  Julie      1
1  Julie    Ben   (Julie, Ben)    Ben  Julie      1
2   Jack   Nina   (Jack, Nina)   Nina   Jack      0
3   Jack  Julie  (Jack, Julie)  Julie   Jack      1