删除反向重复

Question

My dataframe looks like this:

df_in = pd.DataFrame(data={'mol1':['cpd1','cpd2', 'cpd3'], 'mol2': ['cpd2','cpd1', 'cpd4'], 'sim': [0.8,0.8,0.9]})

print(df_in)

   mol1  mol2  sim
0  cpd1  cpd2  0.8
1  cpd2  cpd1  0.8
2  cpd3  cpd4  0.9

The pair (cpd1, cpd2) occurs twice although each element does not belong to the same column.

I would like to get rid of these duplicates to end up with this:

df_out = pd.DataFrame(data={'mol1':['cpd1', 'cpd3'], 'mol2': ['cpd2', 'cpd4'], 'sim': [0.8,0.9]})

print(df_out)

   mol1  mol2  sim
0  cpd1  cpd2  0.8
1  cpd3  cpd4  0.9

If I ignore the third column, there is a solution describes in Pythonic way of removing reversed duplicates in list , but I have to preserve this column.

Answer 1

You can use sorted with apply for columns from list cols and then drop_duplicates :

cols = ['mol1','mol2']
df[cols] = df[cols].apply(sorted, axis=1)
df = df.drop_duplicates()
print (df)
   mol1  mol2  sim
0  cpd1  cpd2  0.8
2  cpd3  cpd4  0.9

Similar solution with numpy.sort :

cols = ['mol1','mol2']
df[cols] = np.sort(df[cols].values, axis=1)
df = df.drop_duplicates()
print (df)
   mol1  mol2  sim
0  cpd1  cpd2  0.8
2  cpd3  cpd4  0.9

If need check duplicates only in cols add parameter subset :

df = pd.DataFrame(
{'mol1':['cpd1','cpd2', 'cpd3'], 
'mol2': ['cpd2', 'cpd1', 'cpd4'], 
'sim': [0.7,0.8,0.9]})
print (df)
   mol1  mol2  sim
0  cpd1  cpd2  0.7
1  cpd2  cpd1  0.8
2  cpd3  cpd4  0.9

cols = ['mol1','mol2']
df[cols] = np.sort(df[cols].values, axis=1)
df = df.drop_duplicates(subset=cols)
print (df)
   mol1  mol2  sim
0  cpd1  cpd2  0.7
2  cpd3  cpd4  0.9