根据另一个数据框中的数据将值分配给pandas列

Question

I have 2 dataframes

df1
ID ID2 NUMBER
1 2 null

df2
ID ID2 NUMBER 
1 2 1
1 2 2
1 2 3

So when doing merge between df1 and df2 usin ID and ID2 I get duplicated columns because df1 has 3 matches in df2. I'd like to assign a random number to df1 and use it for merging, this way I always get 1 to 1 merge. The problem is that my dataset is rather big and sometimes I have only 1 row in df2 (so merge works properly) and sometimes I have 10+ rows in df2. I'd like to assign a number to df1 using:

rand(1,len(df1[(df1.ID=1) & (df1.ID2=2]))

Answer 1

I think I found a solution I'm posting it here so others can tell me if there is a better way.

def select_random_row(grp):
    ID= grp.ID.iloc[0]
    ID2= grp.ID2.iloc[0] 
    return random.randint(1, len(df1[(df1.ID== ID) & (df1.ID2 == ID2)]))

df2['g'] = df2.groupby(['ID','ID2']).apply(select_random_row)

EDIT: This is way to slow to do on large dataset... I decided to just use drop_duplicates before merging and keep 1st record. It isn't randomly like I wanted but it is better than nothing