how to concat sets when using groupby in pandas dataframe?

Question

This is my dataframe:

> df
       a             b
    0  1         set([2, 3])
    1  2         set([2, 3])
    2  3      set([4, 5, 6])
    3  1  set([1, 34, 3, 2])

Now when I groupby, I want to update sets. If it was a list there was no problem. But the output of my command is:

> df.groupby('a').sum()

a         b                
1             NaN
2     set([2, 3])
3  set([4, 5, 6])

What should I do in groupby to update sets? The output I'm looking for is as below:

a         b                
1     set([2, 3, 1, 34])
2     set([2, 3])
3     set([4, 5, 6])

matt_s · Accepted Answer · 2015-10-06 11:12:12Z

17

This might be close to what you want

df.groupby('a').apply(lambda x: set.union(*x.b))

In this case it takes the union of the sets.

If you need to keep the column names you could use:

df.groupby('a').agg({'b':lambda x: set.union(*x)}).reset_index('a')

Result:

    a   b
0   1   set([1, 2, 3, 34])
1   2   set([2, 3])
2   3   set([4, 5, 6])

answered Oct 6, 2015 at 10:39

matt_s

1,0751 gold badge10 silver badges18 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Thanks, It solved set problem, but column name renamed to 0. Why that happened?

It's because the result is a Series so no column name. I've added a method for keeping the column name if you need it.