AttributeError: 'Series' object has no attribute 'columns'

I have a dataframe that looks like this:

id userid int1 int2 string string2 string3 1 90 5067 1000 aaa 100 qqq 1 90 6945 1000 bbb 101 qqq 1 90 9596 1010 ccc 102 qqq

The desired output:

id userid int1 int2 string string2 string3
1 90 [5067,6945,9596] [1000,1000,1010] [aaa,bbb,ccc] [100,101,102] qqq

I have tried to create a new dataframe for int1,int2,string,string2 and then merge them but this solution was very slow as I have a big dataset. A different solution is to use 'groupby':

df = df.groupby(['id','userid','string3']).[['int1'],['int2'],['string'],['string2']].apply(list).reset_index()

but this gives me this error:

AttributeError: 'Series' object has no attribute 'columns'

Any help is appreciated.

2 Answers

Use GroupBy.agg with as_index=False + DataFrame.reindex to return the columns in the initial order:

new_df=( df.groupby(['id','userid','string3'],as_index=False) .agg(list) .reindex(columns=df.columns) )
print(new_df)

If you want you could select the columns:

cols=['int1','int2','string','string2']
new_df=( df.groupby(['id','userid','string3'],as_index=False)[cols] .agg(list) .reindex(columns=df.columns) )

Output

 id userid int1 int2 string \
0 1 90 [5067, 6945, 9596] [1000, 1000, 1010] [aaa, bbb, ccc] string2 string3
0 [100, 101, 102] qqq 

You are passing individual values (lists) within a list. You don't need to use doble brackets like you do when filtering a dataframe and expect a dataframe output. Hopefully this will work better for you:

df = df.groupby(['id','userid','string3'])['int1','int2','string','string2'].apply(list).reset_index()