Comma separated values from pandas GroupBy

问题

i trying to find out if there is away to remove duplicate in my data frame while concatenating the value

example:

df
   key  v1  v2
0  1   n/a  a
1  2   n/a  b
2  3   n/a  c
3  2   n/a  d
4  3   n/a  e

the out put should be like:

 df_out
   key v1   v2
0  1   n/a  a
1  2   n/a  b,d
2  3   n/a  c,e

I try using df.drop_duplicates() and some loop to save the v2 column value and nothing yet. i'm trying to do it nice and clean with out loop by using Pandas.

some one know a way pandas can do it?

回答1:

This should be easy, assuming you have two columns. Use groupby + agg. v1 should be aggregated by first, and v2 should be aggregated by ','.join.

df
   key  v1 v2
0    1 NaN  a
1    2 NaN  b
2    3 NaN  c
3    2 NaN  d
4    3 NaN  e

(df.groupby('key')
   .agg({'v1' : 'first', 'v2' : ','.join})
   .reset_index()
   .reindex(columns=df.columns))

   key  v1   v2
0    1 NaN    a
1    2 NaN  b,d
2    3 NaN  c,e

If you have multiple such columns requiring the same aggregation, build an agg dict called f and pass it to agg.

回答2:

Using set

df.groupby('key').agg(lambda x : ','.join(set(x)))
Out[1255]: 
      v1   v2
key          
1    n/a    a
2    n/a  b,d
3    n/a  c,e

回答3:

Use apply

pandas.core.groupby.GroupBy.apply

GroupBy.apply(func, *args, **kwargs)[source]

Apply function func group-wise and combine the results together.

df.groupby(["key", "v1"])["v2"].apply(list) # or apply(set) depending on your needs

Output:

key  v1
1    n/a       [a]
2    n/a    [b, d]
3    n/a    [c, e]
Name: v2, dtype: object

来源：https://stackoverflow.com/questions/47980402/comma-separated-values-from-pandas-groupby