CODEWITHSUNDEEP
pythonpandasdataframerowpandas-groupby
H_A
H_A

Reputation: 677

Merge multiple values of a column after group by into one column in python pandas

I am looking to do similar task as in this thread: Merge multiple column values into one column in python pandas

But the difference is that I would like to create a new column that merges all the non-null values in a column after grouping by another column(s). Here is a toy example:

df= pd.DataFrame({'ID1' : [1,1,2,2,3,3,3],'ID2' : ['a','a','b','b','c','c','c'],
             'Status' : pd.Series([np.nan,'1', np.nan,'1','2',np.nan,'1'], 
                                  dtype="category")})

 df
Out[74]: 
     ID1 ID2 Status
 0    1   a    NaN
 1    1   a      1
 2    2   b    NaN
 3    2   b      1
 4    3   c      2
 5    3   c    NaN
 6    3   c      1

I then want to groupby ID1 and ID2:

gr = df.groupby(['ID1','ID2'])

Then, I would like my result look like this:

Out:
   NewCol
0   1
1   1
2   2,1

So it is a new DataFrame that contains the non-null values of Status column grouped by ID1 and ID2.

Thanks in advance.

Upvotes: 5

Views: 3549

Answers (1)

jezrael
jezrael

Reputation: 863611

Use GroupBy.agg with lambda function is most general solution:

df1 = df.groupby(['ID1','ID2'])['Status'].agg(lambda x: ','.join(x.dropna())).reset_index()
print (df1)
   ID1 ID2 Status
0    1   a      1
1    2   b      1
2    3   c    2,1

Another idea is remove duplicated in first step, but if some group with only misisng values it is removed from output, so is necessary next processing like merge:

#first group with only NaNs
df= pd.DataFrame({'ID1' : [1,1,2,2,3,3,3],'ID2' : ['a','a','b','b','c','c','c'],
             'Status' : pd.Series([np.nan,np.nan, np.nan,'1','2',np.nan,'1'], 
                                  dtype="category")})


#first group is dropped
df11 = (df.dropna(subset=['Status'])
          .groupby(['ID1','ID2'])['Status']
          .agg(','.join)
          .reset_index())
print (df11)
   ID1 ID2 Status
0    2   b      1
1    3   c    2,1

#added missing pairs and `NaN`s converted to empty strings:
df2 = df.drop_duplicates(['ID1','ID2'])[['ID1','ID2']].merge(df11, how='left').fillna('')
print (df2)
   ID1 ID2 Status
0    1   a       
1    2   b      1
2    3   c    2,1

First solution:

df1 = df.groupby(['ID1','ID2'])['Status'].agg(lambda x: ','.join(x.dropna())).reset_index()
print (df1)
   ID1 ID2 Status
0    1   a       
1    2   b      1
2    3   c    2,1

Upvotes: 4

Related Questions