I have Pandas DataFrame like below (data types of "ID" and "COL1" is "object"):

ID  | COL1 | COL2 | COL3
----|------|------|----
123 | ABc  | 55   | G4
123 | Abc  | 55   | G4
123 | DD   | 55   | G4
44  | RoR  | 41   | P0
44  | RoR  | 41   | P0
55  | XX   | 456  | RR

And I need to:

1. Create new column "COL1_cum" where will be all values from "COL1" per ID separated by commas
2. Drop duplicated IDs
3. Create new column "COL1_num" where will be information how many different levels is in "COL1" per "ID"

So as a result I need something like below:

ID  | COL1_cum | COL1_num |COL2 | COL3
----|----------|----------|-----|-----
123 | ABc, DD  | 2        | 55  | G4
44  | RoR      | 1        | 41  | P0
55  | XX       | 1        | 456 | RR

Explanation for COL1_num:

• for ID = 123 COL1_num = 2 because for ID = 123 in "COL1" we have 2 different values: "ABc" and "DD"
• for ID = 44 COL1_num = 1 because for ID = 44 in "COL1" we have 1 value: "RoR"
• for ID = 55 COL1_num = 1 because for ID = 5 in "COL1" we have 1 value: "XX"

How can I do that in Python Pandas?

CodePudding user response：

If there are 2 columns in input data use DataFrame.drop_duplicates with aggregate join:

df1 = df.drop_duplicates().groupby('ID')['COL1'].agg(','.join).reset_index(name='COL1_cum')

If possible multiple columns is possible specify them:

df1 = (df.drop_duplicates(['ID','COL1'])
         .groupby('ID')['COL1']
         .agg(','.join)
         .reset_index(name='COL1_cum'))

EDIT:

First remove duplciates per all columns:

df1 = df.drop_duplicates()
print (df1)
    ID COL1  COL2 COL3
0  123  ABc    55   G4
2  123   DD    55   G4
3   44  RoR    41   P0
5   55   XX   456   RR

Then aggregate join, size and get first values per another columns (because same values per groups ID):

df2 = (df1.groupby('ID', sort=False, as_index=False)
          .agg(COL1_cum =('COL1',','.join),
               COL1_num=('COL1','size'),
               COL2=('COL2','first'),
                COL3=('COL3','first')))
print (df2)
    ID COL1_cum  COL1_num  COL2 COL3
0  123   ABc,DD         2    55   G4
1   44      RoR         1    41   P0
2   55       XX         1   456   RR

EDIT2: Real data are not duplicated by all columns, possible solution:

df2 = (df.groupby('ID', sort=False, as_index=False)
          .agg(COL1_cum =('COL1',lambda x: ','.join(dict.fromkeys(x))),
               COL1_num=('COL1','nunique'),
               COL2=('COL2','first'),
                COL3=('COL3','first')))
print (df2)
    ID COL1_cum  COL1_num  COL2 COL3
0  123   ABc,DD         2    55   G4
1   44      RoR         1    41   P0
2   55       XX         1   456   RR