python pandas new column categorization based on conditions in other columns

Question

Working with the following python pandas dataframe df:

df = pd.DataFrame({'transaction_id': ['A123','A123','B345','B345','C567','C567','D678','D678'], 
                   'product_id': [255472, 251235, 253764,257344,221577,209809,223551,290678],
                   'product_category': ['X','X','Y','Y','X','Y','Y','X']})

transaction_id | product_id | product_category
A123              255472             X
A123              251235             X
B345              253764             Y
B345              257344             Y
C567              221577             X
C567              209809             Y
D678              223551             Y
D678              290678             X

I need to add another column "transaction_category", which looks at the transaction_id and which product categories are in the transaction_id. This is the output I am looking for:

transaction_id | product_id | product_category | transaction_id
123              255472             X                X only
123              251235             X                X only
345              253764             Y                Y only
345              257344             Y                Y only
567              221577             X                X & Y
567              209809             Y                X & Y
678              223551             Y                X & Y
678              290678             X                X & Y

Please note that I have other columns in my dataframe that I am not using, so I guess I need to start with a grouby?

df2 = df.groupby(['transaction_id','product_category']).reset_index()

Answer 1

IIUC by using transform and join

df.groupby('transaction_id').product_category.transform(lambda x : '&'.join(set(x)))
Out[468]: 
0      X
1      X
2      Y
3      Y
4    X&Y
5    X&Y
6    X&Y
7    X&Y
Name: product_category, dtype: object

From scott match your expected out put :

df['transaction_category']=df.groupby('transaction_id')['product_category'].transform(lambda x: x + ' only' if len(set(x)) < 2 else ' & '.join(set(x)))
df
Out[479]: 
  product_category  product_id transaction_id transaction_category
0                X      255472           A123               X only
1                X      251235           A123               X only
2                Y      253764           B345               Y only
3                Y      257344           B345               Y only
4                X      221577           C567                X & Y
5                Y      209809           C567                X & Y
6                Y      223551           D678                X & Y
7                X      290678           D678                X & Y

Answer 2

the transform method of the groupby object allows your to add full-length columns back to your dataframe via assign:

import pandas

def squeezer(x):
    _x = list(set(x.values))
    if len(_x) == 1:
        return '{} only'.format(_x[0])
    else:
        return ' & '.join(sorted(_x))

df = pandas.DataFrame({
    'transaction_id': ['A123','A123','B345','B345','C567','C567','D678','D678'], 
    'product_id': [255472, 251235, 253764,257344,221577,209809,223551,290678],
    'product_category': ['X','X','Y','Y','X','Y','Y','X']
}).assign(
    products=lambda df:
            df.groupby(by=['transaction_id'])['product_category']
              .transform(squeezer)
)

And we get:

  product_category  product_id transaction_id products
0                X      255472           A123   X only
1                X      251235           A123   X only
2                Y      253764           B345   Y only
3                Y      257344           B345   Y only
4                X      221577           C567    X & Y
5                Y      209809           C567    X & Y
6                Y      223551           D678    X & Y
7                X      290678           D678    X & Y