Custom Grouping For All Possible Groups When Having Missing Values

I have a dictionary which represents a set of products. I need to find all duplicate products within these products. If products have same product_type,color and size -> they are duplicates. I could easily group by ('product_type','color','size') if I did not have a problem: some values are missing. Now I have to find all possible groups of products that might be duplicates between themselves. This means that some elements can appear in multiple groups.

Let me illustrate:

import pandas as pd


def main():
    data= {'product_id': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11],
         'product_type': ['shirt', 'shirt', 'shirt', 'shirt', 'shirt', 'hat', 'hat', 'hat', 'hat', 'hat', 'hat', ],
         'color': [None, None, None, 'red', 'blue', None, 'blue', 'blue', 'blue', 'red', 'red', ],
                       'size': [None, 's', 'xl', None, None, 's', None, 's', 'xl', None, 'xl', ],
                       }
    print(data)

if __name__ == '__main__':
    main()

for this data:

I need this result - list of possibly duplicate products for each possible group (take only the biggest super groups):

So for example, lets take "shirt" with id=1 this product does not have color or size so he can appear in a possible "duplicates group" together with shirt #2 (which has size "s" but does not have color) and shirt #4 (which has color "red" but does not have size). So these three shirts (1,2,4) are possibly duplicates with same color "red" and size "s".

I tried to implement it by looping through all possible combinations of missing values but it feels wrong and complex.

Is there a way to get the desired result?

You can create all possible keys that are not None and then check which item falls into what key - respecting the Nones:

data= {'product_id'  : [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11],
       'product_type': ['shirt', 'shirt', 'shirt', 'shirt', 'shirt', 'hat',
                        'hat', 'hat', 'hat', 'hat', 'hat', ],
       'color'       : [None, None, None, 'red', 'blue', None, 'blue', 
                        'blue', 'blue', 'red', 'red', ],
       'size'        : [None, 's', 'xl', None, None, 's', None, 's', 'xl', None, 'xl', ]}



from itertools import product

# create all keys without None in it     
p = product((t for t in set(data['product_type']) if t), 
            (c for c in set(data['color']) if c), 
            (s for s in set(data['size']) if s))

# create the things you have in stock
inventar = list( zip(data['product_id'],data['product_type'],data['color'],data['size']))
d = {}

# order things into its categories
for cat in p:
    d.setdefault(cat,set())  # uses a set to collect the IDs
    for item in inventar:
        TY, CO, SI = cat
        ID, TYPE, COLOR, SIZE = item

        # the (TYPE or TY) will substitute TY for any TYPE that is None etc.
        if (TYPE or TY)==TY and (COLOR or CO)==CO and (SIZE or SI)==SI:
            d[cat].add(ID)

print(d)

Output:

# category-key            id's that match
{('shirt', 'blue', 's') : {1, 2, 5}, 
 ('shirt', 'blue', 'xl'): {1, 3, 5}, 
 ('shirt', 'red', 's')  : {1, 2, 4}, 
 ('shirt', 'red', 'xl') : {1, 3, 4}, 
 ('hat', 'blue', 's')   : {8, 6, 7}, 
 ('hat', 'blue', 'xl')  : {9, 7}, 
 ('hat', 'red', 's')    : {10, 6},
 ('hat', 'red', 'xl')   : {10, 11}}

Doku:

• itertools.product(*iterables)
• zip(*iterables)
• truth value testing