Adding subtotal / grand total row in pandas muiltiindex produces tuple as an index-CodePudding

I have a df:

df = pd.DataFrame.from_dict({'category': {1050: 'Dining',
  992: 'Dining',
  1054: 'Kitchen',
  1052: 'Kitchen',
  993: 'Living room',
  980: 'Living room',
  996: 'Dining',
  1017: 'Dining',
  1020: 'Bath',
  1001: 'Bath'},
 'subcategory': {1050: 'Chairs',
  992: 'Chairs',
  1054: 'Stool',
  1052: 'Mirror',
  993: 'mirror',
  980: 'chair',
  996: 'Chairs',
  1017: 'Chairs',
  1020: 'Table',
  1001: 'Table'},
 'discount': {1050: '30-40',
  992: '30-40',
  1054: '30-40',
  1052: '30-40',
  993: '30-40',
  980: '30-40',
  996: '30-40',
  1017: '30-40',
  1020: '30-40',
  1001: '30-40'},
 'sales_1': {1050: 9539.86,
  992: 12971.86,
  1054: 6736.53,
  1052: 7163.16,
  993: 8601.16,
  980: 8047.16,
  996: 16322.0,
  1017: 14424.32,
  1020: 6319.58,
  1001: 4551.42},
 'sales_2': {1050: 3226.0,
  992: 11117.0,
  1054: 1613.0,
  1052: 2166.0,
  993: 11117.0,
  980: 3442.0,
  996: 19365.0,
  1017: 3323.0,
  1020: 1411.0,
  1001: 572.0}})

I am trying to add a subtotal in multiindex. I am able to add it with 2 groups like so:

dd =  df_from_dict.groupby(['category', 'subcategory'])[['sales_1', 'sales_2']].sum()

s = dd.groupby(level=0).sum()
s.index = pd.MultiIndex.from_product([s.index, ['Total']])
dd = dd.append(s).sort_index()
dd.loc['Grand Total', :] = dd.sum().values / 2

dd

But when I add 3rd item to the group, discount

dd =  df_from_dict.groupby(['category', 'subcategory','discount'])[['sales_1', 'sales_2']].sum()

s = dd.groupby(level=0).sum()
s.index = pd.MultiIndex.from_product([s.index, ['Total']])
dd = dd.append(s).sort_index()
dd.loc['Grand Total', :] = dd.sum().values / 2

dd

All of a sudden I get tuples instead of normal multiindex. Instead of 3 indexes I get 1 as a tuple:

Whereas I want the same structure as in the first picture but with another level of index. I tried playing with level=1 parameter in group by but it always ends up in a single index as a tuple, I am not sure where is my mistake.

CodePudding user response：

Here is problem in Series called s are 2 levels MultiIndex, in dd are 3 levels, so in append are created tuple.

Solution is set 3 levels MultiIndex in MultiIndex.from_product, so same number levels like dd and solution working correct:

For avoid sorting all another level in DataFrame.sort_index add sort_remaining=False:

dd =  df_from_dict.groupby(['category', 'subcategory','discount'])[['sales_1', 'sales_2']].sum()

s = dd.groupby(level=0).sum()
s.index = pd.MultiIndex.from_product([s.index, ['Total'], ['']])
print (s)

dd = dd.append(s).sort_index(level=0, sort_remaining=False)
dd.loc['Grand Total', :] = dd.sum().values / 2
print (dd)
                                   sales_1  sales_2
category    subcategory discount                   
Bath        Table       30-40     10871.00   1983.0
            Total                 10871.00   1983.0
Dining      Chairs      30-40     53258.04  37031.0
            Total                 53258.04  37031.0
Kitchen     Mirror      30-40      7163.16   2166.0
            Stool       30-40      6736.53   1613.0
            Total                 13899.69   3779.0
Living room chair       30-40      8047.16   3442.0
            mirror      30-40      8601.16  11117.0
            Total                 16648.32  14559.0
Grand Total                       94677.05  57352.0