How to get topN elements from this dict with list values?

Question

For example:

from collections import defaultdict

tag_dict = {'a':[('mary', 0.99, 'f'), ('tom', 0.87), 'm'],
            'b':[('jack', 0.43, 'm')],
            'c':[('martin', 0.987, 'm'), ('alice', 0.973, 'f')]}

I want to get a new defaultdict which contains top 3 elements sorted by the float number in each element. So for the example above, the result would be:

top3_dict = {'a':[('mary', 0.99, 'f')],
             'c':[('martin', 0.987, 'm'), ('alice', 0.973, 'f')]}

Is there an efficient way to do this?

Edit:

My actual question is a little more complicated. There are cases in which the original dict size is less than 3, so in that case the correct way to specify the problem is that:

Get up to topN elements sorted by the float number, if the elements are less n, just get all elements sorted order.

Answer 1

You can flatten the structure out and store the outside key with the pairs. Then sort this, take the top three items and make the dictionary:

from collections import defaultdict

result = defaultdict(list)

sorted_items = sorted(
    [(k, item) for k,v in tag_dict.items() 
     for item in v], 
    key=lambda p: p[1][1], reverse=True)

for k, v in sorted_items[:3]:
    result[k].append(v)
    

Result will be:

defaultdict(list,
            {'a': [('mary', 0.99, 'f')],
             'c': [('martin', 0.987, 'm'), ('alice', 0.973, 'f')]})

Above sorted_items will look like:

[('a', ('mary', 0.99, 'f')),
 ('c', ('martin', 0.987, 'm')) 
 ...
]

at which point making a new dict is simple.

Edit:

My actual question is a little more complicated. There are cases in which the original dict size is less than 3, so in that case the correct way to specify the problem is that:

Get up to topN elements sorted by the float number, if the elements are less n, just get all elements sorted order.

Answer 2

Try the below

from collections import defaultdict
data = []
tag_dict = {'a':[('mary', 0.99, 'f'), ('tom', 0.87,'m') ],
            'b':[('jack', 0.43, 'm')],
            'c':[('martin', 0.987, 'm'), ('alice', 0.973, 'f')]}

for k,v in tag_dict.items():
    for entry in v:
        data.append((k,*entry))
data = sorted(data,key = lambda x: x[2],reverse=True)
result = defaultdict(list)
for i in range(0,3):
    result[data[i][0]].append(data[i][1:])
print(result)

output

defaultdict(<class 'list'>, {'a': [('mary', 0.99, 'f')], 'c': [('martin', 0.987, 'm'), ('alice', 0.973, 'f')]})