[FEATURE] Using defaultdict in *group_by* implementation

[geekypandey]

The current implementation of group_by loops twice over the lst, which significantly reduces the speed when working with large data. Instead using defaultdict will make it possible to loop over the lst only once.

Current implementation:

def group_by(lst, fn):
    return {key: [el for el in lst if fn(el] == key] for key in map(fn, lst)}

Suggested implementation:

from collections import defaultdict

def group_by(lst, fn):
    d = defaultdict(list)
    for el in lst:
        d[fn(el)].append(el)
    return d

[zhangfelix]

Yeah, it looks more efficient that way. The python official website also gives a similar example.

And defaultdict can also be used in the implementation of count_by.

Current implementation:

def count_by(arr, fn=lambda x: x):
  key = {}
  for el in map(fn, arr):
    key[el] = 1 if el not in key else key[el] + 1
  return key

Suggested implementation:

from collections import defaultdict

def count_by(lst, fn):
  d = defaultdict(int)
  for el in lst:
    d[fn(el)] += 1
  return d

[Lattay]

Just want to point out that this change is more important that said.
As I explain in #213 the initial implementation does not loop twice over the elements, it loops n times over all elements, which would be dramatic for as few as 1000 elements.