filter max from iterable for grouped element

[Peter Otten]

Christian wrote:

> as beginner in python , I struggle  somewhat to filter out only the
> maximum in the values for  and get hmax.
 
> h = {'abvjv': ('asyak', 0.9014230420411024),
>  'afqes': ('jarbm', 0.9327883839839753),
>  'aikdj': ('jarbm', 0.9503941616408824),
>  'ajbhn': ('jarbm', 0.9323583083061541),
>  'ajrje': ('jbhdj', 0.9825125732711598),
>  'anbrw': ('jarbm', 0.950801828672098)}
> 
> 
> hmax = {'abvjv': ('asyak', 0.9014230420411024),
>  'ajrje': ('jbhdj', 0.9825125732711598),
>  'anbrw': ('jarbm', 0.950801828672098)}

You can create an intermediate dict:

>>> d = {}
>>> for k, (k2, v) in h.items():
...     d.setdefault(k2, []).append((v, k))
... 
>>> import pprint
>>> pprint.pprint(d)
{'asyak': [(0.9014230420411024, 'abvjv')],
 'jarbm': [(0.9323583083061541, 'ajbhn'),
           (0.950801828672098, 'anbrw'),
           (0.9327883839839753, 'afqes'),
           (0.9503941616408824, 'aikdj')],
 'jbhdj': [(0.9825125732711598, 'ajrje')]}

Now find the maximum values:

>>> for k, pairs in d.items():
...     v, k2 = max(pairs)
...     assert k2 not in hmax
...     hmax[k2] = k, v
... 
>>> pprint.pprint(hmax)
{'abvjv': ('asyak', 0.9014230420411024),
 'ajrje': ('jbhdj', 0.9825125732711598),
 'anbrw': ('jarbm', 0.950801828672098)}

> Maybe it easier when i change the structure of h?

Maybe. Here's one option:

>>> pprint.pprint(data)
[('jarbm', 0.9323583083061541, 'ajbhn'),
 ('jarbm', 0.950801828672098, 'anbrw'),
 ('jarbm', 0.9327883839839753, 'afqes'),
 ('asyak', 0.9014230420411024, 'abvjv'),
 ('jbhdj', 0.9825125732711598, 'ajrje'),
 ('jarbm', 0.9503941616408824, 'aikdj')]
>>> data.sort()
>>> from itertools import groupby
>>> def last(items):
...     for item in items: pass
...     return item
... 
>>> dmax = [last(group) for key, group in groupby(data, key=lambda item: 
item[0])]
>>> pprint.pprint(dmax)
[('asyak', 0.9014230420411024, 'abvjv'),
 ('jarbm', 0.950801828672098, 'anbrw'),
 ('jbhdj', 0.9825125732711598, 'ajrje')]