combining list elements in a list

Question

I am a new python user, and I need help about combining list elements under a condition. I have a list like this:

x = [['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]

I would like to combine list elements which start with the same letter in a list by summing up the other elements. for example, I'd like to obtain this list for x :

x = [['a', 30, 120], ['b', 10, 20]]

How can I achieve this ?

Answer 1

A one-liner using itertools.groupby() :

In [45]: lis=[['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]

In [46]: lis.sort(key=itemgetter(0)) #sort the list first

In [47]: lis
Out[47]: [['a', 10, 20], ['a', 20, 100], ['b', 10, 20]]

In [49]: [[k]+map(sum,zip(*[x[1:] for x in g])) for k,g in groupby(lis,key=itemgetter(0))]
Out[49]: [['a', 30, 120], ['b', 10, 20]]

A simple solution:

In [23]: lis=[['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]

In [24]: ans=[]

In [25]: lis.sort(key=itemgetter(0))   #sort the list according to the first elem

In [26]: lis
Out[26]: [['a', 10, 20], ['a', 20, 100], ['b', 10, 20]]

In [27]: for x in lis:
    if ans:
        if x[0]==ans[-1][0]:  #if the value of the first elem of last element in ans is same as x[0]
            ans[-1][1]+=x[1]
            ans[-1][2]+=x[2]
        else:         
            ans.append(x)
    else:ans.append(x)
   ....:     

In [28]: ans
Out[28]: [['a', 30, 120], ['b', 10, 20]]

Without sorting the list using defaultdict() :

In [69]: dic=defaultdict(list)

In [70]: for x in lis:
    dic[x[0]].append(x[1:])
   ....:     

In [71]: dic
Out[71]: defaultdict(<type 'list'>, {'a': [[10, 20], [20, 100]], 'b': [[10, 20]]})

In [72]: [[k]+map(sum,zip(*i)) for k,i in dic.items()]
Out[72]: [['a', 30, 120], ['b', 10, 20]]

Answer 2

Another approach using dict and map :

>>> x = [['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]
>>> d = {}
>>> from operator import add
>>> for k, v1, v2 in x:
    d[k] = map(add, d.get(k, [0, 0]), [v1, v2])

>>> d
{'a': [30, 120], 'b': [10, 20]}

Answer 3

I'm going to use the answer code for a huge data which include over millons elements. I'd like the reduce the list elements this way.

In such a case you probably don't want to be sorting the data or building a fully copy as you're iterating over it.

The following solution does neither. It can also handle sublists of any length (as long as all lengths are the same):

def add(d, l):
   k = l[0]            # extract the key
   p = d.get(k, None)  # see if we already have a partial sum for this key
   if p:
      d[k] = [x+y for x,y in zip(p, l[1:])] # add to the previous sum
   else:
      d[k] = l[1:]     # create a new sum
   return d

x = [['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]
result = [[k] + v for k,v in reduce(add, x, {}).items()]
print(result)

Alternatively,

import collections, operator

x = [['a', 10, 20], ['b', 10, 20], ['a', 20, 100]]

d = collections.defaultdict(lambda:[0] * (len(x[0]) - 1))
for el in x:
  d[el[0]] = map(operator.add, d[el[0]], el[1:])
result = [[k] + v for k,v in d.items()]
print(result)

This works exactly the same as the first version, but uses defaultdict and explicit iteration.