Python dictionary from list of lists

Question

I am trying to generate a dictionary to count items in a list of lists with conditions. I am trying to get the results as values in a tuple, for example if my initial data is a list of lists like this:

my_list = [[(1,2),5],[(1,2),9],[(2,3),0],[(2,3),2],[(1,2),5],[(1,2),9],[(2,3),5]]

The desired outcome would be a dictionary with key the first element of each lists. The values of the dictionary would be a tuple of two elements, the number of items having that same key (total count) and how many values are 5 for the second item (conditional count).

dict_desired = {(1,2):(4,2),(2,3):(3,1)}

I did it but in a way that seems not very neat, and i was wondering if someone with more python experience could help me out to build it maybe using comprehensions.

Thanks in advance!

Answer 1

The pythonic, efficient way to do this is to use a loop:

my_list = [[(1,2),5],[(1,2),9],[(2,3),0],[(2,3),2],[(1,2),5],[(1,2),9],[(2,3),5]]
result = {}
empty = 0, 0

for key, value in my_list:
    count, fives = result.get(key, empty)
    result[key] = (count + 1, fives + (value == 5))

To do this using comprehensions, you require going over the list on each iteration, something like:

result = {
    key:(
        sum(k == key for k,_ in my_list),
        sum(v == 5 for k,v in my_list if k == key)
    ) for key, value in my_list
}

Which is very unpythonic and inefficient. You can maybe try to make a set out of the keys first, but it still won't be as efficient as the regular for-loop, and perhaps there is some hacky way with assignment expressions to make it work efficiently, but it certainly won't be neater .

So, you have a nice, readable, linear-time for-loop versus a convoluted, unreadable, quadratic time dictionary comprehension. The choice is clear, IMO.