Efficiently counting items in large python lists

user_8517641 0
Viewed 449

I have two very large python lists that look like this:

List A: [0,0,0,0,0,0,0,1,1,1,1,2,2,3,3,3,4.........]
List B: [0,0,0,0,0,0,2,2,2,2,3,3,4,4.........]

These lists go on to very large numbers, but I specify a maximum value, say 100 and after that I can discard the rest.

Now I need to calculate for each value (0,1,2..100) the ratio: occurrences in list A / occurrences in list B. And since this value is not always possible I decided to calculate this value only if there's more than 5 occurrences of the value in each list, and if this condition is not true, then combine the occurrences of the previous value(s) and will give the same ratios for combined values if this condition is correct. For example for the above lists, I want to create a Series that looks like this:

0 : 7/6=1.166 
1 : 9/6 = 1.5
2 : 9/6 = 1.5
3 : 9/6 = 1.5
.
.
.
100 : some_number
pandas list algorithm python
1 Answers

You can use a Counter to count the occurences and takewhile to fill your requirement of stopping at 100.

Instead of discarding values which are not in list b, notice how I used nan.

from collections import Counter
from itertools import takewhile

def get_ratios(a, b, max_=None, min_count=0):
    if max_ is not None:
        a = takewhile(lambda x: x <= max_, a)
        b = takewhile(lambda x: x <= max_, b)

    count_a, count_b = Counter(a), Counter(b)

    return {k: float('nan') if not count_b[k] else count_a[k] / count_b[k]
            for k in set(count_a) | set(count_b)
            if count_a[k] >= min_count <= count_b[k]}

Example

a = [1, 1, 1, 2, 3, 101]
b = [1, 1, 2, 2, 4, 101]

print(get_ratios(a, b, max_=100))

Output

{ 1: 1.5,
  2: 0.5,
  3: nan,
  4: 0.0 }

To ignore some under represented values, you can set min_count to 5 as mentionned in your question.

Notice I didn't fill in empty slots with the ratio of the previous value. Unless you have a very specific use case that requires it, I recommend you do not as this would mix actual data with extrapolated data. It is better to default on the previous value when it is not found, but to not pollute the actual data.

user_50793160
Related