Question

I need to sort a special numpy array, in which blocks of size 19 constitute an element, using a user-defined function to determine the value of such a block.

The first attempt has been to wrap the array in a class and overload the [] operator:

class W:
    def __init__(self, filename="nn.txt"):
        self.nn = array([int(i) for i in open(filename, "r").readlines()[1:]])
        self.size = self.nn.size / 19

    def __getitem__(self, idx):
        return self.nn[idx:idx+19] 

    def __len__(self):
            return self.size

Using this structure I supply a comparison operator, which is passed to sorted():

def avg_cmp(x, y):
   return int(average(x)) - int(average(y))

u = W("nnsmall.txt")
sorted(u, cmp=avg_cmp)

However, this approach is too slow.

Any tips?

Answer 1

have you tried sorted(u, key=average)? this would only calculate the average of each column once.

if the size of the array is always divisible by 19 without remainder:

>>> import numpy as np
>>> n = 2
>>> u = np.array([v for v in range(19*n)])

>>> u = u.reshape(n,19)

>>> sorted(c, key=np.average)

[array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18]),
 array([19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37])]