1

I have a python list of variable length filled with 0s and 1s.

I want to create a new list where all 1s are expanded by a certain offset.

Examples:

offset = 1

l1 = [0,0,1,0]
l1_new = l[0,1,1,1]

l2 = [1,0,0,0,1,0,1,0,0]
l2_new = [1,1,0,1,1,1,1,1,0]

My solution code is not very fast and also does not use any numpy / vectorization / bitwise operations. But I guess some of those methods should be applicable here.

offset = 1
l_old = [0,0,1,0]
l_new = []
for i,l in enumerate(l_old):
    hit = False
    for o in range(offset+1)[1:]:
        if (i+o<len(l_old) and l_old[i+o]) or (i>0 and l_old[i-o]):
            hit = True
            break
    if hit or l_old[i]:
        l_new.append(1)
    else:
        l_new.append(0)

Hint: The solution should be fast and generic for any list of 0s and 1s and for any offset

gustavz
  • 2,964
  • 3
  • 25
  • 47

3 Answers3

2

Here is a linear (O(n+offset)) time solution:

import numpy as np

def symm_dil(a,hw):
    aux = np.zeros(a.size+2*hw+1,a.dtype)
    aux[:a.size] = a
    aux[2*hw+1:] -= a
    return np.minimum(aux.cumsum(),1)[hw:-hw-1]

#example
rng = np.random.default_rng()
a = rng.integers(0,2,10)
print(a)
print(symm_dil(a,2))

Sample output:

[0 0 0 1 0 0 0 0 0 1]
[0 1 1 1 1 1 0 1 1 1]
Paul Panzer
  • 51,835
  • 3
  • 54
  • 99
  • 1
    Just now gave a look at what this is actually doing. Very nice use of cumsum, and efficient – yatu Jun 19 '20 at 09:25
1

You can convolve with an array of ones, and clip the result:

def expand1s(a, offset):
    w = offset*2 +1
    return np.convolve(a, np.ones(w), mode='same').clip(0,1)

expand1s(l1, 1)
# array([0., 1., 1., 1.])

expand1s(l2, 1)
# array([1., 1., 0., 1., 1., 1., 1., 1., 0.])

Or we also have skimage.morphology.binary_dilation:

from skimage.morphology import binary_dilation

a = np.array(l2)
w = offset*2 +1
binary_dilation(a, np.ones(w)).view('i1')
# array([1, 1, 0, 1, 1, 1, 1, 1, 0], dtype=int8)
yatu
  • 86,083
  • 12
  • 84
  • 139
1

Here's a solution that uses a simple comprehension with a slice that's based on offset:

>>> def expand_ones(old: list, offset: int) -> list:
...     return [1 if any(
...         old[max(0, i-offset):min(len(old), i+offset+1)]
...     ) else 0 for i in range(len(old))]
...
>>> expand_ones([1, 0, 0, 0, 1, 0, 1, 0, 0], 1)
[1, 1, 0, 1, 1, 1, 1, 1, 0]
>>> expand_ones([1, 0, 0, 0, 1, 0, 1, 0, 0], 2)
[1, 1, 1, 1, 1, 1, 1, 1, 1]
Samwise
  • 68,105
  • 3
  • 30
  • 44