2048 Pythonic Merge Function

This post goes over how I updated my 2048 merge function code from spaghetti code to being somewhat more readable.

Spaghetti Old Code

I incorporated a few techniques from pretty much all of your suggestions! I realized there were a lot of logical mistakes and redundancy in my code, like why append 0's over and over again when it's not necessary at all, because in the end I back-track, account for missing numbers, and then append 0(s) in their place.

Improvements in new code vs. old code:

1. Use of list comprehensions
2. Use of Python's built-in truthiness
3. Minimal branches (fewer if(s), else(s), etc.)
4. Infusing new and useful methods like extend into my code
5. Improving the code style by enhancing readability
6. Removing redundandcy and flawed logic

def merge(nums):
    '''
    Takes a list as input
    returns merged pairs with
    non zero values shifted to the left.
    fancy interactive doc test below, no output means no problems.
    >>> merge([2, 0, 2, 4])
    [4, 4, 0, 0]
    >>> merge([0, 0, 2, 2])
    [4, 0, 0, 0]
    >>> merge([2, 2, 0, 0])
    [4, 0, 0, 0]
    >>> merge([2, 2, 2, 2, 2])
    [4, 4, 2, 0, 0]
    >>> merge([8, 16, 16, 8])
    [8, 32, 8, 0]
    '''
    slide = [num for num in nums if num]
    pairs = []
    for idx, num in enumerate(slide):
        if idx == len(slide)-1:
            pairs.append(num)
            break
        elif num == slide[idx+1]:
            pairs.append(num*2)
            slide[idx+1] = None
        else:
            pairs.append(num)  # Even if not pair you must append
    slide = [pair for pair in pairs if pair] 
    slide.extend([0] * (len(nums) - len(slide)))
    return slide

if __name__ == '__main__':
    import doctest
    doctest.testmod()

Sidenote: Pylint yells at me for using i instead of idx.