run_length_encoder.py

import timeit
import numpy


def _encode(msg):
    # simple but slow
    ones = numpy.array(msg) == 1
    encoded = []
    offset = 0
    while True:
        first = numpy.argmax(ones)
        if not ones[first]:
            break
        encoded.append(offset + first)
        ones = ones[first:]
        until = numpy.argmin(ones)
        encoded.append(until)
        ones = ones[until:]
        offset += first + until
    return encoded


def encode(msg):
    length = len(msg)
    front = numpy.zeros(length + 1)
    back = numpy.zeros(length + 1)
    front[:-1] = msg
    back[1:] = msg
    ones = numpy.arange(len(front))[(front - back) == 1]
    zeros = numpy.arange(len(front))[(back - front) == 1]
    diff = zeros - ones
    return numpy.stack((ones, diff)).T.flatten().tolist()


def test():
    pairs = [(
        [1, 1, 1, 1, 1],
        [0, 5],
    ), (
        [0, 0, 0, 0, 0],
        []
    ), (
        [0, 0, 1, 1, 1, 0, 0, 1, 1],
        [2, 3, 7, 2]
    ), (
        [0, 0, 1, 1, 1, 0, 0, 1, 1, 0],
        [2, 3, 7, 2]
    )]
    for msg, expected in pairs:
        assert encode(msg) == expected


if __name__ == '__main__':
    size = 100000
    msg = numpy.random.random_integers(0, 1, size)
    print(timeit.timeit(lambda: encode(msg), number=1000))
    print(timeit.timeit(lambda: _encode(msg), number=1000))