Run Length Encoding, Two Ways

rle.py

import numpy as np
import itertools

def rle(x):
    where = np.flatnonzero
    x = np.asarray(x)
    n = len(x)
    if n == 0:
        return np.array([], dtype=int)
    starts = np.r_[0, where(~np.isclose(x[1:], x[:-1], equal_nan=True)) + 1]                                                                                                                                                                                                    
    lengths = np.diff(np.r_[starts, n])
    values = x[starts]
    return np.dstack((lengths, values))[0]


def rld(runs):
    runs_t = np.transpose(runs)
    lengths = runs_t[0].astype(int)
    values = runs_t[1].astype(int)
    starts = np.concatenate(([0],np.cumsum(lengths)[:-1]))
    starts, lengths, values = map(np.asarray, (starts, lengths, values))
    ends = starts + lengths
    n = ends[-1]
    x = np.full(n, 0)
    for lo, hi, val in zip(starts, ends, values):
        x[lo:hi] = val
    return x

ilen = lambda it: sum(1 for _ in it) 
rle = lambda xs: ((ilen(gp), x) for x, gp in itertools.groupby(xs))
rld = lambda xs: itertools.chain.from_iterable(itertools.repeat(x, n) for n, x in xs)

the fancy numpy rt is slightly faster...