Erroneous detection of burst on- and offsets in burst.utils.compute_burst_stats()

[JanCBrammer]

I believe that the detection of burst on- and offsets fails if the boolean burst vector starts and/or ends with bursts.

Here's an illustration:

import numpy as np


def find_starts_ends_A(bursting):    # current implementation

    starts = np.array([])
    ends = np.array([])

    for ii, index in enumerate(np.where(np.diff(bursting) != 0)[0]):

        if (ii % 2) == 0:
            starts = np.append(starts, index)
        else:
            ends = np.append(ends, index)

    return starts, ends


def find_starts_ends_B(bursting):    # suggested alternative implementation

    # Find the indices of changes.
    change = np.diff(bursting)
    idcs, = change.nonzero()

    idcs += 1    # Get indices following the change.

    if bursting[0]:
        # If the first sample fulfills the condition, prepend a zero.
        idcs = np.r_[0, idcs]

    if bursting[-1]:
        # If the last sample fulfills the condition, append an index
        # corresponding to the length of signal
        idcs = np.r_[idcs, bursting.size]

    starts = idcs[0::2]
    ends = idcs[1::2]

    return starts, ends


bursts = np.array([True, True, False, False, True, True, True, False, True])

startsA, endsA = find_starts_ends_A(bursts)
startsB, endsB = find_starts_ends_B(bursts)

print(f"starts A: {startsA}, ends A: {endsA}")
print(f"starts B: {startsB}, ends B: {endsB}")

starts A: [1. 6.], ends A: [3. 7.]
starts B: [0 4 8], ends B: [2 7 9]

Happy to make a PR. Thanks for the amazing package btw.

[ryanhammonds]

Hi @JanCBrammer, thanks for reporting the bug! I tested your solution and it looks great. Feel free to open a PR if you have time. Otherwise, I can implement the change soon.

[JanCBrammer]

Feel free to open a PR if you have time

Great, I should have some time next week.