array size mismatch when using seismology.estimate_numax()

Problem description

For some combinations of values for the keyword arguments window_width and spacing supplied to the acf2d function, estimate_numax() fails with an array size mismatch error. This seems to be because the preallocated array size for the output of utils.autocorrelate does not actually match the size of the array returned by it.

In particular, this seems to happen for red giants observed at short cadence. While I can get the acf2d method to work by masking out frequencies above an (artificial) maximum frequency, this seems like a more subtle issue: just because the code doesn’t break doesn’t mean it’s necessarily not otherwise faulty.

Example

import numpy as np
from astropy import units as u
import lightkurve as lk

def get_PG():
        
    data = np.loadtxt("PSD_all.pow")
    ν = data[..., 0] * u.uHz
    P = u.Quantity(data[..., -1])
    
#     m = ν.value < 500
    m = slice(None, None)
    
    return lk.Periodogram(ν[m], P[m])

pg = get_PG()
seis = pg.flatten().to_seismology()
seis.estimate_numax(window_width=25, spacing=1)

For the provided power spectrum (see here: PSD_all.zip), this fails with spacing < 1.49945671858…, which I don’t have a good explanation for:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-78-76ccb8c22ce8> in <module>
----> 1 seis.estimate_numax(window_width=25, spacing=1.499456718580)

~/.local/lib/python3.8/site-packages/lightkurve/seismology/core.py in estimate_numax(self, method, **kwargs)
    544         if method == "acf2d":
    545             from .numax_estimators import estimate_numax_acf2d
--> 546             result = estimate_numax_acf2d(self.periodogram, **kwargs)
    547         self.numax = result
    548         return result

~/.local/lib/python3.8/site-packages/lightkurve/seismology/numax_estimators.py in estimate_numax_acf2d(periodogram, numaxs, window_width, spacing)
    150     for idx, numax in enumerate(numaxs):
    151         acf = utils.autocorrelate(periodogram, numax, window_width=window_width, frequency_spacing=fs)      #Return the acf at this numax
--> 152         acf2d[:, idx] = acf                                     #Store the 2D acf
    153         metric[idx] = (np.sum(np.abs(acf)) - 1 ) / len(acf)  #Store the max acf power normalised by the length
    154 

ValueError: could not broadcast input array from shape (554) into shape (584)

(the line numbers don’t match up because I added a bunch of print statements to see what’s going on)

Expected behavior

It should not return a ValueError.

Environment

• Linux (Arch)
• lightkurve version 1.11.1
• installed with pip

[darthoctopus]

This seems a satisfactory resolution to me. Thanks for helping out!

[barentsen]

A solution was merged in #780.

Thank you @darthoctopus for opening the issue, and thank you @ojhall94 for fixing!! 👍 👍