get_interp_val is not a drop-in replacement for healpy.get_interp_val

[jvavrek]

The docs state

def get_interp_val(m, theta, phi, nest=False, lonlat=False):
    """
    Drop-in replacement for healpy `~healpy.pixelfunc.get_interp_val`.
    """

Here is a MWE showing differences between the two when using phi from -pi to pi:

import numpy as np
import healpy as hpy
from astropy_healpix import healpy as hpx
import matplotlib.pyplot as plt


step_deg = 1

grid_azimuth, grid_polar = np.meshgrid(
    np.deg2rad(np.arange(-180., 180 + 0.1, step_deg)),
    np.deg2rad(np.arange(0.0, 180 + 0.1, step_deg)),
)
polar, azimuth = grid_polar.ravel(), grid_azimuth.ravel()

npix = 192
m = np.arange(npix)

grid_hpx = hpx.get_interp_val(m, polar, azimuth).reshape(*grid_azimuth.shape)
grid_hpy = hpy.get_interp_val(m, polar, azimuth).reshape(*grid_azimuth.shape)

Results:

np.any(np.isnan(grid_hpx)), np.any(np.isnan(grid_hpy))  # (True, False)

[lpsinger]

Looks like it works fine if you change the range of the polar angle from (-180, 180) to (0, 360). Taking a closer look.

[jvavrek]

Looks like it works fine if you change the range of the polar angle from (-180, 180) to (0, 360). Taking a closer look.

Azimuthal, right? But yes, agreed, that works for me as well.