match_to_catalog_sky should raise a nicer error when the match or catalog coordinates have nan values

I believe this is related to this SO question filed a few months ago.

Basically, match_to_catalog_sky() is throwing an IndexError and I can’t figure out why. I’m using Python 3.7.3 and astropy 3.1.2.

Here’s the code to reproduce it (data files):

from astropy.io import ascii
from astropy.coordinates import SkyCoord
from astropy import units as u

d1 = ascii.read('c1.dat')
ra_obs, dec_obs = d1['ra'], d1['dec']
d2 = ascii.read('c2.dat')
ra_qry, dec_qry = d2['ra'], d2['dec']

c1 = SkyCoord(ra_obs, dec_obs, unit=(u.degree, u.degree))
c2 = SkyCoord(ra_qry, dec_qry, unit=(u.degree, u.degree))

idx, d2d, d3d = c1.match_to_catalog_sky(c2)

The full traceback is:

$ python test.py
/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/angles.py:642: RuntimeWarning: invalid value encountered in less
  if np.any(self_angle < wrap_angle_floor) or np.any(self_angle >= wrap_angle):
/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/angles.py:642: RuntimeWarning: invalid value encountered in greater_equal
  if np.any(self_angle < wrap_angle_floor) or np.any(self_angle >= wrap_angle):
/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/angles.py:529: RuntimeWarning: invalid value encountered in less
  if np.any(angles.value < lower) or np.any(angles.value > upper):
/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/angles.py:529: RuntimeWarning: invalid value encountered in greater
  if np.any(angles.value < lower) or np.any(angles.value > upper):
Traceback (most recent call last):
  File "test.py", line 15, in <module>
    idx, d2d, d3d = c1.match_to_catalog_sky(c2)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/sky_coordinate.py", line 1045, in match_to_catalog_sky
    storekdtree='_kdtree_sky')
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/matching.py", line 154, in match_coordinates_sky
    idx, sep2d, sep3d = match_coordinates_3d(newmatch_u, newcat_u, nthneighbor, storekdtree)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/matching.py", line 83, in match_coordinates_3d
    sep2d = catalogcoord[idx].separation(matchcoord)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/utils/misc.py", line 940, in __getitem__
    return self._apply('__getitem__', item)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/baseframe.py", line 1476, in _apply
    new._data = apply_method(self.data)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/baseframe.py", line 1447, in apply_method
    return value._apply(method, *args, **kwargs)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/representation.py", line 780, in _apply
    rep = super()._apply(method, *args, **kwargs)
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/coordinates/representation.py", line 263, in _apply
    apply_method(getattr(self, component)))
  File "/home/gabriel/miniconda3/envs/py3/lib/python3.7/site-packages/astropy/units/quantity.py", line 1035, in __getitem__
    out = super().__getitem__(key)
IndexError: index 22670 is out of bounds for axis 0 with size 22670

[smoh]

I think it’s because you have nan entries in c2. Not sure if this was also the cause for the SO question but it seems to work fine once you clean nans in c2.

clean = (~np.isnan(c2.ra) & ~np.isnan(c2.dec))
print(len(c2), clean.sum(), len(c2)-clean.sum())
c2_clean = c2[clean]

22670 12660 10010

idx, d2d, d3d = c1.match_to_catalog_sky(c2_clean)

The nan entries seems to completely mess up KDTree built from them making it return dist=inf and idx=(lenth of catalogcoord) for any query. Not sure if this was handled in previous versions of astropy.

[smoh]

(Oops, just saw @adrn mentioned _3d methods already)