Open dhomeier opened 6 months ago
Attention: Patch coverage is 12.50000%
with 21 lines
in your changes are missing coverage. Please review.
Project coverage is 72.04%. Comparing base (
98dfcfd
) to head (ac7b53b
).
Files | Patch % | Lines |
---|---|---|
specutils/io/default_loaders/wcs_fits.py | 12.50% | 21 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Opening this both as a WIP and issue for tracking incompatibilities of the
wcs1d-fits
loader with the IRAF 'equispec' and 'multispec' (Echelle?) formats. These are largely from reports on the list or even off-list or in the Facebook group etc., so I have yet to find out if I can link to publicly available example files.The "fix" implemented here so far basically just works around a WCS error, loading a MULTISPEC file without a valid spectral WCS; this seems to have been ignored for some time, but with recent
specutils
versions results in aThe underlying problem is that the files are using a WCS defined as
which to the best of my knowledge is simply not supported by WCSLIB (I think spatial WCS of similar construction are, but I don't see a way to define a valid header for a spectral WCS). The kludge here is catching that error and working around the WCS failure by just constructing a
spectral_axis
from the pixel scale, but of course without having solved the WCS provides no wavelength calibration – it merely allows to read in the spectrum at all, similar to olderspecutils
versions.As it does not seem that Astropy WCS will be able to do this, a real fix will probably require solving for the spectral axis in
specutils
, as is e.g. implemented in nonlinearwave.