Data loaders need prioritization to resolve multiple matches

[sean-lockwood]

When trying to associate an identifier with my particular type of FITS file, the loader cannot decide between the generic FITS loader and my STIS-specific FITS loader.

IORegistryError: Format is ambiguous - options are: cubetest1, stis-format, tabular-fits

Location of error in AstroPy source code

My identifier:

# Define an optional identifier. If made specific enough, this circumvents the
# need to add `format="my-format"` in the `Spectrum1D.read` call.
def identify_stis_fits(origin, *args, **kwargs):
    from astropy.io import fits
    import os
    import six
    
    if isinstance(args[0], six.string_types) and (os.path.splitext(args[0].lower())[1] == '.fits'):
        with fits.open( args[0]) as f:
            return f[0].header.get('INSTRUME', '') == 'STIS'
    else:
        return False

Glue gets around this by allowing the specification of loader priorities.

[SaOgaz]

Do we want to use the loader priorities for this? Also, should we map this to an actual Astropy issues, since it seems like that's where this would need to be addressed? (@nmearl)

[aragilar]

I'm also interested in this, the current priority system sorts the loaders, but astropy always finds all possible loaders, and then assumes there will be only one possible option (unless a format option is provided). If astropy had a use_first option on read, then the priority system would work. I'll raise an issue on the astropy issue tracker, and link back to this.