Uncertainty handling in 'find_lines_threshold' leads to unit conversion errors

[DavidT3]

Setup
MacOS v12.1
Python v3.10
SpecUtils v1.15.0

File required to reproduce
cutdown_6dF_g2344182-042249.fits.fits.zip

Problem
I'm taking a look at this 6dFGS spectrum, loading it in with the Spectrum1D read method:

bcg1_spec_6df = Spectrum1D.read('spec/cutdown_6dF_g2344182-042249.fits.fits')

Everything loads in fine, I can access wavelength, flux, meta['sky'], etc. properties - however when I come to use a line-finding method:

find_lines_threshold(bcg1_spec_fast, 3)

An exception occurs:

UnitConversionError Traceback (most recent call last)
Cell In[118], line 1
----> 1 find_lines_threshold(bcg1_spec_6df, 3)
File /opt/anaconda3/envs/astropyspec/lib/python3.10/site-packages/specutils/analysis/flux.py:276, in warn_continuum_below_threshold..actual_decorator..wrapper(*args, **kwargs)
274 if conf.do_continuum_function_check:
275 spectrum = args[0]
--> 276 if not is_continuum_below_threshold(spectrum, threshold):
277 if hasattr(threshold, 'unit'):
278 levelorsnr = 'value'
File /opt/anaconda3/envs/astropyspec/lib/python3.10/site-packages/specutils/analysis/flux.py:257, in is_continuum_below_threshold(spectrum, threshold)
253 # If threshold does not have a unit, ie it is not a quantity, then
254 # we are going to calculate based on the S/N if the uncertainty
255 # exists.
256 if uncertainty and uncertainty.uncertainty_type != 'std':
--> 257 return np.median(flux / uncertainty.quantity) < threshold
258 else:
259 return np.median(flux) / mad_std(flux) < threshold
File /opt/anaconda3/envs/astropyspec/lib/python3.10/site-packages/astropy/units/quantity.py:696, in Quantity.array_ufunc(self, function, method, *inputs, **kwargs)
694 return NotImplemented
695 else:
--> 696 raise e
File /opt/anaconda3/envs/astropyspec/lib/python3.10/site-packages/astropy/units/quantity.py:641, in Quantity.array_ufunc(self, function, method, *inputs, **kwargs)
636 # Determine required conversion functions -- to bring the unit of the
637 # input to that expected (e.g., radian for np.sin), or to get
638 # consistent units between two inputs (e.g., in np.add) --
639 # and the unit of the result (or tuple of units for nout > 1).
640 try:
--> 641 converters, unit = converters_and_unit(function, method, *inputs)
643 out = kwargs.get("out", None)
644 # Avoid loop back by turning any Quantity output into array views.
File /opt/anaconda3/envs/astropyspec/lib/python3.10/site-packages/astropy/units/quantity_helper/converters.py:199, in converters_and_unit(function, method, *args)
197 converters[i] = None
198 else:
--> 199 raise UnitConversionError(
200 f"Can only apply '{function.name}' function to "
201 "dimensionless quantities when other argument is not "
202 "a quantity (unless the latter is all zero/infinity/nan)."
203 )
204 except TypeError:
205 # _can_have_arbitrary_unit failed: arg could not be compared
206 # with zero or checked to be finite. Then, ufunc will fail too.
207 raise TypeError(
208 "Unsupported operand type(s) for ufunc {}: '{}'".format(
209 function.name,
210 ",".join([arg.class.name for arg in args]),
211 )
212 )
UnitConversionError: Can only apply 'less' function to dimensionless quantities when other argument is not a quantity (unless the latter is all zero/infinity/nan).

Which seems to be down to the flux-to-uncertainty ratio still having units when compared to the integer/float threshold - I think this is because the 6dF read-in decorator declares the uncertainty as a 'VarianceUncertainty' instance (which as far as I know is the correct type of uncertainty) - but then the uncertainty.quantity object has units of count/angstrom squared, so when the flux quantity is divided by it there remains a unit.

I'm also working with some FAST spectroscopy data, for which I've written my own read-in function, and the same problem occurs because of the VarianceUncertainty, but if I change the declaration of uncertainty to 'StdDevUncertainty' before setting up the Spectrum1D instance (just as a test), then the line-finding functionality works fine.