BUG: Mask cube is ignored in Cubeviz spectral extraction

[pllim]

The current documentation (at the time of original posting this issue) at https://jdaviz.readthedocs.io/en/latest/cubeviz/plugins.html#spectral-extraction does not mention at all how uncertainty and mask arrays are used in the extraction process. The uncertainty is yet another issue (#2713); here I am only focusing on mask:

jdaviz/jdaviz/configs/cubeviz/plugins/spectral_extraction/spectral_extraction.py

Lines 214 to 216 in b563ed3

	nddata = spectral_cube.get_subset_object(
	subset_id=self.aperture.selected, cls=NDDataArray
	)

jdaviz/jdaviz/configs/cubeviz/plugins/spectral_extraction/spectral_extraction.py

Line 231 in b563ed3

mask = nddata.mask

This was originally implemented by @bmorris3 via #2039 . In the code above, mask only takes account of the selected aperture (if any). However, our parser reads in a whole cube array that has the bitmask plane:

jdaviz/jdaviz/configs/cubeviz/plugins/parsers.py

Line 21 in b563ed3

mask=['mask', 'dq', 'quality'])

While there is code to stuff the mask into Spectrum1D object, not all parser logic routes seem to trigger it (how Spectrum1D handle a bitplane is another story):

jdaviz/jdaviz/configs/cubeviz/plugins/parsers.py

Line 184 in b563ed3

mask=sc.mask

As a result, aperture seems to solely define the mask, while the actual mask cube is ignored. How is a user supposed to mask out certain bad pixels from extraction (in addition to the aperture)?

ADDITIONAL REQUIREMENT: When MASK cube does not exist, the extraction should not crash and treat all pixels as "good".

🐱

[bmorris3]

The subset choice of the selected aperture was implemented in #2664. I think the simplest solution for now is to translate to NDDataArray with

nddata = spectral_cube.get_object(cls=NDDataArray)

without specifying the subset, to get the original mask cube for the full dataset. Then we can call to_mask on the aperture's glue subset, and take the logical-and with the full mask, something like:

nddata.mask &= ~aperture_subset.to_mask()  # the ~ may be needed b/c glue subset masks are "backwards"

How does that sound?

[pllim]

Maybe related?

#2679 (comment)