WCS information lost in unit conversion of LowerDimensionalObject

[keflavich]

A similar custom to attribute should be defined (as is already done for SpectralCube).

[e-koch]

It seems this is true for slicing as well.

[keflavich]

Yeah, I've run into this. It shouldn't be too hard. PR may be forthcoming.

[keflavich]

@e-koch care to review this?

[e-koch]

I'm not sure what should be returned here. If a Quantity is returned, then you still lose all of the WCS info. If you have a PV slice, you would want a OneDSpectrum when slicing along the spectral axis, but something else when slicing in the spatial dimension.

Could we check for the axis type when a slice reduces ndim<2? A OneDSpectrum would be returned when the remaining axis is spectral. I'm not sure how to treat 1D spatial slices. Treat it like a mock-projection with a 1 pixel second dimension or define a spatial 1D slice class? Is there a use case for such a thing?

[keflavich]

There might be. I chose to ignore it for now because it's not trivial, but I think we could inspect the WCS and see if one of the axes is spectral; if so, we'd then return the OneDSpectrum

[keflavich]

But I think for single pixels, we really want to drop all the way to a Quantity with no wcs information

[e-koch]

@keflavich - Apart from the one comment, it looks good! I tested on the use case the brought this up and everything checks out.

Depending on what you think about the comment above, test_slices_of_projections_not_projections may need to be altered/become more specific on the returned object.

[keflavich]

There is a bug in WCS that prevents creating 1D slices from 2D WCSs that are half celestial, half spectral:

In [15]: w2 = cube[:,0,:].wcs
WARNING: FITSFixedWarning: 'spcfix' made the change 'Changed CTYPE3 from 'VELO-LSR' to 'VOPT', and SPECSYS to 'LSRK''. [astropy.wcs.wcs]

In [16]: w2.sub([wcs.WCSSUB_SPECTRAL])
Traceback (most recent call last):
  File "<ipython-input-16-0b9117b7127a>", line 1, in <module>
    w2.sub([wcs.WCSSUB_SPECTRAL])
  File "/Users/adam/repos/astropy/astropy/wcs/wcs.py", line 541, in sub
    copy = self.deepcopy()
  File "/Users/adam/repos/astropy/astropy/wcs/wcs.py", line 538, in deepcopy
    return copy.deepcopy(self)
  File "/Users/adam/anaconda/envs/astropy35/lib/python3.5/copy.py", line 166, in deepcopy
    y = copier(memo)
  File "/Users/adam/repos/astropy/astropy/wcs/wcs.py", line 514, in __deepcopy__
    copy.deepcopy(self.wcs, memo),
  File "/Users/adam/anaconda/envs/astropy35/lib/python3.5/copy.py", line 166, in deepcopy
    y = copier(memo)
InconsistentAxisTypesError: ERROR 4 in wcs_types() at line 2516 of file cextern/wcslib/C/wcs.c:
Unmatched celestial axes.

I'll make a separate issue for this.