test_ndcube_components_after_slicing was skipped due to a bug

[nabobalis]

So the test that is failing is:

sunraster/tests/test_spectrogram.py::test_ndcube_components_after_slicing
at

sunraster/spectrogram.py:258: in __str__
    if self.time.isscalar:

with TypeError: Unsupported operand type(s) for ufunc add: 'Time,Quantity',

Situation before it fails

function = <ufunc 'add'>, method = '__call__', args = (<Time object: scale='utc' format='iso' value=2017-01-01 00:00:00.000>, <Quantity [nan] s>), ufunc_helper = <function get_converters_and_unit at 0x7f6ee0f7e040>, units = [None, Unit("s")], converters = [False, None]

The test is using:

(Pdb)  spectrogram_instrument_axes
<sunraster.spectrogram.SpectrogramCube object at 0x7ff87753a970>
SpectrogramCube
---------------
Time Period: ['2017-01-01 00:00:00.000' '2017-01-01 00:00:01.000']
Instrument axes: ['a' 'b' 'c']
Pixel dimensions: [2 2 3] pix
Longitude range: [2159.96856373 3600.        ] arcsec
Latitude range: [4.56894119e-02 1.80000000e+03] arcsec
Spectral range: [1.02e-09 1.06e-09] m
Data unit: ct

The constructed test Cube also has the same problem:

(Pdb) spectrogram_instrument_axes.data[item]
array([[[0.563]]])
(Pdb) spectrogram_instrument_axes.uncertainty[item]
UnknownUncertainty([[[0.75033326]]])
(Pdb) spectrogram_instrument_axes.mask[item]
array([[[False]]])
(Pdb) ec0
['time', 0, <Time object: scale='utc' format='iso' value=['2017-01-01 00:00:00.000']>]
(Pdb) spectrogram_instrument_axes.wcs[item]
WCS Keywords

Number of WCS axes: 3
CTYPE : 'WAVE'  'HPLT-TAN'  'HPLN-TAN'  
CRVAL : 1e-09  0.5  1.0  
CRPIX : 0.0  2.0  2.0  
PC1_1 PC1_2 PC1_3  : 1.0  0.0  0.0  
PC2_1 PC2_2 PC2_3  : 0.0  1.0  0.0  
PC3_1 PC3_2 PC3_3  : 0.0  0.0  1.0  
CDELT : 2.0000000000000002e-11  0.5  0.4  
NAXIS : 1  1  1

It goes to:

Time(self._get_axis_coord(self._time_name, self._time_loc))

and errors on self._get_axis_coord, which is from self.axis_world_coords(wcs=self.extra_coords[axis_name])[0]

Here the wcs is ExtraCoords(time ((0,)) None: TimeTableCoordinate ['time'] [None]: ['2017-01-01 00:00:00.000'])

Inside that, it errors in axis_world_coords from ndcube, at

        axes_coords = values_to_high_level_objects(
            *axes_coords, low_level_wcs=wcs)

with the inputs to that being:

(Pdb) axes_coords
[<Quantity [nan] s>]
(Pdb) wcs
<WCS(output_frame=TemporalFrame, input_frame=PixelFrame, forward_transform=Model: Tabular1D
N_inputs: 1
N_outputs: 1
Parameters: 
  points: (<Quantity [0.] pix>,)
  lookup_table: [0.] s
  method: linear
  fill_value: nan
  bounds_error: False)>

and axes_coords comes from self._generate_world_coords(pixel_corners, wcs) with the input being:

(Pdb) pixel_corners, wcs
(False, <ndcube.extra_coords.extra_coords.ExtraCoords object at 0x7fae6699d760>
ExtraCoords(time ((0,)) None: TimeTableCoordinate ['time'] [None]: ['2017-01-01 00:00:00.000']))

#184 (comment)

The problem is that the sliced extra_coords wcs return nan for the world value of the 0th pixel:

>>> sliced_cube.extra_coords["time"].wcs.pixel_to_world_values(0)
nan

where

>>> sliced_cube.extra_coords["time"].wcs
<WCS(output_frame=TemporalFrame, input_frame=PixelFrame, forward_transform=Model: Tabular1D
N_inputs: 1
N_outputs: 1
Parameters: 
  points: (<Quantity [0.] pix>,)
  lookup_table: [0.] s
  method: linear
  fill_value: nan
  bounds_error: False)>

The question is why is this happening?

[Cadair]

kinda related to spacetelescope/gwcs#333

[Cadair]

This appears to be an astropy bug with length-1 tables:

In [5]: t = Tabular1D([0]*u.pix, lookup_table=[0]*u.s)

In [6]: t
Out[6]: <Tabular1D(points=(<Quantity [0.] pix>,), lookup_table=[0.] s)>

In [7]: t(0)
/home/stuart/.virtualenvs/sunpy-dev/lib/python3.9/site-packages/scipy/interpolate/interpolate.py:2532: RuntimeWarning: invalid value encountered in true_divide
  norm_distances.append((x - grid[i]) /
Out[7]: nan

In [8]: t = Tabular1D([0,1]*u.pix, lookup_table=[0,1]*u.s)

In [9]: t(0)
Out[9]: 0.0

[nabobalis]

I recall we patched this in ndcube.