Spectrum1D WCS has wrong number of dimensions as an NDCube

[gterreran]

As per title, it looks like that the dimensions of the WCS of Spectrum1D objects are not parsed correctly into the NDCube data class, as shown by the following snippet:

import numpy as np
import astropy.units as u
from specutils import Spectrum1D
from astropy.modeling import models

wl = np.linspace(4000,12000,128)*u.angstrom
flux_1 = models.BlackBody(temperature=5800*u.K)(wl)
flux_2 = models.BlackBody(temperature=4800*u.K)(wl)
flux_3 = models.BlackBody(temperature=6800*u.K)(wl)
flux = np.array([flux_1, flux_2, flux_3])*flux_1.unit

spec = Spectrum1D(spectral_axis=wl, flux=flux)
print(spec.wcs.pixel_n_dim)

which returns

1

while it should be 3.

[simontorres]

Actually this behavior is correct, the spectral_axis which is wl is of dimension 1, in order to create the wcs instance the Spectrum1D class uses the function gwcs_from_array from utils/wcs_utils.py which is hard coded to create a 1D coordinate frame.

def gwcs_from_array(array):
    """
    Create a new WCS from provided tabular data. This defaults to being
    a GWCS object.
    """
    orig_array = u.Quantity(array)

    coord_frame = cf.CoordinateFrame(naxes=1,
                                     axes_type=('SPECTRAL',),
                                     axes_order=(0,))

I think there might be an option of getting the desired behavior by defining the flux as a NDCube instance in which case you also need to define the wcs by yourself.

Taking into consideration that this issue originated from the discussion regarding #1106 and that we are talking about Spectrum1D (emphasis on 1D) we should move forward with merging #1106 because the changes required to make the "NDCube.plot" method working are a lot more complex.