__pow__ method incompatible with downstream subclasses

[rosteen]

Describe the bug

The use of _new_instance in the NDCube.__pow__ method was causing a traceback when doing arithmetic with downstream specutils.Spectrum objects that inherit from NDCube. See here for a description/traceback. I worked around this in astropy/specutils#1291 but it might be worth updating the method here to avoid this.

To Reproduce

See linked specutils issue above.

Screenshots

No response

System Details

• ndcube 2.3.4

Installation method

No response

[Cadair]

Hi @rosteen thanks for the bug report.

I think this should be fixed in the upcoming 2.4 release as we added the psf argument to our constructor: https://github.com/sunpy/ndcube/blob/main/ndcube/ndcube.py#L381-L385

I'm in the process of cutting the 2.4 release, is there any chance you could test it works with main?

[rosteen]

Hi @rosteen thanks for the bug report.

I think this should be fixed in the upcoming 2.4 release as we added the psf argument to our constructor: https://github.com/sunpy/ndcube/blob/main/ndcube/ndcube.py#L381-L385

I'm in the process of cutting the 2.4 release, is there any chance you could test it works with main?

Sure, I'll revert the compatibility patch locally to see if it works.

[rosteen]

Looks like I still get the same ValueError with ndcube main:

>       inverse_r = 1/r  # This used to raise an error.
                    ^^^

tests/test_arithmetic.py:108:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../ndcube/ndcube/ndcube.py:1113: in __rtruediv__
    return self.__pow__(-1).__mul__(value)
           ^^^^^^^^^^^^^^^^
../../ndcube/ndcube/ndcube.py:1134: in __pow__
    return self._new_instance(data=new_data, unit=new_unit, uncertainty=new_uncertainty)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
../../ndcube/ndcube/ndcube.py:961: in _new_instance
    new_cube = type(self)(**full_kwargs)
               ^^^^^^^^^^^^^^^^^^^^^^^^^
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <[AttributeError("'Spectrum' object has no attribute '_data'") raised in repr()] Spectrum object at 0x11bbb1db0>
flux = None, spectral_axis = None, spectral_axis_index = None
wcs = <WCS(output_frame=SpectralFrame, input_frame=detector, forward_transform=Model: SpectralTabular1D
N_inputs: 1
N_output...:
  points: (array([0, 1, 2]),)
  lookup_table: [1. 2. 3.] m
  method: linear
  fill_value: nan
  bounds_error: True)>
velocity_convention = None, rest_value = None, redshift = None, radial_velocity = None
bin_specification = None, move_spectral_axis = None
kwargs = {'data': array([-1.33333333, -1.66666667, -2.        ]), 'mask': None, 'meta': OrderedDict(), 'psf': None, ...}
unknown_kwargs = {'psf'}

    def __init__(self, flux=None, spectral_axis=None, spectral_axis_index=None,
                 wcs=None, velocity_convention=None, rest_value=None,
                 redshift=None, radial_velocity=None, bin_specification=None,
                 move_spectral_axis=None, **kwargs):

        if spectral_axis_index == -1:
            spectral_axis_index = flux.ndim - 1

        # If the flux (data) argument is already a Spectrum (as it would
        # be for internal arithmetic operations), avoid setup entirely.
        if isinstance(flux, Spectrum):
            self._spectral_axis_index = flux.spectral_axis_index
            self._spectral_axis = flux.spectral_axis
            super().__init__(flux)
            return

        self._spectral_axis_index = spectral_axis_index
        # Might as well handle this right away
        if spectral_axis_index is None and flux is not None:
            if flux.ndim == 1:
                self._spectral_axis_index = 0
        elif flux is None:
            self._spectral_axis_index = 0

        # Check for pre-defined entries in the kwargs dictionary.
        unknown_kwargs = set(kwargs).difference(
            {'data', 'unit', 'uncertainty', 'meta', 'mask', 'copy',
             'extra_coords'})

        if len(unknown_kwargs) > 0:
>           raise ValueError("Initializer contains unknown arguments(s): {}."
                             "".format(', '.join(map(str, unknown_kwargs))))
E           ValueError: Initializer contains unknown arguments(s): psf.

spectra/spectrum.py:114: ValueError