Add support for the serialization of the ndcube WCS wrappers.

[ViciousEagle03]

This PR adds the support for the serialization of the following wrapper class:

ResampledLowLevelWCS
CompoundLowLevelWCS
ReorderedLowLevelWCS

Currently, this PR does not have tests written with underlying WCS attributes as the astropy.wcs.WCS and would be added after the asdf-astropy PR astropy/asdf-astropy#235 is merged.

Fixes: #750

---

Note: This PR Should be merged after #708 is merged

[ViciousEagle03]

Hi, @braingram, during the Tox CI/CD testing, the test for the test_serialization_compoundwcs is [failing]

It is throwing an asdf.exceptions.AsdfSerializationError indicating that the object of type astropy.modeling.tabular.Tabular1D is not serializable.

I checked the env created by the Tox, and it has asdf-astropy version as 0.6.1, which supports the serialization of this object by the tag: tag:stsci.edu:asdf/transform/tabular-1*.

[braingram]

It looks like _generate_tabular calls tabular_model:

ndcube/ndcube/extra_coords/table_coord.py

Line 139 in 2ef42ae

TabularND = tabular_model(ndim, name=f"Tabular{ndim}D")

This creates a new class, one that hasn't been registered with the asdf-astropy Tabular1D converter. So in this case, even though the class "looks" the same it's unknown to asdf. A minimal example that shows this is:

>> import astropy.modeling.tabular
>> t1d = astropy.modeling.tabular.tabular_model(1, name="Tabular1D")
>> t1d
<class 'astropy.modeling.tabular.Tabular1D'>
Name: Tabular1D
N_inputs: 1
N_outputs: 1
>> astropy.modeling.tabular.Tabular1D
<class 'astropy.modeling.tabular.Tabular1D'>
Name: Tabular1D
N_inputs: 1
N_outputs: 1
>> t1d is not astropy.modeling.tabular.Tabular1D
True

Let me think about how to fix this. There are issues supporting dynamic classes in asdf and the extension machinery is not set up to support this pattern. If it's possible to update table_coord.py to use the existing Tabular1D and Tabular2D for those ndim values the wcses can be supported by the existing machinery.

[Cadair]

If it's possible to update table_coord.py to use the existing Tabular1D and Tabular2D for those ndim values the wcses can be supported by the existing machinery.

This should be easy enough.

[Cadair]

Other than the test issue, I think the code looks generally good 👍

[Cadair]

These can just be numbers?

[ViciousEagle03]

Yes, these values can be numbers, but since there isn't any public method to retrieve the values of factor and offset, we have to access the private attributes _factor and _offset. These attributes are instances of the np.array class(factor and offset are converted to numpy array here), so the schema checks if they are of type "object."

However, it might be better to check against the specific tag: "tag:stsci.edu:asdf/core/ndarray-1.0.0". Thanks for pointing it out.

[ViciousEagle03]

If it's possible to update table_coord.py to use the existing Tabular1D and Tabular2D for those ndim values the wcses can be supported by the existing machinery.

This should be easy enough.

I am on it.

[ViciousEagle03]

Hey @Cadair and @braingram, is the PR ready to be merged, or is there anything else that needs to be addressed besides extending the test suite when the astropy/asdf-astropy#235 PR is merged?