Remove vendored obsgeo_to_frame in favor of using Astropy

[ahmedhosssam]

Hi,
I found that some tests in sunpy/coordinates/tests/test_wcs_utils.py should be removed after Astropy 5.0, but Astropy is now 6.0 so I think it should be removed, right?
If it's right, I think it will make the whole testing process faster, now pytest runs in less than 120 seconds after it used to take more than 180 seconds on my computer.

Please let me know if I'm missing something.

[ayshih]

The vendored version of sunpy.coordinates wcs_utils.obsgeo_to_frame() should be removed at the same time (this all came in with #5315).

It'd be mind-boggling if these three tests take over a minute to run in your environment. I suspect that something went awry in your comparison.

[ahmedhosssam]

The vendored version of sunpy.coordinates wcs_utils.obsgeo_to_frame() should be removed at the same time (this all came in with #5315).

Done.
Should I remove these tests also?

@pytest.mark.parametrize("obsgeo", ([np.nan] * 6, None, [0] * 6, [54] * 5))
def test_obsgeo_invalid(obsgeo):

    with pytest.raises(ValueError):
        obsgeo_to_frame(obsgeo, None)
def test_observer_hgln_crln_priority():
    """
    When extracting the observer information from a FITS header, ensure
    Stonyhurst (HG) coordinates are preferred over Carrington (CR) if present.
    Note that `Map` creates a custom FITS header with a sanitized observer
    location, so it is separately tested in the map module.
    """
    header = {'CRVAL1': 0,
              'CRVAL2': 0,
              'CRPIX1': 5,
              'CRPIX2': 5,
              'CDELT1': 10,
              'CDELT2': 10,
              'CUNIT1': 'arcsec',
              'CUNIT2': 'arcsec',
              'PC1_1': 0,
              'PC1_2': -1,
              'PC2_1': 1,
              'PC2_2': 0,
              'NAXIS1': 6,
              'NAXIS2': 6,
              'CTYPE1': 'HPLN-TAN',
              'CTYPE2': 'HPLT-TAN',
              'date-obs': '1970-01-01T00:00:00',
              'mjd-obs': 40587,
              'hglt_obs': 0,
              'hgln_obs': 0,
              'crlt_obs': 2,
              'crln_obs': 2,
              'dsun_obs': 10,
              'rsun_ref': 690000000}
    wcs = WCS(header)
    c = wcs.pixel_to_world(0, 0)
    assert c.observer.lon == 0 * u.deg
    # Note: don't test whether crlt or hglt is used---according to
    # _set_wcs_aux_obs_coord, those are expected to always be the same and so
    # the same one is always used

[nabobalis]

Do you mean test_obsgeo_invalid? I think that should go but I don't see why the test after that would go?

[ahmedhosssam]

All of the tests that I removed already exist in astropy.wcs.tests.test_utils.py now, except test_observer_hgln_crln_priority, it tests astropy.wcs.WCS. As far as I understand, the test should belong to astropy rather than sunpy. Or the data in header are related to sunpy?

[nabobalis]

I would need Albert to double check but that test should stay in sunpy.

[ayshih]

All of the tests that I removed already exist in astropy.wcs.tests.test_utils.py now, except test_observer_hgln_crln_priority, it tests astropy.wcs.WCS. As far as I understand, the test should belong to astropy rather than sunpy. Or the data in header are related to sunpy?

It's not readily apparent from the test code, but it internally calls a sunpy function to convert the WCS header to a SunPy coordinate frame. The test should stay.

[ayshih]

Please make the changes flagged by pre-commit, otherwise looks good to me

[ahmedhosssam]

Please make the changes flagged by pre-commit, otherwise looks good to me

Done.

[ayshih]

Thanks for the PR!