Closed astrofrog closed 2 years ago
Input has (4, 4, 5)
shape but somehow data['flux']
has (4, 5, 4)
shape. π€―
Specutils does horrible things to input arrays π
Actually I think the test is wrong since we deliberately re-order the axes between Spectrum1D and glue data objects, so should be a simple fix. I'll do it.
But why is it not failing for other jobs?
Because that test is conditional on dev version of dependencies:
pytest.importorskip("ndcube", minversion="1.99")
Sorry for not commenting on this sooner. It started failing on dev because https://github.com/astropy/specutils/pull/822 finally got merged, changing the order that the axes end up after automatically moving the spectral axis to be last. Acknowledged that this is...not ideal π
Previous to that fix on dev the WCS and actual data were getting misaligned when automatically reordering the axes for the spectral axis to be last. So "less than ideal" is better than "actually just wrong".
Actually I believe that astropy/specutils#822 actually fixed the test. I was seeing failures now because astropy dropped Python 3.7 support in main
so I've bumped the Python version used to do the dev CI - let's see if that works now.
Yes, the CI is all fine now. I've released glue-astronomy v0.3!
Merci! Are you going to update pin on jdaviz too? π
Yes in 733. Going to push commits soon.
The following test failure exists in main when testing with e.g.
py38-test-dev
:@rosteen - since you added this test, can you comment on why this might be failing?