Prevent warnings about HIERARCH with CompImageHeader class

[timj]

Description

Compressed image headers warn when you try to assign a HIERARCH keyword but normal image headers don't.

The underlying problem is that Card.keyword strips the HIERARCH and CompImageHeader tries to reconstruct a Card from the keyword of that card. I fix it by checking if the Card itself starts with HIERARCH already and reapply it.

This pull request is to address #11403

Fixes #11403

[timj]

I'm not entirely sure how to test for "this thing did not warn". There are plenty of examples for warnings being issued. Do I do something like:

with assertWarns:
    run code that does not warn
    issue a warning myself

assert that I only got the warning I issued

?

[pllim]

The global warning config for pytest is here:

astropy/setup.cfg

Lines 115 to 141 in aed5138

	filterwarnings =
	error
	ignore:Distutils was imported before Setuptools
	ignore:unclosed file:ResourceWarning
	ignore:unclosed <socket:ResourceWarning
	ignore:unclosed <ssl.SSLSocket:ResourceWarning
	ignore:numpy.ufunc size changed:RuntimeWarning
	ignore:numpy.ndarray size changed:RuntimeWarning
	ignore:Importing from numpy:DeprecationWarning:scipy
	ignore:Conversion of the second argument:FutureWarning:scipy
	ignore:Using a non-tuple sequence:FutureWarning:scipy
	ignore:Using or importing the ABCs from 'collections':DeprecationWarning
	ignore:can't resolve package from __spec__
	ignore:::astropy.tests.plugins.display
	ignore:::astropy.tests.disable_internet
	ignore:direct construction of Asdf:pytest.PytestDeprecationWarning
	ignore:Unknown pytest.mark.mpl_image_compare:pytest.PytestUnknownMarkWarning
	ignore:Unknown config option:pytest.PytestConfigWarning
	ignore:matplotlibrc text.usetex:UserWarning:matplotlib
	# Triggered by ProgressBar > ipykernel.iostream
	ignore:the imp module is deprecated:DeprecationWarning
	# toolz internal deprecation warning https://github.com/pytoolz/toolz/issues/500
	ignore:The toolz.compatibility module is no longer needed:DeprecationWarning
	# Ignore a warning we emit about not supporting the parallel
	# reading option for now, can be removed once the issue is fixed
	ignore:parallel reading does not currently work, so falling back to serial
	ignore:'datfix' made the change:astropy.wcs.wcs.FITSFixedWarning

With this setup, if you do nothing, all unexpected warnings, except those globally ignored, will turn into an exception. So, if you want to test that something should not raise a warning, no special code is needed. But be careful if the warning you are interested in is already ignored by pytest above.

If you want to make sure your code is raising the correct warning, use pytest.warns. More info at https://docs.pytest.org/en/stable/warnings.html . Hope this helps!

[timj]

I've pushed a test. The failure is the samp one again.

[pllim]

Don't worry about the samp failure. If it keeps happening across the PRs, I'll think of something. Thanks for letting us know!

[saimn]

In that case I think it's better to test explicitly that no warning were raised, which can be done with pytest.warns(None) (you can find examples in astropy if you need).

[timj]

I've updated the test as requested above. I assume the documentation failure is nothing to do with this change.

[pllim]

Yeah, matplotlib 3.4.0 release last night broke RTD.

[embray]

Thanks for the fix. Looks good to me!

[saimn]

Looks good to me too. @timj - could you rebase on master and add a changelog entry as described in https://github.com/astropy/astropy/blob/master/docs/changes/README.rst ? Thanks.

[timj]

I've rebased and added a change log entry.

[saimn]

Thanks @timj!