Add milli- prefix to Angstrom unit - #11788

Shouldn't there be a test too?

The automatic going through all units should be good enough. But I note there is a failure in the doctests, plus a real error because mÅ was used in a test to check that it failed.

@byrdie - do you think you'll be able to fix this? If you feel it is too tricky, I can do it for you.

This works in some unexpected way in that it enables u.mAngstrom, u.milliAA and u.milliangstrom rather than u.mAA and u.milliAngstrom - the reason being evidently that for def_unit

the first element is the canonical (short) name, and the rest of the elements are aliases.

It might seem more logical then to define it as

def_unit(['AA', 'Angstrom', 'angstrom'], ...

but that would break e.g. the expected fits format outputs.
As a kind of alternative workaround you might instead define both prefix versions as aliases for both the short and long form:

    prefixes=[(['m', 'milli'], ['milli', 'm'], 1.e-3)],

You will also have to update the matches for test_units:test_suggestions to

    ('ångström', 'Angstrom, angstrom, mAngstrom or mangstrom')

@dhomeier, thanks for explaining that, I had not yet figured out why that test was failing.
I'll add the workaround you posted and update the other test you mentioned.

@mhvk this does raise some concerns that similar problems might pop up downstream if looking for the expected exception. But I don't really see a good alternative if we want to accept u.mAA (and I definitely think we should).
One option could be to allow only 'milli' as 'short' prefix, backwards as it may seem
prefixes=[(['milli'], ['milli', 'm'], 1.e-3)],
but that would still only get rid of mAngstrom, not mangstrom, so maybe not much point in it.

@mhvk, thanks for updating the doctest. I didn't know if I was supposed to update the .rst file directly or if there was a docstring in
a .py file somewhere that I had to fix.

@byrdie - yes, I thought this was getting a bit beyond what a supposedly quick PR should do, especially as with those doctests it is often a bit trickier than thought. Anyway, tests pass now and @dhomeier also agrees this is all OK, so I'll merge. Thanks again, and of course other suggestions for making astropy just a little bit better are very welcome!

p.s. @dhomeier - I agree that the mAngstrom is really not all that great, but ended up also agreeing that if mangstrom is in there, we should just not perfect be the enemy of good enough. It would have been nicer if the names could just have been swapped, so that AA was first.

Thanks @byrdie and congrats to completing the first PR into Astropy!

Thanks @mhvk, @dhomeier and everyone else for your help and allowing me the opportunity to contribute!