Remove code deprecated in 0.3 and 0.4

[astrofrog]

My understanding of the deprecation process is that we are deprecating functions/classes/methods in one stable version and removing it in the next. This PR removes code marked as deprecated in 0.3 and 0.4 since it should be removed before 1.0. I was thinking it would make sense to do this sooner rather than later so that packages that test against 1.0 won't see these functions disappear at the last minute before the 1.0 release. We shouldn't backport this to v0.4.x though.

@embray - there is actually a fair bit of deprecated code in astropy.io.fits - shall we remove it soon so that it doesn't get included in 1.0?

cc @eteq @embray @mdboom @taldcroft @nhmc @aconley (since this includes changes in one of your subpackages)

[astrofrog]

@nhmc @aconley - luminosity_distance was deprecated in 0.4, but Distance.compute_z still uses it, hence the failure. How should we deal with this?

[astrofrog]

By the way, we can also decide to not merge this, I don't feel strongly about it, but we should remove code that we agree should be removed to avoid clutter.

[taldcroft]

👍 for removing the deprecated items.

[astrofrog]

@nhmc @aconley @eteq - the failure is due to an issue in astropy.coordinates in the Distance class - see #2991

[nhmc]

This looks great to me, once #2991 is merged.

[aconley]

Looks like I missed the entire discussion and resolution, but, yes, looks good to me, assuming #2991 fixes the Travis fail.

[mdboom]

👍

[mhvk]

Great!

[embray]

👍

[astrofrog]

@embray - shall I also remove deprecated code from io.fits or will you see to that separately?

[embray]

I'll see to that separately (I have already done that in PyFITS so it's just a matter of merging over).

[eteq]

👍 once #2991 is merged and makes the tests pass.

[astrofrog]

@nhmc @aconley - I just removed some of the tests for astropy.cosmology that I think are no longer relevant - could you double check that I did not remove too much?

[astrofrog]

@nhmc @aconley - I found I had to update test_z_at_value_roundtrip so can you review that carefully? In particular, I had to mark all the methods as being skipped:

    skip = ('Ok', 'angular_diameter_distance_z1z2', 'clone', 'de_density_scale', 'w')

Shouldn't Ok, de_density_scale, and w work though? (they fail currently, hence why I skip them).

[astrofrog]

@nhmc @aconley - the above function is what I was referring to - I'm not sure if I should be skipping Ok, de_density_scale and w - maybe they should just work? (in which case it indicates a bug)

[aconley]

For the Planck 13 cosmology (flat, w=-1), Ok, de_density_scale, and w are redshift independent. So z_at_value can't invert the function. If you want to test those three, you will need to create a more 'interesting' cosmology.

[aconley]

@astrofrog -- other than the round trip test, looks good. And for the round trip test, I don't think there's a problem, it's just that you have to decide if you want to explicitly test those functions (which will require a different test cosmology).

[astrofrog]

@aconley - ok, thanks for the clarification!

Since there were no objections above, I'll go ahead and merge :)

[embray]

Changes that remove more code than they add are always the best.