Docstring return statements

[mauicv]

what is this

Some return statements in alibi detect are formatted incorrectly. This PR fixes this issue #759. These changes are all taken from #748 which will be closed. As an example of the kind of thing this fixes compare this to this.

• Add Docstrings section from alibi CONTRIBUTING.md here
• Check this behaviour

[codecov]

Codecov Report

Merging #789 (e517b8c) into master (5a38fc2) will not change coverage.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #789   +/-   ##
=======================================
  Coverage   81.45%   81.45%           
=======================================
  Files         154      154           
  Lines       10048    10048           
=======================================
  Hits         8185     8185           
  Misses       1863     1863           

Impacted Files	Coverage Δ
alibi_detect/ad/adversarialae.py	63.39% <ø> (ø)
alibi_detect/ad/model_distillation.py	91.80% <ø> (ø)
alibi_detect/base.py	85.45% <ø> (ø)
alibi_detect/cd/base.py	90.97% <ø> (ø)
alibi_detect/cd/base_online.py	88.23% <ø> (ø)
alibi_detect/cd/classifier.py	100.00% <ø> (ø)
alibi_detect/cd/context_aware.py	97.43% <ø> (ø)
alibi_detect/cd/keops/learned_kernel.py	94.16% <ø> (ø)
alibi_detect/cd/keops/mmd.py	98.21% <ø> (ø)
alibi_detect/cd/learned_kernel.py	100.00% <ø> (ø)
... and 44 more

[jklaise]

Since we're squashing commits, there should only be one entry per PR. Can you check if it's the case that the hash of the first commit is used after the squash?

[mauicv]

I was actually wondering about this and meant to check how it works, and it seems the above does not. It seems that we'll have to do two separate PRs one with the changes and one with the relevant commits in .git-blame-ignore-revs.

[jklaise]

Or squash locally before merging so the PR only has one commit? Edit: on second though, you'd need to know the commit hash before it's created so probably not...

[mauicv]

Well I could squash the PR commits to one and then add another commit before merging into master instead of squashing?

[jklaise]

Are these explicit line breaks actually needed?

[mauicv]

Its the difference between this and this. The issues are covered here. I need to add the CONTRIBUTING.md docstring section from alibi to this PR as well btw

[ascillitoe]

Didn't realise this, good spot @mauicv !

[jklaise]

LGTM!

[ascillitoe]

Nitpick:

and the type, if provided, would be written

Could we be a little more specific here? Not detailing types in the general case (since already done by sphinx-autodoc-typehints), but giving types when multiple objects are contained in another object (i.e. a dict), makes a lot of sense IMO. However, in the latter case, are we saying this should always be done? Or its optional?

[mauicv]

I think we're hinting that it's better to give types however it's not something we've done historically. I've opened an issue to change this across the codebase and then perhaps we can change the language here to be a little stronger.

[ascillitoe]

One minor nitpick (or question even), but otherwise LGTM!