Use numpy.typing.DTypeLike

[tomwhite]

This fixes half of #452. NumPy's ArrayLike isn't a drop-in replacement for ours, since it represents something that can be converted to a NumPy array (including lists, for example), whereas we use it in the sense of "something that looks like an array" (which could be a Dask array, for example).

[tomwhite]

The test is failing due to some interaction between numpy.typing and scanpydoc.

[tomwhite]

The test is failing due to some interaction between numpy.typing and scanpydoc.

I think it is failing only on Python 3.7, since SupportsIndex is not there (introduced in Python 3.8), and Python 3.7 when called from scanpydoc doesn't use typing-extensions.

One fix might be to require Python >=3.8 for the doc build, so we could skip "Check for Sphinx doc warnings" in build.yml when running Python 3.7. (Not sure how easy it is to do that though.)

[ravwojdyla]

👍

[tomwhite]

@ravwojdyla thanks for your (offline) suggestion to do this. The workflow syntax works perfectly. But the build is failing on 3.8 too, so I need to look into that!

[codecov-commenter]

Codecov Report

Merging #594 (4be3bf1) into main (70a91ba) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main      #594   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           35        35           
  Lines         2749      2762   +13     
=========================================
+ Hits          2749      2762   +13     

Impacted Files	Coverage Δ
sgkit/io/bgen/bgen_reader.py	100.00% <100.00%> (ø)
sgkit/io/utils.py	100.00% <100.00%> (ø)
sgkit/io/vcf/vcf_reader.py	100.00% <100.00%> (ø)
sgkit/stats/ld.py	100.00% <100.00%> (ø)
sgkit/stats/pca.py	100.00% <100.00%> (ø)
sgkit/typing.py	100.00% <100.00%> (ø)
sgkit/utils.py	100.00% <100.00%> (ø)
sgkit/window.py	100.00% <100.00%> (ø)
sgkit/variables.py	100.00% <0.00%> (ø)
sgkit/stats/popgen.py	100.00% <0.00%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 70a91ba...4be3bf1. Read the comment docs.

[tomwhite]

Various extra changes were needed to get Sphinx to build the docs. The type annotation is very off-putting for users though:

I was expecting the annotation to be DTypeLike, linked to https://numpy.org/devdocs/reference/typing.html#numpy.typing.DTypeLike.

This problem was discussed in Pandas in pandas-dev/pandas#33025. It might be worth investigating further to see if we can improve this.

[ravwojdyla]

@tomwhite nice, should we merge this and investigate/improve in a separate PR/issue?

[tomwhite]

@ravwojdyla I'd rather wait and see if there's a way to avoid the off-putting type signatures in docs, as I don' think this is an improvement as it stands.

[mergify]

This PR has conflicts, @tomwhite please rebase and push updated version 🙏