Remove deprecated blosc code

[dstansby]

This was deprecated in 0.15.0, as code that is either experimental or not intended to be public, so remove it for 0.16.0. This reduces the amount of Python interface we need to maintain for blosc, and enables the option to switch to using the third party blosc python package in the future.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.96%. Comparing base (725cf25) to head (42f9051).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #712      +/-   ##
==========================================
- Coverage   99.96%   99.96%   -0.01%     
==========================================
  Files          63       63              
  Lines        2778     2738      -40     
==========================================
- Hits         2777     2737      -40     
  Misses          1        1              

Files with missing lines	Coverage Δ
numcodecs/__init__.py	100.00% <100.00%> (ø)
numcodecs/tests/common.py	100.00% <ø> (ø)
numcodecs/tests/test_blosc.py	100.00% <ø> (ø)
numcodecs/zarr3.py	99.50% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[flying-sheep]

Hm, this breaks old zarr versions. Is there at least some Zarr 2 version that pins numcodecs to <0.16?

[d-v-b]

looks like this was a breaking change for some implementations. we might want to revert.

[dstansby]

Sorry about this - there's a pin over at zarr-developers/zarr-python#2965, which I'll try and get released ASAP

[d-v-b]

I'm trying to understand why we didn't revert these changes before issuing a new release, as the changes in this PR are breaking for zarr-python 2.18.

[dstansby]

It was my mistake that this broke zarr-python 2.18 - I think the easiest and most pragmatic fix is to put a numcodecs pin in zarr-python 2.18, which I've proposed over at zarr-developers/zarr-python#2965

[d-v-b]

But given that this PR requires changes to zarr-python, why did we issue a new numcodecs release before the preparatory changes in zarr-python were in?

[dstansby]

I guess because we don't proactively test zarr-python v2 against development versions of numcodecs, so this wasn't caught before the release?

[dstansby]

I definitely think we should discuss what process/development changes to make so this doesn't happen again, but in the short term it would be great to get a quick review and merge on zarr-developers/zarr-python#2965 so we can unbreak zarr 2.18

[d-v-b]

maybe i'm getting the timeline wrong, but it looks like the problems with this PR were reported yesterday, but the numcodecs release was today? In theory we could have deferred releasing these changes specifically until zarr-python was prepared for it.

[dstansby]

Oh, I made the "release" entry on GitHub this morning, but the actual release/tagging was yesterday.

[d-v-b]

gotcha, so my timeline was inaccurate

[flying-sheep]

Over at scverse, we use some nice release automation to prevent anything from going wrong with coordinating GitHub/PyPI releases: https://github.com/scverse/anndata/blob/main/.github/workflows/publish.yml

Of course, if you hand-craft the release notes differently for the docs and GitHub, that’s not possible, but if you use the same text, it’s worth it I’d say.