default buffer logic

[d-v-b]

This PR builds on #3638 to add routines to the store ABC, and thereby all other stores, for handling additional types for the prototype parameter passed to store.get and the methods that call get. This change makes it easier to use store methods without unnecessary imports.

The type of prototype is widened to type[Buffer] | BufferPrototype | None. type[Buffer] | BufferPrototype is compacted to the type alias BufferLike.

When prototype is type[Buffer], the store will use that class directly. Since the stores never use the BufferPrototype.nd_buffer, passing a Buffer class should be preferred over using BufferPrototype.

When prototype is None, the store class retrieves a Buffer class by calling its _get_default_buffer_class method. The default implementation returns the default_buffer_prototype which makes our ABC less abstract but keeps these changes non-breaking.

[codecov]

Codecov Report

❌ Patch coverage is 60.89744% with 61 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.99%. Comparing base (1c05c1a) to head (6bba68f).

Files with missing lines	Patch %	Lines
src/zarr/storage/_fsspec.py	0.00%	17 Missing ⚠️
src/zarr/abc/store.py	42.85%	12 Missing ⚠️
src/zarr/testing/store.py	69.69%	10 Missing ⚠️
src/zarr/storage/_local.py	75.00%	7 Missing ⚠️
src/zarr/storage/_memory.py	70.83%	7 Missing ⚠️
src/zarr/storage/_common.py	25.00%	3 Missing ⚠️
src/zarr/storage/_obstore.py	81.25%	3 Missing ⚠️
src/zarr/storage/_logging.py	0.00%	1 Missing ⚠️
src/zarr/storage/_wrapper.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3644      +/-   ##
==========================================
+ Coverage   60.89%   60.99%   +0.10%     
==========================================
  Files          86       86              
  Lines       10182    10293     +111     
==========================================
+ Hits         6200     6278      +78     
- Misses       3982     4015      +33     

Files with missing lines	Coverage Δ
src/zarr/experimental/cache_store.py	100.00% <100.00%> (ø)
src/zarr/storage/_zip.py	72.98% <100.00%> (+1.13%)	⬆️
src/zarr/storage/_logging.py	61.94% <0.00%> (ø)
src/zarr/storage/_wrapper.py	43.58% <0.00%> (ø)
src/zarr/storage/_common.py	66.00% <25.00%> (ø)
src/zarr/storage/_obstore.py	62.28% <81.25%> (+1.45%)	⬆️
src/zarr/storage/_local.py	72.10% <75.00%> (+1.51%)	⬆️
src/zarr/storage/_memory.py	61.34% <70.83%> (+4.90%)	⬆️
src/zarr/testing/store.py	74.92% <69.69%> (-0.08%)	⬇️
src/zarr/abc/store.py	37.73% <42.85%> (+0.85%)	⬆️
... and 1 more

🚀 New features to boost your workflow:

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

[normanrz]

@d-v-b Could you elaborate on the motivation for widening the Store API like this?

[d-v-b]

Stores don't do anything with NDBuffer classes. They just use a Buffer class. So prototype: BufferPrototype is indirect -- no store will use the BufferPrototype.nd_buffer element. That's why prototype can be a Buffer class in this PR.

And multiple stores have already implemented logic for selecting a default buffer class. The changes here propagate that sensible design to all store classes.

[codspeed-hq]

CodSpeed Performance Report

Merging this PR will not alter performance

_{Comparing d-v-b:feat/default-buffer (b205b35) with main (e94504e)}

Summary

✅ 30 untouched benchmarks