TESTS: don't force rectilinear config true in the test

[ianhi]

The new array strategy for stateful testing was ignoring the config setting and forcibly changing it - which gives no way for a downstream consumer to opt of out. This makes it respect the config. There is a small worry that now one of our strategies depends on a global state (potential cause of flakyiness) but I'm not really sure there's a clean way around that.

@dcherian

TODO:

• Add unit tests and/or doctests in docstrings
• Add docstrings and API docs for any new/modified user-facing classes and functions
• New/modified features documented in docs/user-guide/*.md
• Changes documented as a new file in changes/
• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.08%. Comparing base (2fbd30a) to head (2e49ae3).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3909      +/-   ##
==========================================
- Coverage   93.08%   93.08%   -0.01%     
==========================================
  Files          85       85              
  Lines       11230    11225       -5     
==========================================
- Hits        10454    10449       -5     
  Misses        776      776              

Files with missing lines	Coverage Δ
src/zarr/testing/strategies.py	92.91% <100.00%> (-0.13%)	⬇️

🚀 New features to boost your workflow:

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

[dcherian]

huh?

[ianhi]

That was all me - so any confusion is my fault.

This makes this strategy dependent on global state, i.e. a recipe for a flaky strategy down the line. I wanted to leave a note for the future in case that happens. Don't think I really hit the mark here though

[dcherian]

ok let's write in a longer comment then :)

[ianhi]

Suggested change

	Depending on whether the config is set (WARN: look here if we get a flaky error)
	This strategy depends on the global state of the config having rectilinear chunk grids enabled or not.
	This means that it may be a possible source of a hypothesis FlakyStrategy error due dependence
	on global state. However, in practice this seems unlikely to happen.

[ianhi]

@dcherian how is that?

[d-v-b]

i'm confused -- don't we want to turn on rectilinear chunks explicitly before this test?

[d-v-b]

or is the idea to leave that detail to the configuration at test time

[d-v-b]

also... we should rethink putting the config check inside the chunk grid class itself. If someone wants to import the class and construct it, that should be allowed.

[ianhi]

i'm confused -- don't we want to turn on rectilinear chunks explicitly before this test?

Yes for the zarr-python tests, but not within the the strategy. The strategy is public API and might be run on a system that doesn't support rectilinear chunks. For example icechunk format 1 just straight up doesn't support it, so I need to be able to turn it off via the config and have that be respected.

In the actual zarr-python stateful tests it gets turned on by default:

zarr-python/tests/test_store/test_stateful.py

Lines 21 to 25 in 2fbd30a

	@pytest.fixture(autouse=True)
	def _enable_rectilinear_chunks() -> Generator[None, None, None]:
	"""Enable rectilinear chunks since strategies may generate them."""
	with zarr.config.set({"array.rectilinear_chunks": True}):
	yield