(feat): io submodule

[ilan-gold]

A brief summary of how this was handled and why:

1. I moved everything into io related to i/o
2. I left read_zarr and read_h5ad at the top-level for now, although I think that should be deprecated in favor of everything going through io in the future despite their "high" status above the others. They are now documented as being from io to begin to encourage that behavior although we aren't throwing warnings for e.g., anndata.read_h5ad yet
3. I also made a backwards compat for people who use from anndata._io import blah since uhhhhhhh yikes by leaving that module alone. However, at least if anyone tries to import from the top-level from anndata._io import read_h5ad, it will warn. Also, I added a public note about this issue since it seems pervasive.
4. I added tests for the deprecations. For now, we need to check the scanpy imports for warnings but we can open the PR for scanpy before its release to adapt. I think the imports are cached or something so pytest.warns can't run reliably every time because the warning doesn't come every time....not sure what's up with that

• Closes io submodule #1681
• Tests added
• Release note added (or unnecessary)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.49%. Comparing base (34e9783) to head (a947d3e).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1682      +/-   ##
==========================================
- Coverage   86.87%   84.49%   -2.39%     
==========================================
  Files          39       40       +1     
  Lines        6036     6044       +8     
==========================================
- Hits         5244     5107     -137     
- Misses        792      937     +145     

Flag	Coverage Δ
	84.49% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/anndata/__init__.py	93.54% <100.00%> (+0.95%)	⬆️
src/anndata/_core/anndata.py	83.72% <100.00%> (ø)
src/anndata/_core/sparse_dataset.py	92.43% <ø> (ø)
src/anndata/_io/__init__.py	100.00% <100.00%> (ø)
src/anndata/_io/read.py	82.60% <ø> (+1.55%)	⬆️
src/anndata/_io/specs/lazy_methods.py	100.00% <ø> (ø)
src/anndata/_types.py	85.29% <ø> (ø)
src/anndata/experimental/__init__.py	100.00% <100.00%> (ø)
src/anndata/io.py	100.00% <100.00%> (ø)
src/anndata/typing.py	100.00% <100.00%> (ø)
... and 1 more

... and 7 files with indirect coverage changes

[flying-sheep]

Very nice and clean! We should add a release note fragment for this change.

We do turn FutureWarnings into errors right? So it’s impossible that you missed one in here, right?

[ilan-gold]

We do turn FutureWarnings into errors right? So it’s impossible that you missed one in here, right?

Could you elaborate? When I search FutureWarning in our code base, I don't see anything that might signal they are treated as errors, and a lot of pytest.warns.

[flying-sheep]

One of our test job runs with --strict-warnings, which turns all warnings into errors except the ones from an allowlist (that exists because we haven’t gotten around to dealing with them).

These are the default filters, applied in every test job:

anndata/src/testing/anndata/_pytest.py

Lines 26 to 31 in 34e9783

	config.addinivalue_line(
	"filterwarnings", "ignore::anndata._warnings.OldFormatWarning"
	)
	config.addinivalue_line(
	"filterwarnings", "ignore::anndata._warnings.ExperimentalFeatureWarning"
	)

This is the allowlist:

anndata/pyproject.toml

Lines 148 to 154 in 34e9783

	filterwarnings_when_strict = [
	"default::anndata._warnings.ImplicitModificationWarning",
	"default:Transforming to str index:UserWarning",
	"default:(Observation|Variable) names are not unique. To make them unique:UserWarning",
	"default::scipy.sparse.SparseEfficiencyWarning",
	"default::dask.array.core.PerformanceWarning",
	]

This is the code that turns all warnings into errors except the above.

anndata/src/testing/anndata/_pytest.py

Lines 97 to 108 in 34e9783

	def pytest_collection_modifyitems(
	session: pytest.Session, config: pytest.Config, items: Iterable[pytest.Item]
	):
	if not config.getoption("--strict-warnings"):
	return
	warning_filters = [
	"error",
	*_config_get_strlist(config, "filterwarnings"),
	*_config_get_strlist(config, "filterwarnings_when_strict"),
	]
	warning_marks = [pytest.mark.filterwarnings(f) for f in warning_filters]

[ilan-gold]

Sorry at @flying-sheep yes I knew that about stricto-warnings, but I meant about FutureWarning specifically, which is what confused me. In any case I misread "impossible" as "possible" hence my confusion. So yes, it's impossible we missed

[flying-sheep]

Looks great! Two questions

[flying-sheep]

Any reason why we shouldn’t just stop doing that?

[ilan-gold]

Would need to investigate what it's doing there or what that comment even means. Maybe an issue?

[ilan-gold]

Well, I deleted it and nothing happened so....

[flying-sheep]

I put that import and comment there because Isaac made the tests access this by attribute:

import anndata as ad

do_stuff_with(ad._io.specs.foobar)

instead of importing that submodule. It happened to work because some other code somewhere else caused anndata._io.specs to be imported, but “happens to work because of side effects” is not what I want to build upon.

Now that the tests no longer access stuff that way, this import is no longer necessary.

[flying-sheep]

I got rid of a few more obsolete import hacks: a947d3e

Why breaking change ‼️ A breaking change that needs a major version update ?