(fix): skip X writing when X is None

[flying-sheep]

• Closes Writing to H5AD has changed in v0.12.1 #2052
• Tests added
• Release note added (or unnecessary)

The removal of the str stuff is unrelated, I just didn’t want to step over it anymore without need (we don’t support the zarr version anymore that’s mentioned in the comment that addresses the str conversion)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.60%. Comparing base (40b6ccf) to head (a3b23fb).
⚠️ Report is 47 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2054   +/-   ##
=======================================
  Coverage   87.60%   87.60%           
=======================================
  Files          46       46           
  Lines        7049     7052    +3     
=======================================
+ Hits         6175     6178    +3     
  Misses        874      874           

Files with missing lines	Coverage Δ
src/anndata/_io/h5ad.py	93.39% <100.00%> (+0.22%)	⬆️
src/anndata/_io/specs/methods.py	91.07% <100.00%> (+0.01%)	⬆️
src/anndata/_io/zarr.py	82.92% <ø> (-0.99%)	⬇️

[flying-sheep]

adding this pop branch make Ruff complain about too high complexity, so I extracted _write_x and _write_raw.

The logic should otherwise be the same for raw, and almost the same for X:

1. check if X is to be written sparse-to-dense, if yes, do it, else
2. check if X is backed and the file path matches, if yes we’re done with X, else
3. (new) if X is None, remove "X" from the HDF5 group if it’s there, else
4. write X normally to the file

[ilan-gold]

does it make sense to add a bytes-for-bytes test for the underlying file?

[flying-sheep]

minimal AnnData files are byte-identical between 0.11.4 and this branch:

$ uvx --with=anndata==0.11.4 python -c 'from anndata import AnnData; adata = AnnData(shape=(10, 20)); adata.write_h5ad("0.11.h5py")'
$ hatch -e hatch-test.stable run python -c 'from anndata import read_h5ad; read_h5ad("0.11.h5py").write_h5ad("0.12.h5py")'
$ md5sum 0.11.h5py 0.12.h5py 
a8cdc9539050b6523f529073636733cc  0.11.h5py
a8cdc9539050b6523f529073636733cc  0.12.h5py

maximalist anndata (with X, so regardless of this PR) isn’t:

$ uvx --with=anndata[test]==0.11.4 --with=numba>=0.61 python -c 'from anndata.tests.helpers import gen_adata; adata = gen_adata((10, 20)); adata.write_h5ad("0.11.h5py")'
$ hatch -e hatch-test.stable run python -c 'from anndata import read_h5ad; read_h5ad("0.11.h5py").write_h5ad("0.12.h5py")'
$ md5sum 0.11.h5py 0.12.h5py 
74cbadb983b65dccd116c2342abdde30  0.11.h5py
a33f2f9244501426ca941a9b3299245e  0.12.h5py

[flying-sheep]

The HDF5 diff test works locally, I’m requesting your review now, and if it fails on CI, I’ll of course install it somehow and I’ll make sure it runs before merging.