Fix pre-release test failures for AnnData and read-only sparse matrices

[marcovarrone]

Summary by CodeRabbit

• Bug Fixes
  • Prevented unintended in-place changes to adjacency matrices during intra-cluster link removal.
  • Improved robustness of sparse matrix handling to avoid side effects and ensure consistent zero-elimination behavior.
  • Refined diagonal masking for enrichment calculations to use a safer helper approach.
• Tests
  • Updated the non_visium_adata fixture to cast coordinates to integers explicitly for more reliable AnnData construction.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 2a0b9b9f-3d5b-492d-b53f-d6d996b42d73

📥 Commits

Reviewing files that changed from the base of the PR and between fa37f70 and 44cc65c.

📒 Files selected for processing (1)

• src/cellcharter/gr/_nhood.py

---

📝 Walkthrough

Walkthrough

Three bug fixes address unsafe in-place mutation of sparse adjacency matrices: _remove_intra_cluster_links now copies non-writeable matrices before mutating, remove_intra_cluster_links uses explicit separate assignments for conns and dists, and _nhood_enrichment passes a copy of adj. A new _with_diagonal helper replaces direct diagonal masking operations. The test fixture adds an explicit .astype(int) cast.

Changes

Sparse adjacency matrix mutation safety

Layer / File(s)	Summary
Writeable check and explicit conns/dists assignment src/cellcharter/gr/_build.py	_remove_intra_cluster_links detects non-writeable sparse matrices and copies them before mutating data. remove_intra_cluster_links replaces a generator-based combined computation with two separate _remove_intra_cluster_links calls and individual adata.obsp key assignments.
Diagonal masking helper and caller-side copy src/cellcharter/gr/_nhood.py	Introduces _with_diagonal(df, value) helper to safely set diagonal values in DataFrames. _nhood_enrichment passes adj.copy() to avoid side effects on the original matrix and uses the new helper for diagonal-nan masking of observed, expected, and enrichment.
Fixture dtype correction tests/conftest.py	The non_visium_adata fixture casts coordinates using .astype(int) instead of passing dtype=int to the AnnData constructor.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐇 A matrix once frozen, too rigid to write,
Now gets a fresh copy before we take flight.
The diagonal masking, no longer in-place,
A helper ensures data stays safe and sound.
adj.copy() called — side effects are gone!
And fixtures cast cleanly with .astype(int). ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 14.29% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main changes: fixes for test failures related to AnnData and read-only sparse matrices, which aligns with the modifications to handle sparse matrix writeability and AnnData construction changes.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix-tests-v037

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}