fix inplaces for log1p and scale

[Intron7]

Summary by CodeRabbit

• Bug Fixes
  • log1p now preserves input data correctly when run without in-place updates, and no longer writes unnecessary metadata.
  • Scaling embeddings no longer adds summary statistics to feature metadata when the operation targets embedding data.
  • scale now updates feature statistics only when scaling the main matrix or a layer, improving consistency across data locations.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.31%. Comparing base (481c60c) to head (a2981ca).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #719   +/-   ##
=======================================
  Coverage   90.31%   90.31%           
=======================================
  Files         106      106           
  Lines        9581     9582    +1     
=======================================
+ Hits         8653     8654    +1     
  Misses        928      928           

Files with missing lines	Coverage Δ
src/rapids_singlecell/preprocessing/_normalize.py	89.03% <100.00%> (ø)
src/rapids_singlecell/preprocessing/_scale.py	91.66% <100.00%> (+0.04%)	⬆️

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 5ee4d206-0108-4224-aaad-e553e7be6287

📥 Commits

Reviewing files that changed from the base of the PR and between bd53304 and 30f4b24.

📒 Files selected for processing (5)

• docs/release-notes/0.16.0.md
• src/rapids_singlecell/preprocessing/_normalize.py
• src/rapids_singlecell/preprocessing/_scale.py
• tests/test_normalization.py
• tests/test_scaling.py

---

📝 Walkthrough

Walkthrough

This PR fixes two bugs in preprocessing functions: log1p now writes the transformed matrix to adata before setting adata.uns["log1p"] metadata, and scale now only writes mean/std to adata.var when obsm is not provided. Tests and release notes are added accordingly.

Changes

Preprocessing metadata fixes

Layer / File(s)	Summary
log1p side-effect ordering src/rapids_singlecell/preprocessing/_normalize.py, tests/test_normalization.py	log1p now calls _set_obs_rep before setting adata.uns["log1p"] when inplace=True; a new test confirms inplace=False doesn't mutate adata or write uns["log1p"].
scale obsm var-statistics guard src/rapids_singlecell/preprocessing/_scale.py, tests/test_scaling.py, docs/release-notes/0.16.0.md	scale only writes mean/std into adata.var when obsm is None; a new test verifies adata.var is untouched when scaling obsm, and release notes document the fix.

Estimated code review effort: 2 (Simple) | ~10 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Caller
  participant log1p
  participant AnnData

  Caller->>log1p: log1p(adata, inplace=True)
  log1p->>AnnData: _set_obs_rep(transformed matrix)
  log1p->>AnnData: uns["log1p"] = {base}
  log1p-->>Caller: return

Loading

Possibly related PRs

• scverse/rapids-singlecell#699: Modifies the same log1p and scale functions, changing how/when metadata like adata.uns["log1p"] and adata.var["mean"/"std"] are written back to AnnData.

✨ 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-meta-inconsitancies

---

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.}