fix dense int input

[Intron7]

Summary by CodeRabbit

• Bug Fixes

  • Integer-valued dense inputs are now automatically converted to floating point for normalization and scaling, preventing dtype-related issues and improving result consistency.
  • Normalization now preserves expected totals after conversion, and scaling produces correctly centered and scaled output for integer inputs.

• Tests

  • Added coverage for dense integer input handling in normalization and scaling.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.31%. Comparing base (481c60c) to head (383b239).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #718   +/-   ##
=======================================
  Coverage   90.31%   90.31%           
=======================================
  Files         106      106           
  Lines        9581     9585    +4     
=======================================
+ Hits         8653     8657    +4     
  Misses        928      928           

Files with missing lines	Coverage Δ
src/rapids_singlecell/preprocessing/_normalize.py	89.13% <100.00%> (+0.09%)	⬆️
src/rapids_singlecell/preprocessing/_scale.py	91.71% <100.00%> (+0.09%)	⬆️

[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: 760e2e58-7fe9-4ad2-b40c-5ffb3de10a7e

📥 Commits

Reviewing files that changed from the base of the PR and between 908e02b and b931f0d.

📒 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 modifies normalize_total and _scale_dispatch to detect integer-dtype dense inputs and cast them to floating point (cp.float32 and cp.float64 respectively) before processing. Corresponding tests and a release-note entry were added.

Changes

Dense integer promotion

Layer / File(s)	Summary
normalize_total integer cast src/rapids_singlecell/preprocessing/_normalize.py, tests/test_normalization.py	Integer X is cast to cp.float32 before normalization, skipping the prior conditional copy in that path; a new test verifies dtype promotion and correct row sums.
scale integer cast src/rapids_singlecell/preprocessing/_scale.py, tests/test_scaling.py	_scale_dispatch upcasts integer X to cp.float64 before backend dispatch; a new parametrized test verifies dtype promotion and centered output.
Release notes docs/release-notes/0.16.0.md	A bug-fix bullet documents the promotion of dense integer inputs to floating point for both functions.

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

Sequence Diagram(s)

Not applicable — changes are localized dtype-check additions without multi-component interaction flows.

Suggested reviewers: Intramuros, flying-sheep

Poem
A rabbit hopped through code so neat,
found ints where floats should meet,
cast them up to float32, float64 too,
now sums add right, as they should do,
a tidy patch, a release note complete. 🐇

✨ 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 promote-int-to-float

---

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.}