[thirumurugan-2006]

<!--
Thanks for opening a PR to scanpy!
Please be sure to follow the guidelines in our contribution guide (https://scanpy.scverse.org/page/dev/) to familiarize yourself with our workflow and speed up review.
-->

LLM usage

• LLM or coding agent used: GitHub Copilot via VS Code

• If one was used, I, a human, have reviewed, understood and tested every changed line

• Closes highly_variable_genes(flavor="seurat_v3") flags more than n_top_genes genes when var_names contain duplicates #4405

• Tests included

• Release notes not necessary because: this is a focused bug fix with no API change

• No milestones

Summary

scanpy.pp.highly_variable_genes(flavor="seurat_v3") could select more than n_top_genes when adata.var_names contained duplicate gene names.

The previous implementation used label-based .loc assignment after sorting. Duplicate labels could cause multiple rows to be marked as highly variable, exceeding the requested number.

This change uses positional row selection for the final assignment, ensuring that selection is based on individual rows rather than gene-name labels.

Tests

• Added regression coverage for unique and duplicate gene names.
• Verified that both cases select exactly 30 highly variable genes when n_top_genes=30.
• Ran the relevant seurat_v3 and selection-consistency tests.
• Ran Ruff linting and formatting checks.
• Ran git diff --check and Python syntax checks.

Release notes

No release note is necessary because this is a focused bug fix with no public API change.

[thirumurugan-2006]

can i merge it

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 0.00%. Comparing base (0f9179e) to head (450357a).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #4417       +/-   ##
==========================================
- Coverage   83.78%       0   -83.79%     
==========================================
  Files         134       0      -134     
  Lines       12691       0    -12691     
==========================================
- Hits        10633       0    -10633     
+ Misses       2058       0     -2058     

Flag	Coverage Δ
hatch-test.low-vers	?
hatch-test.pre	?

Flags with carried forward coverage won't be shown. Click here to find out more.
see 134 files with indirect coverage changes

[thirumurugan-2006]

is the benchmark will take longer time

[thirumurugan-2006]

The failing CI run appears to be unrelated to my documentation changes. The failure is in tests/test_preprocessing_distributed.py::test_write_zarr, where an AnnData/Dask/Zarr auto-sharding warning is promoted to an error. Since this run is two weeks old, could you confirm whether the issue persists on the current branch or whether the latest CI result can be used for merge eligibility?