Backport PR #4062 on branch 1.12.x (perf: numba based aggregations for sparse data)

[meeseeksmachine]

Backport PR #4062: perf: numba based aggregations for sparse data

[codecov]

Codecov Report

❌ Patch coverage is 97.22222% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (1.12.x@f2268d9). Learn more about missing BASE report.
⚠️ Report is 1 commits behind head on 1.12.x.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/scanpy/get/_aggregated.py	96.77%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff            @@
##             1.12.x    #4064   +/-   ##
=========================================
  Coverage          ?   78.48%           
=========================================
  Files             ?      119           
  Lines             ?    12624           
  Branches          ?        0           
=========================================
  Hits              ?     9908           
  Misses            ?     2716           
  Partials          ?        0           

Flag	Coverage Δ
hatch-test.low-vers	77.78% <97.22%> (?)
hatch-test.pre	78.36% <97.22%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/scanpy/get/_kernels.py	100.00% <100.00%> (ø)
src/scanpy/get/_aggregated.py	93.30% <96.77%> (ø)

[scverse-benchmark]

Benchmark changes

Change	Before [f2268d9] <1.12.1>	After [39a29de]	Ratio	Benchmark (Parameter)
-	4.91G	4.05G	0.82	preprocessing_counts.Agg.peakmem_agg('mean')
-	4.91G	4.05G	0.82	preprocessing_counts.Agg.peakmem_agg('sum')
-	5.83G	4.06G	0.7	preprocessing_counts.Agg.peakmem_agg('var')
-	884±2ms	640±2ms	0.72	preprocessing_counts.Agg.time_agg('count_nonzero')
-	550±2ms	90.2±0.5ms	0.16	preprocessing_counts.Agg.time_agg('mean')
-	547±0.7ms	89.0±1ms	0.16	preprocessing_counts.Agg.time_agg('sum')
-	1.37±0s	132±2ms	0.1	preprocessing_counts.Agg.time_agg('var')
-	453±3μs	387±4μs	0.85	preprocessing_counts.FastSuite.time_log1p('pbmc68k_reduced', 'counts-off-axis')

Comparison: https://github.com/scverse/scanpy/compare/f2268d9ef44ae4cd8811a0d7d7c8e7c0690b5f43..39a29de5342434e4b9317adcbf63603010bd763c
Last changed: Thu, 16 Apr 2026 17:11:40 +0000

More details: https://github.com/scverse/scanpy/pull/4064/checks?check_run_id=71665939915