[timtreis]

test_evaluate_integration_baseline_solver_auto_matches_arpack flaked because of kBET, not the PCA solver. scib's kbet_per_label calls scipy.sparse.linalg.eigsh with an unseeded random start vector. On some runners, 3 of 60 unseeded calls on the same kNN graph gave 0.991667 where the rest gave 0.941667. Start-vector seeds 4, 11 and 19 always give 0.991667.

The test now patches eigsh with a fixed v0, so both runs use the same start vector. The tolerance stays at 1e-4. The two solvers' baselines agree to a subspace angle below 1e-5.

Verified on CI with a temporary diagnostic, since removed: on all three hatch envs, a fixed start vector gave the same kBET in 60 of 60 calls, and the patched test failed 0 of 8 times. tests/test_metrics.py, ruff, ruff format and mypy pass locally.

[read-the-docs-community]

Documentation build overview

📚 mantispy | 🛠️ Build #35898918 | 📁 Comparing 0bc66c3 against latest (d19f209)

  🔍 Preview build  

1 file changed

± api/generated/mantispy.settings.html

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.60%. Comparing base (d19f209) to head (0bc66c3).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #337   +/-   ##
=======================================
  Coverage   79.60%   79.60%           
=======================================
  Files          97       97           
  Lines       12354    12354           
=======================================
  Hits         9835     9835           
  Misses       2519     2519           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.