Sparse test

[henrydingliu]

Summary of Changes

Related GitHub Issue(s)

Additional Context for Reviewers

• I passed tests locally for both code (uv run pytest) and documentation changes (uv run jb build docs --builder=custom --custom-builder=doctest)

---

Note

Medium Risk
Changes Mack standard-error recursion and sparse reduction helpers used in production paths, not only tests; incorrect backend alignment could affect numerical results on sparse triangles.

Overview
Improves sparse-array test coverage and fixes backend mismatches when triangles mix numpy and sparse after auto_sparse or reductions.

Test harness: Sample fixtures now set backend per triangle via set_backend() instead of toggling global ARRAY_BACKEND. The full prism fixture runs sparse-only; a smaller prism_convert slice backs the auto-sparse test.

Sparse utilities: nan_to_num accepts a configurable NaN fill value. New nanquantile / nanmedian on COO arrays support median/quantile on sparse-backed triangles.

Runtime fixes: Mack recursion and summary_ force matching backends before sums and concatenation (e.g. after sum triggers auto-sparse). DevelopmentML._prep_w_ml densifies weights for scikit-learn. Several tests assert Triangle equality or numpy arrays instead of backend-specific xp.testing.

Test tweaks: Cape Cod zero-triangle uses .at; Mack malformed test uses the raa fixture.

^{Reviewed by Cursor Bugbot for commit ac9e1b5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want reviews to match your repository better? Bugbot Learning can learn team-specific rules from PR activity. A team admin can enable Learning in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit ac9e1b5. Configure here.}

[cursor]

Empty sparse quantile crashes

Medium Severity

When a multi-dimensional COO has no stored values (nnz == 0), group_ids is empty but the loop still runs with starts=[0], so group_ids[start] raises IndexError. An all-missing reduction should return NaNs, matching np.nanquantile.

 

^{Reviewed by Cursor Bugbot for commit ac9e1b5. Configure here.}

[github-actions]

Pyright Type Completeness

View the full pyright --verifytypes output for this commit

Project (full chainladder package, at this PR's head): 15.1% of exported symbols fully typed (195 / 1294)

	Known	Ambiguous	Unknown	Total
Project (head)	195	110	989	1294

Other symbols referenced but not exported by chainladder: 13

	Known	Ambiguous	Unknown	Total
Other (head)	3	1	9	13

Symbols without documentation:

• Functions without docstring: 315
• Functions without default param: 0
• Classes without docstring: 10

Patch (exported symbols added or changed by this PR): 28.6% fully typed (2 / 7); 2 no longer exported

	Known	Ambiguous	Unknown	Total
Patch	2	0	5	7

Patch symbol details

Symbol	Status	Change
chainladder.utils.sparse.nanmedian	❌ unknown	new
chainladder.utils.sparse.nanquantile	❌ unknown	new
chainladder.utils.tests.test_sparse.test_1D_nanquantile	✅ known	new
chainladder.utils.tests.test_sparse.test_keepdims_nanquantile	✅ known	new
chainladder.utils.tests.test_utilities.test_maximum_2	❌ unknown	new
chainladder.utils.tests.test_utilities.test_maximum_minimum_1	❌ unknown	new
chainladder.utils.tests.test_utilities.test_minimum_2	❌ unknown	new
chainladder.utils.tests.test_utilities.test_maximum	—	no longer exported (was ❌ unknown)
chainladder.utils.tests.test_utilities.test_minimum	—	no longer exported (was ❌ unknown)