FIX Raise TypeError when sparse input is not supported

[betatim]

This makes cuml more compatible with scikit-learn by raising the exception type that it expects.

Tackling one of the many xfail'ed checks in the compatibility testing.

This changes the type from NotImplementedError to TypeError, which is technically a breaking change. However, I don't think this will effect many people as it is only a problem for those who are passing sparse data to an estimator that doesn't support it and then rely on catching the exception. I can't think of a way to handle this via a "deprecation cycle".

It is a small change that fixes several estimators are the same time and gives us a better overview of those that truly fail this check.

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes

  • Passing sparse inputs now raises a clear TypeError indicating dense data is required.

• Tests

  • Compatibility tests updated: many prior sparse expected-fail markers removed so sparse checks now run; two projection estimators’ expected messages adjusted to reflect observed ValueError behavior on small/sparse inputs.

Walkthrough

Sparse-input handling in internals changed: from_input now raises TypeError with an explicit "sparse matrix passed, dense required" message; tests updated accordingly. Many per-estimator xfail entries for sklearn sparse-data compatibility were removed and two random-projection xfail messages were adjusted.

Changes

Cohort / File(s)	Summary
Array & exception tests python/cuml/cuml/internals/array.py, python/cuml/tests/test_exceptions.py	Changed sparse-input error from NotImplementedError to TypeError with a message indicating sparse matrices are unsupported; updated tests to expect TypeError and match on "sparse".
Sklearn compatibility tests python/cuml/tests/test_sklearn_compatibility.py	Removed many check_estimator_sparse_tag / sparse-related xfail entries from PER_ESTIMATOR_XFAIL_CHECKS; adjusted GaussianRandomProjection and SparseRandomProjection entries to map sparse-tag failures to ValueError messages about small-dataset failures.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• CI Switch to strict xfail policy for common estimator checks #7753 — Overlapping edits to sklearn compatibility xfail mappings in test_sklearn_compatibility.py.
• CI Update xfail list #7768 — Also modifies PER_ESTIMATOR_XFAIL_CHECKS entries; related adjustments to xfail handling.

Suggested reviewers

• csadorf
• viclafargue
• jcrist

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'FIX Raise TypeError when sparse input is not supported' directly and clearly summarizes the main change: converting the exception type from NotImplementedError to TypeError for sparse input handling.
Description check	✅ Passed	The description explains the rationale for the change (scikit-learn compatibility), acknowledges it as a breaking change, notes the impact scope, and describes the benefit of fixing multiple estimators simultaneously.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

No actionable comments were generated in the recent review. 🎉

🧹 Recent nitpick comments

python/cuml/tests/test_sklearn_compatibility.py (1)

572-572: Consider aligning the SparseRandomProjection rationale with GaussianRandomProjection's.

Line 543 says "...raises ValueError with small datasets even though it supports sparse inputs" which clearly explains why check_estimator_sparse_tag fails despite sparse being supported. Line 572 omits the "even though it supports sparse inputs" clarification. Since SparseRandomProjection also inherits SparseInputTagMixin and declares sparse support, the same tag-behavior nuance applies.

Proposed fix

-        "check_estimator_sparse_tag": "SparseRandomProjection raises ValueError with small datasets",
+        "check_estimator_sparse_tag": "SparseRandomProjection raises ValueError with small datasets even though it supports sparse inputs",

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

---

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 and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@python/cuml/tests/test_sklearn_compatibility.py`:
- Around line 702-704: The xfail rationale masks a tag-behavior mismatch:
scikit-learn's check_estimator_sparse_tag detects that input_tags.sparse=True
(via SparseInputTagMixin on GaussianRandomProjection and SparseRandomProjection)
but fit() rejects CSR sparse input; update the PR to either remove
SparseInputTagMixin (so input_tags.sparse is not declared) if sparse inputs
truly aren't supported, or amend the xfail rationale text to explicitly state
that the failure is due to the mismatch between input_tags.sparse and actual
fit() behavior (i.e., tag declares sparse support but fit() raises ValueError
for small datasets) so reviewers know the check is catching a tag inconsistency
rather than just the Johnson–Lindenstrauss small-dataset constraint.

[jcrist]

/merge