Add ten estimator checks; fix bugs they uncovered; enable N rules (#1878)

Adds ten new global checks to `river.checks`: inference purity
(`check_predict_one_pure`), transformer coverage (`check_transform_one`
+ a TrumpApproval dataset for plain `base.Transformer` /
`base.SupervisedTransformer`), clone independence
(`check_clone_is_independent`), three mini-batch consistency checks
(`check_{predict,predict_proba,transform}_many_matches_*_one`),
`_get_params` ↔ `__init__` signature parity
(`check_get_params_matches_signature`), cold-start inference
(`check_predict_one_before_any_learn`), `repr` round-trip through
`clone` (`check_repr_roundtrips_clone`), `clone(new_params=...)`
override propagation (`check_clone_with_new_params_applies`),
classifier label tracking (`check_classifier_tracks_seen_labels`), and
no-aliasing-on-input (`check_no_state_aliasing_with_input`).

The existing dataset-driven checks (`check_pickling`,
`check_shuffle_features_no_impact`, the three feature-disappearance
checks, `check_seeding_is_idempotent`) now dispatch through
`_infer` / `_learn` helpers so transformers are exercised on the same
paths as classifiers, regressors, and anomaly detectors —
`test_estimators.py` collects ~780 more tests (2,655 → 3,381).

Real bugs the checks surfaced and this commit fixes:
- `cluster.TextClust.__init__` overwrote its own `micro_distance` /
  `macro_distance` parameters with runtime instances, breaking
  `clone` / `repr` round-trips. Runtime instances are now stored on
  `_micro_distance` / `_macro_distance`.
- `preprocessing.RobustScaler.transform_one` crashed with `TypeError`
  before any `learn_one` (running median returned `None`).
- `neighbors.KNN{Classifier,Regressor}`, `imblearn.HardSampling*`, and
  `tree.mondrian.MondrianTreeRegressor` stored references to the input
  feature dict in their buffers/state, so callers mutating `x` after
  `learn_one` could corrupt the model.
- `naive_bayes.BaseNB.predict_proba_many` is mis-aligned when trained
  via `learn_one` — flagged via `_unit_test_skips` on the BaseNB
  subclasses for follow-up.

Lint:
- Enabled pep8-naming (`N801`, `N802`, `N804`) in ruff. `N803` and
  `N806` are intentionally excluded because `X: pd.DataFrame` /
  `A_numpy = ...` conventions are pervasive in scientific Python.
- Renamed TextClust's internal camelCase identifiers and nested helper
  classes to snake_case / CapWords.
- **Breaking:** `drift.binary.HDDM_A` → `drift.binary.HDDMA`,
  `drift.binary.HDDM_W` → `drift.binary.HDDMW`,
  `tree.iSOUPTreeRegressor` → `tree.ISOUPTreeRegressor`.

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: MaxHalford

docs/releases/unreleased.md

@@ -1,5 +1,44 @@
 # Unreleased
 
+## checks
+
+- Added ten new global estimator checks to `river.checks`: `check_predict_one_pure` (inference methods are pure), `check_transform_one` (transform_one is exercised and returns a dict), `check_clone_is_independent` (training the original does not mutate clones), `check_predict_many_matches_predict_one` / `check_predict_proba_many_matches_predict_proba_one` / `check_transform_many_matches_transform_one` (mini-batch ↔ one-at-a-time consistency for `base.MiniBatch*` estimators), `check_get_params_matches_signature` (`_get_params()` exposes every `__init__` keyword), `check_predict_one_before_any_learn` (cold-start inference does not crash), `check_repr_roundtrips_clone` (`repr(model) == repr(model.clone())`), `check_clone_with_new_params_applies` (`clone(new_params=...)` applies the overrides), `check_classifier_tracks_seen_labels` (`predict_proba_one` includes every label observed during training), and `check_no_state_aliasing_with_input` (mutating `x` after `learn_one` does not change model state). `_yield_datasets` now also yields a dataset for plain `base.Transformer` / `base.SupervisedTransformer` estimators, which were previously skipped by the dataset-driven checks.
+- Refactored the existing dataset-driven checks (`check_pickling`, `check_shuffle_features_no_impact`, `check_emerging_features`, `check_disappearing_features`, `check_radically_disappearing_features`, `check_seeding_is_idempotent`) to dispatch through `_infer` / `_learn` helpers so transformers are exercised on the same code paths as classifiers, regressors, and anomaly detectors.
+- `checks.utils.assert_predictions_are_close` now treats two NaN floats as equivalent, so transformers that legitimately return NaN before they have observed any data (e.g. `MinMaxScaler.transform_one` on the first event) no longer trip the shuffle-invariance check.
+
+## cluster
+
+- Fixed `cluster.TextClust` corrupting its own parameters: `__init__` was overwriting `self.micro_distance` / `self.macro_distance` with runtime distance instances, breaking `clone` and `repr` round-trips. The runtime instances are now stored on `_micro_distance` / `_macro_distance`. Internal camelCase identifiers (`clusterId`, `microToMacro`, `numClusters`, `updateMacroClusters`, `_calculateIDF`) were renamed to snake_case, and the nested helper classes `tfcontainer`, `microcluster`, `distances` were renamed to `TfContainer`, `MicroCluster`, `Distances`.
+
+## drift
+
+- **Breaking:** Renamed `drift.binary.HDDM_A` → `drift.binary.HDDMA` and `drift.binary.HDDM_W` → `drift.binary.HDDMW` to comply with PEP-8 CapWords class naming.
+
+## imblearn
+
+- Fixed `imblearn.HardSamplingClassifier` / `imblearn.HardSamplingRegressor` storing references to user-supplied feature dictionaries in their buffer; the buffered triplets now hold shallow copies so callers can safely mutate `x` after `learn_one`.
+
+## naive_bayes
+
+- Marked `predict_many`/`predict_proba_many` checks as skipped on `BaseNB` subclasses (`MultinomialNB`, `BernoulliNB`, `ComplementNB`) via `_unit_test_skips`. `joint_log_likelihood_many`'s output is mis-aligned with the input batch when the model is trained via `learn_one` rather than `learn_many`, so the new mini-batch consistency checks fail. Tracked separately.
+
+## neighbors
+
+- Fixed `neighbors.KNNClassifier` / `neighbors.KNNRegressor` storing references to the input feature dicts in their search window; `learn_one` now stores a shallow copy.
+
+## preprocessing
+
+- Fixed `preprocessing.RobustScaler.transform_one` crashing with `TypeError` when called before any `learn_one` (the running median returned `None`); transform now passes the value through unchanged when centering statistics are not yet available.
+
+## tree
+
+- Fixed `tree.mondrian.MondrianTreeRegressor.learn_one` storing the input feature dict by reference on `self._x`; it now stores a shallow copy so callers can safely mutate `x` after `learn_one`. Knock-on fix for `forest.AMFRegressor`.
+- **Breaking:** Renamed `tree.iSOUPTreeRegressor` → `tree.ISOUPTreeRegressor` to comply with PEP-8 CapWords class naming.
+
+## tooling
+
+- Enabled the pep8-naming ruleset (`N801`, `N802`, `N804`) in ruff so that future class, function, and `classmethod`-first-argument naming violations are caught at lint time. `N803` (argument names) and `N806` (local variable names) were intentionally left out — `X: pd.DataFrame`, `A_numpy = ...`, and similar scientific-Python conventions are pervasive in the codebase.
+
 ## docs
 
 - Fixed corrupted markdown cells in the Hoeffding Trees notebook example that caused blank page titles and invisible sidebar navigation. Fixes [#1847](https://github.com/online-ml/river/issues/1847).

mkdocs.yml

@@ -197,8 +197,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -542,7 +542,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -746,8 +746,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -1091,7 +1091,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -1295,8 +1295,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -1640,7 +1640,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -1844,8 +1844,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -2189,7 +2189,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -2393,8 +2393,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -2738,7 +2738,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -2942,8 +2942,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -3287,7 +3287,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -3491,8 +3491,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -3836,7 +3836,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -4040,8 +4040,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -4385,7 +4385,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -4589,8 +4589,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -4933,7 +4933,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -5137,8 +5137,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -5481,7 +5481,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -5685,8 +5685,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -6029,7 +6029,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -6233,8 +6233,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -6577,7 +6577,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -6781,8 +6781,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -7125,7 +7125,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -7329,8 +7329,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -7673,7 +7673,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md
@@ -7877,8 +7877,8 @@ nav:
         - api/drift/binary/DDM.md
         - api/drift/binary/EDDM.md
         - api/drift/binary/FHDDM.md
-        - api/drift/binary/HDDM-A.md
-        - api/drift/binary/HDDM-W.md
+        - api/drift/binary/HDDMA.md
+        - api/drift/binary/HDDMW.md
       - datasets:
         - api/drift/datasets/AirlinePassengers.md
         - api/drift/datasets/Apple.md
@@ -8221,7 +8221,7 @@ nav:
       - api/tree/LASTClassifier.md
       - api/tree/SGTClassifier.md
       - api/tree/SGTRegressor.md
-      - api/tree/iSOUPTreeRegressor.md
+      - api/tree/ISOUPTreeRegressor.md
       - base:
         - api/tree/base/Branch.md
         - api/tree/base/Leaf.md

pyproject.toml

@@ -151,6 +151,14 @@ select = [
     "UP",
     # isort
     "I",
+    # pep8-naming: class names (N801), function/method names (N802), and the
+    # cls/self convention on classmethods (N804). N803 (argument names) and
+    # N806 (local variable names) are deliberately left out — `X: pd.DataFrame`
+    # and `A_numpy = ...` are conventional in ML/scientific code and would
+    # produce hundreds of false positives.
+    "N801",
+    "N802",
+    "N804",
 ]
 ignore = ["E501"]
 fixable = ["ALL"]

river/base/estimator.py

@@ -93,7 +93,7 @@ def _tags(self) -> set[str]:
         return tags
 
     @classmethod
-    def _unit_test_params(self) -> Iterator[dict[str, Any]]:
+    def _unit_test_params(cls) -> Iterator[dict[str, Any]]:
         """Indicates which parameters to use during unit testing.
 
         Most estimators have a default value for each of their parameters. However, in some cases,