yohou.testing

Systematic check functions for testing custom estimators. Use these to validate that your forecasters, transformers, scorers, and splitters conform to Yohou's API contracts.

Forecaster checks

Name	Description
check_clone_preserves_forecaster_params	Check sklearn's clone() preserves init parameters.
check_fit_sets_forecaster_attributes	Check fit() sets required forecaster attributes.
check_fit_predict_without_exogenous	Check forecaster behavior when X=None at fit time.
check_forecaster_methods_call_check_is_fitted	Check all forecaster methods (except fit) raise NotFittedError when unfitted.
check_forecaster_not_fitted_error	Check accessing fitted attributes before fit() raises NotFittedError.
check_forecaster_tags_accessible_before_fit	Check __sklearn_tags__() is accessible before fit().
check_forecaster_tags_match_capabilities	Check forecaster tags accurately reflect capabilities.
check_forecaster_tags_static_after_fit	Check forecaster tags remain static after fit().
check_forecasting_horizon_validation	Check forecasting_horizon < 1 raises ValueError.
check_observe_extends_observations	Check observe() extends observation buffers correctly.
check_predict_time_columns	Check predictions have vintage_time and time columns.
check_prediction_types_property	Check forecaster_type tag is set correctly.
check_rewind_propagates_to_transformers	Check rewind() propagates to transformers in forecaster.
check_rewind_replaces_observations	Check rewind() replaces observation buffers correctly.

Point forecaster checks

Name	Description
check_point_prediction_structure	Check point predictions have correct column structure.
check_point_prediction_types	Check point forecaster has 'point' in forecaster_type tag.

Interval forecaster checks

Name	Description
check_coverage_rates_parameter	Check coverage_rates is list of floats in (0, 1).
check_coverage_rates_validation	Check invalid coverage_rates raise ValueError during fit and predict.
check_interval_bounds	Check upper >= lower for all coverage rates and time steps.
check_interval_prediction_columns	Check interval predictions have {col}lower} and {colupper format.
check_interval_prediction_types	Check interval forecaster has 'interval' in forecaster_type tag.

Reduction forecaster checks

Name	Description
check_estimator_parameter	Check estimator parameter is sklearn BaseEstimator.
check_reduction_strategy	Check reduction_strategy parameter is valid.

Panel data checks

Name	Description
check_panel_data	Check cross-learning with panel data predicts all groups by default.
check_panel_invalid_group_raises	Check that invalid panel_group raises ValueError.
check_panel_single_group	Check cross-learning filters to specified panel group.

Transformer checks

Name	Description
check_fit_idempotent	Check that fit(X).fit(X) equals fit(X).
check_fit_sets_attributes	Check fit() sets required attributes.
check_fit_transform_equivalence	Check fit_transform(X) == fit(X).transform(X).
check_feature_names_out_match	Check get_feature_names_out() matches transform() output columns.
check_transform_output_structure	Check transform() output has "time" column and valid structure.
check_transform_drops_warmup_rows	Check stateful transformers drop exactly observation_horizon rows.
check_transformer_methods_call_check_is_fitted	Check all transformer methods (except fit) raise NotFittedError when unfitted.
check_transformer_preserve_dtypes	Check transformer preserves input dtypes.
check_transformers_unfitted_stateless	Check stateless transformers work without fitting.
check_observation_horizon_after_fit	Check observation_horizon is valid after fit().
check_observation_horizon_not_fitted	Check accessing observation_horizon before fit() raises NotFittedError.
check_observe_concatenates_memory	Check observe() appends new data and maintains horizon size.
check_observe_transform_equivalence	Check observe().transform() == fit().transform() for same final state.
check_observe_transform_sequential_consistency	Check observe_transform(A) then observe_transform(B) == observe_transform(A+B).
check_rewind_transform_behavior	Check rewind_transform() behavior and contract.
check_rewind_updates_memory	Check rewind() updates _X_observed to last observation_horizon rows.
check_memory_bounded	Check memory doesn't grow unbounded with sequential updates.
check_insufficient_data_raises	Check behavior when data length < observation_horizon.
check_inverse_transform_identity	Check inverse_transform(transform(X)) ≈ X.
check_inverse_transform_round_trip	Check inverse_transform(transform(X)) ≈ X with shape validation.
check_inverse_observe_transform_identity	Check inverse_transform(observe_transform(X)) ≈ X.
check_panel_data_support	Check transformer handles panel columns (panel data) correctly.
check_panel_group_preservation	Check that transformers preserve panel group names after transformation.
check_tags_accessible_before_fit	Check __sklearn_tags__() is accessible before fit().
check_tags_match_capabilities	Check tags accurately reflect transformer capabilities.
check_tags_static_after_fit	Check tags remain static (don't change) after fit().

Scorer checks

Name	Description
check_scorer_aggregation_methods	Check all aggregation_method combinations produce valid output.
check_scorer_component_subselection	Check components filtering works correctly.
check_scorer_coverage_rate_subselection	Check coverage_rates parameter filters interval predictions correctly.
check_scorer_lower_is_better	Check lower_is_better convention matches scoring direction.
check_scorer_methods_call_check_is_fitted	Check all scorer methods (except fit) raise NotFittedError when unfitted.
check_scorer_panel_subselection	Check groups filtering works correctly.
check_scorer_parameter_validation	Check parameter validation raises ValueError for invalid inputs.
check_scorer_prediction_type_compatibility	Check scorer works with correct forecaster output type.
check_scorer_tags_accessible_before_fit	Check __sklearn_tags__() is callable on scorer instance.
check_scorer_tags_match_capabilities	Check tag values match actual scorer behavior.
check_scorer_tags_static_after_fit	Check tags remain unchanged after fit.

Splitter checks

Name	Description
check_splitter_n_splits_consistency	Check get_n_splits() matches actual split count.
check_splitter_non_overlapping_tests	Check test sets don't overlap if produces_non_overlapping_tests=True.
check_splitter_panel_data_support	Check splitter handles panel data if supports_panel_data=True.
check_splitter_parameter_constraints	Check parameter constraints are enforced via sklearn validation.
check_splitter_produces_valid_indices	Check all train/test indices are valid row positions.
check_splitter_tags_accessible_before_fit	Check __sklearn_tags__() is callable on splitter instance.
check_splitter_tags_match_capabilities	Check tag values match actual splitter behavior.
check_splitter_tags_static_after_fit	Check tags remain unchanged after fit.

Search checks

Name	Description
check_search_clone_preserves_params	Check sklearn clone() preserves search CV parameters.
check_search_fit_sets_attributes	Check fit() sets required search CV attributes.
check_search_not_fitted_error	Check accessing fitted attributes before fit() raises NotFittedError.
check_search_cv_results_structure	Check cv_results_ has required structure.
check_search_method_availability	Check @available_if decorator logic with refit=True/False.
check_search_predict_delegates	Check predict() delegates to best_forecaster_.predict() correctly.
check_search_observe_delegates	Check observe() delegates to best_forecaster_.observe() correctly.
check_search_rewind_delegates	Check rewind() delegates to best_forecaster_.rewind() correctly.
check_search_multimetric_scoring	Check multi-metric scoring with dict scorer works correctly.
check_search_error_score_handling	Check error_score parameter handles failing fits correctly.
check_search_refit_false_no_forecaster	Check refit=False doesn't create best_forecaster_.
check_search_return_train_score	Check return_train_score=True adds train score keys to cv_results_.
check_search_panel_data	Check groups parameter propagates correctly.
check_grid_search_exhaustive	Check GridSearchCV evaluates all parameter combinations.
check_grid_search_param_grid_validation	Check param_grid format is validated (dict or list of dicts).
check_randomized_search_distributions	Check scipy.stats distributions work for parameter sampling.
check_randomized_search_n_iter	Check n_iter controls number of parameter combinations evaluated.
check_randomized_search_reproducibility	Check random_state produces same parameter samples.

Metadata routing

Name	Description
check_metadata_routing_default_request	Check that by default metadata routing request is empty.
check_metadata_routing_get_metadata_routing	Check that get_metadata_routing() is implemented correctly.
assert_request_equal	Assert metadata request matches expected dictionary.
assert_request_is_empty	Check if a metadata request dict is empty.
check_recorded_metadata	Check whether the expected metadata is passed to the object's method.
record_metadata	Utility function to store passed metadata to a method of obj.