[MNT] add differential testing to utils module (#6620)

This PR adds differntial testing decorators in the `utils` module.

sktime/utils/estimators/tests/test_base.py

@@ -10,6 +10,7 @@
 from sktime.datasets import load_airline
 from sktime.forecasting.base import BaseForecaster
 from sktime.forecasting.naive import NaiveForecaster
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.transformations.base import BaseTransformer
 from sktime.transformations.series.boxcox import BoxCoxTransformer
 from sktime.utils.estimators import make_mock_estimator
@@ -18,6 +19,10 @@
 y_series = load_airline().iloc[:-5]
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 @pytest.mark.parametrize(
     "base", [BaseForecaster, BaseClassifier, BaseClusterer, BaseTransformer]
 )
@@ -46,6 +51,10 @@ def _score(self):
     assert hasattr(dummy_instance, "_MockEstimatorMixin__log")
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 def test_add_log_item():
     """Test _MockEstimatorMixin.add_log_item behaviour."""
     mixin = _MockEstimatorMixin()
@@ -55,6 +64,10 @@ def test_add_log_item():
     assert mixin.log[1] == 2
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 def test_log_is_property():
     """Test _MockEstimatorMixin.log can't be overwritten."""
     mixin = _MockEstimatorMixin()
@@ -63,6 +76,10 @@ def test_log_is_property():
         assert "can't set attribute" in str(excinfo.value)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 def test_method_logger_exception():
     """Test that _method_logger only works for _MockEstimatorMixin subclasses."""
 
@@ -82,6 +99,10 @@ def _method(self):
         assert "Estimator is not a Mock Estimator" in str(excinfo.value)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 def test_method_logger():
     """Test that method logger returns the correct output."""
 
@@ -143,6 +164,10 @@ def _method3(self):
     ]
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 @pytest.mark.parametrize(
     "estimator_class, method_regex, logged_methods",
     [
@@ -161,6 +186,10 @@ def test_make_mock_estimator(estimator_class, method_regex, logged_methods):
     assert set(methods_called) >= set(logged_methods)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed("sktime.utils.estimators"),
+    reason="Run test if estimators module has changed.",
+)
 @pytest.mark.parametrize(
     "estimator_class, estimator_kwargs",
     [

sktime/utils/numba/tests/test_general.py

@@ -6,13 +6,15 @@
 import pytest
 from numpy.testing import assert_array_equal
 
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils.dependencies import _check_soft_dependencies
 
 DATATYPES = ["int32", "int64", "float32", "float64"]
 
 
 @pytest.mark.skipif(
-    not _check_soft_dependencies("numba", severity="none"),
+    not run_test_module_changed(["sktime.utils.numba"])
+    or not _check_soft_dependencies("numba", severity="none"),
     reason="skip test if required soft dependency not available",
 )
 @pytest.mark.parametrize("type", DATATYPES)

sktime/utils/sklearn/tests/test_sklearn_df_adapt.py

@@ -6,9 +6,14 @@
 import pandas as pd
 import pytest
 
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils.sklearn._adapt_df import prep_skl_df
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.sklearn"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("copy_df", [True, False])
 def test_prep_skl_df_coercion(copy_df):
     """Test that prep_skl_df behaves correctly on the coercion case."""
@@ -22,6 +27,10 @@ def test_prep_skl_df_coercion(copy_df):
         assert res is mixed_example
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.sklearn"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("copy_df", [True, False])
 def test_prep_skl_df_non_coercion(copy_df):
     """Test that prep_skl_df behaves correctly on the non-coercion case."""

sktime/utils/sklearn/tests/test_sklearn_typing.py

@@ -10,6 +10,7 @@
 
 from sktime.classification.feature_based import SummaryClassifier
 from sktime.forecasting.naive import NaiveForecaster
+from sktime.tests.test_switch import run_test_for_class
 from sktime.utils.sklearn import is_sklearn_estimator, sklearn_scitype
 
 CORRECT_SCITYPES = {
@@ -23,6 +24,10 @@
 sktime_estimators = [SummaryClassifier, NaiveForecaster]
 
 
+@pytest.mark.skipif(
+    not run_test_for_class(is_sklearn_estimator),
+    reason="Run if utilities have changed.",
+)
 @pytest.mark.parametrize("estimator", sklearn_estimators)
 def test_is_sklearn_estimator_positive(estimator):
     """Test that is_sklearn_estimator recognizes positive examples correctly."""
@@ -33,6 +38,10 @@ def test_is_sklearn_estimator_positive(estimator):
     assert is_sklearn_estimator(estimator), msg
 
 
+@pytest.mark.skipif(
+    not run_test_for_class(is_sklearn_estimator),
+    reason="Run if utilities have changed.",
+)
 @pytest.mark.parametrize("estimator", sktime_estimators)
 def test_is_sklearn_estimator_negative(estimator):
     """Test that is_sklearn_estimator recognizes negative examples correctly."""
@@ -43,6 +52,10 @@ def test_is_sklearn_estimator_negative(estimator):
     assert not is_sklearn_estimator(estimator), msg
 
 
+@pytest.mark.skipif(
+    not run_test_for_class(sklearn_scitype),
+    reason="Run if utilities have changed.",
+)
 @pytest.mark.parametrize("estimator", sklearn_estimators)
 def test_sklearn_scitype(estimator):
     """Test that sklearn_scitype returns the correct scitype string."""

sktime/utils/tests/test_datetime.py

@@ -6,10 +6,12 @@
 
 import numpy as np
 import pandas as pd
+import pytest
 
 from sktime.datasets import load_airline
 from sktime.datatypes import VectorizedDF
 from sktime.datatypes._utilities import get_time_index
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils._testing.hierarchical import _bottom_hier_datagen
 from sktime.utils.datetime import (
     _coerce_duration_to_int,
@@ -20,6 +22,10 @@
 from sktime.utils.dependencies import _check_soft_dependencies
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils", "sktime.datatypes"]),
+    reason="Run if utils or datatypes module has changed.",
+)
 def test_get_freq():
     """Test whether get_freq runs without error."""
     x = pd.Series(
@@ -45,6 +51,10 @@ def test_get_freq():
     assert _get_freq(x4) is None
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils", "sktime.datatypes"]),
+    reason="Run if utils or datatypes module has changed.",
+)
 def test_coerce_duration_to_int() -> None:
     """Test _coerce_duration_to_int."""
     assert _coerce_duration_to_int(duration=0) == 0
@@ -69,6 +79,10 @@ def test_coerce_duration_to_int() -> None:
     )
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils", "sktime.datatypes"]),
+    reason="Run if utils or datatypes module has changed.",
+)
 def test_infer_freq() -> None:
     """Test frequency inference."""
     assert infer_freq(None) is None
@@ -97,6 +111,10 @@ def test_infer_freq() -> None:
     assert infer_freq(y) in ["M", "ME"]
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils", "sktime.datatypes"]),
+    reason="Run if utils or datatypes module has changed.",
+)
 def test_set_freq_hier():
     """Test that setting frequency on a DatetimeIndex MultiIndex works."""
     # from pandas 2.1.0 on, freq is preserved correctly,

sktime/utils/tests/test_multiindex.py

@@ -3,9 +3,14 @@
 import pandas as pd
 import pytest
 
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils.multiindex import flatten_multiindex, rename_multiindex
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.multiindex"]),
+    reason="Run if multiindex module has changed.",
+)
 def test_flatten_multiindex():
     """Test flatten_multiindex contract."""
     mi = pd.MultiIndex.from_product([["a", "b"], [0, 42]])
@@ -18,6 +23,10 @@ def test_flatten_multiindex():
     assert (expected == flat).all()
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.multiindex"]),
+    reason="Run if multiindex module has changed.",
+)
 def test_rename_multiindex():
     """Test rename_multiindex contract."""
     mi = pd.MultiIndex.from_tuples([("a", 1), ("a", 42), ("b", 1), ("c", 0)])

sktime/utils/tests/test_sampling.py

@@ -5,12 +5,17 @@
 
 from sktime.datasets import load_unit_test
 from sktime.datatypes import check_is_scitype
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils.sampling import random_partition, stratified_resample
 
 NK_FIXTURES = [(10, 3), (15, 5), (19, 6), (3, 1), (1, 2)]
 SEED_FIXTURES = [42, 0, 100, -5]
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.sampling", "sktime.datatypes"]),
+    reason="Run if multiindex module has changed.",
+)
 @pytest.mark.parametrize("n, k", NK_FIXTURES)
 def test_partition(n, k):
     """Test that random_partition returns a disjoint partition."""
@@ -35,6 +40,10 @@ def test_partition(n, k):
                 assert len(set(x).intersection(y)) == 0
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.sampling", "sktime.datatypes"]),
+    reason="Run if multiindex module has changed.",
+)
 @pytest.mark.parametrize("seed", SEED_FIXTURES)
 @pytest.mark.parametrize("n, k", NK_FIXTURES)
 def test_seed(n, k, seed):
@@ -47,6 +56,10 @@ def test_seed(n, k, seed):
     assert deep_equals(part, part2)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils.sampling", "sktime.datatypes"]),
+    reason="Run if multiindex module has changed.",
+)
 def test_stratified_resample():
     """Test resampling returns valid data structure and maintains class distribution."""
     trainX, trainy = load_unit_test(split="TRAIN")

sktime/utils/tests/test_seasonal.py

@@ -6,10 +6,15 @@
 import pandas as pd
 import pytest
 
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils._testing.series import _make_series
 from sktime.utils.seasonality import _pivot_sp, _unpivot_sp
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("n_timepoints", [49, 1])
 @pytest.mark.parametrize("index_type", ["period", "datetime", "range", "int"])
 @pytest.mark.parametrize("sp", [2, 10])
@@ -53,6 +58,10 @@ def test_pivot_sp(sp, index_type, n_timepoints, anchor_side):
             assert not np.isnan(df_pivot.iloc[-1, -1])
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("n_timepoints", [49, 1])
 @pytest.mark.parametrize("index_type", ["period", "datetime", "range", "int"])
 @pytest.mark.parametrize("sp", [2, 10])
@@ -81,6 +90,10 @@ def test_unpivot_sp(sp, index_type, n_timepoints, anchor_side):
     assert np.all(df_unpivot == df)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("n_timepoints", [50, 2])
 @pytest.mark.parametrize("index_type", ["period", "datetime", "range", "int"])
 @pytest.mark.parametrize("sp", [3, 10])

sktime/utils/tests/test_utils_time_series.py

@@ -3,10 +3,15 @@
 import pytest
 from scipy.stats import linregress
 
+from sktime.tests.test_switch import run_test_module_changed
 from sktime.utils._testing.forecasting import _generate_polynomial_series
 from sktime.utils.slope_and_trend import _fit_trend, _slope
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("trend_order", [0, 3])
 def test_time_series_slope_against_scipy_linregress(trend_order):
     """Test time series slope against scipy lingress."""
@@ -21,6 +26,10 @@ def test_time_series_slope_against_scipy_linregress(trend_order):
 
 
 # Check linear and constant cases
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("slope", [-1, 0, 1])
 def test_time_series_slope_against_simple_cases(slope):
     """Test time series slope against simple cases."""
@@ -29,6 +38,10 @@ def test_time_series_slope_against_simple_cases(slope):
     np.testing.assert_almost_equal(_slope(y), slope, decimal=10)
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.utils"]),
+    reason="Run if utils module has changed.",
+)
 @pytest.mark.parametrize("order", [0, 1, 2])  # polynomial order
 @pytest.mark.parametrize("n_timepoints", [1, 10])  # number of time series observations
 @pytest.mark.parametrize("n_instances", [1, 10])  # number of samples

sktime/utils/validation/tests/test_check_window_length.py

@@ -2,9 +2,14 @@
 """Tests for window length."""
 import pytest
 
+from sktime.tests.test_switch import run_test_for_class
 from sktime.utils.validation import check_window_length
 
 
+@pytest.mark.skipif(
+    not run_test_for_class(check_window_length),
+    reason="Run if tested function has changed.",
+)
 @pytest.mark.parametrize(
     "window_length, n_timepoints, expected",
     [
@@ -22,6 +27,10 @@ def test_check_window_length(window_length, n_timepoints, expected):
     assert check_window_length(window_length, n_timepoints) == expected
 
 
+@pytest.mark.skipif(
+    not run_test_for_class(check_window_length),
+    reason="Run if tested function has changed.",
+)
 @pytest.mark.parametrize(
     "window_length, n_timepoints",
     [