[MNT] 0.15.0 deprecation actions (#3952)

Deprecation actions for the 0.15.0 release:

* changed `check_is_scitype` parameter `msg_legacy_interface` to `False`
* `ExpandingWindowSplitter` had `start_with_window` argument removed. From now on, `initial_window=0` should be used instead of `start_with_window=False`. Contains bugfix #3953 as the deprecation logic was incomplete.
* Test `test_methods_do_not_change_state` renamed to `test_non_state_changing_method_contract`
* row transformers have been removed, replaced by base class vectorization/broadcasting
* changed default for `default_fc_parameters` in `tsfresh` transformers to follow the default of `tsfresh`, which currently is `comprehensive`
* removed module `transformations.series.compose`. All contents were moved to `transformations.compose`.

Delayed to 0.16.0:
* base class of `ProbabilityThresholdEarlyClassifier` change to `BaseEarlyClassifier`. Unclear what to do here, deprecation action specified breaks the interface compliance tests. Attempt to fix here #3962 but fails `fit_idempotent` test (unclear why, will drop for this release).
* change of forecaster `cutoff` attribute to `pd.Index` type

examples/01_forecasting.ipynb

@@ -6526,8 +6526,8 @@
     "    SlidingWindowSplitter,\n",
     ")\n",
     "from sktime.forecasting.naive import NaiveForecaster\n",
+    "from sktime.transformations.compose import OptionalPassthrough\n",
     "from sktime.transformations.series.adapt import TabularToSeriesAdaptor\n",
-    "from sktime.transformations.series.compose import OptionalPassthrough\n",
     "from sktime.transformations.series.detrend import Deseasonalizer"
    ]
   },

sktime/classification/early_classification/_probability_threshold.py

@@ -22,10 +22,10 @@
 from sktime.utils.validation.panel import check_X
 
 
-# TODO: remove message in v0.15.0 and change base class
+# TODO: remove message in v0.16.0 and change base class
 @deprecated(
     version="0.13.0",
-    reason="The base class of ProbabilityThresholdEarlyClassifier will be changed to BaseEarlyClassifier in v0.15.0. This will change how classification safety decisions are made and returned, see BaseEarlyClassifier or TEASER for the new interface.",  # noqa: E501
+    reason="The base class of ProbabilityThresholdEarlyClassifier will be changed to BaseEarlyClassifier in v0.16.0. This will change how classification safety decisions are made and returned, see BaseEarlyClassifier or TEASER for the new interface.",  # noqa: E501
     category=FutureWarning,
 )
 class ProbabilityThresholdEarlyClassifier(BaseClassifier):

sktime/datatypes/_check.py

@@ -333,15 +333,14 @@ def mtype(
     return mtypes_positive[0]
 
 
-# todo 0.15.0: change msg_legacy_interface default to False
 # todo 0.16.0: remove msg_legacy_interface arg, and remove msg_legacy variable inside
 def check_is_scitype(
     obj,
     scitype: Union[str, List[str]],
     return_metadata=False,
     var_name="obj",
     exclude_mtypes=AMBIGUOUS_MTYPES,
-    msg_legacy_interface=True,
+    msg_legacy_interface=False,
 ):
     """Check object for compliance with scitype specification, return metadata.
 
@@ -461,8 +460,7 @@ def check_is_scitype(
             warn(
                 "return msg (2nd argument) of check_is_scitype will change to "
                 "dict from list type. Set msg_legacy_interface=False for "
-                "post-deprecation behaviour. Default msg_legacy_interface "
-                "will change to True in 0.15.0. Argument msg_legacy_interface "
+                "post-deprecation behaviour. Argument msg_legacy_interface "
                 "will be removed in 0.16.0.",
                 DeprecationWarning,
             )

sktime/forecasting/base/_fh.py

@@ -696,7 +696,7 @@ def _to_relative(fh: ForecastingHorizon, cutoff=None) -> ForecastingHorizon:
         # Out: Index([<0 * Hours>, <4 * Hours>, <8 * Hours>], dtype = 'object')
         # [v - periods[0] for v in periods]
         # Out: Index([<0 * Hours>, <2 * Hours>, <4 * Hours>], dtype='object')
-        # TODO: 0.15.0: Check if this comment below can be removed,
+        # TODO: 0.16.0: Check if this comment below can be removed,
         # so check if pandas has released the fix to PyPI:
         # This bug was reported: https://github.com/pandas-dev/pandas/issues/45999
         # and fixed: https://github.com/pandas-dev/pandas/pull/46006

sktime/forecasting/compose/tests/test_pipeline.py

@@ -28,10 +28,10 @@
 from sktime.forecasting.naive import NaiveForecaster
 from sktime.forecasting.sarimax import SARIMAX
 from sktime.forecasting.trend import PolynomialTrendForecaster
+from sktime.transformations.compose import OptionalPassthrough
 from sktime.transformations.hierarchical.aggregate import Aggregator
 from sktime.transformations.series.adapt import TabularToSeriesAdaptor
 from sktime.transformations.series.boxcox import LogTransformer
-from sktime.transformations.series.compose import OptionalPassthrough
 from sktime.transformations.series.detrend import Detrender
 from sktime.transformations.series.difference import Differencer
 from sktime.transformations.series.exponent import ExponentTransformer
@@ -110,8 +110,8 @@ def test_skip_inverse_transform():
 def test_nesting_pipelines():
     """Test that nesting of pipelines works."""
     from sktime.forecasting.ets import AutoETS
+    from sktime.transformations.compose import OptionalPassthrough
     from sktime.transformations.series.boxcox import LogTransformer
-    from sktime.transformations.series.compose import OptionalPassthrough
     from sktime.transformations.series.detrend import Detrender
     from sktime.utils._testing.scenarios_forecasting import (
         ForecasterFitPredictUnivariateWithX,

sktime/tests/_config.py

@@ -103,7 +103,7 @@
     # #2 amd #3 are due to predict/predict_proba returning two items and that breaking
     #   assert_array_equal
     "TEASER": [
-        "test_methods_do_not_change_state",
+        "test_non_state_changing_method_contract",
         "test_fit_idempotent",
         "test_persistence_via_pickle",
         "test_save_estimators_to_file",

sktime/tests/test_all_estimators.py

@@ -13,7 +13,6 @@
 from copy import deepcopy
 from inspect import getfullargspec, isclass, signature
 from tempfile import TemporaryDirectory
-from warnings import warn
 
 import joblib
 import numpy as np
@@ -1151,7 +1150,7 @@ def test_fit_does_not_overwrite_hyper_params(self, estimator_instance, scenario)
                 % (estimator.__class__.__name__, param_name, original_value, new_value)
             )
 
-    def test_methods_do_not_change_state(
+    def test_non_state_changing_method_contract(
         self, estimator_instance, scenario, method_nsc
     ):
         """Check that non-state-changing methods behave as per interface contract.
@@ -1164,14 +1163,6 @@ def test_methods_do_not_change_state(
             list of BaseEstimator methdos tested: get_fitted_params
             scitype specific method outputs are tested in TestAll[estimatortype] class
         """
-        warn(
-            "name of test_methods_do_not_change_state will change to "
-            "test_non_state_changing_method_contract in 0.15.0. "
-            "For a safe transition in a case where the old name "
-            "has been used as part of an argument in `check_estimator`, use "
-            "both the new and the old name in test/fixture exclusion or inclusion. ",
-            DeprecationWarning,
-        )
         estimator = estimator_instance
         set_random_state(estimator)
 

sktime/transformations/compose.py

@@ -1283,7 +1283,7 @@ class OptionalPassthrough(_DelegatedTransformer):
     --------
     >>> from sktime.datasets import load_airline
     >>> from sktime.forecasting.naive import NaiveForecaster
-    >>> from sktime.transformations.series.compose import OptionalPassthrough
+    >>> from sktime.transformations.compose import OptionalPassthrough
     >>> from sktime.transformations.series.detrend import Deseasonalizer
     >>> from sktime.transformations.series.adapt import TabularToSeriesAdaptor
     >>> from sktime.forecasting.compose import TransformedTargetForecaster
@@ -1416,7 +1416,7 @@ class ColumnwiseTransformer(BaseTransformer):
     --------
     >>> from sktime.datasets import load_longley
     >>> from sktime.transformations.series.detrend import Detrender
-    >>> from sktime.transformations.series.compose import ColumnwiseTransformer
+    >>> from sktime.transformations.compose import ColumnwiseTransformer
     >>> _, X = load_longley()
     >>> transformer = ColumnwiseTransformer(Detrender())
     >>> Xt = transformer.fit_transform(X)

sktime/transformations/panel/compose.py

@@ -4,28 +4,17 @@
 This module has meta-transformations that is build using the pre-existing
 transformations as building blocks.
 """
-from warnings import warn
-
 import numpy as np
 import pandas as pd
-from deprecated.sphinx import deprecated
 from scipy import sparse
-from sklearn.base import clone
 from sklearn.compose import ColumnTransformer as _ColumnTransformer
 
 from sktime.transformations.base import BaseTransformer, _PanelToPanelTransformer
-from sktime.transformations.series.adapt import TabularToSeriesAdaptor
 from sktime.utils.multiindex import flatten_multiindex
-from sktime.utils.sklearn import is_sklearn_estimator
 from sktime.utils.validation.panel import check_X
 
 __author__ = ["mloning", "sajaysurya", "fkiraly"]
-__all__ = [
-    "ColumnTransformer",
-    "SeriesToPrimitivesRowTransformer",
-    "SeriesToSeriesRowTransformer",
-    "ColumnConcatenator",
-]
+__all__ = ["ColumnTransformer", "ColumnConcatenator"]
 
 
 class ColumnTransformer(_ColumnTransformer, _PanelToPanelTransformer):
@@ -272,81 +261,3 @@ def _transform(self, X, y=None):
         Xt.index = pd.MultiIndex.from_arrays([inst_idx, t_idx])
         Xt.index.names = X.index.names
         return Xt
-
-
-row_trafo_deprec_msg = (
-    "All row transformers are deprecated since 0.14.0 and will be removed "
-    "in 0.15.0. Vectorization functionality from Series to Panel is natively "
-    "integrated to all transformers via the base class. Simply use fit "
-    "or transform on Panel data, no row transformer is necessary anymore."
-)
-
-
-class _RowTransformer(BaseTransformer):
-    """Base class for RowTransformer."""
-
-    _tags = {"fit_is_empty": True}
-
-    def __init__(self, transformer, check_transformer=None):
-
-        warn(row_trafo_deprec_msg)
-
-        self.transformer = transformer
-        transformer_ = clone(transformer)
-        # safer wrapping: coerce to sktime transformer
-        if is_sklearn_estimator(transformer_):
-            transformer_ = TabularToSeriesAdaptor(transformer_)
-        self.transformer_ = transformer_
-
-        self.check_transformer = check_transformer
-        super(_RowTransformer, self).__init__()
-        self.clone_tags(transformer_)
-        # fit needs to be run, or the internal fit may not be updated
-        self.set_tags(**{"fit_is_empty": False})
-
-    def _fit(self, *args, **kwargs):
-        """Fit to the data."""
-        return self.transformer_.fit(*args, **kwargs)
-
-    def _transform(self, *args, **kwargs):
-        """Transform the data."""
-        return self.transformer_.transform(*args, **kwargs)
-
-    def _inverse_transform(self, *args, **kwargs):
-        """Inverse transform the data."""
-        return self.transformer_.inverse_transform(*args, **kwargs)
-
-    def _update(self, *args, **kwargs):
-        """Update with the data."""
-        return self.transformer_.update(*args, **kwargs)
-
-    @classmethod
-    def get_test_params(cls):
-        """Return testing parameter settings for the estimator."""
-        from sktime.transformations.series.exponent import ExponentTransformer
-
-        params = {"transformer": ExponentTransformer()}
-        return params
-
-
-class SeriesToPrimitivesRowTransformer(_RowTransformer, BaseTransformer):
-    """Series-to-primitives row transformer."""
-
-
-class SeriesToSeriesRowTransformer(_RowTransformer, BaseTransformer):
-    """Series-to-series row transformer."""
-
-
-@deprecated(version="0.14.0", reason=row_trafo_deprec_msg, category=FutureWarning)
-def make_row_transformer(transformer, transformer_type=None, **kwargs):
-    """Old vectorization utility for transformers for panel data.
-
-    This is now integrated into BaseTransformer, so no longer needed.
-
-    Deprecated from version 0.14.0, will be removed in 0.15.0.
-
-    Returns
-    -------
-    transformer, reference to input `transformer` (unchanged)
-    """
-    return transformer

sktime/transformations/panel/tsfresh.py

@@ -125,17 +125,6 @@ def _get_extraction_params(self):
                 )
             else:
                 fc_parameters = fc_param_lookup[self.default_fc_parameters]
-        # TODO: remove elif in v0.15.0
-        # this will pass None to tsfresh, and hence default to "comprehensive"
-        elif self.default_fc_parameters is None:
-            fc_parameters = fc_param_lookup["efficient"]
-            warn(
-                "Passing None to default_fc_parameters currently defaults "
-                "to 'efficient', this behaviour has been is deprecated. From 0.15.0, "
-                "this will change to passing default_fc_parameters directly to tsfresh,"
-                " which in turn defaults to 'comprehensive'.",
-                DeprecationWarning,
-            )
         else:
             fc_parameters = self.default_fc_parameters
         extraction_params["default_fc_parameters"] = fc_parameters
@@ -150,7 +139,6 @@ def _get_extraction_params(self):
         return extraction_params
 
 
-# todo 0.15.0: change default_fc_parameters docstring to "comprehensive" default"
 class TSFreshFeatureExtractor(_TSFreshFeatureExtractor):
     """Transformer for extracting time series features via `tsfresh.extract_features`.
 
@@ -159,7 +147,7 @@ class TSFreshFeatureExtractor(_TSFreshFeatureExtractor):
     Parameters
     ----------
     default_fc_parameters : str, FCParameters object or None,
-        default=None = "efficient"
+        default=None = tsfresh default = "comprehensive"
         Specifies pre-defined feature sets to be extracted
         If str, should be in ["minimal", "efficient", "comprehensive"]
         See [3] for more details.
@@ -358,7 +346,6 @@ def get_test_params(cls, parameter_set="default"):
         ]
 
 
-# todo 0.15.0: change default_fc_parameters docstring to "comprehensive" default"
 class TSFreshRelevantFeatureExtractor(_TSFreshFeatureExtractor):
     """Transformer for extracting time series features via `tsfresh.extract_features`.
 
@@ -368,7 +355,7 @@ class TSFreshRelevantFeatureExtractor(_TSFreshFeatureExtractor):
     Parameters
     ----------
     default_fc_parameters : str, FCParameters object or None,
-        default=None = "efficient"
+        default=None = tsfresh default = "comprehensive"
         Specifies pre-defined feature sets to be extracted
         If str, should be in ["minimal", "efficient", "comprehensive"]
         See [3] for more details.

sktime/transformations/series/tests/test_featureizer.py

@@ -6,7 +6,7 @@
 
 from sktime.datasets import load_longley
 from sktime.forecasting.model_selection import temporal_train_test_split
-from sktime.transformations.series.compose import YtoX
+from sktime.transformations.compose import YtoX
 from sktime.transformations.series.exponent import ExponentTransformer
 from sktime.transformations.series.lag import Lag
 

sktime/transformations/series/tests/test_optional_passthrough.py

@@ -5,8 +5,8 @@
 import pytest
 from pandas.testing import assert_series_equal
 
+from sktime.transformations.compose import OptionalPassthrough
 from sktime.transformations.series.boxcox import BoxCoxTransformer
-from sktime.transformations.series.compose import OptionalPassthrough
 from sktime.utils._testing.series import _make_series
 
 

sktime/transformations/tests/test_compose.py

@@ -13,11 +13,11 @@
 from sktime.transformations.compose import (
     FeatureUnion,
     InvertTransform,
+    OptionalPassthrough,
     TransformerPipeline,
 )
 from sktime.transformations.panel.padder import PaddingTransformer
 from sktime.transformations.series.boxcox import LogTransformer
-from sktime.transformations.series.compose import OptionalPassthrough
 from sktime.transformations.series.exponent import ExponentTransformer
 from sktime.transformations.series.impute import Imputer
 from sktime.transformations.series.subset import ColumnSelect

sktime/transformations/tests/test_optionalpassthrough.py

@@ -24,8 +24,8 @@ def test_optionalpassthrough():
         SlidingWindowSplitter,
     )
     from sktime.forecasting.naive import NaiveForecaster
+    from sktime.transformations.compose import OptionalPassthrough
     from sktime.transformations.series.adapt import TabularToSeriesAdaptor
-    from sktime.transformations.series.compose import OptionalPassthrough
     from sktime.transformations.series.detrend import Deseasonalizer
 
     # create pipeline