microsoft · debonte · Jun 20, 2025 · Jun 20, 2025 · Jun 20, 2025 · Jun 20, 2025
diff --git a/pyproject.toml b/pyproject.toml
@@ -26,7 +26,7 @@ tests = [
     # Needed for stubtest and downloads their dependencies to get known import symbols
     "networkx",
     "scikit-image",
-    "scikit-learn",
+    "scikit-learn <1.7.0", # TODO: Update stubs for sklearn
     "sympy",
     "vispy",
 ]

diff --git a/stubs/networkx/algorithms/approximation/dominating_set.pyi b/stubs/networkx/algorithms/approximation/dominating_set.pyi
@@ -1,10 +1,6 @@
 from ...classes.graph import Graph
-from ...utils import not_implemented_for
-from ..matching import maximal_matching
 
 __all__ = ["min_weighted_dominating_set", "min_edge_dominating_set"]
 
-# TODO Why doesn't this algorithm work for directed graphs?
-
 def min_weighted_dominating_set(G: Graph, weight: str | None = None) -> set: ...
 def min_edge_dominating_set(G: Graph) -> set: ...
diff --git a/stubs/networkx/algorithms/clique.pyi b/stubs/networkx/algorithms/clique.pyi
@@ -1,10 +1,6 @@
-from collections import defaultdict, deque
-from itertools import chain, combinations, islice
-
 from numpy.typing import ArrayLike
 
 from ..classes.graph import Graph
-from ..utils import not_implemented_for
 
 __all__ = [
     "find_cliques",
@@ -22,8 +18,6 @@ __all__ = [
 
 def enumerate_all_cliques(G: Graph): ...
 def find_cliques(G: Graph, nodes=None): ...
-
-# TODO Should this also be not implemented for directed graphs?
 def find_cliques_recursive(G: Graph, nodes=None): ...
 def make_max_clique_graph(G: Graph, create_using=None): ...
 def make_clique_bipartite(G: Graph, fpos: bool | None = None, create_using=None, name=None): ...

diff --git a/stubs/networkx/algorithms/cuts.pyi b/stubs/networkx/algorithms/cuts.pyi
@@ -1,4 +1,3 @@
-from itertools import chain
 from typing import Any
 
 from ..classes.graph import Graph
@@ -14,19 +13,11 @@ __all__ = [
     "volume",
 ]
 
-# TODO STILL NEED TO UPDATE ALL THE DOCUMENTATION!
-
 def cut_size(G: Graph, S, T=None, weight: Any = None): ...
 def volume(G: Graph, S, weight: Any = None): ...
 def normalized_cut_size(G: Graph, S, T=None, weight: Any = None): ...
 def conductance(G: Graph, S, T=None, weight: Any = None): ...
 def edge_expansion(G: Graph, S, T=None, weight: Any = None): ...
 def mixing_expansion(G: Graph, S, T=None, weight: Any = None): ...
-
-# TODO What is the generalization to two arguments, S and T? Does the
-# denominator become `min(len(S), len(T))`?
 def node_expansion(G: Graph, S): ...
-
-# TODO What is the generalization to two arguments, S and T? Does the
-# denominator become `min(len(S), len(T))`?
 def boundary_expansion(G: Graph, S): ...
diff --git a/stubs/networkx/algorithms/distance_regular.pyi b/stubs/networkx/algorithms/distance_regular.pyi
@@ -3,8 +3,6 @@ from collections.abc import Iterable
 from numpy.typing import ArrayLike
 
 from ..classes.graph import Graph
-from ..utils import not_implemented_for
-from .distance_measures import diameter
 
 __all__ = [
     "is_distance_regular",
@@ -16,7 +14,4 @@ __all__ = [
 def is_distance_regular(G: Graph) -> bool: ...
 def global_parameters(b: ArrayLike, c: ArrayLike) -> Iterable: ...
 def intersection_array(G: Graph): ...
-
-# TODO There is a definition for directed strongly regular graphs.
-
 def is_strongly_regular(G: Graph) -> bool: ...
diff --git a/stubs/networkx/algorithms/tree/branchings.pyi b/stubs/networkx/algorithms/tree/branchings.pyi
@@ -1,27 +1,3 @@
-# TODO: Implement method from Gabow, Galil, Spence and Tarjan:
-#
-# @article{
-#    year={1986},
-#    issn={0209-9683},
-#    journal={Combinatorica},
-#    volume={6},
-#    number={2},
-#    doi={10.1007/BF02579168},
-#    title={Efficient algorithms for finding minimum spanning trees in
-#        undirected and directed graphs},
-#    url={https://doi.org/10.1007/BF02579168},
-#    publisher={Springer-Verlag},
-#    keywords={68 B 15; 68 C 05},
-#    author={Gabow, Harold N. and Galil, Zvi and Spencer, Thomas and Tarjan,
-#        Robert E.},
-#    pages={109-122},
-#    language={English}
-# }
-import string
-from dataclasses import dataclass, field
-from enum import Enum
-from operator import itemgetter
-from queue import PriorityQueue
 from typing import Literal
 
 from ...classes.digraph import DiGraph

diff --git a/stubs/networkx/classes/coreviews.pyi b/stubs/networkx/classes/coreviews.pyi
@@ -82,9 +82,6 @@ class FilterAtlas(Mapping):  # nodedict, nbrdict, keydict
     def __len__(self): ...
     def __iter__(self): ...
     def __getitem__(self, key): ...
-
-    # FIXME should this just be removed? we don't use it, but someone might
-    def copy(self): ...
     def __str__(self): ...
     def __repr__(self): ...
 
@@ -93,21 +90,12 @@ class FilterAdjacency(Mapping):  # edgedict
     def __len__(self): ...
     def __iter__(self): ...
     def __getitem__(self, node): ...
-
-    # FIXME should this just be removed? we don't use it, but someone might
-    def copy(self): ...
     def __str__(self): ...
     def __repr__(self): ...
 
 class FilterMultiInner(FilterAdjacency):  # muliedge_seconddict
     def __iter__(self): ...
     def __getitem__(self, nbr): ...
 
-    # FIXME should this just be removed? we don't use it, but someone might
-    def copy(self): ...
-
 class FilterMultiAdjacency(FilterAdjacency):  # multiedgedict
     def __getitem__(self, node): ...
-
-    # FIXME should this just be removed? we don't use it, but someone might
-    def copy(self): ...
diff --git a/stubs/networkx/conftest.pyi b/stubs/networkx/conftest.pyi
@@ -1,14 +1,8 @@
-import sys
-import warnings
-
-import networkx
 import pytest
 
 def pytest_addoption(parser): ...
 def pytest_configure(config): ...
 def pytest_collection_modifyitems(config, items): ...
-
-# TODO: The warnings below need to be dealt with, but for now we silence them.
 @pytest.fixture(autouse=True)
 def set_warnings(): ...
 @pytest.fixture(autouse=True)

diff --git a/stubs/sklearn/cluster/_agglomerative.pyi b/stubs/sklearn/cluster/_agglomerative.pyi
@@ -49,8 +49,7 @@ class AgglomerativeClustering(ClusterMixin, BaseEstimator):
         self,
         n_clusters: None | int = 2,
         *,
-        affinity: str | Callable = "deprecated",  # TODO(1.4): Remove
-        metric: None | str | Callable = None,  # TODO(1.4): Set to "euclidean"
+        metric: str | Callable = "euclidean",
         memory: None | Memory | str = None,
         connectivity: None | ArrayLike | Callable = None,
         compute_full_tree: Literal["auto"] | bool = "auto",
@@ -77,8 +76,7 @@ class FeatureAgglomeration(ClassNamePrefixFeaturesOutMixin, AgglomerativeCluster
         self,
         n_clusters: None | int = 2,
         *,
-        affinity: str | Callable = "deprecated",  # TODO(1.4): Remove
-        metric: None | str | Callable = None,  # TODO(1.4): Set to "euclidean"
+        metric: str | Callable = "euclidean",
         memory: None | Memory | str = None,
         connectivity: None | ArrayLike | Callable = None,
         compute_full_tree: Literal["auto"] | bool = "auto",

diff --git a/stubs/sklearn/covariance/_robust_covariance.pyi b/stubs/sklearn/covariance/_robust_covariance.pyi
@@ -7,17 +7,6 @@ from numpy.random import RandomState
 from .._typing import Float, Int, MatrixLike
 from . import EmpiricalCovariance
 
-# Author: Virgile Fritsch <virgile.fritsch@inria.fr>
-#
-# License: BSD 3 clause
-
-# Minimum Covariance Determinant
-#   Implementing of an algorithm by Rousseeuw & Van Driessen described in
-#   (A Fast Algorithm for the Minimum Covariance Determinant Estimator,
-#   1999, American Statistical Association and the American Society
-#   for Quality, TECHNOMETRICS)
-# XXX Is this really a public function? It's not listed in the docs or
-# exported by sklearn.covariance. Deprecate?
 def c_step(
     X: MatrixLike,
     n_support: Int,

diff --git a/stubs/sklearn/datasets/_samples_generator.pyi b/stubs/sklearn/datasets/_samples_generator.pyi
@@ -110,9 +110,6 @@ def make_low_rank_matrix(
     tail_strength: Float = 0.5,
     random_state: RandomState | None | Int = None,
 ) -> ndarray: ...
-
-# TODO(1.3): Change argument `data_transposed` default from True to False.
-# TODO(1.3): Deprecate data_transposed, always return data not transposed.
 def make_sparse_coded_signal(
     n_samples: Int,
     *,

diff --git a/stubs/sklearn/ensemble/_gb.pyi b/stubs/sklearn/ensemble/_gb.pyi
@@ -70,7 +70,6 @@ class GradientBoostingClassifier(ClassifierMixin, BaseGradientBoosting):
     feature_importances_: ndarray = ...
     n_estimators_: int = ...
 
-    # TODO(1.3): remove "deviance"
     _parameter_constraints: ClassVar[dict] = ...
 
     def __init__(

diff --git a/stubs/sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.pyi b/stubs/sklearn/ensemble/_hist_gradient_boosting/gradient_boosting.pyi
@@ -96,7 +96,6 @@ class HistGradientBoostingClassifier(ClassifierMixin, BaseHistGradientBoosting):
     do_early_stopping_: bool = ...
     classes_: ndarray = ...
 
-    # TODO(1.3): Remove "binary_crossentropy", "categorical_crossentropy", "auto"
     _parameter_constraints: ClassVar[dict] = ...
 
     def __init__(

diff --git a/stubs/sklearn/feature_selection/_univariate_selection.pyi b/stubs/sklearn/feature_selection/_univariate_selection.pyi
@@ -92,8 +92,6 @@ class SelectFwe(_BaseFilter):
 # Generic filter
 ######################################################################
 
-# TODO this class should fit on either p-values or scores,
-# depending on the mode.
 class GenericUnivariateSelect(_BaseFilter):
     feature_names_in_: ndarray = ...
     n_features_in_: int = ...

diff --git a/stubs/sklearn/linear_model/_stochastic_gradient.pyi b/stubs/sklearn/linear_model/_stochastic_gradient.pyi
@@ -86,7 +86,6 @@ def fit_binary(
 ) -> tuple[ndarray, float, int]: ...
 
 class BaseSGDClassifier(LinearClassifierMixin, BaseSGD, metaclass=ABCMeta):
-    # TODO(1.3): Remove "log""
     loss_functions: ClassVar[dict] = ...
 
     _parameter_constraints: ClassVar[dict] = ...

diff --git a/stubs/sklearn/metrics/_scorer.pyi b/stubs/sklearn/metrics/_scorer.pyi
@@ -90,10 +90,6 @@ adjusted_mutual_info_scorer = ...
 normalized_mutual_info_scorer = ...
 fowlkes_mallows_scorer = ...
 
-# TODO(1.3) Remove
-class _DeprecatedScorers(dict):
-    def __getitem__(self, item): ...
-
 _SCORERS = ...
 
 def get_scorer_names() -> list[str] | ndarray: ...

diff --git a/stubs/sklearn/neighbors/__init__.pyi b/stubs/sklearn/neighbors/__init__.pyi
@@ -5,7 +5,6 @@ from ._base import (
     sort_graph_by_row_values as sort_graph_by_row_values,
 )
 from ._classification import KNeighborsClassifier as KNeighborsClassifier, RadiusNeighborsClassifier as RadiusNeighborsClassifier
-from ._distance_metric import DistanceMetric as DistanceMetric
 from ._graph import (
     KNeighborsTransformer as KNeighborsTransformer,
     RadiusNeighborsTransformer as RadiusNeighborsTransformer,
@@ -22,7 +21,6 @@ from ._unsupervised import NearestNeighbors as NearestNeighbors
 
 __all__ = [
     "BallTree",
-    "DistanceMetric",
     "KDTree",
     "KNeighborsClassifier",
     "KNeighborsRegressor",

diff --git a/stubs/sklearn/neighbors/_distance_metric.pyi b/stubs/sklearn/neighbors/_distance_metric.pyi
diff --git a/stubs/sklearn/neighbors/_kde.pyi b/stubs/sklearn/neighbors/_kde.pyi
@@ -8,14 +8,10 @@ from .._typing import ArrayLike, Float, Int, MatrixLike
 from ..base import BaseEstimator
 from ._binary_tree import BinaryTree
 
-# Author: Jake Vanderplas <jakevdp@cs.washington.edu>
-
 VALID_KERNELS: list = ...
 
 TREE_DICT: dict = ...
 
-# TODO: implement a brute force version for testing purposes
-# TODO: create a density estimation base class?
 class KernelDensity(BaseEstimator):
     bandwidth_: float = ...
     feature_names_in_: ndarray = ...

diff --git a/stubs/sklearn/preprocessing/_polynomial.pyi b/stubs/sklearn/preprocessing/_polynomial.pyi
@@ -34,8 +34,6 @@ class PolynomialFeatures(TransformerMixin, BaseEstimator):
     def fit(self, X: MatrixLike | ArrayLike, y: Any = None) -> Self: ...
     def transform(self, X: MatrixLike | ArrayLike) -> ndarray | spmatrix: ...
 
-# TODO:
-# - sparse support (either scipy or own cython solution)?
 class SplineTransformer(TransformerMixin, BaseEstimator):
     n_features_out_: int = ...
     feature_names_in_: ndarray = ...

diff --git a/stubs/sklearn/stubtest_allowlist.txt b/stubs/sklearn/stubtest_allowlist.txt
@@ -1017,7 +1017,6 @@ sklearn.neighbors.NeighborhoodComponentsAnalysis.__sklearn_tags__
 sklearn.neighbors.RadiusNeighborsClassifier.__sklearn_tags__
 sklearn.neighbors.RadiusNeighborsClassifier.set_score_request
 sklearn.neighbors.RadiusNeighborsRegressor.set_score_request
-sklearn.neighbors.__all__
 sklearn.neighbors._ball_tree.BallTree32
 sklearn.neighbors._ball_tree.BallTree64
 sklearn.neighbors._ball_tree.__all__
@@ -1387,7 +1386,6 @@ sklearn.utils.estimator_checks.check_valid_tag_types
 sklearn.utils.estimator_checks.estimator_checks_generator
 sklearn.utils.extmath.safe_sqr
 sklearn.utils.get_tags
-sklearn.utils.metaestimators.__all__
 sklearn.utils.murmurhash.__pyx_capi__
 sklearn.utils.murmurhash.__test__
 sklearn.utils.murmurhash.murmurhash3_bytes_s32
@@ -1508,16 +1506,12 @@ sklearn.calibration.fmin_bfgs
 sklearn.calibration.is_classifier
 sklearn.calibration.partial
 sklearn.calibration.xlogy
-sklearn.cluster.AgglomerativeClustering.__init__
 sklearn.cluster.Birch.__init__
-sklearn.cluster.FeatureAgglomeration.__init__
 sklearn.cluster.FeatureAgglomeration.fit_predict
 sklearn.cluster.KMeans.__init__
 sklearn.cluster.MiniBatchKMeans.__init__
 sklearn.cluster.SpectralBiclustering.biclusters_
 sklearn.cluster.SpectralCoclustering.biclusters_
-sklearn.cluster._agglomerative.AgglomerativeClustering.__init__
-sklearn.cluster._agglomerative.FeatureAgglomeration.__init__
 sklearn.cluster._agglomerative.FeatureAgglomeration.fit_predict
 sklearn.cluster._agglomerative._TREE_BUILDERS
 sklearn.cluster._bicluster.SpectralBiclustering.biclusters_
@@ -2105,14 +2099,12 @@ sklearn.naive_bayes.Hidden
 sklearn.naive_bayes.MultinomialNB.__init__
 sklearn.naive_bayes.StrOptions
 sklearn.naive_bayes._BaseDiscreteNB.__init__
-sklearn.neighbors.DistanceMetric
 sklearn.neighbors.NearestCentroid.__init__
 sklearn.neighbors.VALID_METRICS
 sklearn.neighbors.VALID_METRICS_SPARSE
 sklearn.neighbors._base.VALID_METRICS
 sklearn.neighbors._base.VALID_METRICS_SPARSE
 sklearn.neighbors._binary_tree
-sklearn.neighbors._distance_metric
 sklearn.neighbors._nearest_centroid.NearestCentroid.__init__
 sklearn.pipeline.FeatureUnion.__init__
 sklearn.pipeline.Pipeline.__init__
@@ -2316,7 +2308,6 @@ sklearn.utils.is_scalar_nan
 sklearn.utils.islice
 sklearn.utils.issparse
 sklearn.utils.metaestimators.attrgetter
-sklearn.utils.metaestimators.if_delegate_has_method
 sklearn.utils.multiclass.Sequence.index
 sklearn.utils.multiclass.check_array
 sklearn.utils.multiclass.dok_matrix

diff --git a/stubs/sklearn/svm/_base.pyi b/stubs/sklearn/svm/_base.pyi
@@ -13,9 +13,6 @@ LIBSVM_IMPL: list = ...
 class BaseLibSVM(BaseEstimator, metaclass=ABCMeta):
     _parameter_constraints: ClassVar[dict] = ...
 
-    # The order of these must match the integer values in LibSVM.
-    # XXX These are actually the same in the dense case. Need to factor
-    # this out.
     _sparse_kernels: ClassVar[list] = ...
 
     @abstractmethod

diff --git a/stubs/sklearn/utils/metaestimators.pyi b/stubs/sklearn/utils/metaestimators.pyi
@@ -1,26 +1,15 @@
-import warnings
 from abc import ABCMeta, abstractmethod
-from collections.abc import Sequence
 from contextlib import suppress as suppress
 from operator import attrgetter as attrgetter
-from typing import Any, Callable, ClassVar
-
-import numpy as np
+from typing import Any, ClassVar
 
 from ..base import BaseEstimator
-from ._available_if import _AvailableIfDescriptor, available_if
+from ._available_if import available_if
 
-__all__ = ["available_if", "if_delegate_has_method"]
+__all__ = ["available_if"]
 
 class _BaseComposition(BaseEstimator, metaclass=ABCMeta):
     steps: ClassVar[list[Any]] = ...
 
     @abstractmethod
     def __init__(self) -> None: ...
-
-# TODO(1.3) remove
-class _IffHasAttrDescriptor(_AvailableIfDescriptor):
-    def __init__(self, fn, delegate_names, attribute_name) -> None: ...
-
-# TODO(1.3) remove
-def if_delegate_has_method(delegate: tuple[str, ...] | str | Sequence[str]) -> Callable: ...