Update pre-commit hooks

.pre-commit-config.yaml

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 ci:
@@ -30,11 +30,11 @@ repos:
             ^python/cudf/cudf/tests/data/vocab.txt
           )
   - repo: https://github.com/MarcoGorelli/cython-lint
-    rev: v0.18.1
+    rev: v0.19.0
     hooks:
       - id: cython-lint
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: 'v1.13.0'
+    rev: 'v2.1.0'
     hooks:
       - id: mypy
         additional_dependencies: [
@@ -202,7 +202,7 @@ repos:
         pass_filenames: false
         verbose: true
   - repo: https://github.com/codespell-project/codespell
-    rev: v2.4.1
+    rev: v2.4.2
     hooks:
       - id: codespell
         additional_dependencies: [tomli]
@@ -213,13 +213,13 @@ repos:
             ^CHANGELOG.md$
           )
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.9.3
+    rev: v0.15.17
     hooks:
       - id: ruff
         args: ["--fix"]
       - id: ruff-format
   - repo: https://github.com/rapidsai/pre-commit-hooks
-    rev: v1.4.1
+    rev: v1.5.1
     hooks:
       - id: verify-copyright
         name: verify-copyright-cudf
@@ -317,7 +317,7 @@ repos:
           (?x)
             ^pyproject[.]toml$
   - repo: https://github.com/rapidsai/dependency-file-generator
-    rev: v1.20.2
+    rev: v1.20.3
     hooks:
       - id: rapids-dependency-file-generator
         args: ["--clean", "--warn-all", "--strict"]
@@ -327,7 +327,7 @@ repos:
       - id: shellcheck
         args: ["--severity=warning"]
   - repo: https://github.com/zizmorcore/zizmor-pre-commit
-    rev: v1.24.1
+    rev: v1.25.2
     hooks:
       - id: zizmor
 

docs/cudf/source/_ext/PandasCompat.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2021-2025, NVIDIA CORPORATION
+# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 # This file is adapted from official sphinx tutorial for `todo` extension:
@@ -119,7 +119,7 @@ def __init__(self, app, doctree, docname):
         self.config = app.config
         self.env = app.env
         self.domain = cast(
-            PandasCompatDomain, app.env.get_domain("pandascompat")
+            "PandasCompatDomain", app.env.get_domain("pandascompat")
         )
         self.document = new_document("")
         self.process(doctree, docname)

pyproject.toml

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 [tool.mypy]
@@ -103,7 +103,7 @@ select = [
     # overload-with-docstring
     "D418",
     # flake8-type-checking
-    "TCH",
+    "TC",
     # flake8-future-annotations
     "FA",
     # non-pep585-annotation
@@ -138,10 +138,15 @@ ignore = [
     # Mutable class attributes should be annotated with
     # `typing.ClassVar`
     "RUF012",
+    # import not at top of file (deferred/conditional imports are intentional)
+    "PLC0415",
 ]
 fixable = ["ALL"]
 exclude = [
     "cpp/scripts/gdb-pretty-printers.py",
+    # .pyi stubs use PEP 695 (Python 3.12+) syntax; pylibcudf's own config
+    # sets target-version = "py312" for these
+    "*.pyi",
 ]
 extend-unsafe-fixes = [
     # zip-without-explicit-strict's autofix sets strict to False, but we want to enforce True

python/cudf/cudf/api/types.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 """Define common type operations."""
@@ -252,7 +252,7 @@ def _union_categoricals(
 
     result_col = cudf.core.column.CategoricalColumn._concat(
         [
-            cast(cudf.core.column.CategoricalColumn, obj._column)
+            cast("cudf.core.column.CategoricalColumn", obj._column)
             for obj in to_union
         ]
     )

python/cudf/cudf/core/_internals/copying.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
@@ -66,7 +66,7 @@ def scatter(
         *target_columns, mode="write", scope="internal"
     ) as target_columns:
         plc_tbl = plc.copying.scatter(
-            cast(list[plc.Scalar], sources)
+            cast("list[plc.Scalar]", sources)
             if isinstance(sources[0], plc.Scalar)
             else plc.Table(
                 [col.plc_column for col in cast("list[ColumnBase]", sources)]

python/cudf/cudf/core/accessors/string.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -3851,7 +3851,7 @@ def isempty(self) -> Series | Index:
             # mypy can't deduce that the return value of
             # StringColumn.__eq__ is ColumnBase because the binops are
             # dynamically added by a mixin class
-            cast(ColumnBase, self._column == "").fillna(False)
+            cast("ColumnBase", self._column == "").fillna(False)
         )
 
     def isspace(self) -> Series | Index:
@@ -4886,7 +4886,7 @@ def character_ngrams(
         if isinstance(result, cudf.Series) and not as_list:
             # before exploding, removes those lists which have 0 length
             result = result[result.list.len() > 0]
-            return result.explode()  # type: ignore[union-attr]
+            return result.explode()
         return result
 
     def hash_character_ngrams(

python/cudf/cudf/core/column/categorical.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -254,7 +254,7 @@ def _binaryop(self, other: ColumnBinaryOperand, op: str) -> ColumnBase:
             )
             plc_col = plc.Column.from_scalar(plc_scalar, len(self))
             other = cast(
-                CategoricalColumn,
+                "CategoricalColumn",
                 ColumnBase.create(plc_col, self.dtype),
             )
         equality_ops = {"__eq__", "__ne__", "NULL_EQUALS", "NULL_NOT_EQUALS"}

python/cudf/cudf/core/column/column.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -3479,7 +3479,7 @@ def as_column(
                 )
             elif inferred_dtype == "boolean":
                 if cudf.get_option("mode.pandas_compatible"):
-                    if dtype.kind != "b" or pd.isna(arbitrary).any():
+                    if dtype.kind != "b" or pd.isna(arbitrary).any():  # type: ignore[union-attr]  # (dtype is required for boolean compatibility)
                         raise MixedTypeError(
                             f"Cannot have mixed values with {inferred_dtype}"
                         )
@@ -3563,7 +3563,7 @@ def as_column(
         if is_arrow_null_dtype(dtype):
             if is_na_like(arbitrary):
                 return column_empty(length, dtype=dtype)
-            pa.scalar(arbitrary, type=dtype.pyarrow_dtype)
+            pa.scalar(arbitrary, type=dtype.pyarrow_dtype)  # type: ignore[union-attr]  # (arrow null dtype has pyarrow_dtype)
 
         pa_type = None
         if isinstance(arbitrary, pd.Interval) or _is_categorical_dtype(dtype):
@@ -3610,17 +3610,18 @@ def as_column(
                     np.dtype(f"{arbitrary.dtype.kind}8[s]")
                 )
 
-        pa_scalar = pa.scalar(arbitrary, type=pa_type)
+        pa_scalar = pa.scalar(arbitrary, type=pa_type)  # type: ignore[type-var]  # (pyarrow accepts normalized scalar inputs)
         if length == 0:
             if dtype is None:
-                dtype = cudf_dtype_from_pa_type(pa_scalar.type)
+                dtype = cudf_dtype_from_pa_type(pa_scalar.type)  # type: ignore[arg-type]  # (pyarrow scalar exposes a DataType)
             return column_empty(length, dtype=dtype)
         else:
             plc_col = plc.Column.from_scalar(
                 pa_scalar_to_plc_scalar(pa_scalar), length
             )
             col = ColumnBase.create(
-                plc_col, cudf_dtype_from_pa_type(pa_scalar.type)
+                plc_col,
+                cudf_dtype_from_pa_type(pa_scalar.type),  # type: ignore[arg-type]  # (pyarrow scalar exposes a DataType)
             )
             if dtype is not None:
                 col = col.astype(dtype)
@@ -3776,18 +3777,20 @@ def as_column(
                 ser = pd.Series(arbitrary).astype(dtype)
             else:
                 ser = pd.Series(
-                    arbitrary, dtype=pd.CategoricalDtype(ordered=dtype.ordered)
+                    arbitrary,
+                    dtype=pd.CategoricalDtype(ordered=dtype.ordered),  # type: ignore[union-attr]  # (categorical dtype converted from cudf dtype)
                 )
-                if dtype.categories is not None:
+                if dtype.categories is not None:  # type: ignore[union-attr]  # (categorical dtype converted from cudf dtype)
                     ser = ser.cat.set_categories(
-                        dtype.categories, ordered=dtype.ordered
+                        dtype.categories,  # type: ignore[union-attr]  # (categorical dtype converted from cudf dtype)
+                        ordered=dtype.ordered,  # type: ignore[union-attr]  # (categorical dtype converted from cudf dtype)
                     )
         else:
             ser = pd.Series(arbitrary, dtype=dtype)
         return as_column(ser, nan_as_null=nan_as_null)
     elif isinstance(dtype, (StructDtype, ListDtype)):
         try:
-            data = pa.array(arbitrary, type=dtype.to_arrow())
+            data = pa.array(arbitrary, type=dtype.to_arrow())  # type: ignore[arg-type]  # (cudf nested dtype converts to Arrow type)
         except (pa.ArrowInvalid, pa.ArrowTypeError):
             if isinstance(dtype, ListDtype):
                 # e.g. test_cudf_list_struct_write
@@ -3800,7 +3803,7 @@ def as_column(
         if is_arrow_null_dtype(dtype):
             arbitrary = pa.array(
                 arbitrary,
-                type=dtype.pyarrow_dtype,
+                type=dtype.pyarrow_dtype,  # type: ignore[union-attr]  # (arrow null dtype has pyarrow_dtype)
                 from_pandas=True,
             )
             return as_column(arbitrary, nan_as_null=nan_as_null, dtype=dtype)

python/cudf/cudf/core/column/datetime.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -602,7 +602,7 @@ def strftime(self, format: str, dtype: DtypeObj) -> StringColumn:
             )
         with self.access(mode="read", scope="internal"):
             return cast(
-                cudf.core.column.string.StringColumn,
+                "cudf.core.column.string.StringColumn",
                 ColumnBase.create(
                     plc.strings.convert.convert_datetime.from_timestamps(
                         self.plc_column,
@@ -777,9 +777,9 @@ def _preserve_tz(result_dtype: np.dtype) -> DtypeObj:
             # When the operation should yield another datetime and self
             # is tz-aware, carry the tz forward with the resolved unit.
             if self_is_tz and result_dtype.kind == "M":
-                tz_dtype = cast(pd.DatetimeTZDtype, self.dtype)
+                tz_dtype = cast("pd.DatetimeTZDtype", self.dtype)
                 unit = np.datetime_data(
-                    cast(np.dtype[np.datetime64], result_dtype)
+                    cast("np.dtype[np.datetime64]", result_dtype)
                 )[0]
                 return pd.DatetimeTZDtype(unit, tz_dtype.tz)
             return result_dtype
@@ -988,7 +988,7 @@ def tz_localize(
         if is_arrow:
             dtype = pd.ArrowDtype(pa.timestamp(self.time_unit, tz))  # type: ignore[call-overload]
         result = cast(
-            DatetimeTZColumn, ColumnBase.create(gmt_data.plc_column, dtype)
+            "DatetimeTZColumn", ColumnBase.create(gmt_data.plc_column, dtype)
         )
         # Avoid re-computing local times from UTC times
         result._local_time = localized
@@ -1070,7 +1070,7 @@ def _local_time(self) -> DatetimeColumn:
         # Perform the add on the tz-naive UTC view so the result is naive
         # (``self + offsets`` now preserves tz by default). The local-time
         # column must be tz-naive to represent the wall-clock values.
-        return cast(DatetimeColumn, self._utc_time + offsets_from_utc)
+        return cast("DatetimeColumn", self._utc_time + offsets_from_utc)
 
     def as_string_column(self, dtype: DtypeObj) -> StringColumn:
         return self._local_time.as_string_column(dtype)
@@ -1105,7 +1105,7 @@ def as_datetime_column(
                     .plc_column
                 )
                 casted = cast(
-                    DatetimeTZColumn, ColumnBase.create(casted_plc, dtype)
+                    "DatetimeTZColumn", ColumnBase.create(casted_plc, dtype)
                 )
             else:
                 casted = self
@@ -1118,7 +1118,7 @@ def as_datetime_column(
                     .plc_column
                 )
                 casted = cast(
-                    DatetimeTZColumn, ColumnBase.create(casted_plc, dtype)
+                    "DatetimeTZColumn", ColumnBase.create(casted_plc, dtype)
                 )
             else:
                 casted = self
@@ -1192,6 +1192,6 @@ def tz_convert(self, tz: str | None) -> DatetimeColumn | DatetimeTZColumn:
                 pd.DatetimeTZDtype(self.time_unit, tz)
             )
         return cast(
-            DatetimeTZColumn,
+            "DatetimeTZColumn",
             ColumnBase.create(self.plc_column, target_dtype),
         )

python/cudf/cudf/core/column/decimal.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -131,12 +131,12 @@ def as_string_column(self, dtype: DtypeObj) -> StringColumn:
                     )
                 )
                 return cast(
-                    cudf.core.column.string.StringColumn,
+                    "cudf.core.column.string.StringColumn",
                     ColumnBase.create(plc_column, dtype),
                 )
         else:
             return cast(
-                cudf.core.column.StringColumn,
+                "cudf.core.column.StringColumn",
                 cudf.core.column.column_empty(0, dtype=dtype),
             )
 
@@ -239,7 +239,7 @@ def _binaryop(self, other: ColumnBinaryOperand, op: str) -> ColumnBase:
             if isinstance(lhs, (int, Decimal)):
                 lhs_binop = _to_plc_scalar(lhs, new_lhs_dtype)
             else:
-                lhs_binop = lhs.astype(new_lhs_dtype)
+                lhs_binop = lhs.astype(new_lhs_dtype)  # type: ignore[union-attr]  # (lhs is decimal column here)
             if isinstance(rhs, (int, Decimal)):
                 rhs_binop = _to_plc_scalar(rhs, new_rhs_dtype)
             else:
@@ -256,7 +256,7 @@ def _binaryop(self, other: ColumnBinaryOperand, op: str) -> ColumnBase:
         }:
             lhs_comp: plc.Scalar | ColumnBase = lhs  # type: ignore[assignment]
             rhs_comp: plc.Scalar | ColumnBase = (
-                _to_plc_scalar(rhs, self.dtype)  # type: ignore[arg-type]
+                _to_plc_scalar(rhs, self.dtype)
                 if isinstance(rhs, (int, Decimal))
                 else rhs
             )

python/cudf/cudf/core/column/interval.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2018-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
@@ -133,7 +133,7 @@ def set_closed(
         )
 
     def _binaryop(self, other: ColumnBinaryOperand, op: str) -> ColumnBase:
-        reflect, op = self._check_reflected_op(op)
+        _reflect, op = self._check_reflected_op(op)
         if not isinstance(other, type(self)):
             return NotImplemented
         if op in {"__eq__", "__ne__", "NULL_EQUALS", "NULL_NOT_EQUALS"}:

python/cudf/cudf/core/column/lists.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 from __future__ import annotations
@@ -77,7 +77,7 @@ def _cast_setitem_value(self, value: Any) -> plc.Scalar:
 
     def _binaryop(self, other: ColumnBinaryOperand, op: str) -> ColumnBase:
         # Lists only support __add__, which concatenates lists.
-        reflect, op = self._check_reflected_op(op)
+        _reflect, op = self._check_reflected_op(op)
         if not isinstance(other, type(self)):
             return NotImplemented
         if is_dtype_obj_list(other.dtype):