Consolidate cudf-classic dtype validators to their own file and make them stricter

python/cudf/cudf/core/accessors/lists.py

@@ -11,8 +11,8 @@
 from cudf.api.types import is_scalar
 from cudf.core.accessors.base_accessor import BaseAccessor
 from cudf.core.column.column import as_column
+from cudf.core.dtype.validators import is_dtype_obj_list, is_dtype_obj_numeric
 from cudf.core.dtypes import ListDtype, dtype as cudf_dtype
-from cudf.utils.dtypes import is_dtype_obj_list, is_dtype_obj_numeric
 from cudf.utils.scalar import pa_scalar_to_plc_scalar
 
 if TYPE_CHECKING:

python/cudf/cudf/core/accessors/struct.py

@@ -7,8 +7,9 @@
 
 from cudf.core.accessors.base_accessor import BaseAccessor
 from cudf.core.column.struct import StructColumn
+from cudf.core.dtype.validators import is_dtype_obj_struct
 from cudf.core.dtypes import StructDtype
-from cudf.utils.dtypes import get_dtype_of_same_kind, is_dtype_obj_struct
+from cudf.utils.dtypes import get_dtype_of_same_kind
 
 if TYPE_CHECKING:
     from cudf.core.dataframe import DataFrame

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

@@ -79,6 +79,15 @@ class CategoricalColumn(column.ColumnBase):
         plc.TypeId.UINT64,
     }
 
+    @classmethod
+    def _validate_args(  # type: ignore[override]
+        cls, plc_column: plc.Column, dtype: CategoricalDtype
+    ) -> tuple[plc.Column, CategoricalDtype]:
+        plc_column, dtype = super()._validate_args(plc_column, dtype)  # type: ignore[assignment]
+        if not isinstance(dtype, CategoricalDtype):
+            raise ValueError(f"{dtype=} must be a CategoricalDtype instance")
+        return plc_column, dtype
+
     def __contains__(self, item: ScalarLike) -> bool:
         try:
             encoded = self._encode(item)

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

@@ -50,6 +50,13 @@
 )
 from cudf.core.column.utils import access_columns
 from cudf.core.copy_types import GatherMap
+from cudf.core.dtype.validators import (
+    is_dtype_obj_decimal,
+    is_dtype_obj_interval,
+    is_dtype_obj_list,
+    is_dtype_obj_numeric,
+    is_dtype_obj_struct,
+)
 from cudf.core.dtypes import (
     CategoricalDtype,
     DecimalDtype,
@@ -72,11 +79,6 @@
     find_common_type,
     get_dtype_of_same_kind,
     is_column_like,
-    is_dtype_obj_decimal,
-    is_dtype_obj_interval,
-    is_dtype_obj_list,
-    is_dtype_obj_numeric,
-    is_dtype_obj_struct,
     is_mixed_with_object_dtype,
     is_pandas_nullable_extension_dtype,
     min_signed_type,
@@ -863,7 +865,7 @@ def _prep_pandas_compat_repr(self) -> StringColumn | Self:
         * null (other types)= str(pd.NA)
         """
         if self.has_nulls():
-            return self.astype(np.dtype("str")).fillna(
+            return self.astype(CUDF_STRING_DTYPE).fillna(
                 str(self._PANDAS_NA_VALUE)
             )
         return self

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

@@ -5,7 +5,7 @@
 
 import warnings
 from decimal import Decimal
-from typing import TYPE_CHECKING, Any, Self, cast
+from typing import TYPE_CHECKING, Any, ClassVar, Self, cast
 
 import numpy as np
 import pandas as pd
@@ -19,14 +19,16 @@
 from cudf.core._internals import binaryop
 from cudf.core.column.column import ColumnBase, as_column
 from cudf.core.column.numerical_base import NumericalBaseColumn
+from cudf.core.dtype.validators import (
+    is_dtype_obj_decimal32,
+    is_dtype_obj_decimal64,
+    is_dtype_obj_decimal128,
+)
 from cudf.core.dtypes import (
     Decimal32Dtype,
     Decimal64Dtype,
     Decimal128Dtype,
     DecimalDtype,
-    is_decimal32_dtype,
-    is_decimal64_dtype,
-    is_decimal128_dtype,
 )
 from cudf.core.mixins import BinaryOperand
 from cudf.utils.dtypes import (
@@ -38,7 +40,7 @@
 from cudf.utils.utils import is_na_like
 
 if TYPE_CHECKING:
-    from collections.abc import Mapping
+    from collections.abc import Callable, Mapping
 
     from cudf._typing import (
         ColumnBinaryOperand,
@@ -70,14 +72,17 @@ class DecimalBaseColumn(NumericalBaseColumn):
     """Base column for decimal32, decimal64 or decimal128 columns"""
 
     _VALID_BINARY_OPERATIONS = BinaryOperand._SUPPORTED_BINARY_OPERATIONS
+    _decimal_type_check: ClassVar[Callable[[DtypeObj], bool]]
 
     @classmethod
     def _validate_args(  # type: ignore[override]
         cls, plc_column: plc.Column, dtype: DecimalDtype
     ) -> tuple[plc.Column, DecimalDtype]:
         plc_column, dtype = super()._validate_args(plc_column, dtype)  # type: ignore[assignment]
-        if not cls._decimal_check(dtype):  # type: ignore[attr-defined]
-            raise ValueError(f"{dtype=} must be a Decimal128Dtype instance")
+        if not cls._decimal_type_check(dtype):
+            raise ValueError(
+                f"{dtype=} must be a valid decimal dtype instance"
+            )
         return plc_column, dtype
 
     def _with_type_metadata(self: Self, dtype: DtypeObj) -> Self:
@@ -379,19 +384,19 @@ def to_pandas(
 class Decimal32Column(DecimalBaseColumn):
     _VALID_PLC_TYPES = {plc.TypeId.DECIMAL32}
     _decimal_cls = Decimal32Dtype
-    _decimal_check = is_decimal32_dtype
+    _decimal_type_check = is_dtype_obj_decimal32
 
 
 class Decimal64Column(DecimalBaseColumn):
     _VALID_PLC_TYPES = {plc.TypeId.DECIMAL64}
     _decimal_cls = Decimal64Dtype
-    _decimal_check = is_decimal64_dtype
+    _decimal_type_check = is_dtype_obj_decimal64
 
 
 class Decimal128Column(DecimalBaseColumn):
     _VALID_PLC_TYPES = {plc.TypeId.DECIMAL128}
     _decimal_cls = Decimal128Dtype
-    _decimal_check = is_decimal128_dtype
+    _decimal_type_check = is_dtype_obj_decimal128
 
     def to_arrow(self) -> pa.Array:
         arrow_array = super().to_arrow()

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

@@ -13,8 +13,8 @@
 
 import cudf
 from cudf.core.column.column import ColumnBase, _handle_nulls, as_column
+from cudf.core.dtype.validators import is_dtype_obj_interval
 from cudf.core.dtypes import IntervalDtype, _dtype_to_metadata
-from cudf.utils.dtypes import is_dtype_obj_interval
 from cudf.utils.scalar import maybe_nested_pa_scalar_to_py
 
 if TYPE_CHECKING:

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

@@ -15,12 +15,10 @@
 
 import cudf
 from cudf.core.column.column import ColumnBase, as_column, column_empty
+from cudf.core.dtype.validators import is_dtype_obj_list
 from cudf.core.dtypes import ListDtype
 from cudf.core.missing import NA
-from cudf.utils.dtypes import (
-    get_dtype_of_same_kind,
-    is_dtype_obj_list,
-)
+from cudf.utils.dtypes import get_dtype_of_same_kind
 from cudf.utils.scalar import (
     maybe_nested_pa_scalar_to_py,
     pa_scalar_to_plc_scalar,

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

@@ -19,14 +19,13 @@
 from cudf.api.types import is_scalar
 from cudf.core._internals import binaryop
 from cudf.core.column.column import ColumnBase, as_column, column_empty
+from cudf.core.dtype.validators import is_dtype_obj_string
 from cudf.core.mixins import Scannable
 from cudf.errors import MixedTypeError
 from cudf.utils.dtypes import (
-    CUDF_STRING_DTYPE,
     cudf_dtype_to_pa_type,
     dtype_to_pylibcudf_type,
     get_dtype_of_same_kind,
-    is_dtype_obj_string,
     is_pandas_nullable_extension_dtype,
 )
 from cudf.utils.scalar import pa_scalar_to_plc_scalar
@@ -117,21 +116,8 @@ def _validate_args(
         cls, plc_column: plc.Column, dtype: np.dtype
     ) -> tuple[plc.Column, np.dtype]:
         plc_column, dtype = super()._validate_args(plc_column, dtype)
-        if (
-            not cudf.get_option("mode.pandas_compatible")
-            and dtype != CUDF_STRING_DTYPE
-            and dtype.kind != "U"
-        ) or (
-            cudf.get_option("mode.pandas_compatible")
-            and not is_dtype_obj_string(dtype)
-        ):
-            raise ValueError(f"dtype must be {CUDF_STRING_DTYPE}")
-        if (
-            cudf.get_option("mode.pandas_compatible")
-            and isinstance(dtype, np.dtype)
-            and dtype.kind == "U"
-        ):
-            dtype = CUDF_STRING_DTYPE
+        if not is_dtype_obj_string(dtype):
+            raise ValueError("dtype must be a valid cuDF string dtype")
         return plc_column, dtype
 
     @property

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

@@ -11,11 +11,9 @@
 
 import cudf
 from cudf.core.column.column import ColumnBase
+from cudf.core.dtype.validators import is_dtype_obj_struct
 from cudf.core.dtypes import StructDtype
-from cudf.utils.dtypes import (
-    dtype_from_pylibcudf_column,
-    is_dtype_obj_struct,
-)
+from cudf.utils.dtypes import dtype_from_pylibcudf_column
 from cudf.utils.scalar import (
     maybe_nested_pa_scalar_to_py,
     pa_scalar_to_plc_scalar,

python/cudf/cudf/core/dataframe.py

@@ -62,6 +62,7 @@
 )
 from cudf.core.column_accessor import ColumnAccessor
 from cudf.core.copy_types import BooleanMask
+from cudf.core.dtype.validators import is_dtype_obj_numeric
 from cudf.core.dtypes import (
     CategoricalDtype,
     Decimal32Dtype,
@@ -107,7 +108,6 @@
     find_common_type,
     get_dtype_of_same_kind,
     is_column_like,
-    is_dtype_obj_numeric,
     is_mixed_with_object_dtype,
     is_pandas_nullable_extension_dtype,
     min_signed_type,

python/cudf/cudf/core/dtype/__init__.py

@@ -0,0 +1,2 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION.
+# SPDX-License-Identifier: Apache-2.0