Merge branch 'branch-25.02' into mukernels_strings

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -55,7 +55,7 @@ dependencies:
 - nbsphinx
 - ninja
 - notebook
-- numba-cuda>=0.0.13,<0.0.18
+- numba-cuda>=0.2.0,<0.3.0
 - numpy>=1.23,<3.0a0
 - numpydoc
 - nvcc_linux-64=11.8
@@ -66,7 +66,7 @@ dependencies:
 - pandas
 - pandas>=2.0,<2.2.4dev0
 - pandoc
-- polars>=1.11,<1.15
+- polars>=1.11,<1.18
 - pre-commit
 - ptxcompiler
 - pyarrow>=14.0.0,<19.0.0a0

conda/environments/all_cuda-125_arch-x86_64.yaml

@@ -54,7 +54,7 @@ dependencies:
 - nbsphinx
 - ninja
 - notebook
-- numba-cuda>=0.0.13,<0.0.18
+- numba-cuda>=0.2.0,<0.3.0
 - numpy>=1.23,<3.0a0
 - numpydoc
 - nvcomp==4.1.0.6
@@ -64,7 +64,7 @@ dependencies:
 - pandas
 - pandas>=2.0,<2.2.4dev0
 - pandoc
-- polars>=1.11,<1.15
+- polars>=1.11,<1.18
 - pre-commit
 - pyarrow>=14.0.0,<19.0.0a0
 - pydata-sphinx-theme!=0.14.2

conda/recipes/cudf-polars/meta.yaml

@@ -1,4 +1,4 @@
-# Copyright (c) 2024, NVIDIA CORPORATION.
+# Copyright (c) 2024-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -43,7 +43,7 @@ requirements:
   run:
     - python
     - pylibcudf ={{ version }}
-    - polars >=1.11,<1.15
+    - polars >=1.11,<1.18
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
 
 test:

conda/recipes/cudf/meta.yaml

@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -80,7 +80,7 @@ requirements:
     - typing_extensions >=4.0.0
     - pandas >=2.0,<2.2.4dev0
     - cupy >=12.0.0
-    - numba-cuda >=0.0.13,<0.0.18
+    - numba-cuda >=0.2.0,<0.3.0
     - numpy >=1.23,<3.0a0
     - pyarrow>=14.0.0,<18.0.0a0
     - libcudf ={{ version }}

cpp/src/io/parquet/decode_fixed.cu

@@ -991,9 +991,6 @@ CUDF_KERNEL void __launch_bounds__(decode_block_size_t, 8)
     return;
   }
 
-  // if we have no work to do (eg, in a skip_rows/num_rows case) in this page.
-  if (s->num_rows == 0) { return; }
-
   bool const should_process_nulls = is_nullable(s) && maybe_has_nulls(s);
 
   // shared buffer. all shared memory is suballocated out of here

dependencies.yaml

@@ -688,7 +688,7 @@ dependencies:
       - output_types: [conda, requirements, pyproject]
         packages:
           - cachetools
-          - &numba-cuda-dep numba-cuda>=0.0.13,<0.0.18
+          - &numba-cuda-dep numba-cuda>=0.2.0,<0.3.0
           - nvtx>=0.2.1
           - packaging
           - rich
@@ -747,7 +747,7 @@ dependencies:
     common:
       - output_types: [conda, requirements, pyproject]
         packages:
-          - polars>=1.11,<1.15
+          - polars>=1.11,<1.18
   run_cudf_polars_experimental:
     common:
       - output_types: [conda, requirements, pyproject]
@@ -810,11 +810,11 @@ dependencies:
         matrices:
           - matrix: {dependencies: "oldest"}
             packages:
-              - *numba-cuda-dep
+              - numba-cuda==0.2.0
               - pandas==2.0.*
           - matrix: {dependencies: "latest"}
             packages:
-              - numba-cuda==0.0.15
+              - *numba-cuda-dep
               - pandas==2.2.3
           - matrix:
             packages:

python/cudf/cudf/utils/_numba.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023-2024, NVIDIA CORPORATION.
+# Copyright (c) 2023-2025, NVIDIA CORPORATION.
 
 import glob
 import os
@@ -130,9 +130,7 @@ def _setup_numba():
             if driver_version < (12, 0):
                 patch_numba_linker_cuda_11()
             else:
-                from pynvjitlink.patch import patch_numba_linker
-
-                patch_numba_linker()
+                numba_config.CUDA_ENABLE_PYNVJITLINK = True
 
 
 class _CUDFNumbaConfig:

python/cudf/pyproject.toml

@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2024, NVIDIA CORPORATION.
+# Copyright (c) 2021-2025, NVIDIA CORPORATION.
 
 [build-system]
 build-backend = "rapids_build_backend.build"
@@ -24,7 +24,7 @@ dependencies = [
     "cupy-cuda11x>=12.0.0",
     "fsspec>=0.6.0",
     "libcudf==25.2.*,>=0.0.0a0",
-    "numba-cuda>=0.0.13,<0.0.18",
+    "numba-cuda>=0.2.0,<0.3.0",
     "numpy>=1.23,<3.0a0",
     "nvtx>=0.2.1",
     "packaging",

python/cudf_polars/cudf_polars/dsl/ir.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2025, NVIDIA CORPORATION & AFFILIATES.
 # SPDX-License-Identifier: Apache-2.0
 """
 DSL nodes for the LogicalPlan of polars.
@@ -34,9 +34,11 @@
 from cudf_polars.utils.versions import POLARS_VERSION_GT_112
 
 if TYPE_CHECKING:
-    from collections.abc import Callable, Hashable, MutableMapping, Sequence
+    from collections.abc import Callable, Hashable, Iterable, MutableMapping, Sequence
     from typing import Literal
 
+    from polars.polars import _expr_nodes as pl_expr
+
     from cudf_polars.typing import Schema
 
 
@@ -1019,7 +1021,27 @@ class ConditionalJoin(IR):
     __slots__ = ("ast_predicate", "options", "predicate")
     _non_child = ("schema", "predicate", "options")
     predicate: expr.Expr
-    options: tuple
+    """Expression predicate to join on"""
+    options: tuple[
+        tuple[
+            str,
+            pl_expr.Operator | Iterable[pl_expr.Operator],
+        ],
+        bool,
+        tuple[int, int] | None,
+        str,
+        bool,
+        Literal["none", "left", "right", "left_right", "right_left"],
+    ]
+    """
+    tuple of options:
+    - predicates: tuple of ir join type (eg. ie_join) and (In)Equality conditions
+    - join_nulls: do nulls compare equal?
+    - slice: optional slice to perform after joining.
+    - suffix: string suffix for right columns if names match
+    - coalesce: should key columns be coalesced (only makes sense for outer joins)
+    - maintain_order: which DataFrame row order to preserve, if any
+    """
 
     def __init__(
         self, schema: Schema, predicate: expr.Expr, options: tuple, left: IR, right: IR
@@ -1029,22 +1051,24 @@ def __init__(
         self.options = options
         self.children = (left, right)
         self.ast_predicate = to_ast(predicate)
-        _, join_nulls, zlice, suffix, coalesce = self.options
+        _, join_nulls, zlice, suffix, coalesce, maintain_order = self.options
         # Preconditions from polars
         assert not join_nulls
         assert not coalesce
+        assert maintain_order == "none"
         if self.ast_predicate is None:
             raise NotImplementedError(
                 f"Conditional join with predicate {predicate}"
             )  # pragma: no cover; polars never delivers expressions we can't handle
-        self._non_child_args = (self.ast_predicate, zlice, suffix)
+        self._non_child_args = (self.ast_predicate, zlice, suffix, maintain_order)
 
     @classmethod
     def do_evaluate(
         cls,
         predicate: plc.expressions.Expression,
         zlice: tuple[int, int] | None,
         suffix: str,
+        maintain_order: Literal["none", "left", "right", "left_right", "right_left"],
         left: DataFrame,
         right: DataFrame,
     ) -> DataFrame:
@@ -1088,6 +1112,7 @@ class Join(IR):
         tuple[int, int] | None,
         str,
         bool,
+        Literal["none", "left", "right", "left_right", "right_left"],
     ]
     """
     tuple of options:
@@ -1096,6 +1121,7 @@ class Join(IR):
     - slice: optional slice to perform after joining.
     - suffix: string suffix for right columns if names match
     - coalesce: should key columns be coalesced (only makes sense for outer joins)
+    - maintain_order: which DataFrame row order to preserve, if any
     """
 
     def __init__(
@@ -1113,6 +1139,9 @@ def __init__(
         self.options = options
         self.children = (left, right)
         self._non_child_args = (self.left_on, self.right_on, self.options)
+        # TODO: Implement maintain_order
+        if options[5] != "none":
+            raise NotImplementedError("maintain_order not implemented yet")
         if any(
             isinstance(e.value, expr.Literal)
             for e in itertools.chain(self.left_on, self.right_on)
@@ -1222,12 +1251,13 @@ def do_evaluate(
             tuple[int, int] | None,
             str,
             bool,
+            Literal["none", "left", "right", "left_right", "right_left"],
         ],
         left: DataFrame,
         right: DataFrame,
     ) -> DataFrame:
         """Evaluate and return a dataframe."""
-        how, join_nulls, zlice, suffix, coalesce = options
+        how, join_nulls, zlice, suffix, coalesce, _ = options
         if how == "cross":
             # Separate implementation, since cross_join returns the
             # result, not the gather maps

python/cudf_polars/cudf_polars/dsl/translate.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2025, NVIDIA CORPORATION & AFFILIATES.
 # SPDX-License-Identifier: Apache-2.0
 
 """Translate polars IR representation to ours."""
@@ -84,7 +84,7 @@ def translate_ir(self, *, n: int | None = None) -> ir.IR:
         # IR is versioned with major.minor, minor is bumped for backwards
         # compatible changes (e.g. adding new nodes), major is bumped for
         # incompatible changes (e.g. renaming nodes).
-        if (version := self.visitor.version()) >= (4, 0):
+        if (version := self.visitor.version()) >= (4, 3):
             e = NotImplementedError(
                 f"No support for polars IR {version=}"
             )  # pragma: no cover; no such version for now.

python/cudf_polars/cudf_polars/testing/plugin.py

@@ -123,6 +123,11 @@ def pytest_configure(config: pytest.Config) -> None:
     "tests/unit/io/test_scan.py::test_scan_include_file_name[False-scan_parquet-write_parquet]": "Need to add include_file_path to IR",
     "tests/unit/io/test_scan.py::test_scan_include_file_name[False-scan_csv-write_csv]": "Need to add include_file_path to IR",
     "tests/unit/io/test_scan.py::test_scan_include_file_name[False-scan_ndjson-write_ndjson]": "Need to add include_file_path to IR",
+    "tests/unit/io/test_write.py::test_write_async[read_parquet-write_parquet]": "Need to add include_file_path to IR",
+    "tests/unit/io/test_write.py::test_write_async[<lambda>-write_csv]": "Need to add include_file_path to IR",
+    "tests/unit/io/test_write.py::test_write_async[read_parquet-<lambda>]": "Need to add include_file_path to IR",
+    "tests/unit/io/test_write.py::test_write_async[<lambda>-<lambda>0]": "Need to add include_file_path to IR",
+    "tests/unit/io/test_write.py::test_write_async[<lambda>-<lambda>2]": "Need to add include_file_path to IR",
     "tests/unit/lazyframe/test_engine_selection.py::test_engine_import_error_raises[gpu]": "Expect this to pass because cudf-polars is installed",
     "tests/unit/lazyframe/test_engine_selection.py::test_engine_import_error_raises[engine1]": "Expect this to pass because cudf-polars is installed",
     "tests/unit/lazyframe/test_lazyframe.py::test_round[dtype1-123.55-1-123.6]": "Rounding midpoints is handled incorrectly",
@@ -140,6 +145,22 @@ def pytest_configure(config: pytest.Config) -> None:
     "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func1-func1-none]": "cudf-polars doesn't nullify division by zero",
     "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func1-func2-none]": "cudf-polars doesn't nullify division by zero",
     "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func1-func3-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr-broadcast_left-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr-broadcast_right-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr-broadcast_both-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr-broadcast_none-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_left-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_right-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_both-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_array.py::test_array_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_none-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr-broadcast_left-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr-broadcast_right-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr-broadcast_both-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr-broadcast_none-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_left-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_right-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_both-none]": "cudf-polars doesn't nullify division by zero",
+    "tests/unit/operations/arithmetic/test_list.py::test_list_arithmetic_values[exec_op_with_expr_no_type_coercion-broadcast_none-none]": "cudf-polars doesn't nullify division by zero",
     "tests/unit/operations/test_abs.py::test_abs_duration": "Need to raise for unsupported uops on timelike values",
     "tests/unit/operations/test_group_by.py::test_group_by_mean_by_dtype[input7-expected7-Float32-Float32]": "Mismatching dtypes, needs cudf#15852",
     "tests/unit/operations/test_group_by.py::test_group_by_mean_by_dtype[input10-expected10-Date-output_dtype10]": "Unsupported groupby-agg for a particular dtype",

python/cudf_polars/pyproject.toml

@@ -1,4 +1,4 @@
-# Copyright (c) 2024, NVIDIA CORPORATION.
+# Copyright (c) 2024-2025, NVIDIA CORPORATION.
 
 [build-system]
 build-backend = "rapids_build_backend.build"
@@ -19,7 +19,7 @@ authors = [
 license = { text = "Apache 2.0" }
 requires-python = ">=3.10"
 dependencies = [
-    "polars>=1.11,<1.15",
+    "polars>=1.11,<1.18",
     "pylibcudf==25.2.*,>=0.0.0a0",
 ] # This list was generated by `rapids-dependency-file-generator`. To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
 classifiers = [

python/cudf_polars/tests/test_join.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2025, NVIDIA CORPORATION & AFFILIATES.
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
@@ -53,6 +53,15 @@ def right():
     )
 
 
+@pytest.mark.parametrize(
+    "maintain_order", ["left", "left_right", "right_left", "right"]
+)
+def test_join_maintain_order_param_unsupported(left, right, maintain_order):
+    q = left.join(right, on=pl.col("a"), how="inner", maintain_order=maintain_order)
+
+    assert_ir_translation_raises(q, NotImplementedError)
+
+
 @pytest.mark.parametrize(
     "join_expr",
     [

python/dask_cudf/pyproject.toml

@@ -47,7 +47,7 @@ cudf = "dask_cudf.backends:CudfBackendEntrypoint"
 [project.optional-dependencies]
 test = [
     "dask-cuda==25.2.*,>=0.0.0a0",
-    "numba-cuda>=0.0.13,<0.0.18",
+    "numba-cuda>=0.2.0,<0.3.0",
     "pytest-cov",
     "pytest-xdist",
     "pytest<8",