diff --git a/.devcontainer/Dockerfile b/.devcontainer/Dockerfile index 8190b5d0297..315a389339a 100644 --- a/.devcontainer/Dockerfile +++ b/.devcontainer/Dockerfile @@ -31,6 +31,6 @@ ENV PYTHONDONTWRITEBYTECODE="1" ENV SCCACHE_REGION="us-east-2" ENV SCCACHE_BUCKET="rapids-sccache-devs" -ENV VAULT_HOST="https://vault.ops.k8s.rapids.ai" +ENV AWS_ROLE_ARN="arn:aws:iam::279114543810:role/nv-gha-token-sccache-devs" ENV HISTFILE="/home/coder/.cache/._bash_history" ENV LIBCUDF_KERNEL_CACHE_PATH="/home/coder/cudf/cpp/build/${PYTHON_PACKAGE_MANAGER}/cuda-${CUDA_VERSION}/latest/jitify_cache" diff --git a/.github/workflows/auto-assign.yml b/.github/workflows/auto-assign.yml new file mode 100644 index 00000000000..673bebd4ecc --- /dev/null +++ b/.github/workflows/auto-assign.yml @@ -0,0 +1,17 @@ +name: "Auto Assign PR" + +on: + pull_request_target: + types: + - opened + - reopened + - synchronize + +jobs: + add_assignees: + runs-on: ubuntu-latest + steps: + - uses: actions-ecosystem/action-add-assignees@v1 + with: + repo_token: "${{ secrets.GITHUB_TOKEN }}" + assignees: ${{ github.actor }} diff --git a/.github/workflows/labeler.yml b/.github/workflows/labeler.yml index 31e78f82a62..f5cb71bfc14 100644 --- a/.github/workflows/labeler.yml +++ b/.github/workflows/labeler.yml @@ -1,4 +1,5 @@ name: "Pull Request Labeler" + on: - pull_request_target diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 174dc72bf02..f5234f58efe 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -16,17 +16,6 @@ repos: ^cpp/cmake/thirdparty/patches/.*| ^python/cudf/cudf/tests/data/subword_tokenizer_data/.* ) - - repo: https://github.com/PyCQA/isort - rev: 5.13.2 - hooks: - - id: isort - # Use the config file specific to each subproject so that each - # project can specify its own first/third-party packages. - args: ["--config-root=python/", "--resolve-all-configs"] - files: python/.* - exclude: | - (?x)^(^python/cudf_polars/.*) - types_or: [python, cython, pyi] - repo: https://github.com/MarcoGorelli/cython-lint rev: v0.16.2 hooks: @@ -150,6 +139,7 @@ repos: rev: v0.4.8 hooks: - id: ruff + args: ["--fix"] files: python/.*$ - id: ruff-format files: python/.*$ @@ -165,7 +155,7 @@ repos: ) - id: verify-alpha-spec - repo: https://github.com/rapidsai/dependency-file-generator - rev: v1.13.11 + rev: v1.16.0 hooks: - id: rapids-dependency-file-generator args: ["--clean"] diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index f9cdde7c2b7..3db1ed35294 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -38,6 +38,7 @@ conduct. More information can be found at: 8. Verify that CI passes all [status checks](https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/collaborating-on-repositories-with-code-quality-features/about-status-checks). Fix if needed. 9. Wait for other developers to review your code and update code as needed. + Changes to any C++ files require at least 2 approvals from the cudf-cpp-codeowners before merging. 10. Once reviewed and approved, a RAPIDS developer will merge your pull request. If you are unsure about anything, don't hesitate to comment on issues and ask for clarification! @@ -293,8 +294,8 @@ In order to run doxygen as a linter on C++/CUDA code, run ./ci/checks/doxygen.sh ``` -Python code runs several linters including [Black](https://black.readthedocs.io/en/stable/), -[isort](https://pycqa.github.io/isort/), and [flake8](https://flake8.pycqa.org/en/latest/). +Python code runs several linters including [Ruff](https://docs.astral.sh/ruff/) +with its various rules for Black-like formatting or Isort. cuDF also uses [codespell](https://github.com/codespell-project/codespell) to find spelling mistakes, and this check is run as a pre-commit hook. To apply the suggested spelling fixes, diff --git a/ci/build_cpp.sh b/ci/build_cpp.sh index e5fcef17a83..3d06eacf9ff 100755 --- a/ci/build_cpp.sh +++ b/ci/build_cpp.sh @@ -15,8 +15,12 @@ rapids-print-env rapids-logger "Begin cpp build" +sccache --zero-stats + # With boa installed conda build forward to boa RAPIDS_PACKAGE_VERSION=$(rapids-generate-version) rapids-conda-retry mambabuild \ conda/recipes/libcudf +sccache --show-adv-stats + rapids-upload-conda-to-s3 cpp diff --git a/ci/build_python.sh b/ci/build_python.sh index 823d7f62290..ed90041cc77 100755 --- a/ci/build_python.sh +++ b/ci/build_python.sh @@ -19,6 +19,8 @@ rapids-logger "Begin py build" CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp) +sccache --zero-stats + # TODO: Remove `--no-test` flag once importing on a CPU # node works correctly # With boa installed conda build forwards to the boa builder @@ -28,12 +30,18 @@ RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION) rapids-conda-retry mambabuild \ --channel "${CPP_CHANNEL}" \ conda/recipes/pylibcudf +sccache --show-adv-stats +sccache --zero-stats + RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION) rapids-conda-retry mambabuild \ --no-test \ --channel "${CPP_CHANNEL}" \ --channel "${RAPIDS_CONDA_BLD_OUTPUT_DIR}" \ conda/recipes/cudf +sccache --show-adv-stats +sccache --zero-stats + RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION) rapids-conda-retry mambabuild \ --no-test \ --channel "${CPP_CHANNEL}" \ @@ -46,6 +54,8 @@ RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION) rapids-conda-retry mambabuild \ --channel "${RAPIDS_CONDA_BLD_OUTPUT_DIR}" \ conda/recipes/cudf_kafka +sccache --show-adv-stats + RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION) rapids-conda-retry mambabuild \ --no-test \ --channel "${CPP_CHANNEL}" \ diff --git a/ci/build_wheel.sh b/ci/build_wheel.sh index bf76f4ed29a..78b8a8a08cf 100755 --- a/ci/build_wheel.sh +++ b/ci/build_wheel.sh @@ -3,7 +3,8 @@ set -euo pipefail -package_dir=$1 +package_name=$1 +package_dir=$2 source rapids-configure-sccache source rapids-date-string @@ -12,4 +13,14 @@ rapids-generate-version > ./VERSION cd "${package_dir}" -python -m pip wheel . -w dist -v --no-deps --disable-pip-version-check +sccache --zero-stats + +rapids-logger "Building '${package_name}' wheel" +python -m pip wheel \ + -w dist \ + -v \ + --no-deps \ + --disable-pip-version-check \ + . + +sccache --show-adv-stats diff --git a/ci/build_wheel_cudf.sh b/ci/build_wheel_cudf.sh index fb93b06dbe2..fef4416a366 100755 --- a/ci/build_wheel_cudf.sh +++ b/ci/build_wheel_cudf.sh @@ -18,7 +18,7 @@ echo "libcudf-${RAPIDS_PY_CUDA_SUFFIX} @ file://$(echo /tmp/libcudf_dist/libcudf echo "pylibcudf-${RAPIDS_PY_CUDA_SUFFIX} @ file://$(echo /tmp/pylibcudf_dist/pylibcudf_*.whl)" >> /tmp/constraints.txt export PIP_CONSTRAINT="/tmp/constraints.txt" -./ci/build_wheel.sh ${package_dir} +./ci/build_wheel.sh cudf ${package_dir} python -m auditwheel repair \ --exclude libcudf.so \ diff --git a/ci/build_wheel_cudf_polars.sh b/ci/build_wheel_cudf_polars.sh index 9c945e11c00..79853cdbdb2 100755 --- a/ci/build_wheel_cudf_polars.sh +++ b/ci/build_wheel_cudf_polars.sh @@ -5,7 +5,7 @@ set -euo pipefail package_dir="python/cudf_polars" -./ci/build_wheel.sh ${package_dir} +./ci/build_wheel.sh cudf-polars ${package_dir} RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})" -RAPIDS_PY_WHEEL_NAME="cudf_polars_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 ${package_dir}/dist +RAPIDS_PY_WHEEL_NAME="cudf_polars_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 python ${package_dir}/dist diff --git a/ci/build_wheel_dask_cudf.sh b/ci/build_wheel_dask_cudf.sh index eb2a91289f7..00c64afa2ef 100755 --- a/ci/build_wheel_dask_cudf.sh +++ b/ci/build_wheel_dask_cudf.sh @@ -5,7 +5,7 @@ set -euo pipefail package_dir="python/dask_cudf" -./ci/build_wheel.sh ${package_dir} +./ci/build_wheel.sh dask-cudf ${package_dir} RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})" -RAPIDS_PY_WHEEL_NAME="dask_cudf_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 ${package_dir}/dist +RAPIDS_PY_WHEEL_NAME="dask_cudf_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 python ${package_dir}/dist diff --git a/ci/build_wheel_libcudf.sh b/ci/build_wheel_libcudf.sh index 91bc071583e..b3d6778ea04 100755 --- a/ci/build_wheel_libcudf.sh +++ b/ci/build_wheel_libcudf.sh @@ -3,10 +3,30 @@ set -euo pipefail +package_name="libcudf" package_dir="python/libcudf" +rapids-logger "Generating build requirements" + +rapids-dependency-file-generator \ + --output requirements \ + --file-key "py_build_${package_name}" \ + --file-key "py_rapids_build_${package_name}" \ + --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION};cuda_suffixed=true" \ +| tee /tmp/requirements-build.txt + +rapids-logger "Installing build requirements" +python -m pip install \ + -v \ + --prefer-binary \ + -r /tmp/requirements-build.txt + +# build with '--no-build-isolation', for better sccache hit rate +# 0 really means "add --no-build-isolation" (ref: https://github.com/pypa/pip/issues/5735) +export PIP_NO_BUILD_ISOLATION=0 + export SKBUILD_CMAKE_ARGS="-DUSE_NVCOMP_RUNTIME_WHEEL=ON" -./ci/build_wheel.sh ${package_dir} +./ci/build_wheel.sh "${package_name}" "${package_dir}" RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})" @@ -16,4 +36,4 @@ python -m auditwheel repair \ -w ${package_dir}/final_dist \ ${package_dir}/dist/* -RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 cpp ${package_dir}/final_dist +RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 cpp "${package_dir}/final_dist" diff --git a/ci/build_wheel_pylibcudf.sh b/ci/build_wheel_pylibcudf.sh index 5e9f7f8a0c4..839d98846fe 100755 --- a/ci/build_wheel_pylibcudf.sh +++ b/ci/build_wheel_pylibcudf.sh @@ -16,7 +16,7 @@ RAPIDS_PY_WHEEL_NAME="libcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-f echo "libcudf-${RAPIDS_PY_CUDA_SUFFIX} @ file://$(echo /tmp/libcudf_dist/libcudf_*.whl)" > /tmp/constraints.txt export PIP_CONSTRAINT="/tmp/constraints.txt" -./ci/build_wheel.sh ${package_dir} +./ci/build_wheel.sh pylibcudf ${package_dir} python -m auditwheel repair \ --exclude libcudf.so \ @@ -24,4 +24,4 @@ python -m auditwheel repair \ -w ${package_dir}/final_dist \ ${package_dir}/dist/* -RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 ${package_dir}/final_dist +RAPIDS_PY_WHEEL_NAME="pylibcudf_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 python ${package_dir}/final_dist diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml index bd5e6c3d569..c3716c4759a 100644 --- a/conda/environments/all_cuda-118_arch-x86_64.yaml +++ b/conda/environments/all_cuda-118_arch-x86_64.yaml @@ -65,7 +65,7 @@ dependencies: - pandas - pandas>=2.0,<2.2.4dev0 - pandoc -- polars>=1.8,<1.9 +- polars>=1.11,<1.12 - pre-commit - ptxcompiler - pyarrow>=14.0.0,<18.0.0a0 diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml index 565a3ebfa3c..38e131e79cb 100644 --- a/conda/environments/all_cuda-125_arch-x86_64.yaml +++ b/conda/environments/all_cuda-125_arch-x86_64.yaml @@ -63,7 +63,7 @@ dependencies: - pandas - pandas>=2.0,<2.2.4dev0 - pandoc -- polars>=1.8,<1.9 +- polars>=1.11,<1.12 - pre-commit - pyarrow>=14.0.0,<18.0.0a0 - pydata-sphinx-theme!=0.14.2 diff --git a/conda/recipes/cudf-polars/meta.yaml b/conda/recipes/cudf-polars/meta.yaml index e8fef715c60..edf92b930d9 100644 --- a/conda/recipes/cudf-polars/meta.yaml +++ b/conda/recipes/cudf-polars/meta.yaml @@ -43,7 +43,7 @@ requirements: run: - python - pylibcudf ={{ version }} - - polars >=1.8,<1.9 + - polars >=1.11,<1.12 - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }} test: diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt index e4b9cbf8921..60132f651d2 100644 --- a/cpp/CMakeLists.txt +++ b/cpp/CMakeLists.txt @@ -369,6 +369,8 @@ add_library( src/filling/sequence.cu src/groupby/groupby.cu src/groupby/hash/compute_groupby.cu + src/groupby/hash/compute_mapping_indices.cu + src/groupby/hash/compute_mapping_indices_null.cu src/groupby/hash/compute_single_pass_aggs.cu src/groupby/hash/create_sparse_results_table.cu src/groupby/hash/flatten_single_pass_aggs.cpp diff --git a/cpp/benchmarks/CMakeLists.txt b/cpp/benchmarks/CMakeLists.txt index e61a8e6e1e6..2a4ac789046 100644 --- a/cpp/benchmarks/CMakeLists.txt +++ b/cpp/benchmarks/CMakeLists.txt @@ -49,7 +49,7 @@ target_compile_options( target_link_libraries( ndsh_data_generator - PUBLIC cudf GTest::gmock GTest::gtest cudf::cudftestutil nvtx3::nvtx3-cpp + PUBLIC cudf cudf::cudftestutil nvtx3::nvtx3-cpp PRIVATE $ ) @@ -345,11 +345,11 @@ ConfigureNVBench(BINARYOP_NVBENCH binaryop/binaryop.cpp binaryop/compiled_binary # ################################################################################################## # * nvtext benchmark ------------------------------------------------------------------- -ConfigureBench(TEXT_BENCH text/ngrams.cpp text/subword.cpp) +ConfigureBench(TEXT_BENCH text/subword.cpp) ConfigureNVBench( - TEXT_NVBENCH text/edit_distance.cpp text/hash_ngrams.cpp text/jaccard.cpp text/minhash.cpp - text/normalize.cpp text/replace.cpp text/tokenize.cpp text/vocab.cpp text/word_minhash.cpp + TEXT_NVBENCH text/edit_distance.cpp text/hash_ngrams.cpp text/jaccard.cpp text/ngrams.cpp + text/normalize.cpp text/replace.cpp text/tokenize.cpp text/vocab.cpp ) # ################################################################################################## diff --git a/cpp/benchmarks/ast/transform.cpp b/cpp/benchmarks/ast/transform.cpp index f44f26e4d2c..7fe61054a26 100644 --- a/cpp/benchmarks/ast/transform.cpp +++ b/cpp/benchmarks/ast/transform.cpp @@ -16,16 +16,29 @@ #include +#include + +#include +#include +#include +#include #include #include +#include +#include #include #include #include +#include #include +#include +#include +#include +#include #include #include #include @@ -86,7 +99,71 @@ static void BM_ast_transform(nvbench::state& state) auto const& expression_tree_root = expressions.back(); // Use the number of bytes read from global memory - state.add_global_memory_reads(table_size * (tree_levels + 1)); + state.add_global_memory_reads(static_cast(table_size) * (tree_levels + 1)); + state.add_global_memory_writes(table_size); + + state.exec(nvbench::exec_tag::sync, + [&](nvbench::launch&) { cudf::compute_column(table, expression_tree_root); }); +} + +template +static void BM_string_compare_ast_transform(nvbench::state& state) +{ + auto const string_width = static_cast(state.get_int64("string_width")); + auto const num_rows = static_cast(state.get_int64("num_rows")); + auto const num_comparisons = static_cast(state.get_int64("num_comparisons")); + auto const hit_rate = static_cast(state.get_int64("hit_rate")); + + CUDF_EXPECTS(num_comparisons > 0, "benchmarks require 1 or more comparisons"); + + // Create table data + auto const num_cols = num_comparisons * 2; + std::vector> columns; + std::for_each( + thrust::make_counting_iterator(0), thrust::make_counting_iterator(num_cols), [&](size_t) { + columns.emplace_back(create_string_column(num_rows, string_width, hit_rate)); + }); + + cudf::table table{std::move(columns)}; + cudf::table_view const table_view = table.view(); + + int64_t const chars_size = std::accumulate( + table_view.begin(), + table_view.end(), + static_cast(0), + [](int64_t size, auto& column) -> int64_t { + return size + cudf::strings_column_view{column}.chars_size(cudf::get_default_stream()); + }); + + // Create column references + auto column_refs = std::vector(); + std::transform(thrust::make_counting_iterator(0), + thrust::make_counting_iterator(num_cols), + std::back_inserter(column_refs), + [](auto const& column_id) { return cudf::ast::column_reference(column_id); }); + + // Create expression trees + std::list expressions; + + // Construct AST tree (a == b && c == d && e == f && ...) + + expressions.emplace_back(cudf::ast::operation(cmp_op, column_refs[0], column_refs[1])); + + std::for_each(thrust::make_counting_iterator(1), + thrust::make_counting_iterator(num_comparisons), + [&](size_t idx) { + auto const& lhs = expressions.back(); + auto const& rhs = expressions.emplace_back( + cudf::ast::operation(cmp_op, column_refs[idx * 2], column_refs[idx * 2 + 1])); + expressions.emplace_back(cudf::ast::operation(reduce_op, lhs, rhs)); + }); + + auto const& expression_tree_root = expressions.back(); + + // Use the number of bytes read from global memory + state.add_element_count(chars_size, "chars_size"); + state.add_global_memory_reads(chars_size); + state.add_global_memory_writes(num_rows); state.exec(nvbench::exec_tag::sync, [&](nvbench::launch&) { cudf::compute_column(table, expression_tree_root); }); @@ -115,3 +192,19 @@ AST_TRANSFORM_BENCHMARK_DEFINE( ast_int32_imbalanced_reuse_nulls, int32_t, TreeType::IMBALANCED_LEFT, true, true); AST_TRANSFORM_BENCHMARK_DEFINE( ast_double_imbalanced_unique_nulls, double, TreeType::IMBALANCED_LEFT, false, true); + +#define AST_STRING_COMPARE_TRANSFORM_BENCHMARK_DEFINE(name, cmp_op, reduce_op) \ + static void name(::nvbench::state& st) \ + { \ + ::BM_string_compare_ast_transform(st); \ + } \ + NVBENCH_BENCH(name) \ + .set_name(#name) \ + .add_int64_axis("string_width", {32, 64, 128, 256}) \ + .add_int64_axis("num_rows", {32768, 262144, 2097152}) \ + .add_int64_axis("num_comparisons", {1, 2, 3, 4}) \ + .add_int64_axis("hit_rate", {50, 100}) + +AST_STRING_COMPARE_TRANSFORM_BENCHMARK_DEFINE(ast_string_equal_logical_and, + cudf::ast::ast_operator::EQUAL, + cudf::ast::ast_operator::LOGICAL_AND); diff --git a/cpp/benchmarks/binaryop/binaryop.cpp b/cpp/benchmarks/binaryop/binaryop.cpp index 7d267a88764..35e41c6c2a4 100644 --- a/cpp/benchmarks/binaryop/binaryop.cpp +++ b/cpp/benchmarks/binaryop/binaryop.cpp @@ -17,12 +17,18 @@ #include #include +#include +#include #include #include +#include + #include #include +#include +#include // This set of benchmarks is designed to be a comparison for the AST benchmarks @@ -44,7 +50,8 @@ static void BM_binaryop_transform(nvbench::state& state) cudf::table_view table{*source_table}; // Use the number of bytes read from global memory - state.add_global_memory_reads(table_size * (tree_levels + 1)); + state.add_global_memory_reads(static_cast(table_size) * (tree_levels + 1)); + state.add_global_memory_writes(table_size); state.exec(nvbench::exec_tag::sync, [&](nvbench::launch&) { // Execute tree that chains additions like (((a + b) + c) + d) @@ -64,11 +71,65 @@ static void BM_binaryop_transform(nvbench::state& state) }); } +template +static void BM_string_compare_binaryop_transform(nvbench::state& state) +{ + auto const string_width = static_cast(state.get_int64("string_width")); + auto const num_rows = static_cast(state.get_int64("num_rows")); + auto const num_comparisons = static_cast(state.get_int64("num_comparisons")); + auto const hit_rate = static_cast(state.get_int64("hit_rate")); + + CUDF_EXPECTS(num_comparisons > 0, "benchmarks require 1 or more comparisons"); + + // Create table data + auto const num_cols = num_comparisons * 2; + std::vector> columns; + std::for_each( + thrust::make_counting_iterator(0), thrust::make_counting_iterator(num_cols), [&](size_t) { + columns.emplace_back(create_string_column(num_rows, string_width, hit_rate)); + }); + + cudf::table table{std::move(columns)}; + cudf::table_view const table_view = table.view(); + + int64_t const chars_size = std::accumulate( + table_view.begin(), table_view.end(), static_cast(0), [](int64_t size, auto& column) { + return size + cudf::strings_column_view{column}.chars_size(cudf::get_default_stream()); + }); + + // Create column references + + // Use the number of bytes read from global memory + state.add_element_count(chars_size, "chars_size"); + state.add_global_memory_reads(chars_size); + state.add_global_memory_writes(num_rows); + + // Construct binary operations (a == b && c == d && e == f && ...) + auto constexpr bool_type = cudf::data_type{cudf::type_id::BOOL8}; + + state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) { + rmm::cuda_stream_view stream{launch.get_stream().get_stream()}; + std::unique_ptr reduction = + cudf::binary_operation(table.get_column(0), table.get_column(1), cmp_op, bool_type, stream); + std::for_each( + thrust::make_counting_iterator(1), + thrust::make_counting_iterator(num_comparisons), + [&](size_t idx) { + std::unique_ptr comparison = cudf::binary_operation( + table.get_column(idx * 2), table.get_column(idx * 2 + 1), cmp_op, bool_type, stream); + std::unique_ptr reduced = + cudf::binary_operation(*comparison, *reduction, reduce_op, bool_type, stream); + stream.synchronize(); + reduction = std::move(reduced); + }); + }); +} + #define BINARYOP_TRANSFORM_BENCHMARK_DEFINE(name, key_type, tree_type, reuse_columns) \ \ static void name(::nvbench::state& st) \ { \ - BM_binaryop_transform(st); \ + ::BM_binaryop_transform(st); \ } \ NVBENCH_BENCH(name) \ .add_int64_axis("tree_levels", {1, 2, 5, 10}) \ @@ -86,3 +147,20 @@ BINARYOP_TRANSFORM_BENCHMARK_DEFINE(binaryop_double_imbalanced_unique, double, TreeType::IMBALANCED_LEFT, false); + +#define STRING_COMPARE_BINARYOP_TRANSFORM_BENCHMARK_DEFINE(name, cmp_op, reduce_op) \ + \ + static void name(::nvbench::state& st) \ + { \ + ::BM_string_compare_binaryop_transform(st); \ + } \ + NVBENCH_BENCH(name) \ + .set_name(#name) \ + .add_int64_axis("string_width", {32, 64, 128, 256}) \ + .add_int64_axis("num_rows", {32768, 262144, 2097152}) \ + .add_int64_axis("num_comparisons", {1, 2, 3, 4}) \ + .add_int64_axis("hit_rate", {50, 100}) + +STRING_COMPARE_BINARYOP_TRANSFORM_BENCHMARK_DEFINE(string_compare_binaryop_transform, + cudf::binary_operator::EQUAL, + cudf::binary_operator::LOGICAL_AND); diff --git a/cpp/benchmarks/binaryop/compiled_binaryop.cpp b/cpp/benchmarks/binaryop/compiled_binaryop.cpp index bc0ff69bce9..cd3c3871a2e 100644 --- a/cpp/benchmarks/binaryop/compiled_binaryop.cpp +++ b/cpp/benchmarks/binaryop/compiled_binaryop.cpp @@ -39,7 +39,7 @@ void BM_compiled_binaryop(nvbench::state& state, cudf::binary_operator binop) // use number of bytes read and written to global memory state.add_global_memory_reads(table_size); state.add_global_memory_reads(table_size); - state.add_global_memory_reads(table_size); + state.add_global_memory_writes(table_size); state.exec(nvbench::exec_tag::sync, [&](nvbench::launch&) { cudf::binary_operation(lhs, rhs, binop, output_dtype); }); diff --git a/cpp/benchmarks/common/generate_input.cu b/cpp/benchmarks/common/generate_input.cu index dc258e32dc5..bdce8a31176 100644 --- a/cpp/benchmarks/common/generate_input.cu +++ b/cpp/benchmarks/common/generate_input.cu @@ -17,13 +17,17 @@ #include "generate_input.hpp" #include "random_distribution_factory.cuh" +#include + #include #include +#include #include #include #include #include #include +#include #include #include #include @@ -918,6 +922,58 @@ std::unique_ptr create_sequence_table(std::vector co return std::make_unique(std::move(columns)); } +std::unique_ptr create_string_column(cudf::size_type num_rows, + cudf::size_type row_width, + int32_t hit_rate) +{ + // build input table using the following data + auto raw_data = cudf::test::strings_column_wrapper( + { + "123 abc 4567890 DEFGHI 0987 5W43", // matches both patterns; + "012345 6789 01234 56789 0123 456", // the rest do not match + "abc 4567890 DEFGHI 0987 Wxyz 123", + "abcdefghijklmnopqrstuvwxyz 01234", + "", + "AbcéDEFGHIJKLMNOPQRSTUVWXYZ 01", + "9876543210,abcdefghijklmnopqrstU", + "9876543210,abcdefghijklmnopqrstU", + "123 édf 4567890 DéFG 0987 X5", + "1", + }) + .release(); + + if (row_width / 32 > 1) { + std::vector columns; + for (int i = 0; i < row_width / 32; ++i) { + columns.push_back(raw_data->view()); + } + raw_data = cudf::strings::concatenate(cudf::table_view(columns)); + } + auto data_view = raw_data->view(); + + // compute number of rows in n_rows that should match + auto const num_matches = (static_cast(num_rows) * hit_rate) / 100; + + // Create a randomized gather-map to build a column out of the strings in data. + data_profile gather_profile = + data_profile_builder().cardinality(0).null_probability(0.0).distribution( + cudf::type_id::INT32, distribution_id::UNIFORM, 1, data_view.size() - 1); + auto gather_table = + create_random_table({cudf::type_id::INT32}, row_count{num_rows}, gather_profile); + gather_table->get_column(0).set_null_mask(rmm::device_buffer{}, 0); + + // Create scatter map by placing 0-index values throughout the gather-map + auto scatter_data = cudf::sequence(num_matches, + cudf::numeric_scalar(0), + cudf::numeric_scalar(num_rows / num_matches)); + auto zero_scalar = cudf::numeric_scalar(0); + auto table = cudf::scatter({zero_scalar}, scatter_data->view(), gather_table->view()); + auto gather_map = table->view().column(0); + table = cudf::gather(cudf::table_view({data_view}), gather_map); + + return std::move(table->release().front()); +} + std::pair create_random_null_mask( cudf::size_type size, std::optional null_probability, unsigned seed) { diff --git a/cpp/benchmarks/common/generate_input.hpp b/cpp/benchmarks/common/generate_input.hpp index 68d3dc492f5..57834fd11d2 100644 --- a/cpp/benchmarks/common/generate_input.hpp +++ b/cpp/benchmarks/common/generate_input.hpp @@ -670,6 +670,18 @@ std::unique_ptr create_random_column(cudf::type_id dtype_id, data_profile const& data_params = data_profile{}, unsigned seed = 1); +/** + * @brief Deterministically generates a large string column filled with data with the given + * parameters. + * + * @param num_rows Number of rows in the output column + * @param row_width Width of each string in the column + * @param hit_rate The hit rate percentage, ranging from 0 to 100 + */ +std::unique_ptr create_string_column(cudf::size_type num_rows, + cudf::size_type row_width, + int32_t hit_rate); + /** * @brief Generate sequence columns starting with value 0 in first row and increasing by 1 in * subsequent rows. diff --git a/cpp/benchmarks/io/cuio_common.cpp b/cpp/benchmarks/io/cuio_common.cpp index fe24fb58728..45b46005c47 100644 --- a/cpp/benchmarks/io/cuio_common.cpp +++ b/cpp/benchmarks/io/cuio_common.cpp @@ -186,7 +186,7 @@ std::string exec_cmd(std::string_view cmd) std::fflush(nullptr); // Switch stderr and stdout to only capture stderr auto const redirected_cmd = std::string{"( "}.append(cmd).append(" 3>&2 2>&1 1>&3) 2>/dev/null"); - std::unique_ptr pipe(popen(redirected_cmd.c_str(), "r"), pclose); + std::unique_ptr pipe(popen(redirected_cmd.c_str(), "r"), pclose); CUDF_EXPECTS(pipe != nullptr, "popen() failed"); std::array buffer; diff --git a/cpp/benchmarks/ndsh/q01.cpp b/cpp/benchmarks/ndsh/q01.cpp index ef709926ae9..485e8e5497c 100644 --- a/cpp/benchmarks/ndsh/q01.cpp +++ b/cpp/benchmarks/ndsh/q01.cpp @@ -104,7 +104,7 @@ } void run_ndsh_q1(nvbench::state& state, - std::unordered_map& sources) + std::unordered_map& sources) { // Define the column projections and filter predicate for `lineitem` table std::vector const lineitem_cols = {"l_returnflag", @@ -124,8 +124,8 @@ void run_ndsh_q1(nvbench::state& state, cudf::ast::ast_operator::LESS_EQUAL, shipdate_ref, shipdate_upper_literal); // Read out the `lineitem` table from parquet file - auto lineitem = - read_parquet(sources["lineitem"].make_source_info(), lineitem_cols, std::move(lineitem_pred)); + auto lineitem = read_parquet( + sources.at("lineitem").make_source_info(), lineitem_cols, std::move(lineitem_pred)); // Calculate the discount price and charge columns and append to lineitem table auto disc_price = @@ -170,7 +170,7 @@ void ndsh_q1(nvbench::state& state) { // Generate the required parquet files in device buffers double const scale_factor = state.get_float64("scale_factor"); - std::unordered_map sources; + std::unordered_map sources; generate_parquet_data_sources(scale_factor, {"lineitem"}, sources); auto stream = cudf::get_default_stream(); diff --git a/cpp/benchmarks/ndsh/q05.cpp b/cpp/benchmarks/ndsh/q05.cpp index 522bc4789c2..1c2d657913e 100644 --- a/cpp/benchmarks/ndsh/q05.cpp +++ b/cpp/benchmarks/ndsh/q05.cpp @@ -89,7 +89,7 @@ } void run_ndsh_q5(nvbench::state& state, - std::unordered_map& sources) + std::unordered_map& sources) { // Define the column projection and filter predicate for the `orders` table std::vector const orders_cols = {"o_custkey", "o_orderkey", "o_orderdate"}; @@ -120,17 +120,17 @@ void run_ndsh_q5(nvbench::state& state, // Read out the tables from parquet files // while pushing down the column projections and filter predicates auto const customer = - read_parquet(sources["customer"].make_source_info(), {"c_custkey", "c_nationkey"}); + read_parquet(sources.at("customer").make_source_info(), {"c_custkey", "c_nationkey"}); auto const orders = - read_parquet(sources["orders"].make_source_info(), orders_cols, std::move(orders_pred)); - auto const lineitem = read_parquet(sources["lineitem"].make_source_info(), + read_parquet(sources.at("orders").make_source_info(), orders_cols, std::move(orders_pred)); + auto const lineitem = read_parquet(sources.at("lineitem").make_source_info(), {"l_orderkey", "l_suppkey", "l_extendedprice", "l_discount"}); auto const supplier = - read_parquet(sources["supplier"].make_source_info(), {"s_suppkey", "s_nationkey"}); + read_parquet(sources.at("supplier").make_source_info(), {"s_suppkey", "s_nationkey"}); auto const nation = - read_parquet(sources["nation"].make_source_info(), {"n_nationkey", "n_regionkey", "n_name"}); + read_parquet(sources.at("nation").make_source_info(), {"n_nationkey", "n_regionkey", "n_name"}); auto const region = - read_parquet(sources["region"].make_source_info(), region_cols, std::move(region_pred)); + read_parquet(sources.at("region").make_source_info(), region_cols, std::move(region_pred)); // Perform the joins auto const join_a = apply_inner_join(region, nation, {"r_regionkey"}, {"n_regionkey"}); @@ -165,7 +165,7 @@ void ndsh_q5(nvbench::state& state) { // Generate the required parquet files in device buffers double const scale_factor = state.get_float64("scale_factor"); - std::unordered_map sources; + std::unordered_map sources; generate_parquet_data_sources( scale_factor, {"customer", "orders", "lineitem", "supplier", "nation", "region"}, sources); diff --git a/cpp/benchmarks/ndsh/q06.cpp b/cpp/benchmarks/ndsh/q06.cpp index 04078547973..e1e56c3622e 100644 --- a/cpp/benchmarks/ndsh/q06.cpp +++ b/cpp/benchmarks/ndsh/q06.cpp @@ -64,7 +64,7 @@ } void run_ndsh_q6(nvbench::state& state, - std::unordered_map& sources) + std::unordered_map& sources) { // Read out the `lineitem` table from parquet file std::vector const lineitem_cols = { @@ -83,8 +83,8 @@ void run_ndsh_q6(nvbench::state& state, cudf::ast::operation(cudf::ast::ast_operator::LESS, shipdate_ref, shipdate_upper_literal); auto const lineitem_pred = std::make_unique( cudf::ast::ast_operator::LOGICAL_AND, shipdate_pred_a, shipdate_pred_b); - auto lineitem = - read_parquet(sources["lineitem"].make_source_info(), lineitem_cols, std::move(lineitem_pred)); + auto lineitem = read_parquet( + sources.at("lineitem").make_source_info(), lineitem_cols, std::move(lineitem_pred)); // Cast the discount and quantity columns to float32 and append to lineitem table auto discout_float = @@ -134,7 +134,7 @@ void ndsh_q6(nvbench::state& state) { // Generate the required parquet files in device buffers double const scale_factor = state.get_float64("scale_factor"); - std::unordered_map sources; + std::unordered_map sources; generate_parquet_data_sources(scale_factor, {"lineitem"}, sources); auto stream = cudf::get_default_stream(); diff --git a/cpp/benchmarks/ndsh/q09.cpp b/cpp/benchmarks/ndsh/q09.cpp index 59218ab8912..2e9a69d9ee2 100644 --- a/cpp/benchmarks/ndsh/q09.cpp +++ b/cpp/benchmarks/ndsh/q09.cpp @@ -112,20 +112,21 @@ } void run_ndsh_q9(nvbench::state& state, - std::unordered_map& sources) + std::unordered_map& sources) { // Read out the table from parquet files auto const lineitem = read_parquet( - sources["lineitem"].make_source_info(), + sources.at("lineitem").make_source_info(), {"l_suppkey", "l_partkey", "l_orderkey", "l_extendedprice", "l_discount", "l_quantity"}); - auto const nation = read_parquet(sources["nation"].make_source_info(), {"n_nationkey", "n_name"}); + auto const nation = + read_parquet(sources.at("nation").make_source_info(), {"n_nationkey", "n_name"}); auto const orders = - read_parquet(sources["orders"].make_source_info(), {"o_orderkey", "o_orderdate"}); - auto const part = read_parquet(sources["part"].make_source_info(), {"p_partkey", "p_name"}); - auto const partsupp = read_parquet(sources["partsupp"].make_source_info(), + read_parquet(sources.at("orders").make_source_info(), {"o_orderkey", "o_orderdate"}); + auto const part = read_parquet(sources.at("part").make_source_info(), {"p_partkey", "p_name"}); + auto const partsupp = read_parquet(sources.at("partsupp").make_source_info(), {"ps_suppkey", "ps_partkey", "ps_supplycost"}); auto const supplier = - read_parquet(sources["supplier"].make_source_info(), {"s_suppkey", "s_nationkey"}); + read_parquet(sources.at("supplier").make_source_info(), {"s_suppkey", "s_nationkey"}); // Generating the `profit` table // Filter the part table using `p_name like '%green%'` @@ -178,7 +179,7 @@ void ndsh_q9(nvbench::state& state) { // Generate the required parquet files in device buffers double const scale_factor = state.get_float64("scale_factor"); - std::unordered_map sources; + std::unordered_map sources; generate_parquet_data_sources( scale_factor, {"part", "supplier", "lineitem", "partsupp", "orders", "nation"}, sources); diff --git a/cpp/benchmarks/ndsh/q10.cpp b/cpp/benchmarks/ndsh/q10.cpp index a520480020a..72edd15083d 100644 --- a/cpp/benchmarks/ndsh/q10.cpp +++ b/cpp/benchmarks/ndsh/q10.cpp @@ -94,7 +94,7 @@ } void run_ndsh_q10(nvbench::state& state, - std::unordered_map& sources) + std::unordered_map& sources) { // Define the column projection and filter predicate for the `orders` table std::vector const orders_cols = {"o_custkey", "o_orderkey", "o_orderdate"}; @@ -122,15 +122,16 @@ void run_ndsh_q10(nvbench::state& state, // Read out the tables from parquet files // while pushing down the column projections and filter predicates auto const customer = read_parquet( - sources["customer"].make_source_info(), + sources.at("customer").make_source_info(), {"c_custkey", "c_name", "c_nationkey", "c_acctbal", "c_address", "c_phone", "c_comment"}); auto const orders = - read_parquet(sources["orders"].make_source_info(), orders_cols, std::move(orders_pred)); + read_parquet(sources.at("orders").make_source_info(), orders_cols, std::move(orders_pred)); auto const lineitem = - read_parquet(sources["lineitem"].make_source_info(), + read_parquet(sources.at("lineitem").make_source_info(), {"l_extendedprice", "l_discount", "l_orderkey", "l_returnflag"}, std::move(lineitem_pred)); - auto const nation = read_parquet(sources["nation"].make_source_info(), {"n_name", "n_nationkey"}); + auto const nation = + read_parquet(sources.at("nation").make_source_info(), {"n_name", "n_nationkey"}); // Perform the joins auto const join_a = apply_inner_join(customer, nation, {"c_nationkey"}, {"n_nationkey"}); @@ -163,7 +164,7 @@ void ndsh_q10(nvbench::state& state) { // Generate the required parquet files in device buffers double const scale_factor = state.get_float64("scale_factor"); - std::unordered_map sources; + std::unordered_map sources; generate_parquet_data_sources( scale_factor, {"customer", "orders", "lineitem", "nation"}, sources); diff --git a/cpp/benchmarks/ndsh/utilities.cpp b/cpp/benchmarks/ndsh/utilities.cpp index 62116ddf661..9f9849860c9 100644 --- a/cpp/benchmarks/ndsh/utilities.cpp +++ b/cpp/benchmarks/ndsh/utilities.cpp @@ -17,6 +17,8 @@ #include "utilities.hpp" #include "common/ndsh_data_generator/ndsh_data_generator.hpp" +#include "common/table_utilities.hpp" +#include "cudf/detail/utilities/integer_utils.hpp" #include #include @@ -30,8 +32,15 @@ #include #include +#include +#include +#include + +#include #include #include +#include +#include namespace { @@ -85,6 +94,15 @@ std::vector const NATION_SCHEMA = { "n_nationkey", "n_name", "n_regionkey", "n_comment"}; std::vector const REGION_SCHEMA = {"r_regionkey", "r_name", "r_comment"}; +std::unordered_map const> const SCHEMAS = { + {"orders", ORDERS_SCHEMA}, + {"lineitem", LINEITEM_SCHEMA}, + {"part", PART_SCHEMA}, + {"partsupp", PARTSUPP_SCHEMA}, + {"supplier", SUPPLIER_SCHEMA}, + {"customer", CUSTOMER_SCHEMA}, + {"nation", NATION_SCHEMA}, + {"region", REGION_SCHEMA}}; } // namespace cudf::table_view table_with_names::table() const { return tbl->view(); } @@ -337,7 +355,7 @@ int32_t days_since_epoch(int year, int month, int day) void write_to_parquet_device_buffer(std::unique_ptr const& table, std::vector const& col_names, - parquet_device_buffer& source) + cuio_source_sink_pair& source) { CUDF_FUNC_RANGE(); auto const stream = cudf::get_default_stream(); @@ -351,55 +369,124 @@ void write_to_parquet_device_buffer(std::unique_ptr const& table, metadata.schema_info = col_name_infos; auto const table_input_metadata = cudf::io::table_input_metadata{metadata}; - // Declare a host and device buffer - std::vector h_buffer; - + auto est_size = static_cast(estimate_size(table->view())); + constexpr auto PQ_MAX_TABLE_BYTES = 8ul << 30; // 8GB + // TODO: best to get this limit from percent_of_free_device_memory(50) of device memory resource. + if (est_size > PQ_MAX_TABLE_BYTES) { + auto builder = cudf::io::chunked_parquet_writer_options::builder(source.make_sink_info()); + builder.metadata(table_input_metadata); + auto const options = builder.build(); + auto num_splits = static_cast( + std::ceil(static_cast(est_size) / (PQ_MAX_TABLE_BYTES))); + std::vector splits(num_splits - 1); + auto num_rows = table->num_rows(); + auto num_row_per_chunk = cudf::util::div_rounding_up_safe(num_rows, num_splits); + std::generate_n(splits.begin(), splits.size(), [num_row_per_chunk, i = 0]() mutable { + return (i += num_row_per_chunk); + }); + std::vector split_tables = cudf::split(table->view(), splits, stream); + auto writer = cudf::io::parquet_chunked_writer(options, stream); + for (auto const& chunk_table : split_tables) { + writer.write(chunk_table); + } + writer.close(); + return; + } // Write parquet data to host buffer - auto builder = - cudf::io::parquet_writer_options::builder(cudf::io::sink_info(&h_buffer), table->view()); + auto builder = cudf::io::parquet_writer_options::builder(source.make_sink_info(), table->view()); builder.metadata(table_input_metadata); auto const options = builder.build(); - cudf::io::write_parquet(options); + cudf::io::write_parquet(options, stream); +} - // Copy host buffer to device buffer - source.d_buffer.resize(h_buffer.size(), stream); - CUDF_CUDA_TRY(cudaMemcpyAsync( - source.d_buffer.data(), h_buffer.data(), h_buffer.size(), cudaMemcpyDefault, stream.value())); +inline auto make_managed_pool() +{ + return rmm::mr::make_owning_wrapper( + std::make_shared(), rmm::percent_of_free_device_memory(50)); } void generate_parquet_data_sources(double scale_factor, std::vector const& table_names, - std::unordered_map& sources) + std::unordered_map& sources) { CUDF_FUNC_RANGE(); - std::for_each(table_names.begin(), table_names.end(), [&](auto const& table_name) { - sources[table_name] = parquet_device_buffer(); - }); - auto [orders, lineitem, part] = cudf::datagen::generate_orders_lineitem_part( - scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + // Set the memory resource to the managed pool + auto old_mr = cudf::get_current_device_resource(); + // if already managed pool or managed, don't create new one. + using managed_pool_mr_t = decltype(make_managed_pool()); + managed_pool_mr_t managed_pool_mr; + bool const is_managed = + dynamic_cast*>(old_mr) or + dynamic_cast(old_mr); + if (!is_managed) { + std::cout << "Creating managed pool just for data generation\n"; + managed_pool_mr = make_managed_pool(); + cudf::set_current_device_resource(managed_pool_mr.get()); + // drawback: if already pool takes 50% of free memory, we are left with 50% of 50% of free + // memory. + } - auto partsupp = cudf::datagen::generate_partsupp( - scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + std::unordered_set const requested_table_names = [&table_names]() { + if (table_names.empty()) { + return std::unordered_set{ + "orders", "lineitem", "part", "partsupp", "supplier", "customer", "nation", "region"}; + } + return std::unordered_set(table_names.begin(), table_names.end()); + }(); + std::for_each( + requested_table_names.begin(), requested_table_names.end(), [&](auto const& table_name) { + sources.emplace(table_name, cuio_source_sink_pair(io_type::HOST_BUFFER)); + }); + std::unordered_map> tables; + + if (sources.count("orders") or sources.count("lineitem") or sources.count("part")) { + auto [orders, lineitem, part] = cudf::datagen::generate_orders_lineitem_part( + scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + if (sources.count("orders")) { + write_to_parquet_device_buffer(orders, SCHEMAS.at("orders"), sources.at("orders")); + orders = {}; + } + if (sources.count("part")) { + write_to_parquet_device_buffer(part, SCHEMAS.at("part"), sources.at("part")); + part = {}; + } + if (sources.count("lineitem")) { + write_to_parquet_device_buffer(lineitem, SCHEMAS.at("lineitem"), sources.at("lineitem")); + lineitem = {}; + } + } + + if (sources.count("partsupp")) { + auto partsupp = cudf::datagen::generate_partsupp( + scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + write_to_parquet_device_buffer(partsupp, SCHEMAS.at("partsupp"), sources.at("partsupp")); + } - auto supplier = cudf::datagen::generate_supplier( - scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + if (sources.count("supplier")) { + auto supplier = cudf::datagen::generate_supplier( + scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + write_to_parquet_device_buffer(supplier, SCHEMAS.at("supplier"), sources.at("supplier")); + } - auto customer = cudf::datagen::generate_customer( - scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + if (sources.count("customer")) { + auto customer = cudf::datagen::generate_customer( + scale_factor, cudf::get_default_stream(), cudf::get_current_device_resource_ref()); + write_to_parquet_device_buffer(customer, SCHEMAS.at("customer"), sources.at("customer")); + } - auto nation = cudf::datagen::generate_nation(cudf::get_default_stream(), - cudf::get_current_device_resource_ref()); + if (sources.count("nation")) { + auto nation = cudf::datagen::generate_nation(cudf::get_default_stream(), + cudf::get_current_device_resource_ref()); + write_to_parquet_device_buffer(nation, SCHEMAS.at("nation"), sources.at("nation")); + } - auto region = cudf::datagen::generate_region(cudf::get_default_stream(), - cudf::get_current_device_resource_ref()); + if (sources.count("region")) { + auto region = cudf::datagen::generate_region(cudf::get_default_stream(), + cudf::get_current_device_resource_ref()); + write_to_parquet_device_buffer(region, SCHEMAS.at("region"), sources.at("region")); + } - write_to_parquet_device_buffer(std::move(orders), ORDERS_SCHEMA, sources["orders"]); - write_to_parquet_device_buffer(std::move(lineitem), LINEITEM_SCHEMA, sources["lineitem"]); - write_to_parquet_device_buffer(std::move(part), PART_SCHEMA, sources["part"]); - write_to_parquet_device_buffer(std::move(partsupp), PARTSUPP_SCHEMA, sources["partsupp"]); - write_to_parquet_device_buffer(std::move(customer), CUSTOMER_SCHEMA, sources["customer"]); - write_to_parquet_device_buffer(std::move(supplier), SUPPLIER_SCHEMA, sources["supplier"]); - write_to_parquet_device_buffer(std::move(nation), NATION_SCHEMA, sources["nation"]); - write_to_parquet_device_buffer(std::move(region), REGION_SCHEMA, sources["region"]); + // Restore the original memory resource + if (!is_managed) { cudf::set_current_device_resource(old_mr); } } diff --git a/cpp/benchmarks/ndsh/utilities.hpp b/cpp/benchmarks/ndsh/utilities.hpp index 762e43deccf..cae07f86a98 100644 --- a/cpp/benchmarks/ndsh/utilities.hpp +++ b/cpp/benchmarks/ndsh/utilities.hpp @@ -14,6 +14,8 @@ * limitations under the License. */ +#include "io/cuio_common.hpp" + #include #include #include @@ -196,24 +198,15 @@ std::tm make_tm(int year, int month, int day); int32_t days_since_epoch(int year, int month, int day); /** - * @brief Struct representing a parquet device buffer - */ -struct parquet_device_buffer { - parquet_device_buffer() : d_buffer{0, cudf::get_default_stream()} {}; - cudf::io::source_info make_source_info() { return cudf::io::source_info(d_buffer); } - rmm::device_uvector d_buffer; -}; - -/** - * @brief Write a `cudf::table` to a parquet device buffer + * @brief Write a `cudf::table` to a parquet cuio sink * * @param table The `cudf::table` to write * @param col_names The column names of the table - * @param parquet_device_buffer The parquet device buffer to write the table to + * @param source The source sink pair to write the table to */ void write_to_parquet_device_buffer(std::unique_ptr const& table, std::vector const& col_names, - parquet_device_buffer& source); + cuio_source_sink_pair& source); /** * @brief Generate NDS-H tables and write to parquet device buffers @@ -224,4 +217,4 @@ void write_to_parquet_device_buffer(std::unique_ptr const& table, */ void generate_parquet_data_sources(double scale_factor, std::vector const& table_names, - std::unordered_map& sources); + std::unordered_map& sources); diff --git a/cpp/benchmarks/string/contains.cpp b/cpp/benchmarks/string/contains.cpp index ae6c8b844c8..a73017dda18 100644 --- a/cpp/benchmarks/string/contains.cpp +++ b/cpp/benchmarks/string/contains.cpp @@ -17,10 +17,6 @@ #include #include -#include - -#include -#include #include #include #include @@ -28,57 +24,6 @@ #include -std::unique_ptr build_input_column(cudf::size_type n_rows, - cudf::size_type row_width, - int32_t hit_rate) -{ - // build input table using the following data - auto raw_data = cudf::test::strings_column_wrapper( - { - "123 abc 4567890 DEFGHI 0987 5W43", // matches both patterns; - "012345 6789 01234 56789 0123 456", // the rest do not match - "abc 4567890 DEFGHI 0987 Wxyz 123", - "abcdefghijklmnopqrstuvwxyz 01234", - "", - "AbcéDEFGHIJKLMNOPQRSTUVWXYZ 01", - "9876543210,abcdefghijklmnopqrstU", - "9876543210,abcdefghijklmnopqrstU", - "123 édf 4567890 DéFG 0987 X5", - "1", - }) - .release(); - - if (row_width / 32 > 1) { - std::vector columns; - for (int i = 0; i < row_width / 32; ++i) { - columns.push_back(raw_data->view()); - } - raw_data = cudf::strings::concatenate(cudf::table_view(columns)); - } - auto data_view = raw_data->view(); - - // compute number of rows in n_rows that should match - auto matches = static_cast(n_rows * hit_rate) / 100; - - // Create a randomized gather-map to build a column out of the strings in data. - data_profile gather_profile = - data_profile_builder().cardinality(0).null_probability(0.0).distribution( - cudf::type_id::INT32, distribution_id::UNIFORM, 1, data_view.size() - 1); - auto gather_table = - create_random_table({cudf::type_id::INT32}, row_count{n_rows}, gather_profile); - gather_table->get_column(0).set_null_mask(rmm::device_buffer{}, 0); - - // Create scatter map by placing 0-index values throughout the gather-map - auto scatter_data = cudf::sequence( - matches, cudf::numeric_scalar(0), cudf::numeric_scalar(n_rows / matches)); - auto zero_scalar = cudf::numeric_scalar(0); - auto table = cudf::scatter({zero_scalar}, scatter_data->view(), gather_table->view()); - auto gather_map = table->view().column(0); - table = cudf::gather(cudf::table_view({data_view}), gather_map); - - return std::move(table->release().front()); -} - // longer pattern lengths demand more working memory per string std::string patterns[] = {"^\\d+ [a-z]+", "[A-Z ]+\\d+ +\\d+[A-Z]+\\d+$", "5W43"}; @@ -94,7 +39,7 @@ static void bench_contains(nvbench::state& state) state.skip("Skip benchmarks greater than size_type limit"); } - auto col = build_input_column(n_rows, row_width, hit_rate); + auto col = create_string_column(n_rows, row_width, hit_rate); auto input = cudf::strings_column_view(col->view()); auto pattern = patterns[pattern_index]; diff --git a/cpp/benchmarks/string/find.cpp b/cpp/benchmarks/string/find.cpp index a9c620e4bf0..996bdcf0332 100644 --- a/cpp/benchmarks/string/find.cpp +++ b/cpp/benchmarks/string/find.cpp @@ -19,7 +19,6 @@ #include -#include #include #include #include @@ -29,10 +28,6 @@ #include -std::unique_ptr build_input_column(cudf::size_type n_rows, - cudf::size_type row_width, - int32_t hit_rate); - static void bench_find_string(nvbench::state& state) { auto const n_rows = static_cast(state.get_int64("num_rows")); @@ -46,7 +41,7 @@ static void bench_find_string(nvbench::state& state) } auto const stream = cudf::get_default_stream(); - auto const col = build_input_column(n_rows, row_width, hit_rate); + auto const col = create_string_column(n_rows, row_width, hit_rate); auto const input = cudf::strings_column_view(col->view()); std::vector h_targets({"5W", "5W43", "0987 5W43"}); diff --git a/cpp/benchmarks/string/like.cpp b/cpp/benchmarks/string/like.cpp index 99cef640dc3..105ae65cbe8 100644 --- a/cpp/benchmarks/string/like.cpp +++ b/cpp/benchmarks/string/like.cpp @@ -18,68 +18,12 @@ #include -#include -#include -#include #include #include #include #include -namespace { -std::unique_ptr build_input_column(cudf::size_type n_rows, - cudf::size_type row_width, - int32_t hit_rate) -{ - // build input table using the following data - auto raw_data = cudf::test::strings_column_wrapper( - { - "123 abc 4567890 DEFGHI 0987 5W43", // matches always; - "012345 6789 01234 56789 0123 456", // the rest do not match - "abc 4567890 DEFGHI 0987 Wxyz 123", - "abcdefghijklmnopqrstuvwxyz 01234", - "", - "AbcéDEFGHIJKLMNOPQRSTUVWXYZ 01", - "9876543210,abcdefghijklmnopqrstU", - "9876543210,abcdefghijklmnopqrstU", - "123 édf 4567890 DéFG 0987 X5", - "1", - }) - .release(); - if (row_width / 32 > 1) { - std::vector columns; - for (int i = 0; i < row_width / 32; ++i) { - columns.push_back(raw_data->view()); - } - raw_data = cudf::strings::concatenate(cudf::table_view(columns)); - } - auto data_view = raw_data->view(); - - // compute number of rows in n_rows that should match - auto matches = static_cast(n_rows * hit_rate) / 100; - - // Create a randomized gather-map to build a column out of the strings in data. - data_profile gather_profile = - data_profile_builder().cardinality(0).null_probability(0.0).distribution( - cudf::type_id::INT32, distribution_id::UNIFORM, 1, data_view.size() - 1); - auto gather_table = - create_random_table({cudf::type_id::INT32}, row_count{n_rows}, gather_profile); - gather_table->get_column(0).set_null_mask(rmm::device_buffer{}, 0); - - // Create scatter map by placing 0-index values throughout the gather-map - auto scatter_data = cudf::sequence( - matches, cudf::numeric_scalar(0), cudf::numeric_scalar(n_rows / matches)); - auto zero_scalar = cudf::numeric_scalar(0); - auto table = cudf::scatter({zero_scalar}, scatter_data->view(), gather_table->view()); - auto gather_map = table->view().column(0); - table = cudf::gather(cudf::table_view({data_view}), gather_map); - - return std::move(table->release().front()); -} - -} // namespace - static void bench_like(nvbench::state& state) { auto const n_rows = static_cast(state.get_int64("num_rows")); @@ -91,7 +35,7 @@ static void bench_like(nvbench::state& state) state.skip("Skip benchmarks greater than size_type limit"); } - auto col = build_input_column(n_rows, row_width, hit_rate); + auto col = create_string_column(n_rows, row_width, hit_rate); auto input = cudf::strings_column_view(col->view()); // This pattern forces reading the entire target string (when matched expected) diff --git a/cpp/benchmarks/text/ngrams.cpp b/cpp/benchmarks/text/ngrams.cpp index 8e48f8e9a05..43d57201b20 100644 --- a/cpp/benchmarks/text/ngrams.cpp +++ b/cpp/benchmarks/text/ngrams.cpp @@ -15,58 +15,45 @@ */ #include -#include -#include -#include #include #include #include -class TextNGrams : public cudf::benchmark {}; +#include -enum class ngrams_type { tokens, characters }; - -static void BM_ngrams(benchmark::State& state, ngrams_type nt) +static void bench_ngrams(nvbench::state& state) { - auto const n_rows = static_cast(state.range(0)); - auto const max_str_length = static_cast(state.range(1)); + auto const num_rows = static_cast(state.get_int64("num_rows")); + auto const row_width = static_cast(state.get_int64("row_width")); + auto const ngram_type = state.get_string("type"); + data_profile const profile = data_profile_builder().distribution( - cudf::type_id::STRING, distribution_id::NORMAL, 0, max_str_length); - auto const column = create_random_column(cudf::type_id::STRING, row_count{n_rows}, profile); + cudf::type_id::STRING, distribution_id::NORMAL, 0, row_width); + auto const column = create_random_column(cudf::type_id::STRING, row_count{num_rows}, profile); cudf::strings_column_view input(column->view()); auto const separator = cudf::string_scalar("_"); - for (auto _ : state) { - cuda_event_timer raii(state, true); - switch (nt) { - case ngrams_type::tokens: nvtext::generate_ngrams(input, 2, separator); break; - case ngrams_type::characters: nvtext::generate_character_ngrams(input); break; - } - } + state.set_cuda_stream(nvbench::make_cuda_stream_view(cudf::get_default_stream().value())); - state.SetBytesProcessed(state.iterations() * input.chars_size(cudf::get_default_stream())); -} + auto chars_size = input.chars_size(cudf::get_default_stream()); + state.add_global_memory_reads(chars_size); + state.add_global_memory_writes(chars_size * 2); -static void generate_bench_args(benchmark::internal::Benchmark* b) -{ - int const min_rows = 1 << 12; - int const max_rows = 1 << 24; - int const row_mult = 8; - int const min_rowlen = 5; - int const max_rowlen = 40; - int const len_mult = 2; - generate_string_bench_args(b, min_rows, max_rows, row_mult, min_rowlen, max_rowlen, len_mult); + if (ngram_type == "chars") { + state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) { + auto result = nvtext::generate_character_ngrams(input); + }); + } else { + state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) { + auto result = nvtext::generate_ngrams(input, 2, separator); + }); + } } -#define NVTEXT_BENCHMARK_DEFINE(name) \ - BENCHMARK_DEFINE_F(TextNGrams, name) \ - (::benchmark::State & st) { BM_ngrams(st, ngrams_type::name); } \ - BENCHMARK_REGISTER_F(TextNGrams, name) \ - ->Apply(generate_bench_args) \ - ->UseManualTime() \ - ->Unit(benchmark::kMillisecond); - -NVTEXT_BENCHMARK_DEFINE(tokens) -NVTEXT_BENCHMARK_DEFINE(characters) +NVBENCH_BENCH(bench_ngrams) + .set_name("ngrams") + .add_int64_axis("num_rows", {131072, 262144, 524288, 1048578}) + .add_int64_axis("row_width", {10, 20, 40, 100}) + .add_string_axis("type", {"chars", "tokens"}); diff --git a/cpp/include/cudf/ast/detail/expression_parser.hpp b/cpp/include/cudf/ast/detail/expression_parser.hpp index a254171ef11..f4cce8e6da6 100644 --- a/cpp/include/cudf/ast/detail/expression_parser.hpp +++ b/cpp/include/cudf/ast/detail/expression_parser.hpp @@ -17,12 +17,8 @@ #include #include -#include #include #include -#include - -#include #include #include diff --git a/cpp/include/cudf/column/column_factories.hpp b/cpp/include/cudf/column/column_factories.hpp index 6bbe32de134..e72661ce49a 100644 --- a/cpp/include/cudf/column/column_factories.hpp +++ b/cpp/include/cudf/column/column_factories.hpp @@ -24,8 +24,6 @@ #include -#include - namespace CUDF_EXPORT cudf { /** * @addtogroup column_factories diff --git a/cpp/include/cudf/column/column_view.hpp b/cpp/include/cudf/column/column_view.hpp index 48f89b8be25..6db5c8b3c7b 100644 --- a/cpp/include/cudf/column/column_view.hpp +++ b/cpp/include/cudf/column/column_view.hpp @@ -16,7 +16,6 @@ #pragma once #include -#include #include #include #include diff --git a/cpp/include/cudf/detail/aggregation/result_cache.hpp b/cpp/include/cudf/detail/aggregation/result_cache.hpp index ec5a511bb7c..486808ebe18 100644 --- a/cpp/include/cudf/detail/aggregation/result_cache.hpp +++ b/cpp/include/cudf/detail/aggregation/result_cache.hpp @@ -19,7 +19,6 @@ #include #include #include -#include #include diff --git a/cpp/include/cudf/detail/is_element_valid.hpp b/cpp/include/cudf/detail/is_element_valid.hpp index 4b74d12f306..26b1bec2ced 100644 --- a/cpp/include/cudf/detail/is_element_valid.hpp +++ b/cpp/include/cudf/detail/is_element_valid.hpp @@ -17,7 +17,6 @@ #pragma once #include -#include #include #include diff --git a/cpp/include/cudf/detail/iterator.cuh b/cpp/include/cudf/detail/iterator.cuh index 4349e1b70fd..30f36d6a5da 100644 --- a/cpp/include/cudf/detail/iterator.cuh +++ b/cpp/include/cudf/detail/iterator.cuh @@ -38,18 +38,19 @@ #include #include +#include +#include #include #include #include #include -#include - namespace cudf { namespace detail { /** * @brief Convenience wrapper for creating a `thrust::transform_iterator` over a - * `thrust::counting_iterator`. + * `thrust::counting_iterator` within the range [0, INT_MAX]. + * * * Example: * @code{.cpp} @@ -62,14 +63,21 @@ namespace detail { * iter[n] == n * n * @endcode * - * @param start The starting value of the counting iterator + * @param start The starting value of the counting iterator (must be size_type or smaller type). * @param f The unary function to apply to the counting iterator. * @return A transform iterator that applies `f` to a counting iterator */ -template -CUDF_HOST_DEVICE inline auto make_counting_transform_iterator(cudf::size_type start, +template +CUDF_HOST_DEVICE inline auto make_counting_transform_iterator(CountingIterType start, UnaryFunction f) { + // Check if the `start` for counting_iterator is of size_type or a smaller integral type + static_assert( + cuda::std::is_integral_v and + cuda::std::numeric_limits::digits <= + cuda::std::numeric_limits::digits, + "The `start` for the counting_transform_iterator must be size_type or smaller type"); + return thrust::make_transform_iterator(thrust::make_counting_iterator(start), f); } diff --git a/cpp/include/cudf/detail/null_mask.cuh b/cpp/include/cudf/detail/null_mask.cuh index 482265d633e..025e2ccc3ec 100644 --- a/cpp/include/cudf/detail/null_mask.cuh +++ b/cpp/include/cudf/detail/null_mask.cuh @@ -166,16 +166,9 @@ size_type inplace_bitmask_binop(Binop op, rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref(); cudf::detail::device_scalar d_counter{0, stream, mr}; - rmm::device_uvector d_masks(masks.size(), stream, mr); - rmm::device_uvector d_begin_bits(masks_begin_bits.size(), stream, mr); - - CUDF_CUDA_TRY(cudaMemcpyAsync( - d_masks.data(), masks.data(), masks.size_bytes(), cudaMemcpyDefault, stream.value())); - CUDF_CUDA_TRY(cudaMemcpyAsync(d_begin_bits.data(), - masks_begin_bits.data(), - masks_begin_bits.size_bytes(), - cudaMemcpyDefault, - stream.value())); + + auto d_masks = cudf::detail::make_device_uvector_async(masks, stream, mr); + auto d_begin_bits = cudf::detail::make_device_uvector_async(masks_begin_bits, stream, mr); auto constexpr block_size = 256; cudf::detail::grid_1d config(dest_mask.size(), block_size); diff --git a/cpp/include/cudf/detail/unary.hpp b/cpp/include/cudf/detail/unary.hpp index 18b1e9b2d2e..0f852db0c54 100644 --- a/cpp/include/cudf/detail/unary.hpp +++ b/cpp/include/cudf/detail/unary.hpp @@ -59,7 +59,7 @@ std::unique_ptr true_if(InputIterator begin, auto output_mutable_view = output->mutable_view(); auto output_data = output_mutable_view.data(); - thrust::transform(rmm::exec_policy(stream), begin, end, output_data, p); + thrust::transform(rmm::exec_policy_nosync(stream), begin, end, output_data, p); return output; } diff --git a/cpp/include/cudf/detail/utilities/batched_memset.hpp b/cpp/include/cudf/detail/utilities/batched_memset.hpp index 75f738f7529..78be5b91248 100644 --- a/cpp/include/cudf/detail/utilities/batched_memset.hpp +++ b/cpp/include/cudf/detail/utilities/batched_memset.hpp @@ -53,8 +53,8 @@ void batched_memset(std::vector> const& bufs, cudf::detail::make_device_uvector_async(bufs, stream, cudf::get_current_device_resource_ref()); // get a vector with the sizes of all buffers - auto sizes = cudf::detail::make_counting_transform_iterator( - static_cast(0), + auto sizes = thrust::make_transform_iterator( + thrust::counting_iterator(0), cuda::proclaim_return_type( [gpu_bufs = gpu_bufs.data()] __device__(std::size_t i) { return gpu_bufs[i].size(); })); diff --git a/cpp/include/cudf/dictionary/dictionary_column_view.hpp b/cpp/include/cudf/dictionary/dictionary_column_view.hpp index 5596f78a90b..0a799f27d00 100644 --- a/cpp/include/cudf/dictionary/dictionary_column_view.hpp +++ b/cpp/include/cudf/dictionary/dictionary_column_view.hpp @@ -15,7 +15,6 @@ */ #pragma once -#include #include /** diff --git a/cpp/include/cudf/io/config_utils.hpp b/cpp/include/cudf/io/config_utils.hpp index 1827ba0e3e6..13a76d50346 100644 --- a/cpp/include/cudf/io/config_utils.hpp +++ b/cpp/include/cudf/io/config_utils.hpp @@ -18,7 +18,8 @@ #include namespace CUDF_EXPORT cudf { -namespace io::cufile_integration { +namespace io { +namespace cufile_integration { /** * @brief Returns true if cuFile and its compatibility mode are enabled. @@ -35,9 +36,15 @@ bool is_gds_enabled(); */ bool is_kvikio_enabled(); -} // namespace io::cufile_integration +/** + * @brief Set kvikIO thread pool size according to the environment variable KVIKIO_NTHREADS. If + * KVIKIO_NTHREADS is not set, use 8 threads by default. + */ +void set_thread_pool_nthreads_from_env(); + +} // namespace cufile_integration -namespace io::nvcomp_integration { +namespace nvcomp_integration { /** * @brief Returns true if all nvCOMP uses are enabled. @@ -49,5 +56,6 @@ bool is_all_enabled(); */ bool is_stable_enabled(); -} // namespace io::nvcomp_integration +} // namespace nvcomp_integration +} // namespace io } // namespace CUDF_EXPORT cudf diff --git a/cpp/include/cudf/io/text/detail/bgzip_utils.hpp b/cpp/include/cudf/io/text/detail/bgzip_utils.hpp index 11eb4518210..5659f86b0c4 100644 --- a/cpp/include/cudf/io/text/detail/bgzip_utils.hpp +++ b/cpp/include/cudf/io/text/detail/bgzip_utils.hpp @@ -16,16 +16,10 @@ #pragma once -#include #include #include -#include - -#include -#include #include -#include namespace CUDF_EXPORT cudf { namespace io::text::detail::bgzip { diff --git a/cpp/include/cudf/strings/detail/strings_children.cuh b/cpp/include/cudf/strings/detail/strings_children.cuh index fb0b25cf9f1..de2f1770e28 100644 --- a/cpp/include/cudf/strings/detail/strings_children.cuh +++ b/cpp/include/cudf/strings/detail/strings_children.cuh @@ -65,19 +65,20 @@ rmm::device_uvector make_chars_buffer(column_view const& offsets, auto chars_data = rmm::device_uvector(chars_size, stream, mr); auto const d_offsets = cudf::detail::offsetalator_factory::make_input_iterator(offsets); - auto const src_ptrs = cudf::detail::make_counting_transform_iterator( - 0u, cuda::proclaim_return_type([begin] __device__(uint32_t idx) { + auto const src_ptrs = thrust::make_transform_iterator( + thrust::make_counting_iterator(0), + cuda::proclaim_return_type([begin] __device__(uint32_t idx) { // Due to a bug in cub (https://github.com/NVIDIA/cccl/issues/586), // we have to use `const_cast` to remove `const` qualifier from the source pointer. // This should be fine as long as we only read but not write anything to the source. return reinterpret_cast(const_cast(begin[idx].first)); })); - auto const src_sizes = cudf::detail::make_counting_transform_iterator( - 0u, cuda::proclaim_return_type([begin] __device__(uint32_t idx) { - return begin[idx].second; - })); - auto const dst_ptrs = cudf::detail::make_counting_transform_iterator( - 0u, + auto const src_sizes = thrust::make_transform_iterator( + thrust::make_counting_iterator(0), + cuda::proclaim_return_type( + [begin] __device__(uint32_t idx) { return begin[idx].second; })); + auto const dst_ptrs = thrust::make_transform_iterator( + thrust::make_counting_iterator(0), cuda::proclaim_return_type([offsets = d_offsets, output = chars_data.data()] __device__( uint32_t idx) { return output + offsets[idx]; })); diff --git a/cpp/include/cudf/utilities/default_stream.hpp b/cpp/include/cudf/utilities/default_stream.hpp index 97a42243250..3e740b81cc9 100644 --- a/cpp/include/cudf/utilities/default_stream.hpp +++ b/cpp/include/cudf/utilities/default_stream.hpp @@ -16,10 +16,8 @@ #pragma once -#include #include -#include #include namespace CUDF_EXPORT cudf { diff --git a/cpp/include/cudf/utilities/traits.hpp b/cpp/include/cudf/utilities/traits.hpp index 3f37ae02151..cf8413b597f 100644 --- a/cpp/include/cudf/utilities/traits.hpp +++ b/cpp/include/cudf/utilities/traits.hpp @@ -22,8 +22,6 @@ #include #include -#include - namespace CUDF_EXPORT cudf { /** diff --git a/cpp/include/cudf/utilities/type_dispatcher.hpp b/cpp/include/cudf/utilities/type_dispatcher.hpp index 15b5f921c1b..6351a84e38f 100644 --- a/cpp/include/cudf/utilities/type_dispatcher.hpp +++ b/cpp/include/cudf/utilities/type_dispatcher.hpp @@ -16,7 +16,6 @@ #pragma once -#include #include #include #include diff --git a/cpp/include/nvtext/minhash.hpp b/cpp/include/nvtext/minhash.hpp index 7c909f1a948..42124461cdf 100644 --- a/cpp/include/nvtext/minhash.hpp +++ b/cpp/include/nvtext/minhash.hpp @@ -41,6 +41,8 @@ namespace CUDF_EXPORT nvtext { * * This function uses MurmurHash3_x86_32 for the hash algorithm. * + * @deprecated Deprecated in 24.12 + * * @throw std::invalid_argument if the width < 2 * * @param input Strings column to compute minhash @@ -51,7 +53,7 @@ namespace CUDF_EXPORT nvtext { * @param mr Device memory resource used to allocate the returned column's device memory * @return Minhash values for each string in input */ -std::unique_ptr minhash( +[[deprecated]] std::unique_ptr minhash( cudf::strings_column_view const& input, cudf::numeric_scalar seed = 0, cudf::size_type width = 4, @@ -71,6 +73,8 @@ std::unique_ptr minhash( * * Any null row entries result in corresponding null output rows. * + * @deprecated Deprecated in 24.12 - to be replaced in a future release + * * @throw std::invalid_argument if the width < 2 * @throw std::invalid_argument if seeds is empty * @throw std::overflow_error if `seeds.size() * input.size()` exceeds the column size limit @@ -83,7 +87,7 @@ std::unique_ptr minhash( * @param mr Device memory resource used to allocate the returned column's device memory * @return List column of minhash values for each string per seed */ -std::unique_ptr minhash( +[[deprecated]] std::unique_ptr minhash( cudf::strings_column_view const& input, cudf::device_span seeds, cudf::size_type width = 4, @@ -102,6 +106,8 @@ std::unique_ptr minhash( * The hash function returns 2 uint64 values but only the first value * is used with the minhash calculation. * + * @deprecated Deprecated in 24.12 + * * @throw std::invalid_argument if the width < 2 * * @param input Strings column to compute minhash @@ -112,7 +118,7 @@ std::unique_ptr minhash( * @param mr Device memory resource used to allocate the returned column's device memory * @return Minhash values as UINT64 for each string in input */ -std::unique_ptr minhash64( +[[deprecated]] std::unique_ptr minhash64( cudf::strings_column_view const& input, cudf::numeric_scalar seed = 0, cudf::size_type width = 4, @@ -132,6 +138,8 @@ std::unique_ptr minhash64( * * Any null row entries result in corresponding null output rows. * + * @deprecated Deprecated in 24.12 - to be replaced in a future release + * * @throw std::invalid_argument if the width < 2 * @throw std::invalid_argument if seeds is empty * @throw std::overflow_error if `seeds.size() * input.size()` exceeds the column size limit @@ -144,7 +152,7 @@ std::unique_ptr minhash64( * @param mr Device memory resource used to allocate the returned column's device memory * @return List column of minhash values for each string per seed */ -std::unique_ptr minhash64( +[[deprecated]] std::unique_ptr minhash64( cudf::strings_column_view const& input, cudf::device_span seeds, cudf::size_type width = 4, @@ -164,6 +172,8 @@ std::unique_ptr minhash64( * * Any null row entries result in corresponding null output rows. * + * @deprecated Deprecated in 24.12 + * * @throw std::invalid_argument if seeds is empty * @throw std::overflow_error if `seeds.size() * input.size()` exceeds the column size limit * @@ -173,7 +183,7 @@ std::unique_ptr minhash64( * @param mr Device memory resource used to allocate the returned column's device memory * @return List column of minhash values for each string per seed */ -std::unique_ptr word_minhash( +[[deprecated]] std::unique_ptr word_minhash( cudf::lists_column_view const& input, cudf::device_span seeds, rmm::cuda_stream_view stream = cudf::get_default_stream(), @@ -193,6 +203,8 @@ std::unique_ptr word_minhash( * * Any null row entries result in corresponding null output rows. * + * @deprecated Deprecated in 24.12 + * * @throw std::invalid_argument if seeds is empty * @throw std::overflow_error if `seeds.size() * input.size()` exceeds the column size limit * @@ -202,7 +214,7 @@ std::unique_ptr word_minhash( * @param mr Device memory resource used to allocate the returned column's device memory * @return List column of minhash values for each string per seed */ -std::unique_ptr word_minhash64( +[[deprecated]] std::unique_ptr word_minhash64( cudf::lists_column_view const& input, cudf::device_span seeds, rmm::cuda_stream_view stream = cudf::get_default_stream(), diff --git a/cpp/include/nvtext/replace.hpp b/cpp/include/nvtext/replace.hpp index bbd0503379b..822edcbdb43 100644 --- a/cpp/include/nvtext/replace.hpp +++ b/cpp/include/nvtext/replace.hpp @@ -82,7 +82,7 @@ namespace CUDF_EXPORT nvtext { * The default of empty string will identify tokens using whitespace. * @param stream CUDA stream used for device memory operations and kernel launches * @param mr Device memory resource used to allocate the returned column's device memory - * @return New strings columns of with replaced strings + * @return New strings column with replaced strings */ std::unique_ptr replace_tokens( cudf::strings_column_view const& input, @@ -131,7 +131,7 @@ std::unique_ptr replace_tokens( * The default of empty string will identify tokens using whitespace. * @param stream CUDA stream used for device memory operations and kernel launches * @param mr Device memory resource used to allocate the returned column's device memory - * @return New strings columns of with replaced strings + * @return New strings column of filtered strings */ std::unique_ptr filter_tokens( cudf::strings_column_view const& input, diff --git a/cpp/include/nvtext/stemmer.hpp b/cpp/include/nvtext/stemmer.hpp index 55a4124bfd0..e5b2a4cc21b 100644 --- a/cpp/include/nvtext/stemmer.hpp +++ b/cpp/include/nvtext/stemmer.hpp @@ -51,7 +51,7 @@ enum class letter_type { * * @code{.pseudo} * Example: - * st = ["trouble", "toy", "sygyzy"] + * st = ["trouble", "toy", "syzygy"] * b1 = is_letter(st, VOWEL, 1) * b1 is now [false, true, true] * @endcode @@ -62,7 +62,7 @@ enum class letter_type { * * @code{.pseudo} * Example: - * st = ["trouble", "toy", "sygyzy"] + * st = ["trouble", "toy", "syzygy"] * b2 = is_letter(st, CONSONANT, -1) // last letter checked in each string * b2 is now [false, true, false] * @endcode @@ -99,7 +99,7 @@ std::unique_ptr is_letter( * * @code{.pseudo} * Example: - * st = ["trouble", "toy", "sygyzy"] + * st = ["trouble", "toy", "syzygy"] * ix = [3, 1, 4] * b1 = is_letter(st, VOWEL, ix) * b1 is now [true, true, false] @@ -111,7 +111,7 @@ std::unique_ptr is_letter( * * @code{.pseudo} * Example: - * st = ["trouble", "toy", "sygyzy"] + * st = ["trouble", "toy", "syzygy"] * ix = [3, -2, 4] // 2nd to last character in st[1] is checked * b2 = is_letter(st, CONSONANT, ix) * b2 is now [false, false, true] diff --git a/cpp/src/ast/expression_parser.cpp b/cpp/src/ast/expression_parser.cpp index 3b650d791aa..5815ce33e33 100644 --- a/cpp/src/ast/expression_parser.cpp +++ b/cpp/src/ast/expression_parser.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,9 +16,6 @@ #include #include #include -#include -#include -#include #include #include #include diff --git a/cpp/src/ast/expressions.cpp b/cpp/src/ast/expressions.cpp index b45b9d0c78c..4c2b56dd4f5 100644 --- a/cpp/src/ast/expressions.cpp +++ b/cpp/src/ast/expressions.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,9 +17,6 @@ #include #include #include -#include -#include -#include #include #include diff --git a/cpp/src/binaryop/binaryop.cpp b/cpp/src/binaryop/binaryop.cpp index a6c878efbbc..1b23ea12a5e 100644 --- a/cpp/src/binaryop/binaryop.cpp +++ b/cpp/src/binaryop/binaryop.cpp @@ -27,15 +27,10 @@ #include #include #include -#include #include -#include #include #include -#include -#include #include -#include #include #include diff --git a/cpp/src/column/column_factories.cpp b/cpp/src/column/column_factories.cpp index 482413d0ccb..972f97e8668 100644 --- a/cpp/src/column/column_factories.cpp +++ b/cpp/src/column/column_factories.cpp @@ -15,19 +15,13 @@ */ #include -#include #include #include #include -#include #include -#include #include -#include #include -#include - namespace cudf { namespace { struct size_of_helper { diff --git a/cpp/src/column/column_view.cpp b/cpp/src/column/column_view.cpp index 386c5ebe478..e831aa9645d 100644 --- a/cpp/src/column/column_view.cpp +++ b/cpp/src/column/column_view.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include #include @@ -27,9 +26,7 @@ #include #include -#include #include -#include #include namespace cudf { diff --git a/cpp/src/copying/copy.cpp b/cpp/src/copying/copy.cpp index d60fb5ce110..5e2065ba844 100644 --- a/cpp/src/copying/copy.cpp +++ b/cpp/src/copying/copy.cpp @@ -20,16 +20,11 @@ #include #include #include -#include #include -#include -#include #include #include -#include - #include namespace cudf { diff --git a/cpp/src/copying/pack.cpp b/cpp/src/copying/pack.cpp index 1282eec6c44..a001807c82b 100644 --- a/cpp/src/copying/pack.cpp +++ b/cpp/src/copying/pack.cpp @@ -18,7 +18,6 @@ #include #include #include -#include #include diff --git a/cpp/src/copying/split.cpp b/cpp/src/copying/split.cpp index 832a72ed5b0..116e3516460 100644 --- a/cpp/src/copying/split.cpp +++ b/cpp/src/copying/split.cpp @@ -14,10 +14,8 @@ * limitations under the License. */ -#include #include #include -#include #include #include diff --git a/cpp/src/datetime/timezone.cpp b/cpp/src/datetime/timezone.cpp index 2196ee97fee..f786624680c 100644 --- a/cpp/src/datetime/timezone.cpp +++ b/cpp/src/datetime/timezone.cpp @@ -13,12 +13,10 @@ * See the License for the specific language governing permissions and * limitations under the License. */ -#include #include #include #include #include -#include #include #include diff --git a/cpp/src/groupby/hash/compute_mapping_indices.cu b/cpp/src/groupby/hash/compute_mapping_indices.cu new file mode 100644 index 00000000000..519d7cd2f1c --- /dev/null +++ b/cpp/src/groupby/hash/compute_mapping_indices.cu @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2024, NVIDIA CORPORATION. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "compute_mapping_indices.cuh" +#include "compute_mapping_indices.hpp" + +namespace cudf::groupby::detail::hash { +template cudf::size_type max_occupancy_grid_size>( + cudf::size_type n); + +template void compute_mapping_indices>( + cudf::size_type grid_size, + cudf::size_type num, + hash_set_ref_t global_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* local_mapping_index, + cudf::size_type* global_mapping_index, + cudf::size_type* block_cardinality, + cuda::std::atomic_flag* needs_global_memory_fallback, + rmm::cuda_stream_view stream); +} // namespace cudf::groupby::detail::hash diff --git a/cpp/src/groupby/hash/compute_mapping_indices.cuh b/cpp/src/groupby/hash/compute_mapping_indices.cuh new file mode 100644 index 00000000000..d353830780f --- /dev/null +++ b/cpp/src/groupby/hash/compute_mapping_indices.cuh @@ -0,0 +1,192 @@ +/* + * Copyright (c) 2024, NVIDIA CORPORATION. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#pragma once + +#include "compute_mapping_indices.hpp" +#include "helpers.cuh" + +#include +#include +#include +#include + +#include + +#include +#include +#include +#include + +#include + +namespace cudf::groupby::detail::hash { +template +__device__ void find_local_mapping(cooperative_groups::thread_block const& block, + cudf::size_type idx, + cudf::size_type num_input_rows, + SetType shared_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* cardinality, + cudf::size_type* local_mapping_index, + cudf::size_type* shared_set_indices) +{ + auto const is_valid_input = + idx < num_input_rows and (not skip_rows_with_nulls or cudf::bit_is_set(row_bitmask, idx)); + auto const [result_idx, inserted] = [&]() { + if (is_valid_input) { + auto const result = shared_set.insert_and_find(idx); + auto const matched_idx = *result.first; + auto const inserted = result.second; + // inserted a new element + if (result.second) { + auto const shared_set_index = atomicAdd(cardinality, 1); + shared_set_indices[shared_set_index] = idx; + local_mapping_index[idx] = shared_set_index; + } + return cuda::std::pair{matched_idx, inserted}; + } + return cuda::std::pair{0, false}; // dummy values + }(); + // Syncing the thread block is needed so that updates in `local_mapping_index` are visible to all + // threads in the thread block. + block.sync(); + if (is_valid_input) { + // element was already in set + if (!inserted) { local_mapping_index[idx] = local_mapping_index[result_idx]; } + } +} + +template +__device__ void find_global_mapping(cooperative_groups::thread_block const& block, + cudf::size_type cardinality, + SetRef global_set, + cudf::size_type* shared_set_indices, + cudf::size_type* global_mapping_index) +{ + // for all unique keys in shared memory hash set, stores their matches in + // global hash set to `global_mapping_index` + for (auto idx = block.thread_rank(); idx < cardinality; idx += block.num_threads()) { + auto const input_idx = shared_set_indices[idx]; + global_mapping_index[block.group_index().x * GROUPBY_SHM_MAX_ELEMENTS + idx] = + *global_set.insert_and_find(input_idx).first; + } +} + +/* + * @brief Inserts keys into the shared memory hash set, and stores the block-wise rank for a given + * row index in `local_mapping_index`. If the number of unique keys found in a threadblock exceeds + * `GROUPBY_CARDINALITY_THRESHOLD`, the threads in that block will exit without updating + * `global_set` or setting `global_mapping_index`. Else, we insert the unique keys found to the + * global hash set, and save the row index of the global sparse table in `global_mapping_index`. + */ +template +CUDF_KERNEL void mapping_indices_kernel(cudf::size_type num_input_rows, + SetRef global_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* local_mapping_index, + cudf::size_type* global_mapping_index, + cudf::size_type* block_cardinality, + cuda::std::atomic_flag* needs_global_memory_fallback) +{ + __shared__ cudf::size_type shared_set_indices[GROUPBY_SHM_MAX_ELEMENTS]; + + // Shared set initialization + __shared__ cuco::window windows[window_extent.value()]; + + auto raw_set = cuco::static_set_ref{ + cuco::empty_key{cudf::detail::CUDF_SIZE_TYPE_SENTINEL}, + global_set.key_eq(), + probing_scheme_t{global_set.hash_function()}, + cuco::thread_scope_block, + cuco::aow_storage_ref{ + window_extent, windows}}; + auto shared_set = raw_set.rebind_operators(cuco::insert_and_find); + + auto const block = cooperative_groups::this_thread_block(); + shared_set.initialize(block); + + __shared__ cudf::size_type cardinality; + if (block.thread_rank() == 0) { cardinality = 0; } + block.sync(); + + auto const stride = cudf::detail::grid_1d::grid_stride(); + + for (auto idx = cudf::detail::grid_1d::global_thread_id(); + idx - block.thread_rank() < num_input_rows; + idx += stride) { + find_local_mapping(block, + idx, + num_input_rows, + shared_set, + row_bitmask, + skip_rows_with_nulls, + &cardinality, + local_mapping_index, + shared_set_indices); + + block.sync(); + + if (cardinality >= GROUPBY_CARDINALITY_THRESHOLD) { + if (block.thread_rank() == 0) { needs_global_memory_fallback->test_and_set(); } + break; + } + } + + // Insert unique keys from shared to global hash set if block-cardinality + // doesn't exceed the threshold upper-limit + if (cardinality < GROUPBY_CARDINALITY_THRESHOLD) { + find_global_mapping(block, cardinality, global_set, shared_set_indices, global_mapping_index); + } + + if (block.thread_rank() == 0) { block_cardinality[block.group_index().x] = cardinality; } +} + +template +cudf::size_type max_occupancy_grid_size(cudf::size_type n) +{ + cudf::size_type max_active_blocks{-1}; + CUDF_CUDA_TRY(cudaOccupancyMaxActiveBlocksPerMultiprocessor( + &max_active_blocks, mapping_indices_kernel, GROUPBY_BLOCK_SIZE, 0)); + auto const grid_size = max_active_blocks * cudf::detail::num_multiprocessors(); + auto const num_blocks = cudf::util::div_rounding_up_safe(n, GROUPBY_BLOCK_SIZE); + return std::min(grid_size, num_blocks); +} + +template +void compute_mapping_indices(cudf::size_type grid_size, + cudf::size_type num, + SetRef global_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* local_mapping_index, + cudf::size_type* global_mapping_index, + cudf::size_type* block_cardinality, + cuda::std::atomic_flag* needs_global_memory_fallback, + rmm::cuda_stream_view stream) +{ + mapping_indices_kernel<<>>( + num, + global_set, + row_bitmask, + skip_rows_with_nulls, + local_mapping_index, + global_mapping_index, + block_cardinality, + needs_global_memory_fallback); +} +} // namespace cudf::groupby::detail::hash diff --git a/cpp/src/groupby/hash/compute_mapping_indices.hpp b/cpp/src/groupby/hash/compute_mapping_indices.hpp new file mode 100644 index 00000000000..473ad99e650 --- /dev/null +++ b/cpp/src/groupby/hash/compute_mapping_indices.hpp @@ -0,0 +1,43 @@ +/* + * Copyright (c) 2024, NVIDIA CORPORATION. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#pragma once + +#include + +#include + +#include + +namespace cudf::groupby::detail::hash { +/* + * @brief Computes the maximum number of active blocks of the given kernel that can be executed on + * the underlying device + */ +template +[[nodiscard]] cudf::size_type max_occupancy_grid_size(cudf::size_type n); + +template +void compute_mapping_indices(cudf::size_type grid_size, + cudf::size_type num, + SetRef global_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* local_mapping_index, + cudf::size_type* global_mapping_index, + cudf::size_type* block_cardinality, + cuda::std::atomic_flag* needs_global_memory_fallback, + rmm::cuda_stream_view stream); +} // namespace cudf::groupby::detail::hash diff --git a/cpp/src/groupby/hash/compute_mapping_indices_null.cu b/cpp/src/groupby/hash/compute_mapping_indices_null.cu new file mode 100644 index 00000000000..81c3c9e456f --- /dev/null +++ b/cpp/src/groupby/hash/compute_mapping_indices_null.cu @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2024, NVIDIA CORPORATION. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "compute_mapping_indices.cuh" +#include "compute_mapping_indices.hpp" + +namespace cudf::groupby::detail::hash { +template cudf::size_type +max_occupancy_grid_size>(cudf::size_type n); + +template void compute_mapping_indices>( + cudf::size_type grid_size, + cudf::size_type num, + nullable_hash_set_ref_t global_set, + bitmask_type const* row_bitmask, + bool skip_rows_with_nulls, + cudf::size_type* local_mapping_index, + cudf::size_type* global_mapping_index, + cudf::size_type* block_cardinality, + cuda::std::atomic_flag* needs_global_memory_fallback, + rmm::cuda_stream_view stream); +} // namespace cudf::groupby::detail::hash diff --git a/cpp/src/groupby/hash/flatten_single_pass_aggs.hpp b/cpp/src/groupby/hash/flatten_single_pass_aggs.hpp index 2bf983e5e90..dfad51f27d4 100644 --- a/cpp/src/groupby/hash/flatten_single_pass_aggs.hpp +++ b/cpp/src/groupby/hash/flatten_single_pass_aggs.hpp @@ -17,7 +17,6 @@ #include #include -#include #include #include diff --git a/cpp/src/groupby/sort/aggregate.cpp b/cpp/src/groupby/sort/aggregate.cpp index a9085a1f1fd..3041e261945 100644 --- a/cpp/src/groupby/sort/aggregate.cpp +++ b/cpp/src/groupby/sort/aggregate.cpp @@ -26,7 +26,6 @@ #include #include #include -#include #include #include #include diff --git a/cpp/src/interop/arrow_utilities.cpp b/cpp/src/interop/arrow_utilities.cpp index a99262fb3bf..c69ebe12d2c 100644 --- a/cpp/src/interop/arrow_utilities.cpp +++ b/cpp/src/interop/arrow_utilities.cpp @@ -20,11 +20,6 @@ #include #include -#include - -#include -#include - #include namespace cudf { diff --git a/cpp/src/interop/arrow_utilities.hpp b/cpp/src/interop/arrow_utilities.hpp index 1b79fbf9eda..e4bdedf6603 100644 --- a/cpp/src/interop/arrow_utilities.hpp +++ b/cpp/src/interop/arrow_utilities.hpp @@ -17,7 +17,6 @@ #pragma once #include -#include #include #include diff --git a/cpp/src/interop/dlpack.cpp b/cpp/src/interop/dlpack.cpp index a1be6aade4e..4395b741e53 100644 --- a/cpp/src/interop/dlpack.cpp +++ b/cpp/src/interop/dlpack.cpp @@ -16,11 +16,8 @@ #include #include #include -#include -#include #include #include -#include #include #include #include diff --git a/cpp/src/io/avro/avro.cpp b/cpp/src/io/avro/avro.cpp index d5caa4720ac..b3fcca62314 100644 --- a/cpp/src/io/avro/avro.cpp +++ b/cpp/src/io/avro/avro.cpp @@ -17,7 +17,6 @@ #include "avro.hpp" #include -#include #include namespace cudf { diff --git a/cpp/src/io/avro/avro.hpp b/cpp/src/io/avro/avro.hpp index 2e992546ccc..fd2c781b8a1 100644 --- a/cpp/src/io/avro/avro.hpp +++ b/cpp/src/io/avro/avro.hpp @@ -18,11 +18,9 @@ #include "avro_common.hpp" -#include #include #include #include -#include #include #include #include diff --git a/cpp/src/io/comp/nvcomp_adapter.hpp b/cpp/src/io/comp/nvcomp_adapter.hpp index 583bd6a3523..2e1cda2d6b7 100644 --- a/cpp/src/io/comp/nvcomp_adapter.hpp +++ b/cpp/src/io/comp/nvcomp_adapter.hpp @@ -18,9 +18,7 @@ #include "gpuinflate.hpp" -#include #include -#include #include #include diff --git a/cpp/src/io/comp/uncomp.cpp b/cpp/src/io/comp/uncomp.cpp index 1af45b41d8e..fb8c308065d 100644 --- a/cpp/src/io/comp/uncomp.cpp +++ b/cpp/src/io/comp/uncomp.cpp @@ -24,8 +24,6 @@ #include #include -#include - #include // uncompress #include // memset @@ -538,8 +536,10 @@ size_t decompress_zstd(host_span src, CUDF_EXPECTS(hd_stats[0].status == compression_status::SUCCESS, "ZSTD decompression failed"); // Copy temporary output to `dst` - CUDF_CUDA_TRY(cudaMemcpyAsync( - dst.data(), d_dst.data(), hd_stats[0].bytes_written, cudaMemcpyDefault, stream.value())); + cudf::detail::cuda_memcpy_async( + dst.subspan(0, hd_stats[0].bytes_written), + device_span{d_dst.data(), hd_stats[0].bytes_written}, + stream); return hd_stats[0].bytes_written; } diff --git a/cpp/src/io/csv/reader_impl.cu b/cpp/src/io/csv/reader_impl.cu index 8c32fc85f78..72fca75c56b 100644 --- a/cpp/src/io/csv/reader_impl.cu +++ b/cpp/src/io/csv/reader_impl.cu @@ -21,6 +21,7 @@ #include "csv_common.hpp" #include "csv_gpu.hpp" +#include "cudf/detail/utilities/cuda_memcpy.hpp" #include "io/comp/io_uncomp.hpp" #include "io/utilities/column_buffer.hpp" #include "io/utilities/hostdevice_vector.hpp" @@ -275,11 +276,10 @@ std::pair, selected_rows_offsets> load_data_and_gather auto const read_offset = byte_range_offset + input_pos + previous_data_size; auto const read_size = target_pos - input_pos - previous_data_size; if (data.has_value()) { - CUDF_CUDA_TRY(cudaMemcpyAsync(d_data.data() + previous_data_size, - data->data() + read_offset, - target_pos - input_pos - previous_data_size, - cudaMemcpyDefault, - stream.value())); + cudf::detail::cuda_memcpy_async( + device_span{d_data.data() + previous_data_size, read_size}, + data->subspan(read_offset, read_size), + stream); } else { if (source->is_device_read_preferred(read_size)) { source->device_read(read_offset, @@ -288,12 +288,11 @@ std::pair, selected_rows_offsets> load_data_and_gather stream); } else { auto const buffer = source->host_read(read_offset, read_size); - CUDF_CUDA_TRY(cudaMemcpyAsync(d_data.data() + previous_data_size, - buffer->data(), - buffer->size(), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); // To prevent buffer going out of scope before we copy the data. + // Use sync version to prevent buffer going out of scope before we copy the data. + cudf::detail::cuda_memcpy( + device_span{d_data.data() + previous_data_size, read_size}, + host_span{reinterpret_cast(buffer->data()), buffer->size()}, + stream); } } @@ -311,12 +310,10 @@ std::pair, selected_rows_offsets> load_data_and_gather range_end, skip_rows, stream); - CUDF_CUDA_TRY(cudaMemcpyAsync(row_ctx.host_ptr(), - row_ctx.device_ptr(), - num_blocks * sizeof(uint64_t), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + + cudf::detail::cuda_memcpy(host_span{row_ctx}.subspan(0, num_blocks), + device_span{row_ctx}.subspan(0, num_blocks), + stream); // Sum up the rows in each character block, selecting the row count that // corresponds to the current input context. Also stores the now known input @@ -331,11 +328,9 @@ std::pair, selected_rows_offsets> load_data_and_gather // At least one row in range in this batch all_row_offsets.resize(total_rows - skip_rows, stream); - CUDF_CUDA_TRY(cudaMemcpyAsync(row_ctx.device_ptr(), - row_ctx.host_ptr(), - num_blocks * sizeof(uint64_t), - cudaMemcpyDefault, - stream.value())); + cudf::detail::cuda_memcpy_async(device_span{row_ctx}.subspan(0, num_blocks), + host_span{row_ctx}.subspan(0, num_blocks), + stream); // Pass 2: Output row offsets cudf::io::csv::gpu::gather_row_offsets(parse_opts.view(), @@ -352,12 +347,9 @@ std::pair, selected_rows_offsets> load_data_and_gather stream); // With byte range, we want to keep only one row out of the specified range if (range_end < data_size) { - CUDF_CUDA_TRY(cudaMemcpyAsync(row_ctx.host_ptr(), - row_ctx.device_ptr(), - num_blocks * sizeof(uint64_t), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + cudf::detail::cuda_memcpy(host_span{row_ctx}.subspan(0, num_blocks), + device_span{row_ctx}.subspan(0, num_blocks), + stream); size_t rows_out_of_range = 0; for (uint32_t i = 0; i < num_blocks; i++) { @@ -401,12 +393,9 @@ std::pair, selected_rows_offsets> load_data_and_gather // Remove header rows and extract header auto const header_row_index = std::max(header_rows, 1) - 1; if (header_row_index + 1 < row_offsets.size()) { - CUDF_CUDA_TRY(cudaMemcpyAsync(row_ctx.host_ptr(), - row_offsets.data() + header_row_index, - 2 * sizeof(uint64_t), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + cudf::detail::cuda_memcpy(host_span{row_ctx}.subspan(0, 2), + device_span{row_offsets.data() + header_row_index, 2}, + stream); auto const header_start = input_pos + row_ctx[0]; auto const header_end = input_pos + row_ctx[1]; diff --git a/cpp/src/io/csv/writer_impl.cu b/cpp/src/io/csv/writer_impl.cu index b84446b5f3e..2bbe05ced84 100644 --- a/cpp/src/io/csv/writer_impl.cu +++ b/cpp/src/io/csv/writer_impl.cu @@ -27,6 +27,7 @@ #include #include #include +#include #include #include #include @@ -405,13 +406,8 @@ void write_chunked(data_sink* out_sink, out_sink->device_write(ptr_all_bytes, total_num_bytes, stream); } else { // copy the bytes to host to write them out - thrust::host_vector h_bytes(total_num_bytes); - CUDF_CUDA_TRY(cudaMemcpyAsync(h_bytes.data(), - ptr_all_bytes, - total_num_bytes * sizeof(char), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + auto const h_bytes = cudf::detail::make_host_vector_sync( + device_span{ptr_all_bytes, total_num_bytes}, stream); out_sink->host_write(h_bytes.data(), total_num_bytes); } diff --git a/cpp/src/io/functions.cpp b/cpp/src/io/functions.cpp index a8682e6a760..ceaeb5d8f85 100644 --- a/cpp/src/io/functions.cpp +++ b/cpp/src/io/functions.cpp @@ -32,10 +32,8 @@ #include #include #include -#include #include #include -#include #include #include diff --git a/cpp/src/io/json/host_tree_algorithms.cu b/cpp/src/io/json/host_tree_algorithms.cu index 7ee652e0239..570a00cbfc2 100644 --- a/cpp/src/io/json/host_tree_algorithms.cu +++ b/cpp/src/io/json/host_tree_algorithms.cu @@ -24,6 +24,7 @@ #include #include #include +#include #include #include #include @@ -198,17 +199,13 @@ NodeIndexT get_row_array_parent_col_id(device_span col_ids, bool is_enabled_lines, rmm::cuda_stream_view stream) { - NodeIndexT value = parent_node_sentinel; - if (!col_ids.empty()) { - auto const list_node_index = is_enabled_lines ? 0 : 1; - CUDF_CUDA_TRY(cudaMemcpyAsync(&value, - col_ids.data() + list_node_index, - sizeof(NodeIndexT), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); - } - return value; + if (col_ids.empty()) { return parent_node_sentinel; } + + auto const list_node_index = is_enabled_lines ? 0 : 1; + auto const value = cudf::detail::make_host_vector_sync( + device_span{col_ids.data() + list_node_index, 1}, stream); + + return value[0]; } /** * @brief Holds member data pointers of `d_json_column` @@ -818,11 +815,7 @@ std::pair, hashmap_of_device_columns> build_tree column_categories.cbegin(), expected_types.begin(), [](auto exp, auto cat) { return exp == NUM_NODE_CLASSES ? cat : exp; }); - cudaMemcpyAsync(d_column_tree.node_categories.begin(), - expected_types.data(), - expected_types.size() * sizeof(column_categories[0]), - cudaMemcpyDefault, - stream.value()); + cudf::detail::cuda_memcpy_async(d_column_tree.node_categories, expected_types, stream); return {is_pruned, columns}; } diff --git a/cpp/src/io/json/json_column.cu b/cpp/src/io/json/json_column.cu index 4584f71775f..7e4d975e431 100644 --- a/cpp/src/io/json/json_column.cu +++ b/cpp/src/io/json/json_column.cu @@ -513,16 +513,14 @@ table_with_metadata device_parse_nested_json(device_span d_input, #endif bool const is_array_of_arrays = [&]() { - std::array h_node_categories = {NC_ERR, NC_ERR}; - auto const size_to_copy = std::min(size_t{2}, gpu_tree.node_categories.size()); - CUDF_CUDA_TRY(cudaMemcpyAsync(h_node_categories.data(), - gpu_tree.node_categories.data(), - sizeof(node_t) * size_to_copy, - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + auto const size_to_copy = std::min(size_t{2}, gpu_tree.node_categories.size()); + if (size_to_copy == 0) return false; + auto const h_node_categories = cudf::detail::make_host_vector_sync( + device_span{gpu_tree.node_categories.data(), size_to_copy}, stream); + if (options.is_enabled_lines()) return h_node_categories[0] == NC_LIST; - return h_node_categories[0] == NC_LIST and h_node_categories[1] == NC_LIST; + return h_node_categories.size() >= 2 and h_node_categories[0] == NC_LIST and + h_node_categories[1] == NC_LIST; }(); auto [gpu_col_id, gpu_row_offsets] = diff --git a/cpp/src/io/json/json_tree.cu b/cpp/src/io/json/json_tree.cu index d949635c1cc..e2fe926ea19 100644 --- a/cpp/src/io/json/json_tree.cu +++ b/cpp/src/io/json/json_tree.cu @@ -264,16 +264,13 @@ tree_meta_t get_tree_representation(device_span tokens, error_count > 0) { auto const error_location = thrust::find(rmm::exec_policy(stream), tokens.begin(), tokens.end(), token_t::ErrorBegin); - SymbolOffsetT error_index; - CUDF_CUDA_TRY( - cudaMemcpyAsync(&error_index, - token_indices.data() + thrust::distance(tokens.begin(), error_location), - sizeof(SymbolOffsetT), - cudaMemcpyDefault, - stream.value())); - stream.synchronize(); + auto error_index = cudf::detail::make_host_vector_sync( + device_span{ + token_indices.data() + thrust::distance(tokens.begin(), error_location), 1}, + stream); + CUDF_FAIL("JSON Parser encountered an invalid format at location " + - std::to_string(error_index)); + std::to_string(error_index[0])); } auto const num_tokens = tokens.size(); diff --git a/cpp/src/io/json/nested_json.hpp b/cpp/src/io/json/nested_json.hpp index f6be4539d7f..7b3b04dea16 100644 --- a/cpp/src/io/json/nested_json.hpp +++ b/cpp/src/io/json/nested_json.hpp @@ -19,10 +19,7 @@ #include #include #include -#include -#include #include -#include #include #include diff --git a/cpp/src/io/json/nested_json_gpu.cu b/cpp/src/io/json/nested_json_gpu.cu index 534b30a6089..60e78f4763d 100644 --- a/cpp/src/io/json/nested_json_gpu.cu +++ b/cpp/src/io/json/nested_json_gpu.cu @@ -1448,10 +1448,6 @@ void get_stack_context(device_span json_in, // Number of stack operations in the input (i.e., number of '{', '}', '[', ']' outside of quotes) cudf::detail::device_scalar d_num_stack_ops(stream); - // Sequence of stack symbols and their position in the original input (sparse representation) - rmm::device_uvector stack_ops{json_in.size(), stream}; - rmm::device_uvector stack_op_indices{json_in.size(), stream}; - // Prepare finite-state transducer that only selects '{', '}', '[', ']' outside of quotes constexpr auto max_translation_table_size = to_stack_op::NUM_SYMBOL_GROUPS * to_stack_op::TT_NUM_STATES; @@ -1468,11 +1464,26 @@ void get_stack_context(device_span json_in, // "Search" for relevant occurrence of brackets and braces that indicate the beginning/end // of structs/lists + // Run FST to estimate the sizes of translated buffers + json_to_stack_ops_fst.Transduce(json_in.begin(), + static_cast(json_in.size()), + thrust::make_discard_iterator(), + thrust::make_discard_iterator(), + d_num_stack_ops.data(), + to_stack_op::start_state, + stream); + + auto stack_ops_bufsize = d_num_stack_ops.value(stream); + // Sequence of stack symbols and their position in the original input (sparse representation) + rmm::device_uvector stack_ops{stack_ops_bufsize, stream}; + rmm::device_uvector stack_op_indices{stack_ops_bufsize, stream}; + + // Run bracket-brace FST to retrieve starting positions of structs and lists json_to_stack_ops_fst.Transduce(json_in.begin(), static_cast(json_in.size()), stack_ops.data(), stack_op_indices.data(), - d_num_stack_ops.data(), + thrust::make_discard_iterator(), to_stack_op::start_state, stream); @@ -1508,6 +1519,7 @@ std::pair, rmm::device_uvector> pr device_span token_indices, rmm::cuda_stream_view stream) { + CUDF_FUNC_RANGE(); // Instantiate FST for post-processing the token stream to remove all tokens that belong to an // invalid JSON line token_filter::UnwrapTokenFromSymbolOp sgid_op{}; @@ -1643,21 +1655,28 @@ std::pair, rmm::device_uvector> ge // see a JSON-line delimiter as the very first item SymbolOffsetT const delimiter_offset = (format == tokenizer_pda::json_format_cfg_t::JSON_LINES_RECOVER ? 1 : 0); - rmm::device_uvector tokens{max_token_out_count + delimiter_offset, stream, mr}; - rmm::device_uvector tokens_indices{ - max_token_out_count + delimiter_offset, stream, mr}; + // Run FST to estimate the size of output buffers json_to_tokens_fst.Transduce(zip_in, static_cast(json_in.size()), - tokens.data() + delimiter_offset, - tokens_indices.data() + delimiter_offset, + thrust::make_discard_iterator(), + thrust::make_discard_iterator(), num_written_tokens.data(), tokenizer_pda::start_state, stream); auto const num_total_tokens = num_written_tokens.value(stream) + delimiter_offset; - tokens.resize(num_total_tokens, stream); - tokens_indices.resize(num_total_tokens, stream); + rmm::device_uvector tokens{num_total_tokens, stream, mr}; + rmm::device_uvector tokens_indices{num_total_tokens, stream, mr}; + + // Run FST to translate the input JSON string into tokens and indices at which they occur + json_to_tokens_fst.Transduce(zip_in, + static_cast(json_in.size()), + tokens.data() + delimiter_offset, + tokens_indices.data() + delimiter_offset, + thrust::make_discard_iterator(), + tokenizer_pda::start_state, + stream); if (delimiter_offset == 1) { tokens.set_element(0, token_t::LineEnd, stream); diff --git a/cpp/src/io/json/read_json.cu b/cpp/src/io/json/read_json.cu index c424d2b3b62..8a740ae17ef 100644 --- a/cpp/src/io/json/read_json.cu +++ b/cpp/src/io/json/read_json.cu @@ -315,13 +315,12 @@ device_span ingest_raw_input(device_span buffer, // Reading to host because decompression of a single block is much faster on the CPU sources[0]->host_read(range_offset, remaining_bytes_to_read, hbuffer.data()); auto uncomp_data = decompress(compression, hbuffer); - CUDF_CUDA_TRY(cudaMemcpyAsync(buffer.data(), - reinterpret_cast(uncomp_data.data()), - uncomp_data.size() * sizeof(char), - cudaMemcpyHostToDevice, - stream.value())); - stream.synchronize(); - return buffer.first(uncomp_data.size()); + auto ret_buffer = buffer.first(uncomp_data.size()); + cudf::detail::cuda_memcpy( + ret_buffer, + host_span{reinterpret_cast(uncomp_data.data()), uncomp_data.size()}, + stream); + return ret_buffer; } table_with_metadata read_json(host_span> sources, diff --git a/cpp/src/io/orc/reader_impl_chunking.cu b/cpp/src/io/orc/reader_impl_chunking.cu index 9cc77e8e738..fcaee9c548e 100644 --- a/cpp/src/io/orc/reader_impl_chunking.cu +++ b/cpp/src/io/orc/reader_impl_chunking.cu @@ -516,10 +516,10 @@ void reader_impl::load_next_stripe_data(read_mode mode) _stream.synchronize(); stream_synchronized = true; } - device_read_tasks.push_back( - std::pair(source_ptr->device_read_async( - read_info.offset, read_info.length, dst_base + read_info.dst_pos, _stream), - read_info.length)); + device_read_tasks.emplace_back( + source_ptr->device_read_async( + read_info.offset, read_info.length, dst_base + read_info.dst_pos, _stream), + read_info.length); } else { auto buffer = source_ptr->host_read(read_info.offset, read_info.length); diff --git a/cpp/src/io/orc/reader_impl_decode.cu b/cpp/src/io/orc/reader_impl_decode.cu index c42348a165f..0081ed30d17 100644 --- a/cpp/src/io/orc/reader_impl_decode.cu +++ b/cpp/src/io/orc/reader_impl_decode.cu @@ -23,6 +23,7 @@ #include #include +#include #include #include #include diff --git a/cpp/src/io/orc/reader_impl_helpers.cpp b/cpp/src/io/orc/reader_impl_helpers.cpp index 4c1079cffe8..7e5db4b7617 100644 --- a/cpp/src/io/orc/reader_impl_helpers.cpp +++ b/cpp/src/io/orc/reader_impl_helpers.cpp @@ -16,8 +16,6 @@ #include "reader_impl_helpers.hpp" -#include - namespace cudf::io::orc::detail { std::unique_ptr create_empty_column(size_type orc_col_id, diff --git a/cpp/src/io/orc/reader_impl_helpers.hpp b/cpp/src/io/orc/reader_impl_helpers.hpp index 5528b2ee763..4cded30d89b 100644 --- a/cpp/src/io/orc/reader_impl_helpers.hpp +++ b/cpp/src/io/orc/reader_impl_helpers.hpp @@ -20,9 +20,6 @@ #include "io/orc/orc.hpp" #include "io/utilities/column_buffer.hpp" -#include -#include - #include #include diff --git a/cpp/src/io/orc/writer_impl.cu b/cpp/src/io/orc/writer_impl.cu index 03020eb649f..d432deb8e79 100644 --- a/cpp/src/io/orc/writer_impl.cu +++ b/cpp/src/io/orc/writer_impl.cu @@ -19,6 +19,7 @@ * @brief cuDF-IO ORC writer class implementation */ +#include "cudf/detail/utilities/cuda_memcpy.hpp" #include "io/comp/nvcomp_adapter.hpp" #include "io/orc/orc_gpu.hpp" #include "io/statistics/column_statistics.cuh" @@ -1408,7 +1409,8 @@ encoded_footer_statistics finish_statistic_blobs(Footer const& footer, num_entries_seen += stripes_per_col; } - std::vector file_stats_merge(num_file_blobs); + auto file_stats_merge = + cudf::detail::make_host_vector(num_file_blobs, stream); for (auto i = 0u; i < num_file_blobs; ++i) { auto col_stats = &file_stats_merge[i]; col_stats->col_dtype = per_chunk_stats.col_types[i]; @@ -1418,11 +1420,10 @@ encoded_footer_statistics finish_statistic_blobs(Footer const& footer, } auto d_file_stats_merge = stats_merge.device_ptr(num_stripe_blobs); - CUDF_CUDA_TRY(cudaMemcpyAsync(d_file_stats_merge, - file_stats_merge.data(), - num_file_blobs * sizeof(statistics_merge_group), - cudaMemcpyDefault, - stream.value())); + cudf::detail::cuda_memcpy_async( + device_span{stats_merge.device_ptr(num_stripe_blobs), num_file_blobs}, + file_stats_merge, + stream); auto file_stat_chunks = stat_chunks.data() + num_stripe_blobs; detail::merge_group_statistics( @@ -1573,7 +1574,7 @@ void write_index_stream(int32_t stripe_id, * @param[in] strm_desc Stream's descriptor * @param[in] enc_stream Chunk's streams * @param[in] compressed_data Compressed stream data - * @param[in,out] stream_out Temporary host output buffer + * @param[in,out] bounce_buffer Pinned memory bounce buffer for D2H data transfer * @param[in,out] stripe Stream's parent stripe * @param[in,out] streams List of all streams * @param[in] compression_kind The compression kind @@ -1584,7 +1585,7 @@ void write_index_stream(int32_t stripe_id, std::future write_data_stream(gpu::StripeStream const& strm_desc, gpu::encoder_chunk_streams const& enc_stream, uint8_t const* compressed_data, - uint8_t* stream_out, + host_span bounce_buffer, StripeInformation* stripe, orc_streams* streams, CompressionKind compression_kind, @@ -1604,11 +1605,10 @@ std::future write_data_stream(gpu::StripeStream const& strm_desc, if (out_sink->is_device_write_preferred(length)) { return out_sink->device_write_async(stream_in, length, stream); } else { - CUDF_CUDA_TRY( - cudaMemcpyAsync(stream_out, stream_in, length, cudaMemcpyDefault, stream.value())); - stream.synchronize(); + cudf::detail::cuda_memcpy( + bounce_buffer.subspan(0, length), device_span{stream_in, length}, stream); - out_sink->host_write(stream_out, length); + out_sink->host_write(bounce_buffer.data(), length); return std::async(std::launch::deferred, [] {}); } }(); @@ -2616,7 +2616,7 @@ void writer::impl::write_orc_data_to_sink(encoded_data const& enc_data, strm_desc, enc_data.streams[strm_desc.column_id][segmentation.stripes[stripe_id].first], compressed_data.data(), - bounce_buffer.data(), + bounce_buffer, &stripe, &streams, _compression_kind, diff --git a/cpp/src/io/parquet/arrow_schema_writer.cpp b/cpp/src/io/parquet/arrow_schema_writer.cpp index ddf65e9020f..d15435b2553 100644 --- a/cpp/src/io/parquet/arrow_schema_writer.cpp +++ b/cpp/src/io/parquet/arrow_schema_writer.cpp @@ -27,7 +27,6 @@ #include "ipc/Schema_generated.h" #include "writer_impl_helpers.hpp" -#include #include #include diff --git a/cpp/src/io/parquet/arrow_schema_writer.hpp b/cpp/src/io/parquet/arrow_schema_writer.hpp index 9bc435bf6c8..66810ee163a 100644 --- a/cpp/src/io/parquet/arrow_schema_writer.hpp +++ b/cpp/src/io/parquet/arrow_schema_writer.hpp @@ -22,10 +22,9 @@ #pragma once #include -#include -#include +#include +#include #include -#include namespace cudf::io::parquet::detail { diff --git a/cpp/src/io/parquet/compact_protocol_reader.hpp b/cpp/src/io/parquet/compact_protocol_reader.hpp index 12c24e2b848..b87f2e9c692 100644 --- a/cpp/src/io/parquet/compact_protocol_reader.hpp +++ b/cpp/src/io/parquet/compact_protocol_reader.hpp @@ -22,10 +22,7 @@ #include #include -#include -#include #include -#include namespace CUDF_EXPORT cudf { namespace io::parquet::detail { diff --git a/cpp/src/io/parquet/compact_protocol_writer.hpp b/cpp/src/io/parquet/compact_protocol_writer.hpp index d4778b1ea15..05859d60c03 100644 --- a/cpp/src/io/parquet/compact_protocol_writer.hpp +++ b/cpp/src/io/parquet/compact_protocol_writer.hpp @@ -17,7 +17,6 @@ #pragma once #include "parquet.hpp" -#include "parquet_common.hpp" #include #include diff --git a/cpp/src/io/parquet/page_data.cu b/cpp/src/io/parquet/page_data.cu index b3276c81c1f..0d24fa4236f 100644 --- a/cpp/src/io/parquet/page_data.cu +++ b/cpp/src/io/parquet/page_data.cu @@ -21,6 +21,7 @@ #include +#include #include namespace cudf::io::parquet::detail { @@ -476,9 +477,9 @@ void WriteFinalOffsets(host_span offsets, auto d_src_data = cudf::detail::make_device_uvector_async( offsets, stream, cudf::get_current_device_resource_ref()); // Iterator for the source (scalar) data - auto src_iter = cudf::detail::make_counting_transform_iterator( - static_cast(0), - cuda::proclaim_return_type( + auto src_iter = thrust::make_transform_iterator( + thrust::make_counting_iterator(0), + cuda::proclaim_return_type( [src = d_src_data.begin()] __device__(std::size_t i) { return src + i; })); // Copy buffer addresses to device and create an iterator diff --git a/cpp/src/io/parquet/predicate_pushdown.cpp b/cpp/src/io/parquet/predicate_pushdown.cpp index f0a0bc0b51b..a965f3325d5 100644 --- a/cpp/src/io/parquet/predicate_pushdown.cpp +++ b/cpp/src/io/parquet/predicate_pushdown.cpp @@ -23,7 +23,6 @@ #include #include #include -#include #include #include #include @@ -454,15 +453,18 @@ std::optional>> aggregate_reader_metadata::fi CUDF_EXPECTS(predicate.type().id() == cudf::type_id::BOOL8, "Filter expression must return a boolean column"); - auto num_bitmasks = num_bitmask_words(predicate.size()); - std::vector host_bitmask(num_bitmasks, ~bitmask_type{0}); - if (predicate.nullable()) { - CUDF_CUDA_TRY(cudaMemcpyAsync(host_bitmask.data(), - predicate.null_mask(), - num_bitmasks * sizeof(bitmask_type), - cudaMemcpyDefault, - stream.value())); - } + auto const host_bitmask = [&] { + auto const num_bitmasks = num_bitmask_words(predicate.size()); + if (predicate.nullable()) { + return cudf::detail::make_host_vector_sync( + device_span(predicate.null_mask(), num_bitmasks), stream); + } else { + auto bitmask = cudf::detail::make_host_vector(num_bitmasks, stream); + std::fill(bitmask.begin(), bitmask.end(), ~bitmask_type{0}); + return bitmask; + } + }(); + auto validity_it = cudf::detail::make_counting_transform_iterator( 0, [bitmask = host_bitmask.data()](auto bit_index) { return bit_is_set(bitmask, bit_index); }); diff --git a/cpp/src/io/parquet/reader.cpp b/cpp/src/io/parquet/reader.cpp index dd354b905f3..170c6e8857f 100644 --- a/cpp/src/io/parquet/reader.cpp +++ b/cpp/src/io/parquet/reader.cpp @@ -16,8 +16,6 @@ #include "reader_impl.hpp" -#include - namespace cudf::io::parquet::detail { reader::reader() = default; diff --git a/cpp/src/io/parquet/reader_impl.cpp b/cpp/src/io/parquet/reader_impl.cpp index f0865c715bc..fed1a309064 100644 --- a/cpp/src/io/parquet/reader_impl.cpp +++ b/cpp/src/io/parquet/reader_impl.cpp @@ -21,11 +21,9 @@ #include #include #include -#include #include #include -#include #include #include @@ -78,7 +76,7 @@ void reader::impl::decode_page_data(read_mode mode, size_t skip_rows, size_t num // TODO: This step is somewhat redundant if size info has already been calculated (nested schema, // chunked reader). auto const has_strings = (kernel_mask & STRINGS_MASK) != 0; - std::vector col_string_sizes(_input_columns.size(), 0L); + auto col_string_sizes = cudf::detail::make_host_vector(_input_columns.size(), _stream); if (has_strings) { // need to compute pages bounds/sizes if we lack page indexes or are using custom bounds // TODO: we could probably dummy up size stats for FLBA data since we know the width diff --git a/cpp/src/io/parquet/reader_impl.hpp b/cpp/src/io/parquet/reader_impl.hpp index 62ffc4d3077..3aa9b94ed6b 100644 --- a/cpp/src/io/parquet/reader_impl.hpp +++ b/cpp/src/io/parquet/reader_impl.hpp @@ -284,7 +284,7 @@ class reader::impl { * * @return Vector of total string data sizes for each column */ - std::vector calculate_page_string_offsets(); + cudf::detail::host_vector calculate_page_string_offsets(); /** * @brief Converts the page data and outputs to columns. diff --git a/cpp/src/io/parquet/reader_impl_chunking.hpp b/cpp/src/io/parquet/reader_impl_chunking.hpp index 3a3cdd34a58..a0c2dbd3e44 100644 --- a/cpp/src/io/parquet/reader_impl_chunking.hpp +++ b/cpp/src/io/parquet/reader_impl_chunking.hpp @@ -107,7 +107,7 @@ struct subpass_intermediate_data { * rowgroups may represent less than all of the rowgroups to be read for the file. */ struct pass_intermediate_data { - std::vector> raw_page_data; + std::vector raw_page_data; // rowgroup, chunk and page information for the current pass. bool has_compressed_data{false}; diff --git a/cpp/src/io/parquet/reader_impl_preprocess.cu b/cpp/src/io/parquet/reader_impl_preprocess.cu index 5138a92ac14..f03f1214b9a 100644 --- a/cpp/src/io/parquet/reader_impl_preprocess.cu +++ b/cpp/src/io/parquet/reader_impl_preprocess.cu @@ -218,7 +218,7 @@ void generate_depth_remappings( */ [[nodiscard]] std::future read_column_chunks_async( std::vector> const& sources, - std::vector>& page_data, + cudf::host_span page_data, cudf::detail::hostdevice_vector& chunks, size_t begin_chunk, size_t end_chunk, @@ -251,23 +251,24 @@ void generate_depth_remappings( if (source->is_device_read_preferred(io_size)) { // Buffer needs to be padded. // Required by `gpuDecodePageData`. - auto buffer = + page_data[chunk] = rmm::device_buffer(cudf::util::round_up_safe(io_size, BUFFER_PADDING_MULTIPLE), stream); auto fut_read_size = source->device_read_async( - io_offset, io_size, static_cast(buffer.data()), stream); + io_offset, io_size, static_cast(page_data[chunk].data()), stream); read_tasks.emplace_back(std::move(fut_read_size)); - page_data[chunk] = datasource::buffer::create(std::move(buffer)); } else { auto const read_buffer = source->host_read(io_offset, io_size); // Buffer needs to be padded. // Required by `gpuDecodePageData`. - auto tmp_buffer = rmm::device_buffer( + page_data[chunk] = rmm::device_buffer( cudf::util::round_up_safe(read_buffer->size(), BUFFER_PADDING_MULTIPLE), stream); - CUDF_CUDA_TRY(cudaMemcpyAsync( - tmp_buffer.data(), read_buffer->data(), read_buffer->size(), cudaMemcpyDefault, stream)); - page_data[chunk] = datasource::buffer::create(std::move(tmp_buffer)); + CUDF_CUDA_TRY(cudaMemcpyAsync(page_data[chunk].data(), + read_buffer->data(), + read_buffer->size(), + cudaMemcpyDefault, + stream)); } - auto d_compdata = page_data[chunk]->data(); + auto d_compdata = static_cast(page_data[chunk].data()); do { chunks[chunk].compressed_data = d_compdata; d_compdata += chunks[chunk].compressed_size; @@ -980,7 +981,7 @@ std::pair> reader::impl::read_column_chunks() std::vector chunk_source_map(num_chunks); // Tracker for eventually deallocating compressed and uncompressed data - raw_page_data = std::vector>(num_chunks); + raw_page_data = std::vector(num_chunks); // Keep track of column chunk file offsets std::vector column_chunk_offsets(num_chunks); @@ -1629,10 +1630,10 @@ void reader::impl::allocate_columns(read_mode mode, size_t skip_rows, size_t num get_page_nesting_size{ d_cols_info.data(), max_depth, subpass.pages.size(), subpass.pages.device_begin()}); - // Manually create a int64_t `key_start` compatible counting_transform_iterator to avoid - // implicit casting to size_type. - auto const reduction_keys = thrust::make_transform_iterator( - thrust::make_counting_iterator(key_start), get_reduction_key{subpass.pages.size()}); + // Manually create a size_t `key_start` compatible counting_transform_iterator. + auto const reduction_keys = + thrust::make_transform_iterator(thrust::make_counting_iterator(key_start), + get_reduction_key{subpass.pages.size()}); // Find the size of each column thrust::reduce_by_key(rmm::exec_policy_nosync(_stream), @@ -1695,15 +1696,14 @@ void reader::impl::allocate_columns(read_mode mode, size_t skip_rows, size_t num nullmask_bufs, std::numeric_limits::max(), _stream); } -std::vector reader::impl::calculate_page_string_offsets() +cudf::detail::host_vector reader::impl::calculate_page_string_offsets() { auto& pass = *_pass_itm_data; auto& subpass = *pass.subpass; auto page_keys = make_page_key_iterator(subpass.pages); - std::vector col_sizes(_input_columns.size(), 0L); - rmm::device_uvector d_col_sizes(col_sizes.size(), _stream); + rmm::device_uvector d_col_sizes(_input_columns.size(), _stream); // use page_index to fetch page string sizes in the proper order auto val_iter = thrust::make_transform_iterator(subpass.pages.device_begin(), @@ -1717,7 +1717,7 @@ std::vector reader::impl::calculate_page_string_offsets() page_offset_output_iter{subpass.pages.device_ptr()}); // now sum up page sizes - rmm::device_uvector reduce_keys(col_sizes.size(), _stream); + rmm::device_uvector reduce_keys(d_col_sizes.size(), _stream); thrust::reduce_by_key(rmm::exec_policy_nosync(_stream), page_keys, page_keys + subpass.pages.size(), @@ -1725,14 +1725,7 @@ std::vector reader::impl::calculate_page_string_offsets() reduce_keys.begin(), d_col_sizes.begin()); - cudaMemcpyAsync(col_sizes.data(), - d_col_sizes.data(), - sizeof(size_t) * col_sizes.size(), - cudaMemcpyDeviceToHost, - _stream); - _stream.synchronize(); - - return col_sizes; + return cudf::detail::make_host_vector_sync(d_col_sizes, _stream); } } // namespace cudf::io::parquet::detail diff --git a/cpp/src/io/parquet/writer_impl.cu b/cpp/src/io/parquet/writer_impl.cu index 190f13eb688..f865c9a7643 100644 --- a/cpp/src/io/parquet/writer_impl.cu +++ b/cpp/src/io/parquet/writer_impl.cu @@ -183,7 +183,7 @@ struct aggregate_writer_metadata { std::vector row_groups; std::vector key_value_metadata; std::vector offset_indexes; - std::vector> column_indexes; + std::vector> column_indexes; }; std::vector files; std::optional> column_orders = std::nullopt; @@ -1543,12 +1543,7 @@ void encode_pages(hostdevice_2dvector& chunks, d_chunks.flat_view(), {column_stats, pages.size()}, column_index_truncate_length, stream); } - auto h_chunks = chunks.host_view(); - CUDF_CUDA_TRY(cudaMemcpyAsync(h_chunks.data(), - d_chunks.data(), - d_chunks.flat_view().size_bytes(), - cudaMemcpyDefault, - stream.value())); + chunks.device_to_host_async(stream); if (comp_stats.has_value()) { comp_stats.value() += collect_compression_statistics(comp_in, comp_res, stream); @@ -2559,12 +2554,11 @@ void writer::impl::write_parquet_data_to_sink( } else { CUDF_EXPECTS(bounce_buffer.size() >= ck.compressed_size, "Bounce buffer was not properly initialized."); - CUDF_CUDA_TRY(cudaMemcpyAsync(bounce_buffer.data(), - dev_bfr + ck.ck_stat_size, - ck.compressed_size, - cudaMemcpyDefault, - _stream.value())); - _stream.synchronize(); + cudf::detail::cuda_memcpy( + host_span{bounce_buffer}.subspan(0, ck.compressed_size), + device_span{dev_bfr + ck.ck_stat_size, ck.compressed_size}, + _stream); + _out_sink[p]->host_write(bounce_buffer.data(), ck.compressed_size); } @@ -2600,13 +2594,8 @@ void writer::impl::write_parquet_data_to_sink( auto const& column_chunk_meta = row_group.columns[i].meta_data; // start transfer of the column index - std::vector column_idx; - column_idx.resize(ck.column_index_size); - CUDF_CUDA_TRY(cudaMemcpyAsync(column_idx.data(), - ck.column_index_blob, - ck.column_index_size, - cudaMemcpyDefault, - _stream.value())); + auto column_idx = cudf::detail::make_host_vector_async( + device_span{ck.column_index_blob, ck.column_index_size}, _stream); // calculate offsets while the column index is transferring int64_t curr_pg_offset = column_chunk_meta.data_page_offset; diff --git a/cpp/src/io/text/bgzip_data_chunk_source.cu b/cpp/src/io/text/bgzip_data_chunk_source.cu index badcd3f58f9..06069630685 100644 --- a/cpp/src/io/text/bgzip_data_chunk_source.cu +++ b/cpp/src/io/text/bgzip_data_chunk_source.cu @@ -74,8 +74,8 @@ class bgzip_data_chunk_reader : public data_chunk_reader { // Buffer needs to be padded. // Required by `inflate_kernel`. device.resize(cudf::util::round_up_safe(host.size(), BUFFER_PADDING_MULTIPLE), stream); - CUDF_CUDA_TRY(cudaMemcpyAsync( - device.data(), host.data(), host.size() * sizeof(T), cudaMemcpyDefault, stream.value())); + cudf::detail::cuda_memcpy_async( + device_span{device}.subspan(0, host.size()), host, stream); } struct decompression_blocks { diff --git a/cpp/src/io/text/data_chunk_source_factories.cpp b/cpp/src/io/text/data_chunk_source_factories.cpp index 58faa0ebfe4..f4a2f29026a 100644 --- a/cpp/src/io/text/data_chunk_source_factories.cpp +++ b/cpp/src/io/text/data_chunk_source_factories.cpp @@ -22,10 +22,6 @@ #include #include -#include - -#include - #include namespace cudf::io::text { @@ -87,8 +83,10 @@ class datasource_chunk_reader : public data_chunk_reader { _source->host_read(_offset, read_size, reinterpret_cast(h_ticket.buffer.data())); // copy the host-pinned data on to device - CUDF_CUDA_TRY(cudaMemcpyAsync( - chunk.data(), h_ticket.buffer.data(), read_size, cudaMemcpyDefault, stream.value())); + cudf::detail::cuda_memcpy_async( + device_span{chunk}.subspan(0, read_size), + host_span{h_ticket.buffer}.subspan(0, read_size), + stream); // record the host-to-device copy. CUDF_CUDA_TRY(cudaEventRecord(h_ticket.event, stream.value())); @@ -153,8 +151,10 @@ class istream_data_chunk_reader : public data_chunk_reader { auto chunk = rmm::device_uvector(read_size, stream); // copy the host-pinned data on to device - CUDF_CUDA_TRY(cudaMemcpyAsync( - chunk.data(), h_ticket.buffer.data(), read_size, cudaMemcpyDefault, stream.value())); + cudf::detail::cuda_memcpy_async( + device_span{chunk}.subspan(0, read_size), + host_span{h_ticket.buffer}.subspan(0, read_size), + stream); // record the host-to-device copy. CUDF_CUDA_TRY(cudaEventRecord(h_ticket.event, stream.value())); @@ -193,12 +193,10 @@ class host_span_data_chunk_reader : public data_chunk_reader { auto chunk = rmm::device_uvector(read_size, stream); // copy the host data to device - CUDF_CUDA_TRY(cudaMemcpyAsync( // - chunk.data(), - _data.data() + _position, - read_size, - cudaMemcpyDefault, - stream.value())); + cudf::detail::cuda_memcpy_async( + cudf::device_span{chunk}.subspan(0, read_size), + cudf::host_span{_data}.subspan(_position, read_size), + stream); _position += read_size; diff --git a/cpp/src/io/utilities/column_buffer.cpp b/cpp/src/io/utilities/column_buffer.cpp index 249dc3b5875..6d954753af8 100644 --- a/cpp/src/io/utilities/column_buffer.cpp +++ b/cpp/src/io/utilities/column_buffer.cpp @@ -21,12 +21,12 @@ #include "column_buffer.hpp" +#include #include #include #include #include -#include #include namespace cudf::io::detail { diff --git a/cpp/src/io/utilities/column_buffer.hpp b/cpp/src/io/utilities/column_buffer.hpp index e73b2bc88de..31c8b781e77 100644 --- a/cpp/src/io/utilities/column_buffer.hpp +++ b/cpp/src/io/utilities/column_buffer.hpp @@ -22,12 +22,9 @@ #pragma once #include -#include #include -#include #include #include -#include #include #include @@ -35,6 +32,8 @@ #include +#include + namespace cudf { namespace io { namespace detail { diff --git a/cpp/src/io/utilities/config_utils.cpp b/cpp/src/io/utilities/config_utils.cpp index a3afbd52896..b66742569d9 100644 --- a/cpp/src/io/utilities/config_utils.cpp +++ b/cpp/src/io/utilities/config_utils.cpp @@ -16,11 +16,10 @@ #include "getenv_or.hpp" -#include #include -#include -#include +#include + #include namespace cudf::io { @@ -53,6 +52,14 @@ bool is_gds_enabled() { return is_always_enabled() or get_env_policy() == usage_ bool is_kvikio_enabled() { return get_env_policy() == usage_policy::KVIKIO; } +void set_thread_pool_nthreads_from_env() +{ + static std::once_flag flag{}; + std::call_once(flag, [] { + auto nthreads = getenv_or("KVIKIO_NTHREADS", 8U); + kvikio::defaults::thread_pool_nthreads_reset(nthreads); + }); +} } // namespace cufile_integration namespace nvcomp_integration { @@ -81,5 +88,4 @@ bool is_all_enabled() { return get_env_policy() == usage_policy::ALWAYS; } bool is_stable_enabled() { return is_all_enabled() or get_env_policy() == usage_policy::STABLE; } } // namespace nvcomp_integration - } // namespace cudf::io diff --git a/cpp/src/io/utilities/data_sink.cpp b/cpp/src/io/utilities/data_sink.cpp index 0b76f3d3e8f..a8a275919d8 100644 --- a/cpp/src/io/utilities/data_sink.cpp +++ b/cpp/src/io/utilities/data_sink.cpp @@ -42,6 +42,7 @@ class file_sink : public data_sink { if (!_output_stream.is_open()) { detail::throw_on_file_open_failure(filepath, true); } if (cufile_integration::is_kvikio_enabled()) { + cufile_integration::set_thread_pool_nthreads_from_env(); _kvikio_file = kvikio::FileHandle(filepath, "w"); CUDF_LOG_INFO("Writing a file using kvikIO, with compatibility mode {}.", _kvikio_file.is_compat_mode_on() ? "on" : "off"); diff --git a/cpp/src/io/utilities/datasource.cpp b/cpp/src/io/utilities/datasource.cpp index 2daaecadca6..9668b30e9a9 100644 --- a/cpp/src/io/utilities/datasource.cpp +++ b/cpp/src/io/utilities/datasource.cpp @@ -18,6 +18,7 @@ #include "getenv_or.hpp" #include +#include #include #include #include @@ -32,7 +33,6 @@ #include #include -#include #include namespace cudf { @@ -48,6 +48,7 @@ class file_source : public datasource { { detail::force_init_cuda_context(); if (cufile_integration::is_kvikio_enabled()) { + cufile_integration::set_thread_pool_nthreads_from_env(); _kvikio_file = kvikio::FileHandle(filepath); CUDF_LOG_INFO("Reading a file using kvikIO, with compatibility mode {}.", _kvikio_file.is_compat_mode_on() ? "on" : "off"); @@ -246,17 +247,18 @@ class device_buffer_source final : public datasource { size_t host_read(size_t offset, size_t size, uint8_t* dst) override { auto const count = std::min(size, this->size() - offset); - auto const stream = cudf::get_default_stream(); - CUDF_CUDA_TRY( - cudaMemcpyAsync(dst, _d_buffer.data() + offset, count, cudaMemcpyDefault, stream.value())); - stream.synchronize(); + auto const stream = cudf::detail::global_cuda_stream_pool().get_stream(); + cudf::detail::cuda_memcpy(host_span{dst, count}, + device_span{ + reinterpret_cast(_d_buffer.data() + offset), count}, + stream); return count; } std::unique_ptr host_read(size_t offset, size_t size) override { auto const count = std::min(size, this->size() - offset); - auto const stream = cudf::get_default_stream(); + auto const stream = cudf::detail::global_cuda_stream_pool().get_stream(); auto h_data = cudf::detail::make_host_vector_async( cudf::device_span{_d_buffer.data() + offset, count}, stream); stream.synchronize(); diff --git a/cpp/src/io/utilities/file_io_utilities.cpp b/cpp/src/io/utilities/file_io_utilities.cpp index 98ed9b28f0a..93cdccfbb9f 100644 --- a/cpp/src/io/utilities/file_io_utilities.cpp +++ b/cpp/src/io/utilities/file_io_utilities.cpp @@ -22,8 +22,6 @@ #include #include -#include - #include #include diff --git a/cpp/src/io/utilities/row_selection.cpp b/cpp/src/io/utilities/row_selection.cpp index c0bbca39167..cf252fe63af 100644 --- a/cpp/src/io/utilities/row_selection.cpp +++ b/cpp/src/io/utilities/row_selection.cpp @@ -16,10 +16,7 @@ #include "io/utilities/row_selection.hpp" -#include - #include -#include namespace cudf::io::detail { diff --git a/cpp/src/io/utilities/row_selection.hpp b/cpp/src/io/utilities/row_selection.hpp index 7c607099cdc..e826feff201 100644 --- a/cpp/src/io/utilities/row_selection.hpp +++ b/cpp/src/io/utilities/row_selection.hpp @@ -15,7 +15,7 @@ */ #pragma once -#include +#include #include #include diff --git a/cpp/src/jit/cache.cpp b/cpp/src/jit/cache.cpp index 89c47d246d0..34a0bdce124 100644 --- a/cpp/src/jit/cache.cpp +++ b/cpp/src/jit/cache.cpp @@ -16,11 +16,8 @@ #include -#include - #include -#include #include namespace cudf { diff --git a/cpp/src/jit/util.cpp b/cpp/src/jit/util.cpp index 0585e02a031..d9a29203133 100644 --- a/cpp/src/jit/util.cpp +++ b/cpp/src/jit/util.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2023, NVIDIA CORPORATION. + * Copyright (c) 2019-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -19,8 +19,6 @@ #include #include -#include - namespace cudf { namespace jit { struct get_data_ptr_functor { diff --git a/cpp/src/quantiles/tdigest/tdigest_aggregation.cu b/cpp/src/quantiles/tdigest/tdigest_aggregation.cu index b0a84a6d50c..d27420658d6 100644 --- a/cpp/src/quantiles/tdigest/tdigest_aggregation.cu +++ b/cpp/src/quantiles/tdigest/tdigest_aggregation.cu @@ -1126,12 +1126,8 @@ std::pair, rmm::device_uvector> generate_mer * `max` of 0. * * @param tdv input tdigests. The tdigests within this column are grouped by key. - * @param h_group_offsets a host iterator of the offsets to the start of each group. A group is - * counted as one even when the cluster is empty in it. The offsets should have the same values as - * the ones in `group_offsets`. * @param group_offsets a device iterator of the offsets to the start of each group. A group is - * counted as one even when the cluster is empty in it. The offsets should have the same values as - * the ones in `h_group_offsets`. + * counted as one even when the cluster is empty in it. * @param group_labels a device iterator of the the group label for each tdigest cluster including * empty clusters. * @param num_group_labels the number of unique group labels. @@ -1142,9 +1138,8 @@ std::pair, rmm::device_uvector> generate_mer * * @return A column containing the merged tdigests. */ -template +template std::unique_ptr merge_tdigests(tdigest_column_view const& tdv, - HGroupOffsetIter h_group_offsets, GroupOffsetIter group_offsets, GroupLabelIter group_labels, size_t num_group_labels, @@ -1313,21 +1308,13 @@ std::unique_ptr reduce_merge_tdigest(column_view const& input, if (input.size() == 0) { return cudf::tdigest::detail::make_empty_tdigest_scalar(stream, mr); } - auto group_offsets_ = group_offsets_fn{input.size()}; - auto h_group_offsets = cudf::detail::make_counting_transform_iterator(0, group_offsets_); - auto group_offsets = cudf::detail::make_counting_transform_iterator(0, group_offsets_); - auto group_labels = thrust::make_constant_iterator(0); - return to_tdigest_scalar(merge_tdigests(tdv, - h_group_offsets, - group_offsets, - group_labels, - input.size(), - 1, - max_centroids, - stream, - mr), - stream, - mr); + auto group_offsets_ = group_offsets_fn{input.size()}; + auto group_offsets = cudf::detail::make_counting_transform_iterator(0, group_offsets_); + auto group_labels = thrust::make_constant_iterator(0); + return to_tdigest_scalar( + merge_tdigests(tdv, group_offsets, group_labels, input.size(), 1, max_centroids, stream, mr), + stream, + mr); } std::unique_ptr group_tdigest(column_view const& col, @@ -1376,16 +1363,7 @@ std::unique_ptr group_merge_tdigest(column_view const& input, return cudf::tdigest::detail::make_empty_tdigests_column(num_groups, stream, mr); } - // bring group offsets back to the host - std::vector h_group_offsets(group_offsets.size()); - cudaMemcpyAsync(h_group_offsets.data(), - group_offsets.begin(), - sizeof(size_type) * group_offsets.size(), - cudaMemcpyDefault, - stream); - return merge_tdigests(tdv, - h_group_offsets.begin(), group_offsets.data(), group_labels.data(), group_labels.size(), diff --git a/cpp/src/quantiles/tdigest/tdigest_column_view.cpp b/cpp/src/quantiles/tdigest/tdigest_column_view.cpp index a9f86ac1b5f..17844b6bb0a 100644 --- a/cpp/src/quantiles/tdigest/tdigest_column_view.cpp +++ b/cpp/src/quantiles/tdigest/tdigest_column_view.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2022, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include #include #include diff --git a/cpp/src/reductions/reductions.cpp b/cpp/src/reductions/reductions.cpp index d187375b69f..75ebc078930 100644 --- a/cpp/src/reductions/reductions.cpp +++ b/cpp/src/reductions/reductions.cpp @@ -26,8 +26,6 @@ #include #include #include -#include -#include #include #include #include diff --git a/cpp/src/reductions/scan/scan.cpp b/cpp/src/reductions/scan/scan.cpp index d3c0b54f286..b91ae19b51a 100644 --- a/cpp/src/reductions/scan/scan.cpp +++ b/cpp/src/reductions/scan/scan.cpp @@ -14,13 +14,10 @@ * limitations under the License. */ -#include #include #include #include #include -#include -#include namespace cudf { diff --git a/cpp/src/reductions/segmented/reductions.cpp b/cpp/src/reductions/segmented/reductions.cpp index 40d1d8a0a53..c4f6c135dde 100644 --- a/cpp/src/reductions/segmented/reductions.cpp +++ b/cpp/src/reductions/segmented/reductions.cpp @@ -13,16 +13,12 @@ * See the License for the specific language governing permissions and * limitations under the License. */ -#include -#include #include #include #include #include #include -#include #include -#include #include #include diff --git a/cpp/src/rolling/detail/optimized_unbounded_window.cpp b/cpp/src/rolling/detail/optimized_unbounded_window.cpp index 72c23395a93..7cad31c0658 100644 --- a/cpp/src/rolling/detail/optimized_unbounded_window.cpp +++ b/cpp/src/rolling/detail/optimized_unbounded_window.cpp @@ -18,13 +18,10 @@ #include #include #include -#include #include #include #include #include -#include -#include #include namespace cudf::detail { diff --git a/cpp/src/rolling/detail/range_window_bounds.hpp b/cpp/src/rolling/detail/range_window_bounds.hpp index 8a53e937f98..77cb2a8c7f5 100644 --- a/cpp/src/rolling/detail/range_window_bounds.hpp +++ b/cpp/src/rolling/detail/range_window_bounds.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,10 +16,7 @@ #pragma once #include -#include #include -#include -#include namespace cudf { namespace detail { diff --git a/cpp/src/rolling/range_window_bounds.cpp b/cpp/src/rolling/range_window_bounds.cpp index 69792136c64..7f698dfcd6b 100644 --- a/cpp/src/rolling/range_window_bounds.cpp +++ b/cpp/src/rolling/range_window_bounds.cpp @@ -19,7 +19,6 @@ #include #include #include -#include namespace cudf { namespace { diff --git a/cpp/src/scalar/scalar.cpp b/cpp/src/scalar/scalar.cpp index 31535198c58..4ec2174a96f 100644 --- a/cpp/src/scalar/scalar.cpp +++ b/cpp/src/scalar/scalar.cpp @@ -26,8 +26,6 @@ #include #include -#include - #include namespace cudf { diff --git a/cpp/src/scalar/scalar_factories.cpp b/cpp/src/scalar/scalar_factories.cpp index 656fe61fbbe..9f242bdffe0 100644 --- a/cpp/src/scalar/scalar_factories.cpp +++ b/cpp/src/scalar/scalar_factories.cpp @@ -16,10 +16,8 @@ #include #include -#include #include #include -#include #include #include diff --git a/cpp/src/strings/convert/convert_durations.cu b/cpp/src/strings/convert/convert_durations.cu index 0db1adf1223..f5d052c6657 100644 --- a/cpp/src/strings/convert/convert_durations.cu +++ b/cpp/src/strings/convert/convert_durations.cu @@ -16,6 +16,7 @@ #include #include #include +#include #include #include #include @@ -152,12 +153,8 @@ struct format_compiler { } // create program in device memory - d_items.resize(items.size(), stream); - CUDF_CUDA_TRY(cudaMemcpyAsync(d_items.data(), - items.data(), - items.size() * sizeof(items[0]), - cudaMemcpyDefault, - stream.value())); + d_items = cudf::detail::make_device_uvector_sync( + items, stream, cudf::get_current_device_resource_ref()); } format_item const* compiled_format_items() { return d_items.data(); } diff --git a/cpp/src/strings/regex/regexec.cpp b/cpp/src/strings/regex/regexec.cpp index d1990733e81..60ad714dfec 100644 --- a/cpp/src/strings/regex/regexec.cpp +++ b/cpp/src/strings/regex/regexec.cpp @@ -24,7 +24,6 @@ #include #include -#include #include #include diff --git a/cpp/src/strings/strings_scalar_factories.cpp b/cpp/src/strings/strings_scalar_factories.cpp index 219d1174d42..1cc405234b2 100644 --- a/cpp/src/strings/strings_scalar_factories.cpp +++ b/cpp/src/strings/strings_scalar_factories.cpp @@ -16,7 +16,6 @@ #include #include -#include #include diff --git a/cpp/src/structs/structs_column_view.cpp b/cpp/src/structs/structs_column_view.cpp index b0284e9cb96..e14142a9ad1 100644 --- a/cpp/src/structs/structs_column_view.cpp +++ b/cpp/src/structs/structs_column_view.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include #include #include diff --git a/cpp/src/structs/utilities.cpp b/cpp/src/structs/utilities.cpp index 5df9943303d..4012ee3d21c 100644 --- a/cpp/src/structs/utilities.cpp +++ b/cpp/src/structs/utilities.cpp @@ -21,13 +21,10 @@ #include #include #include -#include #include #include #include -#include #include -#include #include #include diff --git a/cpp/src/table/table.cpp b/cpp/src/table/table.cpp index cb707c94288..41c64c6decb 100644 --- a/cpp/src/table/table.cpp +++ b/cpp/src/table/table.cpp @@ -18,7 +18,6 @@ #include #include #include -#include #include diff --git a/cpp/src/table/table_view.cpp b/cpp/src/table/table_view.cpp index 8a5340dc20d..659beb749af 100644 --- a/cpp/src/table/table_view.cpp +++ b/cpp/src/table/table_view.cpp @@ -20,10 +20,7 @@ #include #include -#include - #include -#include #include namespace cudf { diff --git a/cpp/src/transform/transform.cpp b/cpp/src/transform/transform.cpp index 52b96bc9039..b919ac16956 100644 --- a/cpp/src/transform/transform.cpp +++ b/cpp/src/transform/transform.cpp @@ -23,8 +23,6 @@ #include #include #include -#include -#include #include #include diff --git a/cpp/src/utilities/cuda.cpp b/cpp/src/utilities/cuda.cpp index 53ca0608170..d979bda41d0 100644 --- a/cpp/src/utilities/cuda.cpp +++ b/cpp/src/utilities/cuda.cpp @@ -18,8 +18,6 @@ #include #include -#include - namespace cudf::detail { cudf::size_type num_multiprocessors() diff --git a/cpp/src/utilities/host_memory.cpp b/cpp/src/utilities/host_memory.cpp index 9d8e3cf2fa6..e30806a5011 100644 --- a/cpp/src/utilities/host_memory.cpp +++ b/cpp/src/utilities/host_memory.cpp @@ -18,7 +18,6 @@ #include #include #include -#include #include #include diff --git a/cpp/src/utilities/prefetch.cpp b/cpp/src/utilities/prefetch.cpp index 58971552758..000526723c4 100644 --- a/cpp/src/utilities/prefetch.cpp +++ b/cpp/src/utilities/prefetch.cpp @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include #include diff --git a/cpp/src/utilities/stream_pool.cpp b/cpp/src/utilities/stream_pool.cpp index 8c29182bfb5..7069b59be26 100644 --- a/cpp/src/utilities/stream_pool.cpp +++ b/cpp/src/utilities/stream_pool.cpp @@ -23,7 +23,6 @@ #include #include -#include #include #include diff --git a/cpp/src/utilities/traits.cpp b/cpp/src/utilities/traits.cpp index a68dc84e340..c1e71f5f8f9 100644 --- a/cpp/src/utilities/traits.cpp +++ b/cpp/src/utilities/traits.cpp @@ -19,8 +19,6 @@ #include #include -#include - namespace cudf { namespace { diff --git a/cpp/src/utilities/type_checks.cpp b/cpp/src/utilities/type_checks.cpp index 3095b342748..84c8529641d 100644 --- a/cpp/src/utilities/type_checks.cpp +++ b/cpp/src/utilities/type_checks.cpp @@ -21,8 +21,6 @@ #include #include -#include - #include namespace cudf { diff --git a/cpp/tests/CMakeLists.txt b/cpp/tests/CMakeLists.txt index a4213dcbe94..b78a64d0e55 100644 --- a/cpp/tests/CMakeLists.txt +++ b/cpp/tests/CMakeLists.txt @@ -611,7 +611,6 @@ ConfigureTest( text/bpe_tests.cpp text/edit_distance_tests.cpp text/jaccard_tests.cpp - text/minhash_tests.cpp text/ngrams_tests.cpp text/ngrams_tokenize_tests.cpp text/normalize_tests.cpp diff --git a/cpp/tests/ast/transform_tests.cpp b/cpp/tests/ast/transform_tests.cpp index a4bde50a21e..7af88d8aa34 100644 --- a/cpp/tests/ast/transform_tests.cpp +++ b/cpp/tests/ast/transform_tests.cpp @@ -18,7 +18,6 @@ #include #include #include -#include #include #include @@ -26,14 +25,8 @@ #include #include #include -#include -#include -#include #include #include -#include - -#include #include @@ -41,7 +34,6 @@ #include #include #include -#include #include template diff --git a/cpp/tests/binaryop/binop-compiled-test.cpp b/cpp/tests/binaryop/binop-compiled-test.cpp index aa5b49567e6..3bd67001c16 100644 --- a/cpp/tests/binaryop/binop-compiled-test.cpp +++ b/cpp/tests/binaryop/binop-compiled-test.cpp @@ -26,9 +26,7 @@ #include #include #include -#include #include -#include #include diff --git a/cpp/tests/binaryop/binop-generic-ptx-test.cpp b/cpp/tests/binaryop/binop-generic-ptx-test.cpp index 03cc87a1968..e9a2761db4a 100644 --- a/cpp/tests/binaryop/binop-generic-ptx-test.cpp +++ b/cpp/tests/binaryop/binop-generic-ptx-test.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2023, NVIDIA CORPORATION. + * Copyright (c) 2019-2024, NVIDIA CORPORATION. * * Copyright 2018-2019 BlazingDB, Inc. * Copyright 2018 Christian Noboa Mardini @@ -19,7 +19,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/bitmask/bitmask_tests.cpp b/cpp/tests/bitmask/bitmask_tests.cpp index fe221fb1c48..799bf646e52 100644 --- a/cpp/tests/bitmask/bitmask_tests.cpp +++ b/cpp/tests/bitmask/bitmask_tests.cpp @@ -16,7 +16,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/column/bit_cast_test.cpp b/cpp/tests/column/bit_cast_test.cpp index ab230ab036e..5570a7d498c 100644 --- a/cpp/tests/column/bit_cast_test.cpp +++ b/cpp/tests/column/bit_cast_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include @@ -26,8 +25,6 @@ #include -#include - template struct rep_type_impl { using type = void; diff --git a/cpp/tests/column/column_test.cpp b/cpp/tests/column/column_test.cpp index 631f5150829..d700adaebd5 100644 --- a/cpp/tests/column/column_test.cpp +++ b/cpp/tests/column/column_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include #include diff --git a/cpp/tests/column/column_view_device_span_test.cpp b/cpp/tests/column/column_view_device_span_test.cpp index 6de9121158b..470437f4112 100644 --- a/cpp/tests/column/column_view_device_span_test.cpp +++ b/cpp/tests/column/column_view_device_span_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/column/column_view_shallow_test.cpp b/cpp/tests/column/column_view_shallow_test.cpp index 37ab4b8f387..ad344476332 100644 --- a/cpp/tests/column/column_view_shallow_test.cpp +++ b/cpp/tests/column/column_view_shallow_test.cpp @@ -15,9 +15,7 @@ */ #include -#include #include -#include #include #include diff --git a/cpp/tests/column/factories_test.cpp b/cpp/tests/column/factories_test.cpp index 603187f0330..aa9d508b6aa 100644 --- a/cpp/tests/column/factories_test.cpp +++ b/cpp/tests/column/factories_test.cpp @@ -26,11 +26,8 @@ #include #include #include -#include #include -#include - #include class ColumnFactoryTest : public cudf::test::BaseFixture { diff --git a/cpp/tests/copying/concatenate_tests.cpp b/cpp/tests/copying/concatenate_tests.cpp index 18140c34abd..aedc498964a 100644 --- a/cpp/tests/copying/concatenate_tests.cpp +++ b/cpp/tests/copying/concatenate_tests.cpp @@ -34,8 +34,6 @@ #include #include -#include - #include #include #include diff --git a/cpp/tests/copying/copy_if_else_nested_tests.cpp b/cpp/tests/copying/copy_if_else_nested_tests.cpp index cfbd181f944..e1cdfe9beed 100644 --- a/cpp/tests/copying/copy_if_else_nested_tests.cpp +++ b/cpp/tests/copying/copy_if_else_nested_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/copy_range_tests.cpp b/cpp/tests/copying/copy_range_tests.cpp index 25d93da277b..e2133a546e4 100644 --- a/cpp/tests/copying/copy_range_tests.cpp +++ b/cpp/tests/copying/copy_range_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/copy_tests.cpp b/cpp/tests/copying/copy_tests.cpp index 4124f749012..9c00725d5d2 100644 --- a/cpp/tests/copying/copy_tests.cpp +++ b/cpp/tests/copying/copy_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/gather_list_tests.cpp b/cpp/tests/copying/gather_list_tests.cpp index 247090aac90..93f71345c5c 100644 --- a/cpp/tests/copying/gather_list_tests.cpp +++ b/cpp/tests/copying/gather_list_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,8 +17,6 @@ #include #include #include -#include -#include #include #include diff --git a/cpp/tests/copying/gather_str_tests.cpp b/cpp/tests/copying/gather_str_tests.cpp index 28098878086..795e3f30aa1 100644 --- a/cpp/tests/copying/gather_str_tests.cpp +++ b/cpp/tests/copying/gather_str_tests.cpp @@ -16,7 +16,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/gather_struct_tests.cpp b/cpp/tests/copying/gather_struct_tests.cpp index 1598ab2646a..b2c0f7acc3a 100644 --- a/cpp/tests/copying/gather_struct_tests.cpp +++ b/cpp/tests/copying/gather_struct_tests.cpp @@ -17,20 +17,15 @@ #include #include #include -#include #include #include #include #include #include -#include -#include -#include #include #include #include -#include #include diff --git a/cpp/tests/copying/gather_tests.cpp b/cpp/tests/copying/gather_tests.cpp index 07ce672b14d..908dcd67673 100644 --- a/cpp/tests/copying/gather_tests.cpp +++ b/cpp/tests/copying/gather_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include #include diff --git a/cpp/tests/copying/get_value_tests.cpp b/cpp/tests/copying/get_value_tests.cpp index 90ff97e7355..b2d64dac7c8 100644 --- a/cpp/tests/copying/get_value_tests.cpp +++ b/cpp/tests/copying/get_value_tests.cpp @@ -16,10 +16,8 @@ #include #include -#include #include #include -#include #include #include diff --git a/cpp/tests/copying/purge_nonempty_nulls_tests.cpp b/cpp/tests/copying/purge_nonempty_nulls_tests.cpp index 4f28ff12941..1f76efdc4c3 100644 --- a/cpp/tests/copying/purge_nonempty_nulls_tests.cpp +++ b/cpp/tests/copying/purge_nonempty_nulls_tests.cpp @@ -16,13 +16,10 @@ #include #include #include -#include #include #include #include -#include -#include #include #include #include diff --git a/cpp/tests/copying/reverse_tests.cpp b/cpp/tests/copying/reverse_tests.cpp index e4b2d319ddf..46516436901 100644 --- a/cpp/tests/copying/reverse_tests.cpp +++ b/cpp/tests/copying/reverse_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,17 +17,13 @@ #include #include #include -#include #include #include #include -#include -#include #include #include -#include #include #include #include diff --git a/cpp/tests/copying/sample_tests.cpp b/cpp/tests/copying/sample_tests.cpp index 2f76e3f1fcd..8be5d8c1fbb 100644 --- a/cpp/tests/copying/sample_tests.cpp +++ b/cpp/tests/copying/sample_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -15,12 +15,9 @@ */ #include -#include #include -#include #include -#include #include #include #include diff --git a/cpp/tests/copying/scatter_list_scalar_tests.cpp b/cpp/tests/copying/scatter_list_scalar_tests.cpp index 42d2e004d6b..23faa6e5b86 100644 --- a/cpp/tests/copying/scatter_list_scalar_tests.cpp +++ b/cpp/tests/copying/scatter_list_scalar_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -21,7 +21,6 @@ #include #include -#include using mask_vector = std::vector; using size_column = cudf::test::fixed_width_column_wrapper; diff --git a/cpp/tests/copying/scatter_list_tests.cpp b/cpp/tests/copying/scatter_list_tests.cpp index a82860a3eec..1f87fcfcc99 100644 --- a/cpp/tests/copying/scatter_list_tests.cpp +++ b/cpp/tests/copying/scatter_list_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/scatter_struct_scalar_tests.cpp b/cpp/tests/copying/scatter_struct_scalar_tests.cpp index 78572b0bb37..1d1da8a1b1e 100644 --- a/cpp/tests/copying/scatter_struct_scalar_tests.cpp +++ b/cpp/tests/copying/scatter_struct_scalar_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -19,7 +19,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/copying/scatter_struct_tests.cpp b/cpp/tests/copying/scatter_struct_tests.cpp index c92244d047b..7d88e9af85f 100644 --- a/cpp/tests/copying/scatter_struct_tests.cpp +++ b/cpp/tests/copying/scatter_struct_tests.cpp @@ -21,7 +21,6 @@ #include #include -#include #include using namespace cudf::test::iterators; diff --git a/cpp/tests/copying/scatter_tests.cpp b/cpp/tests/copying/scatter_tests.cpp index 41a753cd0ac..74c04446bdd 100644 --- a/cpp/tests/copying/scatter_tests.cpp +++ b/cpp/tests/copying/scatter_tests.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include #include @@ -23,7 +22,6 @@ #include #include #include -#include #include diff --git a/cpp/tests/copying/segmented_gather_list_tests.cpp b/cpp/tests/copying/segmented_gather_list_tests.cpp index 8881fb344a2..a133ae43872 100644 --- a/cpp/tests/copying/segmented_gather_list_tests.cpp +++ b/cpp/tests/copying/segmented_gather_list_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -22,7 +22,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/copying/shift_tests.cpp b/cpp/tests/copying/shift_tests.cpp index ff6808d9a79..72a8e7357bc 100644 --- a/cpp/tests/copying/shift_tests.cpp +++ b/cpp/tests/copying/shift_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include @@ -30,7 +29,6 @@ #include #include -#include using TestTypes = cudf::test::Types; diff --git a/cpp/tests/copying/slice_tests.cpp b/cpp/tests/copying/slice_tests.cpp index aef0d4ad78a..3868a147fa8 100644 --- a/cpp/tests/copying/slice_tests.cpp +++ b/cpp/tests/copying/slice_tests.cpp @@ -22,12 +22,8 @@ #include #include -#include #include #include -#include -#include -#include #include #include diff --git a/cpp/tests/copying/utility_tests.cpp b/cpp/tests/copying/utility_tests.cpp index 0905f9babdc..90457f8d74c 100644 --- a/cpp/tests/copying/utility_tests.cpp +++ b/cpp/tests/copying/utility_tests.cpp @@ -23,7 +23,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/datetime/datetime_ops_test.cpp b/cpp/tests/datetime/datetime_ops_test.cpp index 603edb27c7c..44f99adc0e9 100644 --- a/cpp/tests/datetime/datetime_ops_test.cpp +++ b/cpp/tests/datetime/datetime_ops_test.cpp @@ -23,14 +23,11 @@ #include #include -#include #include #include #include #include -#include - #define XXX false // stub for null values constexpr cudf::test::debug_output_level verbosity{cudf::test::debug_output_level::ALL_ERRORS}; diff --git a/cpp/tests/dictionary/add_keys_test.cpp b/cpp/tests/dictionary/add_keys_test.cpp index 46bf5468922..ebc8c11e86c 100644 --- a/cpp/tests/dictionary/add_keys_test.cpp +++ b/cpp/tests/dictionary/add_keys_test.cpp @@ -24,8 +24,6 @@ #include #include -#include - struct DictionaryAddKeysTest : public cudf::test::BaseFixture {}; TEST_F(DictionaryAddKeysTest, StringsColumn) diff --git a/cpp/tests/dictionary/encode_test.cpp b/cpp/tests/dictionary/encode_test.cpp index 5db0e9fa1e4..dfa3ede5d46 100644 --- a/cpp/tests/dictionary/encode_test.cpp +++ b/cpp/tests/dictionary/encode_test.cpp @@ -21,8 +21,6 @@ #include #include -#include - struct DictionaryEncodeTest : public cudf::test::BaseFixture {}; TEST_F(DictionaryEncodeTest, EncodeStringColumn) diff --git a/cpp/tests/dictionary/fill_test.cpp b/cpp/tests/dictionary/fill_test.cpp index 18696b66e48..bc7d19201aa 100644 --- a/cpp/tests/dictionary/fill_test.cpp +++ b/cpp/tests/dictionary/fill_test.cpp @@ -18,13 +18,10 @@ #include #include -#include #include #include #include -#include - struct DictionaryFillTest : public cudf::test::BaseFixture {}; TEST_F(DictionaryFillTest, StringsColumn) diff --git a/cpp/tests/dictionary/search_test.cpp b/cpp/tests/dictionary/search_test.cpp index 25501b4fde7..2774173b80a 100644 --- a/cpp/tests/dictionary/search_test.cpp +++ b/cpp/tests/dictionary/search_test.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include diff --git a/cpp/tests/dictionary/slice_test.cpp b/cpp/tests/dictionary/slice_test.cpp index d80f8dee079..8c15d6dbecd 100644 --- a/cpp/tests/dictionary/slice_test.cpp +++ b/cpp/tests/dictionary/slice_test.cpp @@ -19,7 +19,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/filling/fill_tests.cpp b/cpp/tests/filling/fill_tests.cpp index 26badefe698..a5e2db6a005 100644 --- a/cpp/tests/filling/fill_tests.cpp +++ b/cpp/tests/filling/fill_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/filling/repeat_tests.cpp b/cpp/tests/filling/repeat_tests.cpp index 6326765c68b..c856984a4a3 100644 --- a/cpp/tests/filling/repeat_tests.cpp +++ b/cpp/tests/filling/repeat_tests.cpp @@ -17,14 +17,11 @@ #include #include #include -#include #include #include #include #include -#include -#include #include #include @@ -33,7 +30,6 @@ #include #include -#include constexpr cudf::test::debug_output_level verbosity{cudf::test::debug_output_level::ALL_ERRORS}; diff --git a/cpp/tests/filling/sequence_tests.cpp b/cpp/tests/filling/sequence_tests.cpp index 0783b4e5bbb..53782c90c26 100644 --- a/cpp/tests/filling/sequence_tests.cpp +++ b/cpp/tests/filling/sequence_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/fixed_point/fixed_point_tests.cpp b/cpp/tests/fixed_point/fixed_point_tests.cpp index a222289216d..b96c6909e55 100644 --- a/cpp/tests/fixed_point/fixed_point_tests.cpp +++ b/cpp/tests/fixed_point/fixed_point_tests.cpp @@ -18,17 +18,14 @@ #include #include #include -#include #include #include -#include #include #include #include #include -#include #include using namespace numeric; diff --git a/cpp/tests/groupby/collect_list_tests.cpp b/cpp/tests/groupby/collect_list_tests.cpp index a79b6a32916..ba456084a7c 100644 --- a/cpp/tests/groupby/collect_list_tests.cpp +++ b/cpp/tests/groupby/collect_list_tests.cpp @@ -20,8 +20,6 @@ #include #include -#include - template struct groupby_collect_list_test : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/groupby/collect_set_tests.cpp b/cpp/tests/groupby/collect_set_tests.cpp index 61d2838590b..dfd7eb82c4a 100644 --- a/cpp/tests/groupby/collect_set_tests.cpp +++ b/cpp/tests/groupby/collect_set_tests.cpp @@ -19,7 +19,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/groupby/correlation_tests.cpp b/cpp/tests/groupby/correlation_tests.cpp index 26f714632dd..f8cc813e877 100644 --- a/cpp/tests/groupby/correlation_tests.cpp +++ b/cpp/tests/groupby/correlation_tests.cpp @@ -25,7 +25,6 @@ #include #include -#include using namespace cudf::test::iterators; diff --git a/cpp/tests/groupby/covariance_tests.cpp b/cpp/tests/groupby/covariance_tests.cpp index e3eb2da201f..81378bb91e8 100644 --- a/cpp/tests/groupby/covariance_tests.cpp +++ b/cpp/tests/groupby/covariance_tests.cpp @@ -23,10 +23,8 @@ #include #include -#include #include -#include using namespace cudf::test::iterators; diff --git a/cpp/tests/groupby/groupby_test_util.cpp b/cpp/tests/groupby/groupby_test_util.cpp index 5d99d15ae77..df0375d6a09 100644 --- a/cpp/tests/groupby/groupby_test_util.cpp +++ b/cpp/tests/groupby/groupby_test_util.cpp @@ -17,8 +17,8 @@ #include "groupby_test_util.hpp" #include -#include #include +#include #include #include @@ -27,9 +27,6 @@ #include #include #include -#include - -#include void test_single_agg(cudf::column_view const& keys, cudf::column_view const& values, diff --git a/cpp/tests/groupby/groupby_test_util.hpp b/cpp/tests/groupby/groupby_test_util.hpp index 755b0c20f17..9d2e613be3e 100644 --- a/cpp/tests/groupby/groupby_test_util.hpp +++ b/cpp/tests/groupby/groupby_test_util.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,11 +16,8 @@ #pragma once -#include #include -#include #include -#include enum class force_use_sort_impl : bool { NO, YES }; diff --git a/cpp/tests/groupby/histogram_tests.cpp b/cpp/tests/groupby/histogram_tests.cpp index 2d447025919..783cfb17e49 100644 --- a/cpp/tests/groupby/histogram_tests.cpp +++ b/cpp/tests/groupby/histogram_tests.cpp @@ -20,7 +20,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/groupby/max_scan_tests.cpp b/cpp/tests/groupby/max_scan_tests.cpp index d86de798844..6195e0179ec 100644 --- a/cpp/tests/groupby/max_scan_tests.cpp +++ b/cpp/tests/groupby/max_scan_tests.cpp @@ -22,7 +22,6 @@ #include #include -#include using namespace cudf::test::iterators; diff --git a/cpp/tests/groupby/merge_lists_tests.cpp b/cpp/tests/groupby/merge_lists_tests.cpp index 279d71560b4..4481e2dc022 100644 --- a/cpp/tests/groupby/merge_lists_tests.cpp +++ b/cpp/tests/groupby/merge_lists_tests.cpp @@ -21,7 +21,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/groupby/merge_sets_tests.cpp b/cpp/tests/groupby/merge_sets_tests.cpp index 9736bb84dd6..1bfba265478 100644 --- a/cpp/tests/groupby/merge_sets_tests.cpp +++ b/cpp/tests/groupby/merge_sets_tests.cpp @@ -21,7 +21,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/groupby/rank_scan_tests.cpp b/cpp/tests/groupby/rank_scan_tests.cpp index 7f31bc9089f..f2a50248b4a 100644 --- a/cpp/tests/groupby/rank_scan_tests.cpp +++ b/cpp/tests/groupby/rank_scan_tests.cpp @@ -22,8 +22,6 @@ #include #include -#include - using namespace cudf::test::iterators; template diff --git a/cpp/tests/groupby/shift_tests.cpp b/cpp/tests/groupby/shift_tests.cpp index 14c9ceb4508..49f9d7cb10a 100644 --- a/cpp/tests/groupby/shift_tests.cpp +++ b/cpp/tests/groupby/shift_tests.cpp @@ -21,7 +21,6 @@ #include #include -#include #include template diff --git a/cpp/tests/hashing/md5_test.cpp b/cpp/tests/hashing/md5_test.cpp index 69e518cbf8d..b54adb52496 100644 --- a/cpp/tests/hashing/md5_test.cpp +++ b/cpp/tests/hashing/md5_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/murmurhash3_x86_32_test.cpp b/cpp/tests/hashing/murmurhash3_x86_32_test.cpp index c1a6e6ff6e1..b4622f5eb81 100644 --- a/cpp/tests/hashing/murmurhash3_x86_32_test.cpp +++ b/cpp/tests/hashing/murmurhash3_x86_32_test.cpp @@ -17,11 +17,9 @@ #include #include #include -#include #include #include -#include #include constexpr cudf::test::debug_output_level verbosity{cudf::test::debug_output_level::ALL_ERRORS}; diff --git a/cpp/tests/hashing/sha1_test.cpp b/cpp/tests/hashing/sha1_test.cpp index e28e71442a6..1e86751bb4c 100644 --- a/cpp/tests/hashing/sha1_test.cpp +++ b/cpp/tests/hashing/sha1_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/sha224_test.cpp b/cpp/tests/hashing/sha224_test.cpp index 61b584f94df..259e7102ee2 100644 --- a/cpp/tests/hashing/sha224_test.cpp +++ b/cpp/tests/hashing/sha224_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/sha256_test.cpp b/cpp/tests/hashing/sha256_test.cpp index 8bc47c92c6b..a4affc87874 100644 --- a/cpp/tests/hashing/sha256_test.cpp +++ b/cpp/tests/hashing/sha256_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/sha384_test.cpp b/cpp/tests/hashing/sha384_test.cpp index 4c79934f98d..8a5c090eeea 100644 --- a/cpp/tests/hashing/sha384_test.cpp +++ b/cpp/tests/hashing/sha384_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/sha512_test.cpp b/cpp/tests/hashing/sha512_test.cpp index 0eb1c60b8fc..77fc56b5f13 100644 --- a/cpp/tests/hashing/sha512_test.cpp +++ b/cpp/tests/hashing/sha512_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/hashing/xxhash_64_test.cpp b/cpp/tests/hashing/xxhash_64_test.cpp index ab4ed829681..d8694a72d94 100644 --- a/cpp/tests/hashing/xxhash_64_test.cpp +++ b/cpp/tests/hashing/xxhash_64_test.cpp @@ -17,11 +17,8 @@ #include #include #include -#include #include -#include -#include #include using NumericTypesNoBools = diff --git a/cpp/tests/interop/from_arrow_device_test.cpp b/cpp/tests/interop/from_arrow_device_test.cpp index 2151ec6e22f..1ddc33e749a 100644 --- a/cpp/tests/interop/from_arrow_device_test.cpp +++ b/cpp/tests/interop/from_arrow_device_test.cpp @@ -17,17 +17,13 @@ #include "nanoarrow_utils.hpp" #include -#include #include #include -#include #include #include #include #include -#include -#include #include #include #include diff --git a/cpp/tests/interop/from_arrow_host_test.cpp b/cpp/tests/interop/from_arrow_host_test.cpp index ef9936b214c..d93ef28aab8 100644 --- a/cpp/tests/interop/from_arrow_host_test.cpp +++ b/cpp/tests/interop/from_arrow_host_test.cpp @@ -20,7 +20,6 @@ #include #include #include -#include #include #include @@ -28,7 +27,6 @@ #include #include #include -#include #include #include #include diff --git a/cpp/tests/interop/from_arrow_stream_test.cpp b/cpp/tests/interop/from_arrow_stream_test.cpp index 80a2e4b2ffd..3916025bf22 100644 --- a/cpp/tests/interop/from_arrow_stream_test.cpp +++ b/cpp/tests/interop/from_arrow_stream_test.cpp @@ -17,27 +17,14 @@ #include "nanoarrow_utils.hpp" #include -#include -#include #include -#include -#include -#include -#include #include -#include -#include -#include -#include #include #include #include -#include #include -#include - struct VectorOfArrays { std::vector arrays; nanoarrow::UniqueSchema schema; diff --git a/cpp/tests/interop/from_arrow_test.cpp b/cpp/tests/interop/from_arrow_test.cpp index 6e742b9e4cf..18efae75cb1 100644 --- a/cpp/tests/interop/from_arrow_test.cpp +++ b/cpp/tests/interop/from_arrow_test.cpp @@ -25,9 +25,7 @@ #include #include #include -#include #include -#include #include #include #include @@ -37,8 +35,6 @@ #include #include -#include -#include std::unique_ptr get_cudf_table() { diff --git a/cpp/tests/interop/to_arrow_device_test.cpp b/cpp/tests/interop/to_arrow_device_test.cpp index 7ba586461dc..29aa928c277 100644 --- a/cpp/tests/interop/to_arrow_device_test.cpp +++ b/cpp/tests/interop/to_arrow_device_test.cpp @@ -17,21 +17,15 @@ #include "nanoarrow_utils.hpp" #include -#include #include -#include -#include #include #include -#include -#include #include #include #include #include #include -#include #include #include diff --git a/cpp/tests/interop/to_arrow_host_test.cpp b/cpp/tests/interop/to_arrow_host_test.cpp index fcb4433b42e..fa3aa82fee2 100644 --- a/cpp/tests/interop/to_arrow_host_test.cpp +++ b/cpp/tests/interop/to_arrow_host_test.cpp @@ -17,20 +17,14 @@ #include "nanoarrow_utils.hpp" #include -#include #include -#include -#include #include #include #include #include -#include #include #include -#include -#include #include #include #include diff --git a/cpp/tests/interop/to_arrow_test.cpp b/cpp/tests/interop/to_arrow_test.cpp index a6aa4b22eca..86295d8efb1 100644 --- a/cpp/tests/interop/to_arrow_test.cpp +++ b/cpp/tests/interop/to_arrow_test.cpp @@ -19,14 +19,12 @@ #include #include #include -#include #include #include #include #include #include -#include #include #include #include diff --git a/cpp/tests/io/csv_test.cpp b/cpp/tests/io/csv_test.cpp index b265dcf9273..cc1e367d114 100644 --- a/cpp/tests/io/csv_test.cpp +++ b/cpp/tests/io/csv_test.cpp @@ -17,14 +17,12 @@ #include #include #include -#include #include #include #include #include #include -#include #include #include #include @@ -32,18 +30,12 @@ #include #include #include -#include -#include #include -#include #include -#include - #include #include -#include #include #include #include diff --git a/cpp/tests/io/file_io_test.cpp b/cpp/tests/io/file_io_test.cpp index 3c41f21b0a4..1b85541687a 100644 --- a/cpp/tests/io/file_io_test.cpp +++ b/cpp/tests/io/file_io_test.cpp @@ -15,13 +15,10 @@ */ #include -#include #include #include -#include - // Base test fixture for tests struct CuFileIOTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/io/json/json_quote_normalization_test.cpp b/cpp/tests/io/json/json_quote_normalization_test.cpp index d23acf3ae00..c8c2d18903f 100644 --- a/cpp/tests/io/json/json_quote_normalization_test.cpp +++ b/cpp/tests/io/json/json_quote_normalization_test.cpp @@ -20,7 +20,6 @@ #include #include -#include #include #include #include @@ -29,7 +28,6 @@ #include #include -#include #include diff --git a/cpp/tests/io/json/json_test.cpp b/cpp/tests/io/json/json_test.cpp index cb6716f4a18..5f070bd53b9 100644 --- a/cpp/tests/io/json/json_test.cpp +++ b/cpp/tests/io/json/json_test.cpp @@ -39,8 +39,6 @@ #include -#include - #include #include #include diff --git a/cpp/tests/io/json/json_tree.cpp b/cpp/tests/io/json/json_tree.cpp index 15682c6ae6b..887d4fa783f 100644 --- a/cpp/tests/io/json/json_tree.cpp +++ b/cpp/tests/io/json/json_tree.cpp @@ -15,12 +15,8 @@ */ #include "io/json/nested_json.hpp" -#include "io/utilities/hostdevice_vector.hpp" #include -#include -#include -#include #include #include @@ -29,9 +25,9 @@ #include #include -#include #include +#include #include #include #include diff --git a/cpp/tests/io/json/nested_json_test.cpp b/cpp/tests/io/json/nested_json_test.cpp index f32aba0e632..e0e955c4f48 100644 --- a/cpp/tests/io/json/nested_json_test.cpp +++ b/cpp/tests/io/json/nested_json_test.cpp @@ -21,24 +21,16 @@ #include #include #include -#include #include -#include #include -#include #include -#include #include -#include #include #include #include #include -#include - -#include #include #include diff --git a/cpp/tests/io/orc_chunked_reader_test.cu b/cpp/tests/io/orc_chunked_reader_test.cu index 8ad1fea649d..5f1aea71f73 100644 --- a/cpp/tests/io/orc_chunked_reader_test.cu +++ b/cpp/tests/io/orc_chunked_reader_test.cu @@ -1358,10 +1358,11 @@ TEST_F(OrcChunkedReaderInputLimitTest, SizeTypeRowsOverflow) int64_t constexpr total_rows = num_rows * num_reps; static_assert(total_rows > std::numeric_limits::max()); - auto const it = cudf::detail::make_counting_transform_iterator(0l, [num_rows](int64_t i) { - return (i % num_rows) % static_cast(std::numeric_limits::max() / 2); - }); - auto const col = data_col(it, it + num_rows); + auto const it = thrust::make_transform_iterator( + thrust::make_counting_iterator(0), [num_rows](int64_t i) { + return (i % num_rows) % static_cast(std::numeric_limits::max() / 2); + }); + auto const col = data_col(it, it + num_rows); auto const chunk_table = cudf::table_view{{col}}; std::vector data_buffer; diff --git a/cpp/tests/io/orc_test.cpp b/cpp/tests/io/orc_test.cpp index cce0adbf317..fce99187516 100644 --- a/cpp/tests/io/orc_test.cpp +++ b/cpp/tests/io/orc_test.cpp @@ -31,7 +31,6 @@ #include #include #include -#include #include #include #include diff --git a/cpp/tests/io/parquet_common.hpp b/cpp/tests/io/parquet_common.hpp index c90b81ed27a..d66aa3bde9d 100644 --- a/cpp/tests/io/parquet_common.hpp +++ b/cpp/tests/io/parquet_common.hpp @@ -22,13 +22,11 @@ #include #include -#include #include #include #include #include -#include #include #include diff --git a/cpp/tests/io/parquet_misc_test.cpp b/cpp/tests/io/parquet_misc_test.cpp index f1286a00d22..d66f685cd9c 100644 --- a/cpp/tests/io/parquet_misc_test.cpp +++ b/cpp/tests/io/parquet_misc_test.cpp @@ -20,8 +20,6 @@ #include #include -#include -#include #include diff --git a/cpp/tests/io/parquet_reader_test.cpp b/cpp/tests/io/parquet_reader_test.cpp index 7986a3c6d70..177e6163d4f 100644 --- a/cpp/tests/io/parquet_reader_test.cpp +++ b/cpp/tests/io/parquet_reader_test.cpp @@ -29,6 +29,8 @@ #include #include +#include + #include TEST_F(ParquetReaderTest, UserBounds) diff --git a/cpp/tests/io/parquet_test.cpp b/cpp/tests/io/parquet_test.cpp index be2ecd56424..5c3c8342cd2 100644 --- a/cpp/tests/io/parquet_test.cpp +++ b/cpp/tests/io/parquet_test.cpp @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include // NOTE: this file exists to define the parquet test's `main()` function. diff --git a/cpp/tests/io/row_selection_test.cpp b/cpp/tests/io/row_selection_test.cpp index ebadd870091..c40d3bbd299 100644 --- a/cpp/tests/io/row_selection_test.cpp +++ b/cpp/tests/io/row_selection_test.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include diff --git a/cpp/tests/io/text/data_chunk_source_test.cpp b/cpp/tests/io/text/data_chunk_source_test.cpp index 6f46df20633..79ce908f3e0 100644 --- a/cpp/tests/io/text/data_chunk_source_test.cpp +++ b/cpp/tests/io/text/data_chunk_source_test.cpp @@ -15,14 +15,11 @@ */ #include -#include #include #include #include -#include - #include #include diff --git a/cpp/tests/io/text/multibyte_split_test.cpp b/cpp/tests/io/text/multibyte_split_test.cpp index 74d08061df9..60244462e2c 100644 --- a/cpp/tests/io/text/multibyte_split_test.cpp +++ b/cpp/tests/io/text/multibyte_split_test.cpp @@ -19,16 +19,12 @@ #include #include #include -#include -#include #include -#include #include #include #include #include -#include #include using cudf::test::strings_column_wrapper; diff --git a/cpp/tests/iterator/value_iterator.cpp b/cpp/tests/iterator/value_iterator.cpp index 22bc7475dbe..f7f7c0f2721 100644 --- a/cpp/tests/iterator/value_iterator.cpp +++ b/cpp/tests/iterator/value_iterator.cpp @@ -13,7 +13,6 @@ * the License. */ -#include #include CUDF_TEST_PROGRAM_MAIN() diff --git a/cpp/tests/jit/parse_ptx_function.cpp b/cpp/tests/jit/parse_ptx_function.cpp index 6f9dfd06730..c9bb691907a 100644 --- a/cpp/tests/jit/parse_ptx_function.cpp +++ b/cpp/tests/jit/parse_ptx_function.cpp @@ -16,7 +16,6 @@ #include "jit/parser.hpp" -#include #include #include diff --git a/cpp/tests/join/cross_join_tests.cpp b/cpp/tests/join/cross_join_tests.cpp index d87f5e54153..971913443e5 100644 --- a/cpp/tests/join/cross_join_tests.cpp +++ b/cpp/tests/join/cross_join_tests.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include #include diff --git a/cpp/tests/join/distinct_join_tests.cpp b/cpp/tests/join/distinct_join_tests.cpp index 178edc52dd3..9070efa38fe 100644 --- a/cpp/tests/join/distinct_join_tests.cpp +++ b/cpp/tests/join/distinct_join_tests.cpp @@ -15,12 +15,8 @@ */ #include -#include #include -#include #include -#include -#include #include #include @@ -31,7 +27,6 @@ #include #include -#include #include template diff --git a/cpp/tests/join/join_tests.cpp b/cpp/tests/join/join_tests.cpp index 3431e941359..6a8a54c8465 100644 --- a/cpp/tests/join/join_tests.cpp +++ b/cpp/tests/join/join_tests.cpp @@ -20,17 +20,12 @@ #include #include #include -#include #include -#include #include #include -#include -#include #include #include -#include #include #include #include diff --git a/cpp/tests/join/semi_anti_join_tests.cpp b/cpp/tests/join/semi_anti_join_tests.cpp index 554d5754e39..ddc65c3f379 100644 --- a/cpp/tests/join/semi_anti_join_tests.cpp +++ b/cpp/tests/join/semi_anti_join_tests.cpp @@ -22,7 +22,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/json/json_tests.cpp b/cpp/tests/json/json_tests.cpp index 42a574ac5c0..53166e04173 100644 --- a/cpp/tests/json/json_tests.cpp +++ b/cpp/tests/json/json_tests.cpp @@ -19,7 +19,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/large_strings/large_strings_fixture.cpp b/cpp/tests/large_strings/large_strings_fixture.cpp index 7b61be113f9..f1404990354 100644 --- a/cpp/tests/large_strings/large_strings_fixture.cpp +++ b/cpp/tests/large_strings/large_strings_fixture.cpp @@ -16,12 +16,10 @@ #include "large_strings_fixture.hpp" -#include #include #include #include -#include #include #include diff --git a/cpp/tests/large_strings/parquet_tests.cpp b/cpp/tests/large_strings/parquet_tests.cpp index 007c08ce0fb..f47782a2d02 100644 --- a/cpp/tests/large_strings/parquet_tests.cpp +++ b/cpp/tests/large_strings/parquet_tests.cpp @@ -16,8 +16,6 @@ #include "large_strings_fixture.hpp" -#include -#include #include #include diff --git a/cpp/tests/lists/contains_tests.cpp b/cpp/tests/lists/contains_tests.cpp index 8fb2b403051..7ae7a6a7414 100644 --- a/cpp/tests/lists/contains_tests.cpp +++ b/cpp/tests/lists/contains_tests.cpp @@ -22,7 +22,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/lists/extract_tests.cpp b/cpp/tests/lists/extract_tests.cpp index 92dd5df5ec7..2c24f695c29 100644 --- a/cpp/tests/lists/extract_tests.cpp +++ b/cpp/tests/lists/extract_tests.cpp @@ -21,12 +21,8 @@ #include #include -#include -#include #include -#include - #include #include #include diff --git a/cpp/tests/lists/sequences_tests.cpp b/cpp/tests/lists/sequences_tests.cpp index 74545903eb3..dcb906cd2ef 100644 --- a/cpp/tests/lists/sequences_tests.cpp +++ b/cpp/tests/lists/sequences_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/lists/stream_compaction/apply_boolean_mask_tests.cpp b/cpp/tests/lists/stream_compaction/apply_boolean_mask_tests.cpp index 5625b47e7ea..18aa118bb81 100644 --- a/cpp/tests/lists/stream_compaction/apply_boolean_mask_tests.cpp +++ b/cpp/tests/lists/stream_compaction/apply_boolean_mask_tests.cpp @@ -20,8 +20,6 @@ #include #include -#include -#include #include namespace cudf::test { diff --git a/cpp/tests/merge/merge_dictionary_test.cpp b/cpp/tests/merge/merge_dictionary_test.cpp index dd528c19e4e..1d7a31fd797 100644 --- a/cpp/tests/merge/merge_dictionary_test.cpp +++ b/cpp/tests/merge/merge_dictionary_test.cpp @@ -17,9 +17,7 @@ #include #include #include -#include -#include #include #include #include diff --git a/cpp/tests/merge/merge_string_test.cpp b/cpp/tests/merge/merge_string_test.cpp index bea044496b3..d9fdb6099f0 100644 --- a/cpp/tests/merge/merge_string_test.cpp +++ b/cpp/tests/merge/merge_string_test.cpp @@ -17,10 +17,8 @@ #include #include #include -#include #include -#include #include #include #include @@ -30,10 +28,6 @@ #include -#include -#include -#include -#include #include #include diff --git a/cpp/tests/merge/merge_test.cpp b/cpp/tests/merge/merge_test.cpp index 6208d395f0a..fad390105d7 100644 --- a/cpp/tests/merge/merge_test.cpp +++ b/cpp/tests/merge/merge_test.cpp @@ -21,7 +21,6 @@ #include #include #include -#include #include #include @@ -34,7 +33,6 @@ #include #include -#include #include diff --git a/cpp/tests/partitioning/round_robin_test.cpp b/cpp/tests/partitioning/round_robin_test.cpp index 89d23c39dca..3693cfbcc72 100644 --- a/cpp/tests/partitioning/round_robin_test.cpp +++ b/cpp/tests/partitioning/round_robin_test.cpp @@ -17,10 +17,8 @@ #include #include #include -#include #include -#include #include #include #include @@ -30,12 +28,7 @@ #include -#include -#include -#include -#include #include -#include #include using cudf::test::fixed_width_column_wrapper; diff --git a/cpp/tests/quantiles/quantile_test.cpp b/cpp/tests/quantiles/quantile_test.cpp index 6e88365b6e8..23b58618fe1 100644 --- a/cpp/tests/quantiles/quantile_test.cpp +++ b/cpp/tests/quantiles/quantile_test.cpp @@ -22,7 +22,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/quantiles/quantiles_test.cpp b/cpp/tests/quantiles/quantiles_test.cpp index 44d4ec61852..c7e11af8c85 100644 --- a/cpp/tests/quantiles/quantiles_test.cpp +++ b/cpp/tests/quantiles/quantiles_test.cpp @@ -16,7 +16,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/reductions/ewm_tests.cpp b/cpp/tests/reductions/ewm_tests.cpp index 09cec688509..1117b0d1acf 100644 --- a/cpp/tests/reductions/ewm_tests.cpp +++ b/cpp/tests/reductions/ewm_tests.cpp @@ -18,9 +18,7 @@ #include #include -#include -#include #include template diff --git a/cpp/tests/reductions/list_rank_test.cpp b/cpp/tests/reductions/list_rank_test.cpp index f5470f7d881..736b5081d8f 100644 --- a/cpp/tests/reductions/list_rank_test.cpp +++ b/cpp/tests/reductions/list_rank_test.cpp @@ -14,14 +14,9 @@ * limitations under the License. */ -#include - #include #include -#include -#include -#include #include struct ListRankScanTest : public cudf::test::BaseFixture { diff --git a/cpp/tests/reductions/rank_tests.cpp b/cpp/tests/reductions/rank_tests.cpp index 3ab1fc01eaa..19633211192 100644 --- a/cpp/tests/reductions/rank_tests.cpp +++ b/cpp/tests/reductions/rank_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -21,7 +21,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/reductions/reduction_tests.cpp b/cpp/tests/reductions/reduction_tests.cpp index bdb98372836..c09cde8f9e4 100644 --- a/cpp/tests/reductions/reduction_tests.cpp +++ b/cpp/tests/reductions/reduction_tests.cpp @@ -22,9 +22,7 @@ #include #include -#include #include -#include #include #include #include @@ -33,11 +31,9 @@ #include #include -#include #include #include -#include #include #include diff --git a/cpp/tests/reductions/scan_tests.cpp b/cpp/tests/reductions/scan_tests.cpp index c4463d68a68..72d92c5ac53 100644 --- a/cpp/tests/reductions/scan_tests.cpp +++ b/cpp/tests/reductions/scan_tests.cpp @@ -20,13 +20,11 @@ #include #include -#include #include #include #include #include -#include #include #include diff --git a/cpp/tests/reductions/scan_tests.hpp b/cpp/tests/reductions/scan_tests.hpp index 858697d8ef5..c2cce4bbbfa 100644 --- a/cpp/tests/reductions/scan_tests.hpp +++ b/cpp/tests/reductions/scan_tests.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2022, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -20,9 +20,7 @@ #include #include -#include #include -#include #include #include @@ -30,7 +28,6 @@ #include #include -#include template struct TypeParam_to_host_type { diff --git a/cpp/tests/replace/clamp_test.cpp b/cpp/tests/replace/clamp_test.cpp index 239c9ce6ddd..e972ea35ed0 100644 --- a/cpp/tests/replace/clamp_test.cpp +++ b/cpp/tests/replace/clamp_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/replace/normalize_replace_tests.cpp b/cpp/tests/replace/normalize_replace_tests.cpp index 2de17388ee8..c35f385329a 100644 --- a/cpp/tests/replace/normalize_replace_tests.cpp +++ b/cpp/tests/replace/normalize_replace_tests.cpp @@ -19,7 +19,6 @@ #include #include -#include #include // This is the main test fixture diff --git a/cpp/tests/replace/replace_nans_tests.cpp b/cpp/tests/replace/replace_nans_tests.cpp index 35232204db7..1b9fe92066a 100644 --- a/cpp/tests/replace/replace_nans_tests.cpp +++ b/cpp/tests/replace/replace_nans_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/replace/replace_nulls_tests.cpp b/cpp/tests/replace/replace_nulls_tests.cpp index fcee27305f2..0c8ccea52a6 100644 --- a/cpp/tests/replace/replace_nulls_tests.cpp +++ b/cpp/tests/replace/replace_nulls_tests.cpp @@ -20,13 +20,11 @@ #include #include #include -#include #include #include #include #include -#include #include #include #include diff --git a/cpp/tests/replace/replace_tests.cpp b/cpp/tests/replace/replace_tests.cpp index b12bf08520f..ae4041bcfaf 100644 --- a/cpp/tests/replace/replace_tests.cpp +++ b/cpp/tests/replace/replace_tests.cpp @@ -20,20 +20,16 @@ #include #include #include -#include #include #include #include #include -#include -#include #include #include #include #include -#include #include diff --git a/cpp/tests/reshape/byte_cast_tests.cpp b/cpp/tests/reshape/byte_cast_tests.cpp index b3d9b2e2f5f..59585c0e947 100644 --- a/cpp/tests/reshape/byte_cast_tests.cpp +++ b/cpp/tests/reshape/byte_cast_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/reshape/tile_tests.cpp b/cpp/tests/reshape/tile_tests.cpp index ed76b9d2ea5..25cfc5c5108 100644 --- a/cpp/tests/reshape/tile_tests.cpp +++ b/cpp/tests/reshape/tile_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -15,7 +15,6 @@ */ #include -#include #include #include #include diff --git a/cpp/tests/rolling/collect_ops_test.cpp b/cpp/tests/rolling/collect_ops_test.cpp index 165e0347785..e8a36d9ab48 100644 --- a/cpp/tests/rolling/collect_ops_test.cpp +++ b/cpp/tests/rolling/collect_ops_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/rolling/empty_input_test.cpp b/cpp/tests/rolling/empty_input_test.cpp index e7d1e3f0b10..2e1815671a9 100644 --- a/cpp/tests/rolling/empty_input_test.cpp +++ b/cpp/tests/rolling/empty_input_test.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -15,9 +15,7 @@ */ #include -#include #include -#include #include #include diff --git a/cpp/tests/rolling/grouped_rolling_range_test.cpp b/cpp/tests/rolling/grouped_rolling_range_test.cpp index fcfbd0eee78..2cb9b60000b 100644 --- a/cpp/tests/rolling/grouped_rolling_range_test.cpp +++ b/cpp/tests/rolling/grouped_rolling_range_test.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2022-2023, NVIDIA CORPORATION. + * Copyright (c) 2022-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,21 +17,16 @@ #include #include #include -#include #include #include #include #include -#include -#include #include #include #include #include -#include -#include #include #include diff --git a/cpp/tests/rolling/grouped_rolling_test.cpp b/cpp/tests/rolling/grouped_rolling_test.cpp index 78d5daf7e83..78b444bcd93 100644 --- a/cpp/tests/rolling/grouped_rolling_test.cpp +++ b/cpp/tests/rolling/grouped_rolling_test.cpp @@ -19,7 +19,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/rolling/lead_lag_test.cpp b/cpp/tests/rolling/lead_lag_test.cpp index de057e96320..6519b0ed4ee 100644 --- a/cpp/tests/rolling/lead_lag_test.cpp +++ b/cpp/tests/rolling/lead_lag_test.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023, NVIDIA CORPORATION. + * Copyright (c) 2020-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -16,7 +16,6 @@ #include #include #include -#include #include #include @@ -26,7 +25,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/rolling/nth_element_test.cpp b/cpp/tests/rolling/nth_element_test.cpp index 2444992e68f..5f2b383ed55 100644 --- a/cpp/tests/rolling/nth_element_test.cpp +++ b/cpp/tests/rolling/nth_element_test.cpp @@ -17,22 +17,15 @@ #include #include #include -#include #include #include #include -#include -#include #include -#include - #include #include -#include - #include #include diff --git a/cpp/tests/rolling/offset_row_window_test.cpp b/cpp/tests/rolling/offset_row_window_test.cpp index 0eaab0c9f7a..dcaa47e722b 100644 --- a/cpp/tests/rolling/offset_row_window_test.cpp +++ b/cpp/tests/rolling/offset_row_window_test.cpp @@ -17,14 +17,10 @@ #include #include #include -#include #include #include -#include -#include #include -#include template using fwcw = cudf::test::fixed_width_column_wrapper; diff --git a/cpp/tests/rolling/range_rolling_window_test.cpp b/cpp/tests/rolling/range_rolling_window_test.cpp index 461c41025e9..daf5fcc1d96 100644 --- a/cpp/tests/rolling/range_rolling_window_test.cpp +++ b/cpp/tests/rolling/range_rolling_window_test.cpp @@ -17,22 +17,17 @@ #include #include #include -#include #include #include -#include #include #include -#include -#include #include #include #include #include -#include #include #include diff --git a/cpp/tests/rolling/range_window_bounds_test.cpp b/cpp/tests/rolling/range_window_bounds_test.cpp index b77451bf0bc..a67555280f4 100644 --- a/cpp/tests/rolling/range_window_bounds_test.cpp +++ b/cpp/tests/rolling/range_window_bounds_test.cpp @@ -15,9 +15,6 @@ */ #include -#include -#include -#include #include #include @@ -25,8 +22,6 @@ #include -#include - struct RangeWindowBoundsTest : public cudf::test::BaseFixture {}; template diff --git a/cpp/tests/rolling/rolling_test.cpp b/cpp/tests/rolling/rolling_test.cpp index 6e0dc16dca9..72a511fd5f1 100644 --- a/cpp/tests/rolling/rolling_test.cpp +++ b/cpp/tests/rolling/rolling_test.cpp @@ -19,7 +19,6 @@ #include #include #include -#include #include #include #include @@ -30,7 +29,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/scalar/factories_test.cpp b/cpp/tests/scalar/factories_test.cpp index 5f132f3ace9..26987ea1b7b 100644 --- a/cpp/tests/scalar/factories_test.cpp +++ b/cpp/tests/scalar/factories_test.cpp @@ -22,11 +22,8 @@ #include #include -#include #include -#include - class ScalarFactoryTest : public cudf::test::BaseFixture {}; template diff --git a/cpp/tests/search/search_dictionary_test.cpp b/cpp/tests/search/search_dictionary_test.cpp index 78f79ccc648..a3bb1dfda10 100644 --- a/cpp/tests/search/search_dictionary_test.cpp +++ b/cpp/tests/search/search_dictionary_test.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include diff --git a/cpp/tests/search/search_list_test.cpp b/cpp/tests/search/search_list_test.cpp index 7584003e800..fb5d0fcc889 100644 --- a/cpp/tests/search/search_list_test.cpp +++ b/cpp/tests/search/search_list_test.cpp @@ -20,7 +20,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/search/search_struct_test.cpp b/cpp/tests/search/search_struct_test.cpp index c35d359e75c..05b9deb3463 100644 --- a/cpp/tests/search/search_struct_test.cpp +++ b/cpp/tests/search/search_struct_test.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023, NVIDIA CORPORATION. + * Copyright (c) 2021-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -20,7 +20,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/search/search_test.cpp b/cpp/tests/search/search_test.cpp index 7550cc27161..8d750be5677 100644 --- a/cpp/tests/search/search_test.cpp +++ b/cpp/tests/search/search_test.cpp @@ -20,7 +20,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/sort/is_sorted_tests.cpp b/cpp/tests/sort/is_sorted_tests.cpp index 109095192f9..e3c9f8d349e 100644 --- a/cpp/tests/sort/is_sorted_tests.cpp +++ b/cpp/tests/sort/is_sorted_tests.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include #include diff --git a/cpp/tests/sort/rank_test.cpp b/cpp/tests/sort/rank_test.cpp index e08a2105aea..ded46cb1f31 100644 --- a/cpp/tests/sort/rank_test.cpp +++ b/cpp/tests/sort/rank_test.cpp @@ -18,10 +18,8 @@ #include #include #include -#include #include -#include #include #include #include diff --git a/cpp/tests/sort/sort_nested_types_tests.cpp b/cpp/tests/sort/sort_nested_types_tests.cpp index 8ab23936ceb..ce4148a941e 100644 --- a/cpp/tests/sort/sort_nested_types_tests.cpp +++ b/cpp/tests/sort/sort_nested_types_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2023, NVIDIA CORPORATION. + * Copyright (c) 2023-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/sort/sort_test.cpp b/cpp/tests/sort/sort_test.cpp index 6a35e977b46..e1505c7a474 100644 --- a/cpp/tests/sort/sort_test.cpp +++ b/cpp/tests/sort/sort_test.cpp @@ -28,7 +28,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/sort/stable_sort_tests.cpp b/cpp/tests/sort/stable_sort_tests.cpp index 655166e0d62..88de9d51523 100644 --- a/cpp/tests/sort/stable_sort_tests.cpp +++ b/cpp/tests/sort/stable_sort_tests.cpp @@ -25,9 +25,6 @@ #include #include -#include -#include - #include #include diff --git a/cpp/tests/stream_compaction/apply_boolean_mask_tests.cpp b/cpp/tests/stream_compaction/apply_boolean_mask_tests.cpp index 6c0582fb846..1204b019739 100644 --- a/cpp/tests/stream_compaction/apply_boolean_mask_tests.cpp +++ b/cpp/tests/stream_compaction/apply_boolean_mask_tests.cpp @@ -20,9 +20,7 @@ #include #include #include -#include -#include #include #include #include @@ -31,8 +29,6 @@ #include #include -#include -#include #include struct ApplyBooleanMask : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/stream_compaction/distinct_count_tests.cpp b/cpp/tests/stream_compaction/distinct_count_tests.cpp index a2dab649961..ee1bb3ead92 100644 --- a/cpp/tests/stream_compaction/distinct_count_tests.cpp +++ b/cpp/tests/stream_compaction/distinct_count_tests.cpp @@ -15,16 +15,11 @@ */ #include -#include #include #include -#include #include -#include -#include #include -#include #include #include diff --git a/cpp/tests/stream_compaction/distinct_tests.cpp b/cpp/tests/stream_compaction/distinct_tests.cpp index 14d7d8789ac..c618ff68cbb 100644 --- a/cpp/tests/stream_compaction/distinct_tests.cpp +++ b/cpp/tests/stream_compaction/distinct_tests.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include #include @@ -27,8 +26,6 @@ #include #include -#include - auto constexpr null{0}; // null at current level auto constexpr XXX{0}; // null pushed down from parent level auto constexpr NaN = std::numeric_limits::quiet_NaN(); diff --git a/cpp/tests/stream_compaction/drop_nans_tests.cpp b/cpp/tests/stream_compaction/drop_nans_tests.cpp index bf72da5c840..71321361564 100644 --- a/cpp/tests/stream_compaction/drop_nans_tests.cpp +++ b/cpp/tests/stream_compaction/drop_nans_tests.cpp @@ -15,12 +15,9 @@ */ #include -#include #include #include -#include -#include #include #include #include diff --git a/cpp/tests/stream_compaction/drop_nulls_tests.cpp b/cpp/tests/stream_compaction/drop_nulls_tests.cpp index dbac1d58195..d3b45c2323e 100644 --- a/cpp/tests/stream_compaction/drop_nulls_tests.cpp +++ b/cpp/tests/stream_compaction/drop_nulls_tests.cpp @@ -15,12 +15,10 @@ */ #include -#include #include #include #include -#include #include #include #include diff --git a/cpp/tests/stream_compaction/stable_distinct_tests.cpp b/cpp/tests/stream_compaction/stable_distinct_tests.cpp index 6c6c53331d4..cc847da6340 100644 --- a/cpp/tests/stream_compaction/stable_distinct_tests.cpp +++ b/cpp/tests/stream_compaction/stable_distinct_tests.cpp @@ -15,20 +15,16 @@ */ #include -#include #include #include #include #include -#include #include #include #include #include -#include - auto constexpr null{0}; // null at current level auto constexpr XXX{0}; // null pushed down from parent level auto constexpr NaN = std::numeric_limits::quiet_NaN(); diff --git a/cpp/tests/stream_compaction/unique_count_tests.cpp b/cpp/tests/stream_compaction/unique_count_tests.cpp index 640d159fc4f..bad93e92712 100644 --- a/cpp/tests/stream_compaction/unique_count_tests.cpp +++ b/cpp/tests/stream_compaction/unique_count_tests.cpp @@ -15,16 +15,11 @@ */ #include -#include #include #include -#include #include -#include -#include #include -#include #include #include diff --git a/cpp/tests/stream_compaction/unique_tests.cpp b/cpp/tests/stream_compaction/unique_tests.cpp index d5b6915b520..e2b32b898b3 100644 --- a/cpp/tests/stream_compaction/unique_tests.cpp +++ b/cpp/tests/stream_compaction/unique_tests.cpp @@ -15,22 +15,16 @@ */ #include -#include #include #include #include -#include #include -#include #include #include #include #include -#include -#include - using cudf::nan_policy; using cudf::null_equality; using cudf::null_policy; diff --git a/cpp/tests/streams/binaryop_test.cpp b/cpp/tests/streams/binaryop_test.cpp index 2a7b52b1b6b..3dcc6f9e632 100644 --- a/cpp/tests/streams/binaryop_test.cpp +++ b/cpp/tests/streams/binaryop_test.cpp @@ -21,7 +21,6 @@ #include #include -#include #include class BinaryopTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/streams/io/csv_test.cpp b/cpp/tests/streams/io/csv_test.cpp index 42894a0ebcb..a74ee64f8de 100644 --- a/cpp/tests/streams/io/csv_test.cpp +++ b/cpp/tests/streams/io/csv_test.cpp @@ -17,13 +17,9 @@ #include #include #include -#include #include -#include -#include #include -#include #include #include diff --git a/cpp/tests/streams/io/json_test.cpp b/cpp/tests/streams/io/json_test.cpp index f98e685ed0c..d352c6c3b2a 100644 --- a/cpp/tests/streams/io/json_test.cpp +++ b/cpp/tests/streams/io/json_test.cpp @@ -19,9 +19,7 @@ #include #include -#include #include -#include #include #include diff --git a/cpp/tests/streams/io/multibyte_split_test.cpp b/cpp/tests/streams/io/multibyte_split_test.cpp index b0eff1d3340..5bb17226029 100644 --- a/cpp/tests/streams/io/multibyte_split_test.cpp +++ b/cpp/tests/streams/io/multibyte_split_test.cpp @@ -17,7 +17,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/streams/io/orc_test.cpp b/cpp/tests/streams/io/orc_test.cpp index cc43bf15b5d..10722557e6a 100644 --- a/cpp/tests/streams/io/orc_test.cpp +++ b/cpp/tests/streams/io/orc_test.cpp @@ -17,19 +17,11 @@ #include #include #include -#include -#include #include #include -#include #include -#include -#include -#include -#include -#include #include #include diff --git a/cpp/tests/streams/io/parquet_test.cpp b/cpp/tests/streams/io/parquet_test.cpp index 9d2dec2d697..18bb80e64af 100644 --- a/cpp/tests/streams/io/parquet_test.cpp +++ b/cpp/tests/streams/io/parquet_test.cpp @@ -17,13 +17,9 @@ #include #include #include -#include -#include #include #include -#include -#include #include #include diff --git a/cpp/tests/streams/join_test.cpp b/cpp/tests/streams/join_test.cpp index 2811bb676fa..27bd7e080c9 100644 --- a/cpp/tests/streams/join_test.cpp +++ b/cpp/tests/streams/join_test.cpp @@ -19,11 +19,9 @@ #include #include -#include #include #include #include -#include #include #include diff --git a/cpp/tests/streams/null_mask_test.cpp b/cpp/tests/streams/null_mask_test.cpp index e96224003f4..ed37a72545f 100644 --- a/cpp/tests/streams/null_mask_test.cpp +++ b/cpp/tests/streams/null_mask_test.cpp @@ -14,15 +14,12 @@ * limitations under the License. */ -#include - #include #include #include #include #include -#include class NullMaskTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/streams/reduction_test.cpp b/cpp/tests/streams/reduction_test.cpp index b4f013fc960..9ab972302e4 100644 --- a/cpp/tests/streams/reduction_test.cpp +++ b/cpp/tests/streams/reduction_test.cpp @@ -17,11 +17,8 @@ #include #include #include -#include -#include #include -#include #include #include diff --git a/cpp/tests/streams/rolling_test.cpp b/cpp/tests/streams/rolling_test.cpp index b352ad2c0d2..4d9899870b4 100644 --- a/cpp/tests/streams/rolling_test.cpp +++ b/cpp/tests/streams/rolling_test.cpp @@ -17,12 +17,10 @@ #include #include #include -#include #include #include #include -#include class RollingTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/streams/stream_compaction_test.cpp b/cpp/tests/streams/stream_compaction_test.cpp index 07b2d77cc04..e7b282601e1 100644 --- a/cpp/tests/streams/stream_compaction_test.cpp +++ b/cpp/tests/streams/stream_compaction_test.cpp @@ -15,20 +15,16 @@ */ #include -#include #include #include #include -#include #include #include #include #include #include -#include - auto constexpr NaN = std::numeric_limits::quiet_NaN(); auto constexpr KEEP_ANY = cudf::duplicate_keep_option::KEEP_ANY; auto constexpr KEEP_FIRST = cudf::duplicate_keep_option::KEEP_FIRST; diff --git a/cpp/tests/streams/strings/factory_test.cpp b/cpp/tests/streams/strings/factory_test.cpp index 36e595ab9fa..449e0830b0c 100644 --- a/cpp/tests/streams/strings/factory_test.cpp +++ b/cpp/tests/streams/strings/factory_test.cpp @@ -15,7 +15,6 @@ */ #include -#include #include #include diff --git a/cpp/tests/streams/strings/reverse_test.cpp b/cpp/tests/streams/strings/reverse_test.cpp index 4b4d0a7aff5..154e1c1b715 100644 --- a/cpp/tests/streams/strings/reverse_test.cpp +++ b/cpp/tests/streams/strings/reverse_test.cpp @@ -21,7 +21,6 @@ #include #include -#include class StringsReverseTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/streams/transform_test.cpp b/cpp/tests/streams/transform_test.cpp index cf81dc6fb42..9f168abcb31 100644 --- a/cpp/tests/streams/transform_test.cpp +++ b/cpp/tests/streams/transform_test.cpp @@ -15,17 +15,11 @@ */ #include -#include #include #include -#include -#include #include -#include -#include #include -#include #include #include diff --git a/cpp/tests/strings/array_tests.cpp b/cpp/tests/strings/array_tests.cpp index 9c0ecaa52c0..06b9c2fa3c1 100644 --- a/cpp/tests/strings/array_tests.cpp +++ b/cpp/tests/strings/array_tests.cpp @@ -23,10 +23,8 @@ #include #include #include -#include #include #include -#include #include #include diff --git a/cpp/tests/strings/combine/concatenate_tests.cpp b/cpp/tests/strings/combine/concatenate_tests.cpp index bb57d6f5e8a..e53adcf373a 100644 --- a/cpp/tests/strings/combine/concatenate_tests.cpp +++ b/cpp/tests/strings/combine/concatenate_tests.cpp @@ -22,7 +22,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/strings/combine/join_list_elements_tests.cpp b/cpp/tests/strings/combine/join_list_elements_tests.cpp index 00317146088..c92f1cfc8f8 100644 --- a/cpp/tests/strings/combine/join_list_elements_tests.cpp +++ b/cpp/tests/strings/combine/join_list_elements_tests.cpp @@ -22,7 +22,6 @@ #include #include #include -#include using namespace cudf::test::iterators; diff --git a/cpp/tests/strings/concatenate_tests.cpp b/cpp/tests/strings/concatenate_tests.cpp index 5cf4015b9e9..51dcc60d95e 100644 --- a/cpp/tests/strings/concatenate_tests.cpp +++ b/cpp/tests/strings/concatenate_tests.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2023, NVIDIA CORPORATION. + * Copyright (c) 2019-2024, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -20,7 +20,6 @@ #include #include -#include #include diff --git a/cpp/tests/strings/datetime_tests.cpp b/cpp/tests/strings/datetime_tests.cpp index b3dc3010c67..da0db0fc056 100644 --- a/cpp/tests/strings/datetime_tests.cpp +++ b/cpp/tests/strings/datetime_tests.cpp @@ -19,7 +19,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/strings/extract_tests.cpp b/cpp/tests/strings/extract_tests.cpp index 7e0338f1bf4..37b25d9b287 100644 --- a/cpp/tests/strings/extract_tests.cpp +++ b/cpp/tests/strings/extract_tests.cpp @@ -21,7 +21,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/strings/findall_tests.cpp b/cpp/tests/strings/findall_tests.cpp index 4821a7fa999..7eb4b32d078 100644 --- a/cpp/tests/strings/findall_tests.cpp +++ b/cpp/tests/strings/findall_tests.cpp @@ -20,7 +20,6 @@ #include #include #include -#include #include #include @@ -28,8 +27,6 @@ #include -#include - struct StringsFindallTests : public cudf::test::BaseFixture {}; TEST_F(StringsFindallTests, FindallTest) diff --git a/cpp/tests/strings/fixed_point_tests.cpp b/cpp/tests/strings/fixed_point_tests.cpp index 79054551498..b788c05c152 100644 --- a/cpp/tests/strings/fixed_point_tests.cpp +++ b/cpp/tests/strings/fixed_point_tests.cpp @@ -23,8 +23,6 @@ #include #include -#include - struct StringsConvertTest : public cudf::test::BaseFixture {}; template diff --git a/cpp/tests/strings/integers_tests.cpp b/cpp/tests/strings/integers_tests.cpp index 26bcfe8028d..c08effdb969 100644 --- a/cpp/tests/strings/integers_tests.cpp +++ b/cpp/tests/strings/integers_tests.cpp @@ -24,9 +24,6 @@ #include #include -#include -#include - #include #include diff --git a/cpp/tests/structs/structs_column_tests.cpp b/cpp/tests/structs/structs_column_tests.cpp index 219bd6d8b01..a34ff25cb69 100644 --- a/cpp/tests/structs/structs_column_tests.cpp +++ b/cpp/tests/structs/structs_column_tests.cpp @@ -17,28 +17,18 @@ #include #include #include -#include #include #include #include -#include #include -#include #include -#include -#include -#include -#include #include #include #include -#include #include -#include -#include #include #include diff --git a/cpp/tests/structs/utilities_tests.cpp b/cpp/tests/structs/utilities_tests.cpp index c33eedf9bd9..c0df2f01a63 100644 --- a/cpp/tests/structs/utilities_tests.cpp +++ b/cpp/tests/structs/utilities_tests.cpp @@ -14,21 +14,15 @@ * limitations under the License. */ -#include "cudf_test/default_stream.hpp" - #include #include #include -#include #include #include #include -#include -#include #include #include -#include #include #include diff --git a/cpp/tests/table/row_operators_tests.cpp b/cpp/tests/table/row_operators_tests.cpp index 5fa63c47cf0..216c4d7b6bb 100644 --- a/cpp/tests/table/row_operators_tests.cpp +++ b/cpp/tests/table/row_operators_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/table/table_tests.cpp b/cpp/tests/table/table_tests.cpp index 1637ba7d7d3..363f1a0ba5d 100644 --- a/cpp/tests/table/table_tests.cpp +++ b/cpp/tests/table/table_tests.cpp @@ -17,17 +17,14 @@ #include #include #include -#include #include #include #include -#include #include #include #include -#include template using column_wrapper = cudf::test::fixed_width_column_wrapper; diff --git a/cpp/tests/text/minhash_tests.cpp b/cpp/tests/text/minhash_tests.cpp index e23f3f6e7d8..ef35a4472cf 100644 --- a/cpp/tests/text/minhash_tests.cpp +++ b/cpp/tests/text/minhash_tests.cpp @@ -21,13 +21,9 @@ #include #include -#include #include -#include -#include - #include struct MinHashTest : public cudf::test::BaseFixture {}; diff --git a/cpp/tests/text/ngrams_tests.cpp b/cpp/tests/text/ngrams_tests.cpp index 1acb4fc4265..c72c7cfc80e 100644 --- a/cpp/tests/text/ngrams_tests.cpp +++ b/cpp/tests/text/ngrams_tests.cpp @@ -28,8 +28,6 @@ #include -#include - struct TextGenerateNgramsTest : public cudf::test::BaseFixture {}; TEST_F(TextGenerateNgramsTest, Ngrams) diff --git a/cpp/tests/text/normalize_tests.cpp b/cpp/tests/text/normalize_tests.cpp index b0d41004e7e..2515cc917fa 100644 --- a/cpp/tests/text/normalize_tests.cpp +++ b/cpp/tests/text/normalize_tests.cpp @@ -20,7 +20,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/text/stemmer_tests.cpp b/cpp/tests/text/stemmer_tests.cpp index a343913411c..82c4bf53cfc 100644 --- a/cpp/tests/text/stemmer_tests.cpp +++ b/cpp/tests/text/stemmer_tests.cpp @@ -19,7 +19,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/text/subword_tests.cpp b/cpp/tests/text/subword_tests.cpp index a615780c02a..782551ad66e 100644 --- a/cpp/tests/text/subword_tests.cpp +++ b/cpp/tests/text/subword_tests.cpp @@ -19,13 +19,11 @@ #include #include -#include #include #include #include -#include #include // Global environment for temporary files diff --git a/cpp/tests/transform/bools_to_mask_test.cpp b/cpp/tests/transform/bools_to_mask_test.cpp index 2684123c08a..9437440f34d 100644 --- a/cpp/tests/transform/bools_to_mask_test.cpp +++ b/cpp/tests/transform/bools_to_mask_test.cpp @@ -20,10 +20,8 @@ #include #include -#include #include #include -#include #include diff --git a/cpp/tests/transform/nans_to_null_test.cpp b/cpp/tests/transform/nans_to_null_test.cpp index ba16c100e7a..42ca872a936 100644 --- a/cpp/tests/transform/nans_to_null_test.cpp +++ b/cpp/tests/transform/nans_to_null_test.cpp @@ -17,12 +17,10 @@ #include #include #include -#include #include #include #include -#include template struct NaNsToNullTest : public cudf::test::BaseFixture { diff --git a/cpp/tests/transpose/transpose_test.cpp b/cpp/tests/transpose/transpose_test.cpp index 5a88c402b8c..7797b2b2cf8 100644 --- a/cpp/tests/transpose/transpose_test.cpp +++ b/cpp/tests/transpose/transpose_test.cpp @@ -22,7 +22,6 @@ #include #include -#include #include #include diff --git a/cpp/tests/types/traits_test.cpp b/cpp/tests/types/traits_test.cpp index 0d9092c33da..46468af515d 100644 --- a/cpp/tests/types/traits_test.cpp +++ b/cpp/tests/types/traits_test.cpp @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include #include diff --git a/cpp/tests/unary/cast_tests.cpp b/cpp/tests/unary/cast_tests.cpp index 45b89b76070..ed4c1340dbb 100644 --- a/cpp/tests/unary/cast_tests.cpp +++ b/cpp/tests/unary/cast_tests.cpp @@ -20,18 +20,15 @@ #include #include -#include #include #include #include #include #include -#include #include #include -#include #include static auto const test_timestamps_D = std::vector{ diff --git a/cpp/tests/unary/math_ops_test.cpp b/cpp/tests/unary/math_ops_test.cpp index 5bfbf70d5f9..663a919f3f4 100644 --- a/cpp/tests/unary/math_ops_test.cpp +++ b/cpp/tests/unary/math_ops_test.cpp @@ -22,10 +22,6 @@ #include #include #include -#include -#include - -#include #include diff --git a/cpp/tests/unary/unary_ops_test.cpp b/cpp/tests/unary/unary_ops_test.cpp index e7477c34642..3c616461c74 100644 --- a/cpp/tests/unary/unary_ops_test.cpp +++ b/cpp/tests/unary/unary_ops_test.cpp @@ -23,7 +23,6 @@ #include #include -#include #include template diff --git a/cpp/tests/utilities/random_seed.cpp b/cpp/tests/utilities/random_seed.cpp index ab5a31ce161..555d89b7dc5 100644 --- a/cpp/tests/utilities/random_seed.cpp +++ b/cpp/tests/utilities/random_seed.cpp @@ -13,8 +13,9 @@ * See the License for the specific language governing permissions and * limitations under the License. */ +#include -#include +#include namespace cudf { namespace test { diff --git a/cpp/tests/utilities_tests/column_debug_tests.cpp b/cpp/tests/utilities_tests/column_debug_tests.cpp index 7aa05af4591..2a57d678d07 100644 --- a/cpp/tests/utilities_tests/column_debug_tests.cpp +++ b/cpp/tests/utilities_tests/column_debug_tests.cpp @@ -16,12 +16,9 @@ #include #include -#include #include #include -#include - #include #include diff --git a/cpp/tests/utilities_tests/column_utilities_tests.cpp b/cpp/tests/utilities_tests/column_utilities_tests.cpp index 9d6d5ccb9b5..a13ce825d0b 100644 --- a/cpp/tests/utilities_tests/column_utilities_tests.cpp +++ b/cpp/tests/utilities_tests/column_utilities_tests.cpp @@ -17,20 +17,16 @@ #include #include #include -#include #include #include #include #include #include -#include #include #include -#include - template struct ColumnUtilitiesTest : public cudf::test::BaseFixture { cudf::test::UniformRandomGenerator random; diff --git a/cpp/tests/utilities_tests/column_wrapper_tests.cpp b/cpp/tests/utilities_tests/column_wrapper_tests.cpp index 479c6687e75..339678f3be8 100644 --- a/cpp/tests/utilities_tests/column_wrapper_tests.cpp +++ b/cpp/tests/utilities_tests/column_wrapper_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/utilities_tests/lists_column_wrapper_tests.cpp b/cpp/tests/utilities_tests/lists_column_wrapper_tests.cpp index 5e3fda5e6f7..ff50dc39979 100644 --- a/cpp/tests/utilities_tests/lists_column_wrapper_tests.cpp +++ b/cpp/tests/utilities_tests/lists_column_wrapper_tests.cpp @@ -17,7 +17,6 @@ #include #include #include -#include #include #include diff --git a/cpp/tests/utilities_tests/type_check_tests.cpp b/cpp/tests/utilities_tests/type_check_tests.cpp index fecb896f95a..c1c5776be74 100644 --- a/cpp/tests/utilities_tests/type_check_tests.cpp +++ b/cpp/tests/utilities_tests/type_check_tests.cpp @@ -18,7 +18,6 @@ #include #include -#include #include #include #include diff --git a/cpp/tests/utilities_tests/type_list_tests.cpp b/cpp/tests/utilities_tests/type_list_tests.cpp index 849457056e4..6c3a84763a0 100644 --- a/cpp/tests/utilities_tests/type_list_tests.cpp +++ b/cpp/tests/utilities_tests/type_list_tests.cpp @@ -14,7 +14,6 @@ * limitations under the License. */ -#include #include using namespace cudf::test; // this will make reading code way easier @@ -23,6 +22,7 @@ namespace { // Work around to remove parentheses surrounding a type template struct argument_type; + template struct argument_type { using type = U; diff --git a/dependencies.yaml b/dependencies.yaml index ff97b67f0ce..7c7aa43fa41 100644 --- a/dependencies.yaml +++ b/dependencies.yaml @@ -232,7 +232,7 @@ files: key: cudf-pandas-tests includes: - test_python_cudf_pandas - py_rapids_build_cudf_polars: + py_build_cudf_polars: output: pyproject pyproject_dir: python/cudf_polars extras: @@ -727,7 +727,7 @@ dependencies: common: - output_types: [conda, requirements, pyproject] packages: - - polars>=1.8,<1.9 + - polars>=1.11,<1.12 run_dask_cudf: common: - output_types: [conda, requirements, pyproject] diff --git a/docs/cudf/source/conf.py b/docs/cudf/source/conf.py index ecf619ddc44..5942cc16850 100644 --- a/docs/cudf/source/conf.py +++ b/docs/cudf/source/conf.py @@ -342,10 +342,7 @@ def clean_all_xml_files(path): "cudf.Series": ("cudf.core.series.Series", "cudf.Series"), "cudf.Index": ("cudf.core.index.Index", "cudf.Index"), "cupy.core.core.ndarray": ("cupy.ndarray", "cupy.ndarray"), - # TODO: Replace the first entry in a follow-up with rmm.pylibrmm.device_buffer.DeviceBuffer - # when the RMM objects inventory is generated from branch-24.12. The RMM objects inventory - # can be accessed here : https://docs.rapids.ai/api/rmm/nightly/objects.inv - "DeviceBuffer": ("rmm.DeviceBuffer", "rmm.DeviceBuffer"), + "DeviceBuffer": ("rmm.pylibrmm.device_buffer.DeviceBuffer", "rmm.DeviceBuffer"), } diff --git a/docs/cudf/source/developer_guide/contributing_guide.md b/docs/cudf/source/developer_guide/contributing_guide.md index 6fce268f309..f4d2c7319b3 100644 --- a/docs/cudf/source/developer_guide/contributing_guide.md +++ b/docs/cudf/source/developer_guide/contributing_guide.md @@ -15,8 +15,7 @@ Developers are strongly recommended to set up `pre-commit` prior to any developm The `.pre-commit-config.yaml` file at the root of the repo is the primary source of truth linting. Specifically, cuDF uses the following tools: -- [`ruff`](https://beta.ruff.rs/) checks for general code formatting compliance. -- [`isort`](https://pycqa.github.io/isort/) ensures imports are sorted consistently. +- [`ruff`](https://docs.astral.sh/ruff/) checks for general code formatting compliance. - [`mypy`](http://mypy-lang.org/) performs static type checking. In conjunction with [type hints](https://docs.python.org/3/library/typing.html), `mypy` can help catch various bugs that are otherwise difficult to find. diff --git a/docs/cudf/source/user_guide/10min.ipynb b/docs/cudf/source/user_guide/10min.ipynb index 95f5f9734dd..46221b6015b 100644 --- a/docs/cudf/source/user_guide/10min.ipynb +++ b/docs/cudf/source/user_guide/10min.ipynb @@ -38,10 +38,10 @@ "import os\n", "\n", "import cupy as cp\n", + "import dask_cudf\n", "import pandas as pd\n", "\n", "import cudf\n", - "import dask_cudf\n", "\n", "cp.random.seed(12)\n", "\n", diff --git a/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/index.rst b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/index.rst index 3a79c869971..e0735a197fd 100644 --- a/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/index.rst +++ b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/index.rst @@ -10,3 +10,5 @@ nvtext minhash ngrams_tokenize normalize + replace + stemmer diff --git a/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/replace.rst b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/replace.rst new file mode 100644 index 00000000000..04cee972dc1 --- /dev/null +++ b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/replace.rst @@ -0,0 +1,6 @@ +======= +replace +======= + +.. automodule:: pylibcudf.nvtext.replace + :members: diff --git a/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/stemmer.rst b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/stemmer.rst new file mode 100644 index 00000000000..b407ff8451a --- /dev/null +++ b/docs/cudf/source/user_guide/api_docs/pylibcudf/nvtext/stemmer.rst @@ -0,0 +1,6 @@ +======= +stemmer +======= + +.. automodule:: pylibcudf.nvtext.stemmer + :members: diff --git a/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/index.rst b/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/index.rst index c8c0016126d..ae670b5bd8a 100644 --- a/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/index.rst +++ b/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/index.rst @@ -16,6 +16,7 @@ strings regex_flags regex_program repeat + replace_re replace side_type slice diff --git a/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/replace_re.rst b/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/replace_re.rst new file mode 100644 index 00000000000..5bf715ef657 --- /dev/null +++ b/docs/cudf/source/user_guide/api_docs/pylibcudf/strings/replace_re.rst @@ -0,0 +1,6 @@ +========== +replace_re +========== + +.. automodule:: pylibcudf.strings.replace_re + :members: diff --git a/docs/cudf/source/user_guide/guide-to-udfs.ipynb b/docs/cudf/source/user_guide/guide-to-udfs.ipynb index 75eafcc5387..abfe5a1b178 100644 --- a/docs/cudf/source/user_guide/guide-to-udfs.ipynb +++ b/docs/cudf/source/user_guide/guide-to-udfs.ipynb @@ -101,6 +101,8 @@ "outputs": [], "source": [ "# define a scalar function\n", + "\n", + "\n", "def f(x):\n", " return x + 1" ] @@ -247,6 +249,8 @@ "outputs": [], "source": [ "# redefine the same function from above\n", + "\n", + "\n", "def f(x):\n", " return x + 1" ] @@ -1622,6 +1626,8 @@ "outputs": [], "source": [ "# a user defined aggregation function.\n", + "\n", + "\n", "def udaf(df):\n", " return df[\"b\"].max() - df[\"b\"].min() / 2" ] diff --git a/java/src/main/java/ai/rapids/cudf/HostMemoryReservation.java b/java/src/main/java/ai/rapids/cudf/HostMemoryReservation.java deleted file mode 100644 index 72c2e659372..00000000000 --- a/java/src/main/java/ai/rapids/cudf/HostMemoryReservation.java +++ /dev/null @@ -1,32 +0,0 @@ -/* - * - * Copyright (c) 2023, NVIDIA CORPORATION. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - * - */ - -package ai.rapids.cudf; - -/** - * Represents some amount of host memory that has been reserved. A reservation guarantees that one - * or more allocations up to the reserved amount, minus padding for alignment will succeed. A - * reservation typically guarantees the amount can be allocated one, meaning when a buffer - * allocated from a reservation is freed it is not returned to the reservation, but to the pool of - * memory the reservation originally came from. If more memory is allocated from the reservation - * an OutOfMemoryError may be thrown, but it is not guaranteed to happen. - * - * When the reservation is closed any unused reservation will be returned to the pool of memory - * the reservation came from. - */ -public interface HostMemoryReservation extends HostMemoryAllocator, AutoCloseable {} diff --git a/java/src/main/java/ai/rapids/cudf/RegexFlag.java b/java/src/main/java/ai/rapids/cudf/RegexFlag.java index 7ed8e0354c9..68a3856f37d 100644 --- a/java/src/main/java/ai/rapids/cudf/RegexFlag.java +++ b/java/src/main/java/ai/rapids/cudf/RegexFlag.java @@ -28,7 +28,16 @@ public enum RegexFlag { DEFAULT(0), // default MULTILINE(8), // the '^' and '$' honor new-line characters DOTALL(16), // the '.' matching includes new-line characters - ASCII(256); // use only ASCII when matching built-in character classes + ASCII(256), // use only ASCII when matching built-in character classes + /** + * EXT_NEWLINE(512): Extends line delimiters to include the following Unicode characters + * - NEXT_LINE ('\u0085') + * - LINE_SEPARATOR ('\u2028') + * - PARAGRAPH_SEPARATOR ('\u2029') + * - CARRIAGE_RETURN ('\r') + * - NEW_LINE ('\n') + */ + EXT_NEWLINE(512); final int nativeId; // Native id, for use with libcudf. private RegexFlag(int nativeId) { // Only constant values should be used diff --git a/java/src/main/java/ai/rapids/cudf/Schema.java b/java/src/main/java/ai/rapids/cudf/Schema.java index 76b2799aad6..6da591d659f 100644 --- a/java/src/main/java/ai/rapids/cudf/Schema.java +++ b/java/src/main/java/ai/rapids/cudf/Schema.java @@ -29,26 +29,52 @@ public class Schema { public static final Schema INFERRED = new Schema(); private final DType topLevelType; + + /** + * Default value for precision value, when it is not specified or the column type is not decimal. + */ + private static final int UNKNOWN_PRECISION = -1; + + /** + * Store precision for the top level column, only applicable if the column is a decimal type. + *

+ * This variable is not designed to be used by any libcudf's APIs since libcudf does not support + * precisions for fixed point numbers. + * Instead, it is used only to pass down the precision values from Spark's DecimalType to the + * JNI level, where some JNI functions require these values to perform their operations. + */ + private final int topLevelPrecision; + private final List childNames; private final List childSchemas; private boolean flattened = false; private String[] flattenedNames; private DType[] flattenedTypes; + private int[] flattenedPrecisions; private int[] flattenedCounts; private Schema(DType topLevelType, + int topLevelPrecision, List childNames, List childSchemas) { this.topLevelType = topLevelType; + this.topLevelPrecision = topLevelPrecision; this.childNames = childNames; this.childSchemas = childSchemas; } + private Schema(DType topLevelType, + List childNames, + List childSchemas) { + this(topLevelType, UNKNOWN_PRECISION, childNames, childSchemas); + } + /** * Inferred schema. */ private Schema() { topLevelType = null; + topLevelPrecision = UNKNOWN_PRECISION; childNames = null; childSchemas = null; } @@ -104,14 +130,17 @@ private void flattenIfNeeded() { if (flatLen == 0) { flattenedNames = null; flattenedTypes = null; + flattenedPrecisions = null; flattenedCounts = null; } else { String[] names = new String[flatLen]; DType[] types = new DType[flatLen]; + int[] precisions = new int[flatLen]; int[] counts = new int[flatLen]; - collectFlattened(names, types, counts, 0); + collectFlattened(names, types, precisions, counts, 0); flattenedNames = names; flattenedTypes = types; + flattenedPrecisions = precisions; flattenedCounts = counts; } flattened = true; @@ -128,19 +157,20 @@ private int flattenedLength(int startingLength) { return startingLength; } - private int collectFlattened(String[] names, DType[] types, int[] counts, int offset) { + private int collectFlattened(String[] names, DType[] types, int[] precisions, int[] counts, int offset) { if (childSchemas != null) { for (int i = 0; i < childSchemas.size(); i++) { Schema child = childSchemas.get(i); names[offset] = childNames.get(i); types[offset] = child.topLevelType; + precisions[offset] = child.topLevelPrecision; if (child.childNames != null) { counts[offset] = child.childNames.size(); } else { counts[offset] = 0; } offset++; - offset = this.childSchemas.get(i).collectFlattened(names, types, counts, offset); + offset = this.childSchemas.get(i).collectFlattened(names, types, precisions, counts, offset); } } return offset; @@ -226,6 +256,22 @@ public int[] getFlattenedTypeScales() { return ret; } + /** + * Get decimal precisions of the columns' types flattened from all levels in schema by + * depth-first traversal. + *

+ * This is used to pass down the decimal precisions from Spark to only the JNI layer, where + * some JNI functions require precision values to perform their operations. + * Decimal precisions should not be consumed by any libcudf's APIs since libcudf does not + * support precisions for fixed point numbers. + * + * @return An array containing decimal precision of all columns in schema. + */ + public int[] getFlattenedDecimalPrecisions() { + flattenIfNeeded(); + return flattenedPrecisions; + } + /** * Get the types of the columns in schema flattened from all levels by depth-first traversal. * @return An array containing types of all columns in schema. @@ -307,11 +353,13 @@ public HostColumnVector.DataType asHostDataType() { public static class Builder { private final DType topLevelType; + private final int topLevelPrecision; private final List names; private final List types; - private Builder(DType topLevelType) { + private Builder(DType topLevelType, int topLevelPrecision) { this.topLevelType = topLevelType; + this.topLevelPrecision = topLevelPrecision; if (topLevelType == DType.STRUCT || topLevelType == DType.LIST) { // There can be children names = new ArrayList<>(); @@ -322,14 +370,19 @@ private Builder(DType topLevelType) { } } + private Builder(DType topLevelType) { + this(topLevelType, UNKNOWN_PRECISION); + } + /** * Add a new column * @param type the type of column to add * @param name the name of the column to add (Ignored for list types) + * @param precision the decimal precision, only applicable for decimal types * @return the builder for the new column. This should really only be used when the type * passed in is a LIST or a STRUCT. */ - public Builder addColumn(DType type, String name) { + public Builder addColumn(DType type, String name, int precision) { if (names == null) { throw new IllegalStateException("A column of type " + topLevelType + " cannot have children"); @@ -340,21 +393,31 @@ public Builder addColumn(DType type, String name) { if (names.contains(name)) { throw new IllegalStateException("Cannot add duplicate names to a schema"); } - Builder ret = new Builder(type); + Builder ret = new Builder(type, precision); types.add(ret); names.add(name); return ret; } + public Builder addColumn(DType type, String name) { + return addColumn(type, name, UNKNOWN_PRECISION); + } + /** * Adds a single column to the current schema. addColumn is preferred as it can be used * to support nested types. * @param type the type of the column. * @param name the name of the column. + * @param precision the decimal precision, only applicable for decimal types. * @return this for chaining. */ + public Builder column(DType type, String name, int precision) { + addColumn(type, name, precision); + return this; + } + public Builder column(DType type, String name) { - addColumn(type, name); + addColumn(type, name, UNKNOWN_PRECISION); return this; } diff --git a/java/src/test/java/ai/rapids/cudf/ColumnVectorTest.java b/java/src/test/java/ai/rapids/cudf/ColumnVectorTest.java index 708744569df..14c290b300a 100644 --- a/java/src/test/java/ai/rapids/cudf/ColumnVectorTest.java +++ b/java/src/test/java/ai/rapids/cudf/ColumnVectorTest.java @@ -31,6 +31,7 @@ import java.util.ArrayList; import java.util.Arrays; import java.util.Collections; +import java.util.EnumSet; import java.util.List; import java.util.Optional; import java.util.concurrent.atomic.AtomicInteger; @@ -3877,6 +3878,43 @@ void testExtractRe() { } } + @Test +void testExtractReWithMultiLineDelimiters() { + String NEXT_LINE = "\u0085"; + String LINE_SEPARATOR = "\u2028"; + String PARAGRAPH_SEPARATOR = "\u2029"; + String CARRIAGE_RETURN = "\r"; + String NEW_LINE = "\n"; + + try (ColumnVector input = ColumnVector.fromStrings( + "boo:" + NEXT_LINE + "boo::" + LINE_SEPARATOR + "boo:::", + "boo:::" + LINE_SEPARATOR + "zzé" + CARRIAGE_RETURN + "lll", + "boo::", + "", + "boo::" + NEW_LINE, + "boo::" + CARRIAGE_RETURN, + "boo:" + NEXT_LINE + "boo::" + PARAGRAPH_SEPARATOR, + "boo:" + NEW_LINE + "boo::" + LINE_SEPARATOR, + "boo:" + NEXT_LINE + "boo::" + NEXT_LINE); + Table expected_ext_newline = new Table.TestBuilder() + .column("boo:::", null, "boo::", null, "boo::", "boo::", "boo::", "boo::", "boo::") + .build(); + Table expected_default = new Table.TestBuilder() + .column("boo:::", null, "boo::", null, "boo::", null, null, null, null) + .build()) { + + // Regex pattern to match 'boo:' followed by one or more colons at the end of the string + try (Table found = input.extractRe(new RegexProgram("(boo:+)$", EnumSet.of(RegexFlag.EXT_NEWLINE)))) { + assertColumnsAreEqual(expected_ext_newline.getColumns()[0], found.getColumns()[0]); + } + + try (Table found = input.extractRe(new RegexProgram("(boo:+)$", EnumSet.of(RegexFlag.DEFAULT)))) { + assertColumnsAreEqual(expected_default.getColumns()[0], found.getColumns()[0]); + } + } + } + + @Test void testExtractAllRecord() { String pattern = "([ab])(\\d)"; diff --git a/pyproject.toml b/pyproject.toml index 661c68ee62e..6933484f4e7 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -34,6 +34,8 @@ select = [ "F", # pycodestyle Warning "W", + # isort + "I", # no-blank-line-before-function "D201", # one-blank-line-after-class diff --git a/python/cudf/benchmarks/conftest.py b/python/cudf/benchmarks/conftest.py index 7b2b71cf216..0e4afadccf5 100644 --- a/python/cudf/benchmarks/conftest.py +++ b/python/cudf/benchmarks/conftest.py @@ -56,27 +56,23 @@ # into the main repo. sys.path.insert(0, os.path.join(os.path.dirname(__file__), "common")) -from config import cudf # noqa: W0611, E402, F401 -from utils import ( # noqa: E402 - OrderedSet, - collapse_fixtures, - column_generators, - make_fixture, -) - # Turn off isort until we upgrade to 5.8.0 # https://github.com/pycqa/isort/issues/1594 -# isort: off from config import ( # noqa: W0611, E402, F401 NUM_COLS, NUM_ROWS, collect_ignore, + cudf, # noqa: W0611, E402, F401 pytest_collection_modifyitems, pytest_sessionfinish, pytest_sessionstart, ) - -# isort: on +from utils import ( # noqa: E402 + OrderedSet, + collapse_fixtures, + column_generators, + make_fixture, +) @pytest_cases.fixture(params=[0, 1], ids=["AxisIndex", "AxisColumn"]) diff --git a/python/cudf/cudf/_lib/column.pyx b/python/cudf/cudf/_lib/column.pyx index 065655505b8..94dbdf5534d 100644 --- a/python/cudf/cudf/_lib/column.pyx +++ b/python/cudf/cudf/_lib/column.pyx @@ -688,15 +688,18 @@ cdef class Column: # special case for string column is_string_column = (cv.type().id() == libcudf_types.type_id.STRING) if is_string_column: - # get the size from offset child column (device to host copy) - offsets_column_index = 0 - offset_child_column = cv.child(offsets_column_index) - if offset_child_column.size() == 0: + if cv.num_children() == 0: base_nbytes = 0 else: - chars_size = get_element( - offset_child_column, offset_child_column.size()-1).value - base_nbytes = chars_size + # get the size from offset child column (device to host copy) + offsets_column_index = 0 + offset_child_column = cv.child(offsets_column_index) + if offset_child_column.size() == 0: + base_nbytes = 0 + else: + chars_size = get_element( + offset_child_column, offset_child_column.size()-1).value + base_nbytes = chars_size if data_ptr: if data_owner is None: diff --git a/python/cudf/cudf/_lib/copying.pyx b/python/cudf/cudf/_lib/copying.pyx index 30353c4be6c..4221e745e65 100644 --- a/python/cudf/cudf/_lib/copying.pyx +++ b/python/cudf/cudf/_lib/copying.pyx @@ -4,7 +4,7 @@ import pickle from libc.stdint cimport uint8_t, uintptr_t from libcpp cimport bool -from libcpp.memory cimport make_shared, shared_ptr, unique_ptr +from libcpp.memory cimport unique_ptr from libcpp.utility cimport move from libcpp.vector cimport vector @@ -30,10 +30,6 @@ from libcpp.memory cimport make_unique cimport pylibcudf.libcudf.contiguous_split as cpp_contiguous_split from pylibcudf.libcudf.column.column cimport column from pylibcudf.libcudf.column.column_view cimport column_view -from pylibcudf.libcudf.lists.gather cimport ( - segmented_gather as cpp_segmented_gather, -) -from pylibcudf.libcudf.lists.lists_column_view cimport lists_column_view from pylibcudf.libcudf.scalar.scalar cimport scalar from pylibcudf.libcudf.types cimport size_type @@ -339,26 +335,6 @@ def get_element(Column input_column, size_type index): ) -@acquire_spill_lock() -def segmented_gather(Column source_column, Column gather_map): - cdef shared_ptr[lists_column_view] source_LCV = ( - make_shared[lists_column_view](source_column.view()) - ) - cdef shared_ptr[lists_column_view] gather_map_LCV = ( - make_shared[lists_column_view](gather_map.view()) - ) - cdef unique_ptr[column] c_result - - with nogil: - c_result = move( - cpp_segmented_gather( - source_LCV.get()[0], gather_map_LCV.get()[0]) - ) - - result = Column.from_unique_ptr(move(c_result)) - return result - - cdef class _CPackedColumns: @staticmethod diff --git a/python/cudf/cudf/_lib/interop.pyx b/python/cudf/cudf/_lib/interop.pyx index 1dc586bb257..1c9d3a01b80 100644 --- a/python/cudf/cudf/_lib/interop.pyx +++ b/python/cudf/cudf/_lib/interop.pyx @@ -1,49 +1,22 @@ # Copyright (c) 2020-2024, NVIDIA CORPORATION. -from cpython cimport pycapsule -from libcpp.memory cimport unique_ptr -from libcpp.utility cimport move - import pylibcudf -from pylibcudf.libcudf.interop cimport ( - DLManagedTensor, - from_dlpack as cpp_from_dlpack, - to_dlpack as cpp_to_dlpack, -) -from pylibcudf.libcudf.table.table cimport table -from pylibcudf.libcudf.table.table_view cimport table_view - -from cudf._lib.utils cimport ( - columns_from_pylibcudf_table, - columns_from_unique_ptr, - table_view_from_columns, -) +from cudf._lib.utils cimport columns_from_pylibcudf_table from cudf.core.buffer import acquire_spill_lock from cudf.core.dtypes import ListDtype, StructDtype -def from_dlpack(dlpack_capsule): +def from_dlpack(object dlpack_capsule): """ Converts a DLPack Tensor PyCapsule into a list of columns. DLPack Tensor PyCapsule is expected to have the name "dltensor". """ - cdef DLManagedTensor* dlpack_tensor = pycapsule.\ - PyCapsule_GetPointer(dlpack_capsule, 'dltensor') - pycapsule.PyCapsule_SetName(dlpack_capsule, 'used_dltensor') - - cdef unique_ptr[table] c_result - - with nogil: - c_result = move( - cpp_from_dlpack(dlpack_tensor) - ) - - res = columns_from_unique_ptr(move(c_result)) - dlpack_tensor.deleter(dlpack_tensor) - return res + return columns_from_pylibcudf_table( + pylibcudf.interop.from_dlpack(dlpack_capsule) + ) def to_dlpack(list source_columns): @@ -52,39 +25,13 @@ def to_dlpack(list source_columns): DLPack Tensor PyCapsule will have the name "dltensor". """ - if any(column.null_count for column in source_columns): - raise ValueError( - "Cannot create a DLPack tensor with null values. \ - Input is required to have null count as zero." - ) - - cdef DLManagedTensor *dlpack_tensor - cdef table_view source_table_view = table_view_from_columns(source_columns) - - with nogil: - dlpack_tensor = cpp_to_dlpack( - source_table_view + return pylibcudf.interop.to_dlpack( + pylibcudf.Table( + [col.to_pylibcudf(mode="read") for col in source_columns] ) - - return pycapsule.PyCapsule_New( - dlpack_tensor, - 'dltensor', - dlmanaged_tensor_pycapsule_deleter ) -cdef void dlmanaged_tensor_pycapsule_deleter(object pycap_obj) noexcept: - cdef DLManagedTensor* dlpack_tensor = 0 - try: - dlpack_tensor = pycapsule.PyCapsule_GetPointer( - pycap_obj, 'used_dltensor') - return # we do not call a used capsule's deleter - except Exception: - dlpack_tensor = pycapsule.PyCapsule_GetPointer( - pycap_obj, 'dltensor') - dlpack_tensor.deleter(dlpack_tensor) - - def gather_metadata(object cols_dtypes): """ Generates a ColumnMetadata vector for each column. diff --git a/python/cudf/cudf/_lib/lists.pyx b/python/cudf/cudf/_lib/lists.pyx index 7e8710bedb6..12432ac6d5d 100644 --- a/python/cudf/cudf/_lib/lists.pyx +++ b/python/cudf/cudf/_lib/lists.pyx @@ -9,7 +9,7 @@ from pylibcudf.libcudf.types cimport null_order, size_type from cudf._lib.column cimport Column from cudf._lib.utils cimport columns_from_pylibcudf_table -import pylibcudf +import pylibcudf as plc from pylibcudf cimport Scalar @@ -17,7 +17,7 @@ from pylibcudf cimport Scalar @acquire_spill_lock() def count_elements(Column col): return Column.from_pylibcudf( - pylibcudf.lists.count_elements( + plc.lists.count_elements( col.to_pylibcudf(mode="read")) ) @@ -25,8 +25,8 @@ def count_elements(Column col): @acquire_spill_lock() def explode_outer(list source_columns, int explode_column_idx): return columns_from_pylibcudf_table( - pylibcudf.lists.explode_outer( - pylibcudf.Table([c.to_pylibcudf(mode="read") for c in source_columns]), + plc.lists.explode_outer( + plc.Table([c.to_pylibcudf(mode="read") for c in source_columns]), explode_column_idx, ) ) @@ -35,7 +35,7 @@ def explode_outer(list source_columns, int explode_column_idx): @acquire_spill_lock() def distinct(Column col, bool nulls_equal, bool nans_all_equal): return Column.from_pylibcudf( - pylibcudf.lists.distinct( + plc.lists.distinct( col.to_pylibcudf(mode="read"), nulls_equal, nans_all_equal, @@ -46,7 +46,7 @@ def distinct(Column col, bool nulls_equal, bool nans_all_equal): @acquire_spill_lock() def sort_lists(Column col, bool ascending, str na_position): return Column.from_pylibcudf( - pylibcudf.lists.sort_lists( + plc.lists.sort_lists( col.to_pylibcudf(mode="read"), ascending, null_order.BEFORE if na_position == "first" else null_order.AFTER, @@ -58,7 +58,7 @@ def sort_lists(Column col, bool ascending, str na_position): @acquire_spill_lock() def extract_element_scalar(Column col, size_type index): return Column.from_pylibcudf( - pylibcudf.lists.extract_list_element( + plc.lists.extract_list_element( col.to_pylibcudf(mode="read"), index, ) @@ -68,7 +68,7 @@ def extract_element_scalar(Column col, size_type index): @acquire_spill_lock() def extract_element_column(Column col, Column index): return Column.from_pylibcudf( - pylibcudf.lists.extract_list_element( + plc.lists.extract_list_element( col.to_pylibcudf(mode="read"), index.to_pylibcudf(mode="read"), ) @@ -78,7 +78,7 @@ def extract_element_column(Column col, Column index): @acquire_spill_lock() def contains_scalar(Column col, py_search_key): return Column.from_pylibcudf( - pylibcudf.lists.contains( + plc.lists.contains( col.to_pylibcudf(mode="read"), py_search_key.device_value.c_value, ) @@ -88,7 +88,7 @@ def contains_scalar(Column col, py_search_key): @acquire_spill_lock() def index_of_scalar(Column col, object py_search_key): return Column.from_pylibcudf( - pylibcudf.lists.index_of( + plc.lists.index_of( col.to_pylibcudf(mode="read"), py_search_key.device_value.c_value, True, @@ -99,7 +99,7 @@ def index_of_scalar(Column col, object py_search_key): @acquire_spill_lock() def index_of_column(Column col, Column search_keys): return Column.from_pylibcudf( - pylibcudf.lists.index_of( + plc.lists.index_of( col.to_pylibcudf(mode="read"), search_keys.to_pylibcudf(mode="read"), True, @@ -110,8 +110,8 @@ def index_of_column(Column col, Column search_keys): @acquire_spill_lock() def concatenate_rows(list source_columns): return Column.from_pylibcudf( - pylibcudf.lists.concatenate_rows( - pylibcudf.Table([ + plc.lists.concatenate_rows( + plc.Table([ c.to_pylibcudf(mode="read") for c in source_columns ]) ) @@ -121,8 +121,18 @@ def concatenate_rows(list source_columns): @acquire_spill_lock() def concatenate_list_elements(Column input_column, dropna=False): return Column.from_pylibcudf( - pylibcudf.lists.concatenate_list_elements( + plc.lists.concatenate_list_elements( input_column.to_pylibcudf(mode="read"), dropna, ) ) + + +@acquire_spill_lock() +def segmented_gather(Column source_column, Column gather_map): + return Column.from_pylibcudf( + plc.lists.segmented_gather( + source_column.to_pylibcudf(mode="read"), + gather_map.to_pylibcudf(mode="read"), + ) + ) diff --git a/python/cudf/cudf/_lib/nvtext/normalize.pyx b/python/cudf/cudf/_lib/nvtext/normalize.pyx index 633bc902db1..cc45123dd0a 100644 --- a/python/cudf/cudf/_lib/nvtext/normalize.pyx +++ b/python/cudf/cudf/_lib/nvtext/normalize.pyx @@ -11,16 +11,18 @@ from pylibcudf import nvtext @acquire_spill_lock() def normalize_spaces(Column input): - result = nvtext.normalize.normalize_spaces( - input.to_pylibcudf(mode="read") + return Column.from_pylibcudf( + nvtext.normalize.normalize_spaces( + input.to_pylibcudf(mode="read") + ) ) - return Column.from_pylibcudf(result) @acquire_spill_lock() def normalize_characters(Column input, bool do_lower=True): - result = nvtext.normalize.normalize_characters( - input.to_pylibcudf(mode="read"), - do_lower, + return Column.from_pylibcudf( + nvtext.normalize.normalize_characters( + input.to_pylibcudf(mode="read"), + do_lower, + ) ) - return Column.from_pylibcudf(result) diff --git a/python/cudf/cudf/_lib/nvtext/replace.pyx b/python/cudf/cudf/_lib/nvtext/replace.pyx index 61ae3da5782..bec56ade83c 100644 --- a/python/cudf/cudf/_lib/nvtext/replace.pyx +++ b/python/cudf/cudf/_lib/nvtext/replace.pyx @@ -2,20 +2,10 @@ from cudf.core.buffer import acquire_spill_lock -from libcpp.memory cimport unique_ptr -from libcpp.utility cimport move - -from pylibcudf.libcudf.column.column cimport column -from pylibcudf.libcudf.column.column_view cimport column_view -from pylibcudf.libcudf.nvtext.replace cimport ( - filter_tokens as cpp_filter_tokens, - replace_tokens as cpp_replace_tokens, -) -from pylibcudf.libcudf.scalar.scalar cimport string_scalar from pylibcudf.libcudf.types cimport size_type from cudf._lib.column cimport Column -from cudf._lib.scalar cimport DeviceScalar +from pylibcudf import nvtext @acquire_spill_lock() @@ -30,27 +20,14 @@ def replace_tokens(Column strings, provided. """ - cdef DeviceScalar delimiter = py_delimiter.device_value - - cdef column_view c_strings = strings.view() - cdef column_view c_targets = targets.view() - cdef column_view c_replacements = replacements.view() - - cdef const string_scalar* c_delimiter = delimiter\ - .get_raw_ptr() - cdef unique_ptr[column] c_result - - with nogil: - c_result = move( - cpp_replace_tokens( - c_strings, - c_targets, - c_replacements, - c_delimiter[0], - ) + return Column.from_pylibcudf( + nvtext.replace.replace_tokens( + strings.to_pylibcudf(mode="read"), + targets.to_pylibcudf(mode="read"), + replacements.to_pylibcudf(mode="read"), + py_delimiter.device_value.c_value, ) - - return Column.from_unique_ptr(move(c_result)) + ) @acquire_spill_lock() @@ -65,24 +42,11 @@ def filter_tokens(Column strings, character provided. """ - cdef DeviceScalar replacement = py_replacement.device_value - cdef DeviceScalar delimiter = py_delimiter.device_value - - cdef column_view c_strings = strings.view() - cdef const string_scalar* c_repl = replacement\ - .get_raw_ptr() - cdef const string_scalar* c_delimiter = delimiter\ - .get_raw_ptr() - cdef unique_ptr[column] c_result - - with nogil: - c_result = move( - cpp_filter_tokens( - c_strings, - min_token_length, - c_repl[0], - c_delimiter[0], - ) + return Column.from_pylibcudf( + nvtext.replace.filter_tokens( + strings.to_pylibcudf(mode="read"), + min_token_length, + py_replacement.device_value.c_value, + py_delimiter.device_value.c_value, ) - - return Column.from_unique_ptr(move(c_result)) + ) diff --git a/python/cudf/cudf/_lib/nvtext/stemmer.pyx b/python/cudf/cudf/_lib/nvtext/stemmer.pyx index 5bf25562fed..63a389b64d5 100644 --- a/python/cudf/cudf/_lib/nvtext/stemmer.pyx +++ b/python/cudf/cudf/_lib/nvtext/stemmer.pyx @@ -1,24 +1,19 @@ # Copyright (c) 2020-2024, NVIDIA CORPORATION. -from cudf.core.buffer import acquire_spill_lock - -from libcpp.memory cimport unique_ptr -from libcpp.utility cimport move - from enum import IntEnum -from pylibcudf.libcudf.column.column cimport column -from pylibcudf.libcudf.column.column_view cimport column_view +from cudf.core.buffer import acquire_spill_lock + from pylibcudf.libcudf.nvtext.stemmer cimport ( - is_letter as cpp_is_letter, letter_type, - porter_stemmer_measure as cpp_porter_stemmer_measure, underlying_type_t_letter_type, ) from pylibcudf.libcudf.types cimport size_type from cudf._lib.column cimport Column +from pylibcudf import nvtext + class LetterType(IntEnum): CONSONANT = letter_type.CONSONANT @@ -27,43 +22,34 @@ class LetterType(IntEnum): @acquire_spill_lock() def porter_stemmer_measure(Column strings): - cdef column_view c_strings = strings.view() - cdef unique_ptr[column] c_result - - with nogil: - c_result = move(cpp_porter_stemmer_measure(c_strings)) - - return Column.from_unique_ptr(move(c_result)) + return Column.from_pylibcudf( + nvtext.stemmer.porter_stemmer_measure( + strings.to_pylibcudf(mode="read"), + ) + ) @acquire_spill_lock() def is_letter(Column strings, object ltype, size_type index): - cdef column_view c_strings = strings.view() - cdef letter_type c_ltype = ( - ltype + return Column.from_pylibcudf( + nvtext.stemmer.is_letter( + strings.to_pylibcudf(mode="read"), + ltype==LetterType.VOWEL, + index, + ) ) - cdef unique_ptr[column] c_result - - with nogil: - c_result = move(cpp_is_letter(c_strings, c_ltype, index)) - - return Column.from_unique_ptr(move(c_result)) @acquire_spill_lock() def is_letter_multi(Column strings, object ltype, Column indices): - cdef column_view c_strings = strings.view() - cdef column_view c_indices = indices.view() - cdef letter_type c_ltype = ( - ltype + return Column.from_pylibcudf( + nvtext.stemmer.is_letter( + strings.to_pylibcudf(mode="read"), + ltype==LetterType.VOWEL, + indices.to_pylibcudf(mode="read"), + ) ) - cdef unique_ptr[column] c_result - - with nogil: - c_result = move(cpp_is_letter(c_strings, c_ltype, c_indices)) - - return Column.from_unique_ptr(move(c_result)) diff --git a/python/cudf/cudf/_lib/strings/replace_re.pyx b/python/cudf/cudf/_lib/strings/replace_re.pyx index fffc8b7c3f6..462d5c903e8 100644 --- a/python/cudf/cudf/_lib/strings/replace_re.pyx +++ b/python/cudf/cudf/_lib/strings/replace_re.pyx @@ -1,26 +1,11 @@ # Copyright (c) 2020-2024, NVIDIA CORPORATION. -from cython.operator cimport dereference -from libcpp.memory cimport unique_ptr -from libcpp.string cimport string -from libcpp.utility cimport move -from libcpp.vector cimport vector +from pylibcudf.libcudf.types cimport size_type +import pylibcudf as plc from cudf.core.buffer import acquire_spill_lock -from pylibcudf.libcudf.column.column cimport column -from pylibcudf.libcudf.column.column_view cimport column_view -from pylibcudf.libcudf.scalar.scalar cimport string_scalar -from pylibcudf.libcudf.strings.regex_flags cimport regex_flags -from pylibcudf.libcudf.strings.regex_program cimport regex_program -from pylibcudf.libcudf.strings.replace_re cimport ( - replace_re as cpp_replace_re, - replace_with_backrefs as cpp_replace_with_backrefs, -) -from pylibcudf.libcudf.types cimport size_type - from cudf._lib.column cimport Column -from cudf._lib.scalar cimport DeviceScalar @acquire_spill_lock() @@ -34,28 +19,16 @@ def replace_re(Column source_strings, `n` indicates the number of resplacements to be made from start. (-1 indicates all) """ - - cdef DeviceScalar repl = py_repl.device_value - - cdef unique_ptr[column] c_result - cdef column_view source_view = source_strings.view() - - cdef string pattern_string = str(pattern).encode() - cdef const string_scalar* scalar_repl = \ - (repl.get_raw_ptr()) - cdef regex_flags c_flags = regex_flags.DEFAULT - cdef unique_ptr[regex_program] c_prog - - with nogil: - c_prog = move(regex_program.create(pattern_string, c_flags)) - c_result = move(cpp_replace_re( - source_view, - dereference(c_prog), - scalar_repl[0], - n - )) - - return Column.from_unique_ptr(move(c_result)) + plc_column = plc.strings.replace_re.replace_re( + source_strings.to_pylibcudf(mode="read"), + plc.strings.regex_program.RegexProgram.create( + str(pattern), + plc.strings.regex_flags.RegexFlags.DEFAULT + ), + py_repl.device_value.c_value, + n + ) + return Column.from_pylibcudf(plc_column) @acquire_spill_lock() @@ -68,50 +41,29 @@ def replace_with_backrefs( new string with the extracted elements found using `pattern` regular expression in `source_strings`. """ - cdef unique_ptr[column] c_result - cdef column_view source_view = source_strings.view() - - cdef string pattern_string = str(pattern).encode() - cdef string repl_string = str(repl).encode() - cdef regex_flags c_flags = regex_flags.DEFAULT - cdef unique_ptr[regex_program] c_prog - - with nogil: - c_prog = move(regex_program.create(pattern_string, c_flags)) - c_result = move(cpp_replace_with_backrefs( - source_view, - dereference(c_prog), - repl_string - )) - - return Column.from_unique_ptr(move(c_result)) + plc_column = plc.strings.replace_re.replace_with_backrefs( + source_strings.to_pylibcudf(mode="read"), + plc.strings.regex_program.RegexProgram.create( + str(pattern), + plc.strings.regex_flags.RegexFlags.DEFAULT + ), + repl + ) + return Column.from_pylibcudf(plc_column) @acquire_spill_lock() def replace_multi_re(Column source_strings, - object patterns, + list patterns, Column repl_strings): """ Returns a Column after replacing occurrences of multiple regular expressions `patterns` with their corresponding strings in `repl_strings` in `source_strings`. """ - cdef unique_ptr[column] c_result - cdef column_view source_view = source_strings.view() - cdef column_view repl_view = repl_strings.view() - - cdef int pattern_size = len(patterns) - cdef vector[string] patterns_vector - patterns_vector.reserve(pattern_size) - - for pattern in patterns: - patterns_vector.push_back(str.encode(pattern)) - - with nogil: - c_result = move(cpp_replace_re( - source_view, - patterns_vector, - repl_view - )) - - return Column.from_unique_ptr(move(c_result)) + plc_column = plc.strings.replace_re.replace_re( + source_strings.to_pylibcudf(mode="read"), + patterns, + repl_strings.to_pylibcudf(mode="read") + ) + return Column.from_pylibcudf(plc_column) diff --git a/python/cudf/cudf/_typing.py b/python/cudf/cudf/_typing.py index 6e8ad556b08..3b13cc258ab 100644 --- a/python/cudf/cudf/_typing.py +++ b/python/cudf/cudf/_typing.py @@ -1,8 +1,8 @@ # Copyright (c) 2021-2024, NVIDIA CORPORATION. import sys -from collections.abc import Callable -from typing import TYPE_CHECKING, Any, Dict, Iterable, TypeVar, Union +from collections.abc import Callable, Iterable +from typing import TYPE_CHECKING, Any, TypeVar, Union import numpy as np from pandas import Period, Timedelta, Timestamp @@ -42,7 +42,7 @@ SeriesOrSingleColumnIndex = Union["cudf.Series", "cudf.core.index.Index"] # Groupby aggregation -AggType = Union[str, Callable] -MultiColumnAggType = Union[ - AggType, Iterable[AggType], Dict[Any, Iterable[AggType]] +AggType = Union[str, Callable] # noqa: UP007 +MultiColumnAggType = Union[ # noqa: UP007 + AggType, Iterable[AggType], dict[Any, Iterable[AggType]] ] diff --git a/python/cudf/cudf/core/buffer/buffer.py b/python/cudf/cudf/core/buffer/buffer.py index caff019f575..ffa306bf93f 100644 --- a/python/cudf/cudf/core/buffer/buffer.py +++ b/python/cudf/cudf/core/buffer/buffer.py @@ -6,7 +6,7 @@ import pickle import weakref from types import SimpleNamespace -from typing import Any, Literal, Mapping +from typing import TYPE_CHECKING, Any, Literal import numpy from typing_extensions import Self @@ -18,6 +18,9 @@ from cudf.core.abc import Serializable from cudf.utils.string import format_bytes +if TYPE_CHECKING: + from collections.abc import Mapping + def host_memory_allocation(nbytes: int) -> memoryview: """Allocate host memory using NumPy diff --git a/python/cudf/cudf/core/buffer/exposure_tracked_buffer.py b/python/cudf/cudf/core/buffer/exposure_tracked_buffer.py index 0bd8d6054b3..ecf9807cfc2 100644 --- a/python/cudf/cudf/core/buffer/exposure_tracked_buffer.py +++ b/python/cudf/cudf/core/buffer/exposure_tracked_buffer.py @@ -2,13 +2,16 @@ from __future__ import annotations -from typing import Literal, Mapping +from typing import TYPE_CHECKING, Literal from typing_extensions import Self import cudf from cudf.core.buffer.buffer import Buffer, BufferOwner +if TYPE_CHECKING: + from collections.abc import Mapping + class ExposureTrackedBuffer(Buffer): """An exposure tracked buffer. diff --git a/python/cudf/cudf/core/column/__init__.py b/python/cudf/cudf/core/column/__init__.py index 06791df7dc0..a1e87d04bc9 100644 --- a/python/cudf/cudf/core/column/__init__.py +++ b/python/cudf/cudf/core/column/__init__.py @@ -29,4 +29,3 @@ Decimal128Column, DecimalBaseColumn, ) -from cudf.core.column.interval import IntervalColumn # noqa: F401 diff --git a/python/cudf/cudf/core/column/categorical.py b/python/cudf/cudf/core/column/categorical.py index 864e87b5377..087d0ed65f5 100644 --- a/python/cudf/cudf/core/column/categorical.py +++ b/python/cudf/cudf/core/column/categorical.py @@ -4,7 +4,7 @@ import warnings from functools import cached_property -from typing import TYPE_CHECKING, Any, Mapping, Sequence, cast +from typing import TYPE_CHECKING, Any, cast import numpy as np import pandas as pd @@ -26,6 +26,7 @@ if TYPE_CHECKING: from collections import abc + from collections.abc import Mapping, Sequence import numba.cuda diff --git a/python/cudf/cudf/core/column/column.py b/python/cudf/cudf/core/column/column.py index 7674565e2c3..d2cd6e8ac8f 100644 --- a/python/cudf/cudf/core/column/column.py +++ b/python/cudf/cudf/core/column/column.py @@ -4,10 +4,11 @@ import pickle from collections import abc +from collections.abc import MutableSequence, Sequence from functools import cached_property from itertools import chain from types import SimpleNamespace -from typing import TYPE_CHECKING, Any, Literal, MutableSequence, Sequence, cast +from typing import TYPE_CHECKING, Any, Literal, cast import cupy import numpy as np diff --git a/python/cudf/cudf/core/column/datetime.py b/python/cudf/cudf/core/column/datetime.py index 2c9b0baa9b6..b6dc250e64d 100644 --- a/python/cudf/cudf/core/column/datetime.py +++ b/python/cudf/cudf/core/column/datetime.py @@ -8,7 +8,7 @@ import locale import re from locale import nl_langinfo -from typing import TYPE_CHECKING, Literal, Sequence, cast +from typing import TYPE_CHECKING, Literal, cast import numpy as np import pandas as pd @@ -31,6 +31,8 @@ from cudf.utils.utils import _all_bools_with_nulls if TYPE_CHECKING: + from collections.abc import Sequence + from cudf._typing import ( ColumnBinaryOperand, DatetimeLikeScalar, diff --git a/python/cudf/cudf/core/column/decimal.py b/python/cudf/cudf/core/column/decimal.py index 8803ebd6791..8ae06f72d1e 100644 --- a/python/cudf/cudf/core/column/decimal.py +++ b/python/cudf/cudf/core/column/decimal.py @@ -3,8 +3,9 @@ from __future__ import annotations import warnings +from collections.abc import Sequence from decimal import Decimal -from typing import TYPE_CHECKING, Sequence, cast +from typing import TYPE_CHECKING, cast import cupy as cp import numpy as np diff --git a/python/cudf/cudf/core/column/lists.py b/python/cudf/cudf/core/column/lists.py index c6a39199e3b..6b25e568f00 100644 --- a/python/cudf/cudf/core/column/lists.py +++ b/python/cudf/cudf/core/column/lists.py @@ -3,7 +3,7 @@ from __future__ import annotations from functools import cached_property -from typing import TYPE_CHECKING, Sequence, cast +from typing import TYPE_CHECKING, cast import numpy as np import pandas as pd @@ -11,7 +11,6 @@ from typing_extensions import Self import cudf -from cudf._lib.copying import segmented_gather from cudf._lib.lists import ( concatenate_list_elements, concatenate_rows, @@ -22,6 +21,7 @@ extract_element_scalar, index_of_column, index_of_scalar, + segmented_gather, sort_lists, ) from cudf._lib.strings.convert.convert_lists import format_list_column @@ -34,6 +34,8 @@ from cudf.core.missing import NA if TYPE_CHECKING: + from collections.abc import Sequence + from cudf._typing import ColumnBinaryOperand, ColumnLike, Dtype, ScalarLike from cudf.core.buffer import Buffer diff --git a/python/cudf/cudf/core/column/methods.py b/python/cudf/cudf/core/column/methods.py index 05a0ab2e09a..a91c080fe21 100644 --- a/python/cudf/cudf/core/column/methods.py +++ b/python/cudf/cudf/core/column/methods.py @@ -2,9 +2,7 @@ from __future__ import annotations -from typing import Union, overload - -from typing_extensions import Literal +from typing import Literal, Union, overload import cudf import cudf.core.column diff --git a/python/cudf/cudf/core/column/numerical.py b/python/cudf/cudf/core/column/numerical.py index 78d2814ed26..620cae65374 100644 --- a/python/cudf/cudf/core/column/numerical.py +++ b/python/cudf/cudf/core/column/numerical.py @@ -3,7 +3,7 @@ from __future__ import annotations import functools -from typing import TYPE_CHECKING, Any, Sequence, cast +from typing import TYPE_CHECKING, Any, cast import numpy as np import pandas as pd @@ -28,7 +28,7 @@ from .numerical_base import NumericalBaseColumn if TYPE_CHECKING: - from collections.abc import Callable + from collections.abc import Callable, Sequence from cudf._typing import ( ColumnBinaryOperand, diff --git a/python/cudf/cudf/core/column/string.py b/python/cudf/cudf/core/column/string.py index 45d1a8b087b..856ce0f75de 100644 --- a/python/cudf/cudf/core/column/string.py +++ b/python/cudf/cudf/core/column/string.py @@ -5,7 +5,7 @@ import re import warnings from functools import cached_property -from typing import TYPE_CHECKING, Sequence, cast, overload +from typing import TYPE_CHECKING, cast, overload import numpy as np import pandas as pd @@ -35,6 +35,8 @@ def str_to_boolean(column: StringColumn): if TYPE_CHECKING: + from collections.abc import Sequence + import cupy import numba.cuda @@ -998,7 +1000,7 @@ def replace( return self._return_or_inplace( libstrings.replace_multi_re( self._column, - pat, + list(pat), column.as_column(repl, dtype="str"), ) if regex diff --git a/python/cudf/cudf/core/column/timedelta.py b/python/cudf/cudf/core/column/timedelta.py index 6b6f3e517a8..087d6474e7f 100644 --- a/python/cudf/cudf/core/column/timedelta.py +++ b/python/cudf/cudf/core/column/timedelta.py @@ -4,7 +4,7 @@ import datetime import functools -from typing import TYPE_CHECKING, Sequence, cast +from typing import TYPE_CHECKING, cast import numpy as np import pandas as pd @@ -19,6 +19,8 @@ from cudf.utils.utils import _all_bools_with_nulls if TYPE_CHECKING: + from collections.abc import Sequence + from cudf._typing import ColumnBinaryOperand, DatetimeLikeScalar, Dtype _unit_to_nanoseconds_conversion = { diff --git a/python/cudf/cudf/core/column_accessor.py b/python/cudf/cudf/core/column_accessor.py index bc093fdaa9a..496e86ed709 100644 --- a/python/cudf/cudf/core/column_accessor.py +++ b/python/cudf/cudf/core/column_accessor.py @@ -5,8 +5,9 @@ import itertools import sys from collections import abc +from collections.abc import Mapping from functools import cached_property, reduce -from typing import TYPE_CHECKING, Any, Mapping, cast +from typing import TYPE_CHECKING, Any, cast import numpy as np import pandas as pd diff --git a/python/cudf/cudf/core/dataframe.py b/python/cudf/cudf/core/dataframe.py index 7d4d34f5b04..bf1c39b23da 100644 --- a/python/cudf/cudf/core/dataframe.py +++ b/python/cudf/cudf/core/dataframe.py @@ -13,8 +13,8 @@ import textwrap import warnings from collections import abc, defaultdict -from collections.abc import Callable, Iterator -from typing import TYPE_CHECKING, Any, Literal, MutableMapping, cast +from collections.abc import Callable, Iterator, MutableMapping +from typing import TYPE_CHECKING, Any, Literal, cast import cupy import numba diff --git a/python/cudf/cudf/core/df_protocol.py b/python/cudf/cudf/core/df_protocol.py index 5250a741d3d..aa601a2b322 100644 --- a/python/cudf/cudf/core/df_protocol.py +++ b/python/cudf/cudf/core/df_protocol.py @@ -3,7 +3,7 @@ import enum from collections import abc -from typing import Any, Iterable, Mapping, Sequence, Tuple, cast +from typing import TYPE_CHECKING, Any, cast import cupy as cp import numpy as np @@ -20,6 +20,9 @@ build_column, ) +if TYPE_CHECKING: + from collections.abc import Iterable, Mapping, Sequence + # Implementation of interchange protocol classes # ---------------------------------------------- @@ -61,7 +64,7 @@ class _MaskKind(enum.IntEnum): _DtypeKind.BOOL, _DtypeKind.STRING, } -ProtoDtype = Tuple[_DtypeKind, int, str, str] +ProtoDtype = tuple[_DtypeKind, int, str, str] class _CuDFBuffer: diff --git a/python/cudf/cudf/core/frame.py b/python/cudf/cudf/core/frame.py index 37ad6b8fabb..205edd91d9d 100644 --- a/python/cudf/cudf/core/frame.py +++ b/python/cudf/cudf/core/frame.py @@ -6,7 +6,7 @@ import pickle import warnings from collections import abc -from typing import TYPE_CHECKING, Any, Literal, MutableMapping +from typing import TYPE_CHECKING, Any, Literal # TODO: The `numpy` import is needed for typing purposes during doc builds # only, need to figure out why the `np` alias is insufficient then remove. @@ -36,6 +36,7 @@ from cudf.utils.utils import _array_ufunc, _warn_no_dask_cudf if TYPE_CHECKING: + from collections.abc import MutableMapping from types import ModuleType from cudf._typing import Dtype, ScalarLike diff --git a/python/cudf/cudf/core/groupby/groupby.py b/python/cudf/cudf/core/groupby/groupby.py index 81b20488d8d..6630bd96c01 100644 --- a/python/cudf/cudf/core/groupby/groupby.py +++ b/python/cudf/cudf/core/groupby/groupby.py @@ -8,7 +8,7 @@ import warnings from collections import abc from functools import cached_property -from typing import TYPE_CHECKING, Any, Iterable, Literal +from typing import TYPE_CHECKING, Any, Literal import cupy as cp import numpy as np @@ -36,6 +36,8 @@ from cudf.utils.utils import GetAttrGetItemMixin if TYPE_CHECKING: + from collections.abc import Iterable + from cudf._typing import ( AggType, DataFrameOrSeries, diff --git a/python/cudf/cudf/core/index.py b/python/cudf/cudf/core/index.py index cd07c58c5d9..1b90e9f9df0 100644 --- a/python/cudf/cudf/core/index.py +++ b/python/cudf/cudf/core/index.py @@ -5,10 +5,10 @@ import operator import pickle import warnings -from collections.abc import Hashable +from collections.abc import Hashable, MutableMapping from functools import cache, cached_property from numbers import Number -from typing import TYPE_CHECKING, Any, Literal, MutableMapping, cast +from typing import TYPE_CHECKING, Any, Literal, cast import cupy import numpy as np diff --git a/python/cudf/cudf/core/indexed_frame.py b/python/cudf/cudf/core/indexed_frame.py index 5952815deef..e031f2a4e8e 100644 --- a/python/cudf/cudf/core/indexed_frame.py +++ b/python/cudf/cudf/core/indexed_frame.py @@ -10,9 +10,7 @@ from typing import ( TYPE_CHECKING, Any, - Callable, Literal, - MutableMapping, TypeVar, cast, ) @@ -63,6 +61,8 @@ from cudf.utils.utils import _warn_no_dask_cudf if TYPE_CHECKING: + from collections.abc import Callable, MutableMapping + from cudf._typing import ( ColumnLike, DataFrameOrSeries, diff --git a/python/cudf/cudf/core/indexing_utils.py b/python/cudf/cudf/core/indexing_utils.py index 8182e5cede2..ce6a5c960dd 100644 --- a/python/cudf/cudf/core/indexing_utils.py +++ b/python/cudf/cudf/core/indexing_utils.py @@ -3,9 +3,7 @@ from __future__ import annotations from dataclasses import dataclass -from typing import Any, List, Union - -from typing_extensions import TypeAlias +from typing import Any, TypeAlias import cudf from cudf.api.types import _is_scalar_or_zero_d_array, is_integer @@ -46,11 +44,11 @@ class ScalarIndexer: key: GatherMap -IndexingSpec: TypeAlias = Union[ - EmptyIndexer, MapIndexer, MaskIndexer, ScalarIndexer, SliceIndexer -] +IndexingSpec: TypeAlias = ( + EmptyIndexer | MapIndexer | MaskIndexer | ScalarIndexer | SliceIndexer +) -ColumnLabels: TypeAlias = List[str] +ColumnLabels: TypeAlias = list[str] def destructure_iloc_key( diff --git a/python/cudf/cudf/core/multiindex.py b/python/cudf/cudf/core/multiindex.py index 92d094d9de5..bfff62f0a89 100644 --- a/python/cudf/cudf/core/multiindex.py +++ b/python/cudf/cudf/core/multiindex.py @@ -8,7 +8,7 @@ import pickle import warnings from functools import cached_property -from typing import TYPE_CHECKING, Any, MutableMapping +from typing import TYPE_CHECKING, Any import cupy as cp import numpy as np @@ -36,7 +36,7 @@ from cudf.utils.utils import NotIterable, _external_only_api, _is_same_name if TYPE_CHECKING: - from collections.abc import Generator, Hashable + from collections.abc import Generator, Hashable, MutableMapping from typing_extensions import Self diff --git a/python/cudf/cudf/core/series.py b/python/cudf/cudf/core/series.py index 29ed18ac0ce..9b60424c924 100644 --- a/python/cudf/cudf/core/series.py +++ b/python/cudf/cudf/core/series.py @@ -9,7 +9,7 @@ import warnings from collections import abc from shutil import get_terminal_size -from typing import TYPE_CHECKING, Any, Literal, MutableMapping +from typing import TYPE_CHECKING, Any, Literal import cupy import numpy as np @@ -71,6 +71,8 @@ from cudf.utils.performance_tracking import _performance_tracking if TYPE_CHECKING: + from collections.abc import MutableMapping + import pyarrow as pa from cudf._typing import ( diff --git a/python/cudf/cudf/core/tools/datetimes.py b/python/cudf/cudf/core/tools/datetimes.py index 68f34fa28ff..885e7b16644 100644 --- a/python/cudf/cudf/core/tools/datetimes.py +++ b/python/cudf/cudf/core/tools/datetimes.py @@ -4,7 +4,7 @@ import math import re import warnings -from typing import Literal, Sequence +from typing import TYPE_CHECKING, Literal import numpy as np import pandas as pd @@ -20,6 +20,9 @@ from cudf.core import column from cudf.core.index import ensure_index +if TYPE_CHECKING: + from collections.abc import Sequence + # https://github.com/pandas-dev/pandas/blob/2.2.x/pandas/core/tools/datetimes.py#L1112 _unit_map = { "year": "year", diff --git a/python/cudf/cudf/pandas/fast_slow_proxy.py b/python/cudf/cudf/pandas/fast_slow_proxy.py index c364d55e677..73afde407db 100644 --- a/python/cudf/cudf/pandas/fast_slow_proxy.py +++ b/python/cudf/cudf/pandas/fast_slow_proxy.py @@ -10,9 +10,9 @@ import pickle import types import warnings -from collections.abc import Callable, Iterator +from collections.abc import Callable, Iterator, Mapping from enum import IntEnum -from typing import Any, Literal, Mapping +from typing import Any, Literal import numpy as np diff --git a/python/cudf/cudf/pandas/module_accelerator.py b/python/cudf/cudf/pandas/module_accelerator.py index f82e300e83d..38103a71908 100644 --- a/python/cudf/cudf/pandas/module_accelerator.py +++ b/python/cudf/cudf/pandas/module_accelerator.py @@ -17,7 +17,7 @@ from abc import abstractmethod from importlib._bootstrap import _ImportLockContext as ImportLock from types import ModuleType -from typing import Any, ContextManager, NamedTuple +from typing import Any, ContextManager, NamedTuple # noqa: UP035 from typing_extensions import Self diff --git a/python/cudf/cudf/pandas/scripts/analyze-test-failures.py b/python/cudf/cudf/pandas/scripts/analyze-test-failures.py index 8870fbc5c28..bb2fc00d9fc 100644 --- a/python/cudf/cudf/pandas/scripts/analyze-test-failures.py +++ b/python/cudf/cudf/pandas/scripts/analyze-test-failures.py @@ -9,6 +9,7 @@ python analyze-test-failures.py Example: +------- python analyze-test-failures.py log.json frame/* """ diff --git a/python/cudf/cudf/pandas/scripts/conftest-patch.py b/python/cudf/cudf/pandas/scripts/conftest-patch.py index d12d2697729..59966a5ff0c 100644 --- a/python/cudf/cudf/pandas/scripts/conftest-patch.py +++ b/python/cudf/cudf/pandas/scripts/conftest-patch.py @@ -35,7 +35,7 @@ def null_assert_warnings(*args, **kwargs): @pytest.fixture(scope="session", autouse=True) # type: ignore def patch_testing_functions(): - tm.assert_produces_warning = null_assert_warnings + tm.assert_produces_warning = null_assert_warnings # noqa: F821 pytest.raises = replace_kwargs({"match": None})(pytest.raises) diff --git a/python/cudf/cudf/pandas/scripts/summarize-test-results.py b/python/cudf/cudf/pandas/scripts/summarize-test-results.py index 4ea0b3b4413..a0ad872e4c7 100644 --- a/python/cudf/cudf/pandas/scripts/summarize-test-results.py +++ b/python/cudf/cudf/pandas/scripts/summarize-test-results.py @@ -5,7 +5,8 @@ """ Summarizes the test results per module. -Examples: +Examples +-------- python summarize-test-results.py log.json python summarize-test-results.py log.json --output json python summarize-test-results.py log.json --output table diff --git a/python/cudf/cudf_pandas_tests/test_fast_slow_proxy.py b/python/cudf/cudf_pandas_tests/test_fast_slow_proxy.py index a75a20a4681..63fd9601fc1 100644 --- a/python/cudf/cudf_pandas_tests/test_fast_slow_proxy.py +++ b/python/cudf/cudf_pandas_tests/test_fast_slow_proxy.py @@ -387,7 +387,8 @@ def test_dir_bound_method( ): """This test will fail because dir for bound methods is currently incorrect, but we have no way to fix it without materializing the slow - type, which is unnecessarily expensive.""" + type, which is unnecessarily expensive. + """ Fast, FastIntermediate = fast_and_intermediate_with_doc Slow, SlowIntermediate = slow_and_intermediate_with_doc diff --git a/python/cudf/pyproject.toml b/python/cudf/pyproject.toml index feab04ffadc..80201dd84db 100644 --- a/python/cudf/pyproject.toml +++ b/python/cudf/pyproject.toml @@ -81,50 +81,6 @@ cudf-pandas-tests = [ Homepage = "https://github.com/rapidsai/cudf" Documentation = "https://docs.rapids.ai/api/cudf/stable/" -[tool.isort] -line_length = 79 -multi_line_output = 3 -include_trailing_comma = true -force_grid_wrap = 0 -combine_as_imports = true -order_by_type = true -known_dask = [ - "dask", - "distributed", - "dask_cuda", -] -known_rapids = [ - "rmm", - "pylibcudf" -] -known_first_party = [ - "cudf", -] -default_section = "THIRDPARTY" -sections = [ - "FUTURE", - "STDLIB", - "THIRDPARTY", - "DASK", - "RAPIDS", - "FIRSTPARTY", - "LOCALFOLDER", -] -skip = [ - "thirdparty", - ".eggs", - ".git", - ".hg", - ".mypy_cache", - ".tox", - ".venv", - "_build", - "buck-out", - "build", - "dist", - "__init__.py", -] - [tool.pytest.ini_options] addopts = "--tb=native --strict-config --strict-markers" empty_parameter_set_mark = "fail_at_collect" @@ -174,3 +130,18 @@ wheel.packages = ["cudf"] provider = "scikit_build_core.metadata.regex" input = "cudf/VERSION" regex = "(?P.*)" + +[tool.ruff] +extend = "../../pyproject.toml" + +[tool.ruff.lint.isort] +combine-as-imports = true +known-first-party = ["cudf"] +section-order = ["future", "standard-library", "third-party", "dask", "rapids", "first-party", "local-folder"] + +[tool.ruff.lint.isort.sections] +dask = ["dask", "distributed", "dask_cuda"] +rapids = ["rmm", "pylibcudf"] + +[tool.ruff.lint.per-file-ignores] +"__init__.py" = ["E402", "F401"] diff --git a/python/cudf_kafka/pyproject.toml b/python/cudf_kafka/pyproject.toml index 87e19a2bccf..667cd7b1db8 100644 --- a/python/cudf_kafka/pyproject.toml +++ b/python/cudf_kafka/pyproject.toml @@ -32,51 +32,20 @@ test = [ Homepage = "https://github.com/rapidsai/cudf" Documentation = "https://docs.rapids.ai/api/cudf/stable/" -[tool.isort] -line_length = 79 -multi_line_output = 3 -include_trailing_comma = true -force_grid_wrap = 0 -combine_as_imports = true -order_by_type = true -known_dask = [ - "dask", - "distributed", - "dask_cuda", - "streamz", -] -known_rapids = [ - "rmm", - "cudf", - "dask_cudf", -] -known_first_party = [ - "cudf_kafka", -] -default_section = "THIRDPARTY" -sections = [ - "FUTURE", - "STDLIB", - "THIRDPARTY", - "DASK", - "RAPIDS", - "FIRSTPARTY", - "LOCALFOLDER", -] -skip = [ - "thirdparty", - ".eggs", - ".git", - ".hg", - ".mypy_cache", - ".tox", - ".venv", - "_build", - "buck-out", - "build", - "dist", - "__init__.py", -] +[tool.ruff] +extend = "../../pyproject.toml" + +[tool.ruff.lint.isort] +combine-as-imports = true +known-first-party = ["cudf_kafka"] +section-order = ["future", "standard-library", "third-party", "dask", "rapids", "first-party", "local-folder"] + +[tool.ruff.lint.isort.sections] +dask = ["dask", "distributed", "dask_cuda", "streamz"] +rapids = ["rmm", "cudf", "dask_cudf"] + +[tool.ruff.lint.per-file-ignores] +"__init__.py" = ["E402", "F401"] [tool.pytest.ini_options] addopts = "--tb=native --strict-config --strict-markers" diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/aggregation.py b/python/cudf_polars/cudf_polars/dsl/expressions/aggregation.py index b8b18ec5039..41b1defab39 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/aggregation.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/aggregation.py @@ -30,14 +30,13 @@ class Agg(Expr): - __slots__ = ("name", "options", "op", "request", "children") + __slots__ = ("name", "options", "op", "request") _non_child = ("dtype", "name", "options") - children: tuple[Expr, ...] def __init__( self, dtype: plc.DataType, name: str, options: Any, *children: Expr ) -> None: - super().__init__(dtype) + self.dtype = dtype self.name = name self.options = options self.children = children diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/base.py b/python/cudf_polars/cudf_polars/dsl/expressions/base.py index 8d021b0231d..effe8cb2378 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/base.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/base.py @@ -13,9 +13,10 @@ import pylibcudf as plc from cudf_polars.containers import Column +from cudf_polars.dsl.nodebase import Node if TYPE_CHECKING: - from collections.abc import Mapping, Sequence + from collections.abc import Mapping from cudf_polars.containers import Column, DataFrame @@ -32,100 +33,16 @@ class ExecutionContext(IntEnum): ROLLING = enum.auto() -class Expr: - """ - An abstract expression object. +class Expr(Node["Expr"]): + """An abstract expression object.""" - This contains a (potentially empty) tuple of child expressions, - along with non-child data. For uniform reconstruction and - implementation of hashing and equality schemes, child classes need - to provide a certain amount of metadata when they are defined. - Specifically, the ``_non_child`` attribute must list, in-order, - the names of the slots that are passed to the constructor. The - constructor must take arguments in the order ``(*_non_child, - *children).`` - """ - - __slots__ = ("dtype", "_hash_value", "_repr_value") + __slots__ = ("dtype",) dtype: plc.DataType """Data type of the expression.""" - _hash_value: int - """Caching slot for the hash of the expression.""" - _repr_value: str - """Caching slot for repr of the expression.""" - children: tuple[Expr, ...] = () - """Children of the expression.""" + # This annotation is needed because of https://github.com/python/mypy/issues/17981 _non_child: ClassVar[tuple[str, ...]] = ("dtype",) """Names of non-child data (not Exprs) for reconstruction.""" - # Constructor must take arguments in order (*_non_child, *children) - def __init__(self, dtype: plc.DataType) -> None: - self.dtype = dtype - - def _ctor_arguments(self, children: Sequence[Expr]) -> Sequence: - return (*(getattr(self, attr) for attr in self._non_child), *children) - - def get_hash(self) -> int: - """ - Return the hash of this expr. - - Override this in subclasses, rather than __hash__. - - Returns - ------- - The integer hash value. - """ - return hash((type(self), self._ctor_arguments(self.children))) - - def __hash__(self) -> int: - """Hash of an expression with caching.""" - try: - return self._hash_value - except AttributeError: - self._hash_value = self.get_hash() - return self._hash_value - - def is_equal(self, other: Any) -> bool: - """ - Equality of two expressions. - - Override this in subclasses, rather than __eq__. - - Parameter - --------- - other - object to compare to - - Returns - ------- - True if the two expressions are equal, false otherwise. - """ - if type(self) is not type(other): - return False # pragma: no cover; __eq__ trips first - return self._ctor_arguments(self.children) == other._ctor_arguments( - other.children - ) - - def __eq__(self, other: Any) -> bool: - """Equality of expressions.""" - if type(self) is not type(other) or hash(self) != hash(other): - return False - else: - return self.is_equal(other) - - def __ne__(self, other: Any) -> bool: - """Inequality of expressions.""" - return not self.__eq__(other) - - def __repr__(self) -> str: - """String representation of an expression with caching.""" - try: - return self._repr_value - except AttributeError: - args = ", ".join(f"{arg!r}" for arg in self._ctor_arguments(self.children)) - self._repr_value = f"{type(self).__name__}({args})" - return self._repr_value - def do_evaluate( self, df: DataFrame, @@ -311,11 +228,11 @@ class Col(Expr): __slots__ = ("name",) _non_child = ("dtype", "name") name: str - children: tuple[()] def __init__(self, dtype: plc.DataType, name: str) -> None: self.dtype = dtype self.name = name + self.children = () def do_evaluate( self, diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/binaryop.py b/python/cudf_polars/cudf_polars/dsl/expressions/binaryop.py index 19baae3611d..11a47e7ea51 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/binaryop.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/binaryop.py @@ -24,9 +24,8 @@ class BinOp(Expr): - __slots__ = ("op", "children") + __slots__ = ("op",) _non_child = ("dtype", "op") - children: tuple[Expr, Expr] def __init__( self, @@ -35,7 +34,7 @@ def __init__( left: Expr, right: Expr, ) -> None: - super().__init__(dtype) + self.dtype = dtype if plc.traits.is_boolean(self.dtype): # For boolean output types, bitand and bitor implement # boolean logic, so translate. bitxor also does, but the diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/boolean.py b/python/cudf_polars/cudf_polars/dsl/expressions/boolean.py index ff9973a47d5..9c14a8386f3 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/boolean.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/boolean.py @@ -31,9 +31,8 @@ class BooleanFunction(Expr): - __slots__ = ("name", "options", "children") + __slots__ = ("name", "options") _non_child = ("dtype", "name", "options") - children: tuple[Expr, ...] def __init__( self, @@ -42,7 +41,7 @@ def __init__( options: tuple[Any, ...], *children: Expr, ) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.name = name self.children = children diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/datetime.py b/python/cudf_polars/cudf_polars/dsl/expressions/datetime.py index f752a23b628..596e193d8fe 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/datetime.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/datetime.py @@ -25,7 +25,7 @@ class TemporalFunction(Expr): - __slots__ = ("name", "options", "children") + __slots__ = ("name", "options") _COMPONENT_MAP: ClassVar[dict[pl_expr.TemporalFunction, str]] = { pl_expr.TemporalFunction.Year: plc.datetime.DatetimeComponent.YEAR, pl_expr.TemporalFunction.Month: plc.datetime.DatetimeComponent.MONTH, @@ -39,7 +39,6 @@ class TemporalFunction(Expr): pl_expr.TemporalFunction.Nanosecond: plc.datetime.DatetimeComponent.NANOSECOND, } _non_child = ("dtype", "name", "options") - children: tuple[Expr, ...] def __init__( self, @@ -48,7 +47,7 @@ def __init__( options: tuple[Any, ...], *children: Expr, ) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.name = name self.children = children diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/literal.py b/python/cudf_polars/cudf_polars/dsl/expressions/literal.py index 562a2255033..c8aa993b994 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/literal.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/literal.py @@ -16,7 +16,7 @@ from cudf_polars.utils import dtypes if TYPE_CHECKING: - from collections.abc import Mapping + from collections.abc import Hashable, Mapping import pyarrow as pa @@ -31,12 +31,12 @@ class Literal(Expr): __slots__ = ("value",) _non_child = ("dtype", "value") value: pa.Scalar[Any] - children: tuple[()] def __init__(self, dtype: plc.DataType, value: pa.Scalar[Any]) -> None: - super().__init__(dtype) + self.dtype = dtype assert value.type == plc.interop.to_arrow(dtype) self.value = value + self.children = () def do_evaluate( self, @@ -58,19 +58,19 @@ class LiteralColumn(Expr): __slots__ = ("value",) _non_child = ("dtype", "value") value: pa.Array[Any, Any] - children: tuple[()] def __init__(self, dtype: plc.DataType, value: pl.Series) -> None: - super().__init__(dtype) + self.dtype = dtype data = value.to_arrow() self.value = data.cast(dtypes.downcast_arrow_lists(data.type)) + self.children = () - def get_hash(self) -> int: + def get_hashable(self) -> Hashable: """Compute a hash of the column.""" # This is stricter than necessary, but we only need this hash # for identity in groupby replacements so it's OK. And this # way we avoid doing potentially expensive compute. - return hash((type(self), self.dtype, id(self.value))) + return (type(self), self.dtype, id(self.value)) def do_evaluate( self, diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/rolling.py b/python/cudf_polars/cudf_polars/dsl/expressions/rolling.py index f7dcc3c542c..fa68bcb9426 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/rolling.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/rolling.py @@ -17,24 +17,22 @@ class RollingWindow(Expr): - __slots__ = ("options", "children") + __slots__ = ("options",) _non_child = ("dtype", "options") - children: tuple[Expr] def __init__(self, dtype: plc.DataType, options: Any, agg: Expr) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.children = (agg,) raise NotImplementedError("Rolling window not implemented") class GroupedRollingWindow(Expr): - __slots__ = ("options", "children") + __slots__ = ("options",) _non_child = ("dtype", "options") - children: tuple[Expr, ...] def __init__(self, dtype: plc.DataType, options: Any, agg: Expr, *by: Expr) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.children = (agg, *by) raise NotImplementedError("Grouped rolling window not implemented") diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/selection.py b/python/cudf_polars/cudf_polars/dsl/expressions/selection.py index a7a3e68a28c..0247256e507 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/selection.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/selection.py @@ -23,12 +23,11 @@ class Gather(Expr): - __slots__ = ("children",) + __slots__ = () _non_child = ("dtype",) - children: tuple[Expr, Expr] def __init__(self, dtype: plc.DataType, values: Expr, indices: Expr) -> None: - super().__init__(dtype) + self.dtype = dtype self.children = (values, indices) def do_evaluate( @@ -65,12 +64,11 @@ def do_evaluate( class Filter(Expr): - __slots__ = ("children",) + __slots__ = () _non_child = ("dtype",) - children: tuple[Expr, Expr] def __init__(self, dtype: plc.DataType, values: Expr, indices: Expr): - super().__init__(dtype) + self.dtype = dtype self.children = (values, indices) def do_evaluate( diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/sorting.py b/python/cudf_polars/cudf_polars/dsl/expressions/sorting.py index 861b73ce6a0..99512e2ef52 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/sorting.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/sorting.py @@ -23,14 +23,13 @@ class Sort(Expr): - __slots__ = ("options", "children") + __slots__ = ("options",) _non_child = ("dtype", "options") - children: tuple[Expr] def __init__( self, dtype: plc.DataType, options: tuple[bool, bool, bool], column: Expr ) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.children = (column,) @@ -59,9 +58,8 @@ def do_evaluate( class SortBy(Expr): - __slots__ = ("options", "children") + __slots__ = ("options",) _non_child = ("dtype", "options") - children: tuple[Expr, ...] def __init__( self, @@ -70,7 +68,7 @@ def __init__( column: Expr, *by: Expr, ) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.children = (column, *by) diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/string.py b/python/cudf_polars/cudf_polars/dsl/expressions/string.py index 6669669aadc..62b54c63a8d 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/string.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/string.py @@ -28,9 +28,8 @@ class StringFunction(Expr): - __slots__ = ("name", "options", "children", "_regex_program") + __slots__ = ("name", "options", "_regex_program") _non_child = ("dtype", "name", "options") - children: tuple[Expr, ...] def __init__( self, @@ -39,7 +38,7 @@ def __init__( options: tuple[Any, ...], *children: Expr, ) -> None: - super().__init__(dtype) + self.dtype = dtype self.options = options self.name = name self.children = children diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/ternary.py b/python/cudf_polars/cudf_polars/dsl/expressions/ternary.py index c7d7a802ded..d2b5d6bae29 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/ternary.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/ternary.py @@ -26,14 +26,13 @@ class Ternary(Expr): - __slots__ = ("children",) + __slots__ = () _non_child = ("dtype",) - children: tuple[Expr, Expr, Expr] def __init__( self, dtype: plc.DataType, when: Expr, then: Expr, otherwise: Expr ) -> None: - super().__init__(dtype) + self.dtype = dtype self.children = (when, then, otherwise) def do_evaluate( diff --git a/python/cudf_polars/cudf_polars/dsl/expressions/unary.py b/python/cudf_polars/cudf_polars/dsl/expressions/unary.py index 3d4d15be1ce..53f6ed29239 100644 --- a/python/cudf_polars/cudf_polars/dsl/expressions/unary.py +++ b/python/cudf_polars/cudf_polars/dsl/expressions/unary.py @@ -26,12 +26,11 @@ class Cast(Expr): """Class representing a cast of an expression.""" - __slots__ = ("children",) + __slots__ = () _non_child = ("dtype",) - children: tuple[Expr] def __init__(self, dtype: plc.DataType, value: Expr) -> None: - super().__init__(dtype) + self.dtype = dtype self.children = (value,) if not dtypes.can_cast(value.dtype, self.dtype): raise NotImplementedError( @@ -60,7 +59,9 @@ def collect_agg(self, *, depth: int) -> AggInfo: class Len(Expr): """Class representing the length of an expression.""" - children: tuple[()] + def __init__(self, dtype: plc.DataType) -> None: + self.dtype = dtype + self.children = () def do_evaluate( self, @@ -90,9 +91,8 @@ def collect_agg(self, *, depth: int) -> AggInfo: class UnaryFunction(Expr): """Class representing unary functions of an expression.""" - __slots__ = ("name", "options", "children") + __slots__ = ("name", "options") _non_child = ("dtype", "name", "options") - children: tuple[Expr, ...] # Note: log, and pow are handled via translation to binops _OP_MAPPING: ClassVar[dict[str, plc.unary.UnaryOperator]] = { @@ -142,7 +142,7 @@ class UnaryFunction(Expr): def __init__( self, dtype: plc.DataType, name: str, options: tuple[Any, ...], *children: Expr ) -> None: - super().__init__(dtype) + self.dtype = dtype self.name = name self.options = options self.children = children diff --git a/python/cudf_polars/cudf_polars/dsl/ir.py b/python/cudf_polars/cudf_polars/dsl/ir.py index e319c363a23..f79e229d3f3 100644 --- a/python/cudf_polars/cudf_polars/dsl/ir.py +++ b/python/cudf_polars/cudf_polars/dsl/ir.py @@ -13,8 +13,8 @@ from __future__ import annotations -import dataclasses import itertools +import json from functools import cache from pathlib import Path from typing import TYPE_CHECKING, Any, ClassVar @@ -27,10 +27,11 @@ import cudf_polars.dsl.expr as expr from cudf_polars.containers import Column, DataFrame -from cudf_polars.utils import dtypes, sorting +from cudf_polars.dsl.nodebase import Node +from cudf_polars.utils import dtypes if TYPE_CHECKING: - from collections.abc import Callable, MutableMapping + from collections.abc import Callable, Hashable, MutableMapping, Sequence from typing import Literal from cudf_polars.typing import Schema @@ -121,16 +122,27 @@ def broadcast(*columns: Column, target_length: int | None = None) -> list[Column ] -@dataclasses.dataclass -class IR: +class IR(Node["IR"]): """Abstract plan node, representing an unevaluated dataframe.""" + __slots__ = ("schema",) + # This annotation is needed because of https://github.com/python/mypy/issues/17981 + _non_child: ClassVar[tuple[str, ...]] = ("schema",) schema: Schema """Mapping from column names to their data types.""" - def __post_init__(self): - """Validate preconditions.""" - pass # noqa: PIE790 + def get_hashable(self) -> Hashable: + """ + Hashable representation of node, treating schema dictionary. + + Since the schema is a dictionary, even though it is morally + immutable, it is not hashable. We therefore convert it to + tuples for hashing purposes. + """ + # Schema is the first constructor argument + args = self._ctor_arguments(self.children)[1:] + schema_hash = tuple(self.schema.items()) + return (type(self), schema_hash, args) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """ @@ -159,24 +171,50 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: ) # pragma: no cover -@dataclasses.dataclass class PythonScan(IR): """Representation of input from a python function.""" + __slots__ = ("options", "predicate") + _non_child = ("schema", "options", "predicate") options: Any """Arbitrary options.""" predicate: expr.NamedExpr | None """Filter to apply to the constructed dataframe before returning it.""" - def __post_init__(self): - """Validate preconditions.""" + def __init__(self, schema: Schema, options: Any, predicate: expr.NamedExpr | None): + self.schema = schema + self.options = options + self.predicate = predicate + self.children = () raise NotImplementedError("PythonScan not implemented") -@dataclasses.dataclass class Scan(IR): """Input from files.""" + __slots__ = ( + "typ", + "reader_options", + "cloud_options", + "paths", + "with_columns", + "skip_rows", + "n_rows", + "row_index", + "predicate", + ) + _non_child = ( + "schema", + "typ", + "reader_options", + "cloud_options", + "paths", + "with_columns", + "skip_rows", + "n_rows", + "row_index", + "predicate", + ) typ: str """What type of file are we reading? Parquet, CSV, etc...""" reader_options: dict[str, Any] @@ -185,7 +223,7 @@ class Scan(IR): """Cloud-related authentication options, currently ignored.""" paths: list[str] """List of paths to read from.""" - with_columns: list[str] + with_columns: list[str] | None """Projected columns to return.""" skip_rows: int """Rows to skip at the start when reading.""" @@ -196,9 +234,30 @@ class Scan(IR): predicate: expr.NamedExpr | None """Mask to apply to the read dataframe.""" - def __post_init__(self) -> None: - """Validate preconditions.""" - super().__post_init__() + def __init__( + self, + schema: Schema, + typ: str, + reader_options: dict[str, Any], + cloud_options: dict[str, Any] | None, + paths: list[str], + with_columns: list[str] | None, + skip_rows: int, + n_rows: int, + row_index: tuple[str, int] | None, + predicate: expr.NamedExpr | None, + ): + self.schema = schema + self.typ = typ + self.reader_options = reader_options + self.cloud_options = cloud_options + self.paths = paths + self.with_columns = with_columns + self.skip_rows = skip_rows + self.n_rows = n_rows + self.row_index = row_index + self.predicate = predicate + self.children = () if self.typ not in ("csv", "parquet", "ndjson"): # pragma: no cover # This line is unhittable ATM since IPC/Anonymous scan raise # on the polars side @@ -258,6 +317,28 @@ def __post_init__(self) -> None: "Reading only parquet metadata to produce row index." ) + def get_hashable(self) -> Hashable: + """ + Hashable representation of the node. + + The options dictionaries are serialised for hashing purposes + as json strings. + """ + schema_hash = tuple(self.schema.items()) + return ( + type(self), + schema_hash, + self.typ, + json.dumps(self.reader_options), + json.dumps(self.cloud_options), + tuple(self.paths), + tuple(self.with_columns) if self.with_columns is not None else None, + self.skip_rows, + self.n_rows, + self.row_index, + self.predicate, + ) + def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" with_columns = self.with_columns @@ -401,7 +482,6 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return df.filter(mask) -@dataclasses.dataclass class Cache(IR): """ Return a cached plan node. @@ -409,20 +489,25 @@ class Cache(IR): Used for CSE at the plan level. """ + __slots__ = ("key",) + _non_child = ("schema", "key") key: int """The cache key.""" - value: IR - """The unevaluated node to cache.""" + + def __init__(self, schema: Schema, key: int, value: IR): + self.schema = schema + self.key = key + self.children = (value,) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" try: return cache[self.key] except KeyError: - return cache.setdefault(self.key, self.value.evaluate(cache=cache)) + (value,) = self.children + return cache.setdefault(self.key, value.evaluate(cache=cache)) -@dataclasses.dataclass class DataFrameScan(IR): """ Input from an existing polars DataFrame. @@ -430,13 +515,38 @@ class DataFrameScan(IR): This typically arises from ``q.collect().lazy()`` """ + __slots__ = ("df", "projection", "predicate") + _non_child = ("schema", "df", "projection", "predicate") df: Any """Polars LazyFrame object.""" - projection: list[str] + projection: tuple[str, ...] | None """List of columns to project out.""" predicate: expr.NamedExpr | None """Mask to apply.""" + def __init__( + self, + schema: Schema, + df: Any, + projection: Sequence[str] | None, + predicate: expr.NamedExpr | None, + ): + self.schema = schema + self.df = df + self.projection = tuple(projection) if projection is not None else None + self.predicate = predicate + self.children = () + + def get_hashable(self) -> Hashable: + """ + Hashable representation of the node. + + The (heavy) dataframe object is hashed as its id, so this is + not stable across runs, or repeat instances of the same equal dataframes. + """ + schema_hash = tuple(self.schema.items()) + return (type(self), schema_hash, id(self.df), self.projection, self.predicate) + def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" pdf = pl.DataFrame._from_pydf(self.df) @@ -454,28 +564,39 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return df -@dataclasses.dataclass class Select(IR): """Produce a new dataframe selecting given expressions from an input.""" - df: IR - """Input dataframe.""" - expr: list[expr.NamedExpr] + __slots__ = ("exprs", "should_broadcast") + _non_child = ("schema", "exprs", "should_broadcast") + exprs: tuple[expr.NamedExpr, ...] """List of expressions to evaluate to form the new dataframe.""" should_broadcast: bool """Should columns be broadcast?""" + def __init__( + self, + schema: Schema, + exprs: Sequence[expr.NamedExpr], + should_broadcast: bool, # noqa: FBT001 + df: IR, + ): + self.schema = schema + self.exprs = tuple(exprs) + self.should_broadcast = should_broadcast + self.children = (df,) + def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) # Handle any broadcasting - columns = [e.evaluate(df) for e in self.expr] + columns = [e.evaluate(df) for e in self.exprs] if self.should_broadcast: columns = broadcast(*columns) return DataFrame(columns) -@dataclasses.dataclass class Reduce(IR): """ Produce a new dataframe selecting given expressions from an input. @@ -483,36 +604,73 @@ class Reduce(IR): This is a special case of :class:`Select` where all outputs are a single row. """ - df: IR - """Input dataframe.""" - expr: list[expr.NamedExpr] + __slots__ = ("exprs",) + _non_child = ("schema", "exprs") + exprs: tuple[expr.NamedExpr, ...] """List of expressions to evaluate to form the new dataframe.""" + def __init__( + self, schema: Schema, exprs: Sequence[expr.NamedExpr], df: IR + ): # pragma: no cover; polars doesn't emit this node yet + self.schema = schema + self.exprs = tuple(exprs) + self.children = (df,) + def evaluate( self, *, cache: MutableMapping[int, DataFrame] ) -> DataFrame: # pragma: no cover; polars doesn't emit this node yet """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) - columns = broadcast(*(e.evaluate(df) for e in self.expr)) + (child,) = self.children + df = child.evaluate(cache=cache) + columns = broadcast(*(e.evaluate(df) for e in self.exprs)) assert all(column.obj.size() == 1 for column in columns) return DataFrame(columns) -@dataclasses.dataclass class GroupBy(IR): """Perform a groupby.""" - df: IR - """Input dataframe.""" - agg_requests: list[expr.NamedExpr] - """List of expressions to evaluate groupwise.""" - keys: list[expr.NamedExpr] - """List of expressions forming the keys.""" + __slots__ = ( + "agg_requests", + "keys", + "maintain_order", + "options", + "agg_infos", + ) + _non_child = ("schema", "keys", "agg_requests", "maintain_order", "options") + keys: tuple[expr.NamedExpr, ...] + """Grouping keys.""" + agg_requests: tuple[expr.NamedExpr, ...] + """Aggregation expressions.""" maintain_order: bool - """Should the order of the input dataframe be maintained?""" + """Preserve order in groupby.""" options: Any - """Options controlling style of groupby.""" - agg_infos: list[expr.AggInfo] = dataclasses.field(init=False) + """Arbitrary options.""" + + def __init__( + self, + schema: Schema, + keys: Sequence[expr.NamedExpr], + agg_requests: Sequence[expr.NamedExpr], + maintain_order: bool, # noqa: FBT001 + options: Any, + df: IR, + ): + self.schema = schema + self.keys = tuple(keys) + self.agg_requests = tuple(agg_requests) + self.maintain_order = maintain_order + self.options = options + self.children = (df,) + if self.options.rolling: + raise NotImplementedError( + "rolling window/groupby" + ) # pragma: no cover; rollingwindow constructor has already raised + if self.options.dynamic: + raise NotImplementedError("dynamic group by") + if any(GroupBy.check_agg(a.value) > 1 for a in self.agg_requests): + raise NotImplementedError("Nested aggregations in groupby") + self.agg_infos = [req.collect_agg(depth=0) for req in self.agg_requests] @staticmethod def check_agg(agg: expr.Expr) -> int: @@ -542,22 +700,10 @@ def check_agg(agg: expr.Expr) -> int: else: raise NotImplementedError(f"No handler for {agg=}") - def __post_init__(self) -> None: - """Check whether all the aggregations are implemented.""" - super().__post_init__() - if self.options.rolling: - raise NotImplementedError( - "rolling window/groupby" - ) # pragma: no cover; rollingwindow constructor has already raised - if any(GroupBy.check_agg(a.value) > 1 for a in self.agg_requests): - raise NotImplementedError("Nested aggregations in groupby") - self.agg_infos = [req.collect_agg(depth=0) for req in self.agg_requests] - if len(self.keys) == 0: - raise NotImplementedError("dynamic groupby") - def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) keys = broadcast( *(k.evaluate(df) for k in self.keys), target_length=df.num_rows ) @@ -646,23 +792,20 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return DataFrame(broadcasted).slice(self.options.slice) -@dataclasses.dataclass class Join(IR): """A join of two dataframes.""" - left: IR - """Left frame.""" - right: IR - """Right frame.""" - left_on: list[expr.NamedExpr] + __slots__ = ("left_on", "right_on", "options") + _non_child = ("schema", "left_on", "right_on", "options") + left_on: tuple[expr.NamedExpr, ...] """List of expressions used as keys in the left frame.""" - right_on: list[expr.NamedExpr] + right_on: tuple[expr.NamedExpr, ...] """List of expressions used as keys in the right frame.""" options: tuple[ - Literal["inner", "left", "right", "full", "leftsemi", "leftanti", "cross"], + Literal["inner", "left", "right", "full", "semi", "anti", "cross"], bool, tuple[int, int] | None, - str | None, + str, bool, ] """ @@ -674,9 +817,20 @@ class Join(IR): - coalesce: should key columns be coalesced (only makes sense for outer joins) """ - def __post_init__(self) -> None: - """Validate preconditions.""" - super().__post_init__() + def __init__( + self, + schema: Schema, + left_on: Sequence[expr.NamedExpr], + right_on: Sequence[expr.NamedExpr], + options: Any, + left: IR, + right: IR, + ): + self.schema = schema + self.left_on = tuple(left_on) + self.right_on = tuple(right_on) + self.options = options + self.children = (left, right) if any( isinstance(e.value, expr.Literal) for e in itertools.chain(self.left_on, self.right_on) @@ -686,7 +840,7 @@ def __post_init__(self) -> None: @staticmethod @cache def _joiners( - how: Literal["inner", "left", "right", "full", "leftsemi", "leftanti"], + how: Literal["inner", "left", "right", "full", "semi", "anti"], ) -> tuple[ Callable, plc.copying.OutOfBoundsPolicy, plc.copying.OutOfBoundsPolicy | None ]: @@ -708,13 +862,13 @@ def _joiners( plc.copying.OutOfBoundsPolicy.NULLIFY, plc.copying.OutOfBoundsPolicy.NULLIFY, ) - elif how == "leftsemi": + elif how == "semi": return ( plc.join.left_semi_join, plc.copying.OutOfBoundsPolicy.DONT_CHECK, None, ) - elif how == "leftanti": + elif how == "anti": return ( plc.join.left_anti_join, plc.copying.OutOfBoundsPolicy.DONT_CHECK, @@ -777,10 +931,8 @@ def _reorder_maps( def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - left = self.left.evaluate(cache=cache) - right = self.right.evaluate(cache=cache) + left, right = (c.evaluate(cache=cache) for c in self.children) how, join_nulls, zlice, suffix, coalesce = self.options - suffix = "_right" if suffix is None else suffix if how == "cross": # Separate implementation, since cross_join returns the # result, not the gather maps @@ -802,7 +954,7 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: columns[left.num_columns :], right.column_names, strict=True ) ] - return DataFrame([*left_cols, *right_cols]) + return DataFrame([*left_cols, *right_cols]).slice(zlice) # TODO: Waiting on clarity based on https://github.com/pola-rs/polars/issues/17184 left_on = DataFrame(broadcast(*(e.evaluate(left) for e in self.left_on))) right_on = DataFrame(broadcast(*(e.evaluate(right) for e in self.right_on))) @@ -866,20 +1018,30 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return result.slice(zlice) -@dataclasses.dataclass class HStack(IR): """Add new columns to a dataframe.""" - df: IR - """Input dataframe.""" - columns: list[expr.NamedExpr] - """List of expressions to produce new columns.""" + __slots__ = ("columns", "should_broadcast") + _non_child = ("schema", "columns", "should_broadcast") should_broadcast: bool - """Should columns be broadcast?""" + """Should the resulting evaluated columns be broadcast to the same length.""" + + def __init__( + self, + schema: Schema, + columns: Sequence[expr.NamedExpr], + should_broadcast: bool, # noqa: FBT001 + df: IR, + ): + self.schema = schema + self.columns = tuple(columns) + self.should_broadcast = should_broadcast + self.children = (df,) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) columns = [c.evaluate(df) for c in self.columns] if self.should_broadcast: columns = broadcast(*columns, target_length=df.num_rows) @@ -895,20 +1057,36 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return df.with_columns(columns) -@dataclasses.dataclass class Distinct(IR): """Produce a new dataframe with distinct rows.""" - df: IR - """Input dataframe.""" + __slots__ = ("keep", "subset", "zlice", "stable") + _non_child = ("schema", "keep", "subset", "zlice", "stable") keep: plc.stream_compaction.DuplicateKeepOption - """Which rows to keep.""" - subset: set[str] | None - """Which columns to inspect when computing distinct rows.""" + """Which distinct value to keep.""" + subset: frozenset[str] | None + """Which columns should be used to define distinctness. If None, + then all columns are used.""" zlice: tuple[int, int] | None - """Optional slice to perform after compaction.""" + """Optional slice to apply to the result.""" stable: bool - """Should order be preserved?""" + """Should the result maintain ordering.""" + + def __init__( + self, + schema: Schema, + keep: plc.stream_compaction.DuplicateKeepOption, + subset: frozenset[str] | None, + zlice: tuple[int, int] | None, + stable: bool, # noqa: FBT001 + df: IR, + ): + self.schema = schema + self.keep = keep + self.subset = subset + self.zlice = zlice + self.stable = stable + self.children = (df,) _KEEP_MAP: ClassVar[dict[str, plc.stream_compaction.DuplicateKeepOption]] = { "first": plc.stream_compaction.DuplicateKeepOption.KEEP_FIRST, @@ -917,18 +1095,10 @@ class Distinct(IR): "any": plc.stream_compaction.DuplicateKeepOption.KEEP_ANY, } - def __init__(self, schema: Schema, df: IR, options: Any) -> None: - self.schema = schema - self.df = df - (keep, subset, maintain_order, zlice) = options - self.keep = Distinct._KEEP_MAP[keep] - self.subset = set(subset) if subset is not None else None - self.stable = maintain_order - self.zlice = zlice - def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) if self.subset is None: indices = list(range(df.num_columns)) keys_sorted = all(c.is_sorted for c in df.column_map.values()) @@ -967,46 +1137,44 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return result.slice(self.zlice) -@dataclasses.dataclass class Sort(IR): """Sort a dataframe.""" - df: IR - """Input.""" - by: list[expr.NamedExpr] - """List of expressions to produce sort keys.""" - do_sort: Callable[..., plc.Table] - """pylibcudf sorting function.""" + __slots__ = ("by", "order", "null_order", "stable", "zlice") + _non_child = ("schema", "by", "order", "null_order", "stable", "zlice") + by: tuple[expr.NamedExpr, ...] + """Sort keys.""" + order: tuple[plc.types.Order, ...] + """Sort order for each sort key.""" + null_order: tuple[plc.types.NullOrder, ...] + """Null sorting location for each sort key.""" + stable: bool + """Should the sort be stable?""" zlice: tuple[int, int] | None - """Optional slice to apply after sorting.""" - order: list[plc.types.Order] - """Order keys should be sorted in.""" - null_order: list[plc.types.NullOrder] - """Where nulls sort to.""" + """Optional slice to apply to the result.""" def __init__( self, schema: Schema, - df: IR, - by: list[expr.NamedExpr], - options: Any, + by: Sequence[expr.NamedExpr], + order: Sequence[plc.types.Order], + null_order: Sequence[plc.types.NullOrder], + stable: bool, # noqa: FBT001 zlice: tuple[int, int] | None, - ) -> None: + df: IR, + ): self.schema = schema - self.df = df - self.by = by + self.by = tuple(by) + self.order = tuple(order) + self.null_order = tuple(null_order) + self.stable = stable self.zlice = zlice - stable, nulls_last, descending = options - self.order, self.null_order = sorting.sort_order( - descending, nulls_last=nulls_last, num_keys=len(by) - ) - self.do_sort = ( - plc.sorting.stable_sort_by_key if stable else plc.sorting.sort_by_key - ) + self.children = (df,) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) sort_keys = broadcast( *(k.evaluate(df) for k in self.by), target_length=df.num_rows ) @@ -1016,11 +1184,14 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: for i, k in enumerate(sort_keys) if k.name in df.column_map and k.obj is df.column_map[k.name].obj } - table = self.do_sort( + do_sort = ( + plc.sorting.stable_sort_by_key if self.stable else plc.sorting.sort_by_key + ) + table = do_sort( df.table, plc.Table([k.obj for k in sort_keys]), - self.order, - self.null_order, + list(self.order), + list(self.null_order), ) columns: list[Column] = [] for name, c in zip(df.column_map, table.columns(), strict=True): @@ -1037,49 +1208,64 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return DataFrame(columns).slice(self.zlice) -@dataclasses.dataclass class Slice(IR): """Slice a dataframe.""" - df: IR - """Input.""" + __slots__ = ("offset", "length") + _non_child = ("schema", "offset", "length") offset: int """Start of the slice.""" length: int """Length of the slice.""" + def __init__(self, schema: Schema, offset: int, length: int, df: IR): + self.schema = schema + self.offset = offset + self.length = length + self.children = (df,) + def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) return df.slice((self.offset, self.length)) -@dataclasses.dataclass class Filter(IR): """Filter a dataframe with a boolean mask.""" - df: IR - """Input.""" + __slots__ = ("mask",) + _non_child = ("schema", "mask") mask: expr.NamedExpr - """Expression evaluating to a mask.""" + """Expression to produce the filter mask.""" + + def __init__(self, schema: Schema, mask: expr.NamedExpr, df: IR): + self.schema = schema + self.mask = mask + self.children = (df,) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) (mask,) = broadcast(self.mask.evaluate(df), target_length=df.num_rows) return df.filter(mask) -@dataclasses.dataclass class Projection(IR): """Select a subset of columns from a dataframe.""" - df: IR - """Input.""" + __slots__ = () + _non_child = ("schema",) + + def __init__(self, schema: Schema, df: IR): + self.schema = schema + self.children = (df,) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - df = self.df.evaluate(cache=cache) + (child,) = self.children + df = child.evaluate(cache=cache) # This can reorder things. columns = broadcast( *(df.column_map[name] for name in self.schema), target_length=df.num_rows @@ -1087,16 +1273,15 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: return DataFrame(columns) -@dataclasses.dataclass class MapFunction(IR): """Apply some function to a dataframe.""" - df: IR - """Input.""" + __slots__ = ("name", "options") + _non_child = ("schema", "name", "options") name: str - """Function name.""" + """Name of the function to apply""" options: Any - """Arbitrary options, interpreted per function.""" + """Arbitrary name-specific options""" _NAMES: ClassVar[frozenset[str]] = frozenset( [ @@ -1111,9 +1296,11 @@ class MapFunction(IR): ] ) - def __post_init__(self) -> None: - """Validate preconditions.""" - super().__post_init__() + def __init__(self, schema: Schema, name: str, options: Any, df: IR): + self.schema = schema + self.name = name + self.options = options + self.children = (df,) if self.name not in MapFunction._NAMES: raise NotImplementedError(f"Unhandled map function {self.name}") if self.name == "explode": @@ -1127,7 +1314,7 @@ def __post_init__(self) -> None: old, new, _ = self.options # TODO: perhaps polars should validate renaming in the IR? if len(new) != len(set(new)) or ( - set(new) & (set(self.df.schema.keys()) - set(old)) + set(new) & (set(df.schema.keys()) - set(old)) ): raise NotImplementedError("Duplicate new names in rename.") elif self.name == "unpivot": @@ -1136,31 +1323,31 @@ def __post_init__(self) -> None: variable_name = "variable" if variable_name is None else variable_name if len(pivotees) == 0: index = frozenset(indices) - pivotees = [name for name in self.df.schema if name not in index] + pivotees = [name for name in df.schema if name not in index] if not all( - dtypes.can_cast(self.df.schema[p], self.schema[value_name]) - for p in pivotees + dtypes.can_cast(df.schema[p], self.schema[value_name]) for p in pivotees ): raise NotImplementedError( "Unpivot cannot cast all input columns to " f"{self.schema[value_name].id()}" ) - self.options = (indices, pivotees, variable_name, value_name) + self.options = (tuple(indices), tuple(pivotees), variable_name, value_name) def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" + (child,) = self.children if self.name == "rechunk": # No-op in our data model # Don't think this appears in a plan tree from python - return self.df.evaluate(cache=cache) # pragma: no cover + return child.evaluate(cache=cache) # pragma: no cover elif self.name == "rename": - df = self.df.evaluate(cache=cache) + df = child.evaluate(cache=cache) # final tag is "swapping" which is useful for the # optimiser (it blocks some pushdown operations) old, new, _ = self.options return df.rename_columns(dict(zip(old, new, strict=True))) elif self.name == "explode": - df = self.df.evaluate(cache=cache) + df = child.evaluate(cache=cache) ((to_explode,),) = self.options index = df.column_names.index(to_explode) subset = df.column_names_set - {to_explode} @@ -1170,7 +1357,7 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: elif self.name == "unpivot": indices, pivotees, variable_name, value_name = self.options npiv = len(pivotees) - df = self.df.evaluate(cache=cache) + df = child.evaluate(cache=cache) index_columns = [ Column(col, name=name) for col, name in zip( @@ -1209,37 +1396,40 @@ def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: raise AssertionError("Should never be reached") # pragma: no cover -@dataclasses.dataclass class Union(IR): """Concatenate dataframes vertically.""" - dfs: list[IR] - """List of inputs.""" + __slots__ = ("zlice",) + _non_child = ("schema", "zlice") zlice: tuple[int, int] | None - """Optional slice to apply after concatenation.""" + """Optional slice to apply to the result.""" - def __post_init__(self) -> None: - """Validate preconditions.""" - super().__post_init__() - schema = self.dfs[0].schema - if not all(s.schema == schema for s in self.dfs[1:]): + def __init__(self, schema: Schema, zlice: tuple[int, int] | None, *children: IR): + self.schema = schema + self.zlice = zlice + self.children = children + schema = self.children[0].schema + if not all(s.schema == schema for s in self.children[1:]): raise NotImplementedError("Schema mismatch") def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" # TODO: only evaluate what we need if we have a slice - dfs = [df.evaluate(cache=cache) for df in self.dfs] + dfs = [df.evaluate(cache=cache) for df in self.children] return DataFrame.from_table( plc.concatenate.concatenate([df.table for df in dfs]), dfs[0].column_names ).slice(self.zlice) -@dataclasses.dataclass class HConcat(IR): """Concatenate dataframes horizontally.""" - dfs: list[IR] - """List of inputs.""" + __slots__ = () + _non_child = ("schema",) + + def __init__(self, schema: Schema, *children: IR): + self.schema = schema + self.children = children @staticmethod def _extend_with_nulls(table: plc.Table, *, nrows: int) -> plc.Table: @@ -1271,7 +1461,7 @@ def _extend_with_nulls(table: plc.Table, *, nrows: int) -> plc.Table: def evaluate(self, *, cache: MutableMapping[int, DataFrame]) -> DataFrame: """Evaluate and return a dataframe.""" - dfs = [df.evaluate(cache=cache) for df in self.dfs] + dfs = [df.evaluate(cache=cache) for df in self.children] max_rows = max(df.num_rows for df in dfs) # Horizontal concatenation extends shorter tables with nulls dfs = [ diff --git a/python/cudf_polars/cudf_polars/dsl/nodebase.py b/python/cudf_polars/cudf_polars/dsl/nodebase.py new file mode 100644 index 00000000000..228d300f467 --- /dev/null +++ b/python/cudf_polars/cudf_polars/dsl/nodebase.py @@ -0,0 +1,152 @@ +# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. +# SPDX-License-Identifier: Apache-2.0 + +"""Base class for IR nodes, and utilities.""" + +from __future__ import annotations + +from typing import TYPE_CHECKING, Any, ClassVar, Generic, TypeVar + +if TYPE_CHECKING: + from collections.abc import Hashable, Sequence + + from typing_extensions import Self + + +__all__: list[str] = ["Node"] + +T = TypeVar("T", bound="Node[Any]") + + +class Node(Generic[T]): + """ + An abstract node type. + + Nodes are immutable! + + This contains a (potentially empty) tuple of child nodes, + along with non-child data. For uniform reconstruction and + implementation of hashing and equality schemes, child classes need + to provide a certain amount of metadata when they are defined. + Specifically, the ``_non_child`` attribute must list, in-order, + the names of the slots that are passed to the constructor. The + constructor must take arguments in the order ``(*_non_child, + *children).`` + """ + + __slots__ = ("_hash_value", "_repr_value", "children") + _hash_value: int + _repr_value: str + children: tuple[T, ...] + _non_child: ClassVar[tuple[str, ...]] = () + + def _ctor_arguments(self, children: Sequence[T]) -> Sequence[Any | T]: + return (*(getattr(self, attr) for attr in self._non_child), *children) + + def reconstruct( + self, children: Sequence[T] + ) -> Self: # pragma: no cover; not yet used + """ + Rebuild this node with new children. + + Parameters + ---------- + children + New children + + Returns + ------- + New node with new children. Non-child data is shared with the input. + """ + return type(self)(*self._ctor_arguments(children)) + + def get_hashable(self) -> Hashable: + """ + Return a hashable object for the node. + + Returns + ------- + Hashable object. + + Notes + ----- + This method is used by the :meth:`__hash__` implementation + (which does caching). If your node type needs special-case + handling for some of its attributes, override this method, not + :meth:`__hash__`. + """ + return (type(self), self._ctor_arguments(self.children)) + + def __hash__(self) -> int: + """ + Hash of an expression with caching. + + See Also + -------- + get_hashable + """ + try: + return self._hash_value + except AttributeError: + self._hash_value = hash(self.get_hashable()) + return self._hash_value + + def is_equal(self, other: Self) -> bool: + """ + Equality of two nodes of equal type. + + Override this in subclasses, rather than :meth:`__eq__`. + + Parameter + --------- + other + object of same type to compare to. + + Notes + ----- + Since nodes are immutable, this does common subexpression + elimination when two nodes are determined to be equal. + + :meth:`__eq__` handles the case where the objects being + compared are not of the same type, so in this method, we only + need to implement equality of equal types. + + Returns + ------- + True if the two nodes are equal, false otherwise. + """ + if self is other: + return True + result = self._ctor_arguments(self.children) == other._ctor_arguments( + other.children + ) + # Eager CSE for nodes that match. + if result: + self.children = other.children + return result + + def __eq__(self, other: Any) -> bool: + """ + Equality of expressions. + + See Also + -------- + is_equal + """ + if type(self) is not type(other) or hash(self) != hash(other): + return False + else: + return self.is_equal(other) + + def __ne__(self, other: Any) -> bool: + """Inequality of expressions.""" + return not self.__eq__(other) + + def __repr__(self) -> str: + """String representation of an expression with caching.""" + try: + return self._repr_value + except AttributeError: + args = ", ".join(f"{arg!r}" for arg in self._ctor_arguments(self.children)) + self._repr_value = f"{type(self).__name__}({args})" + return self._repr_value diff --git a/python/cudf_polars/cudf_polars/dsl/translate.py b/python/cudf_polars/cudf_polars/dsl/translate.py index a0291037f01..c28f2c2651a 100644 --- a/python/cudf_polars/cudf_polars/dsl/translate.py +++ b/python/cudf_polars/cudf_polars/dsl/translate.py @@ -5,10 +5,11 @@ from __future__ import annotations +import functools import json from contextlib import AbstractContextManager, nullcontext from functools import singledispatch -from typing import Any +from typing import TYPE_CHECKING, Any import pyarrow as pa import pylibcudf as plc @@ -19,8 +20,12 @@ from polars.polars import _expr_nodes as pl_expr, _ir_nodes as pl_ir from cudf_polars.dsl import expr, ir +from cudf_polars.dsl.traversal import make_recursive, reuse_if_unchanged from cudf_polars.typing import NodeTraverser -from cudf_polars.utils import dtypes +from cudf_polars.utils import dtypes, sorting + +if TYPE_CHECKING: + from cudf_polars.typing import ExprTransformer __all__ = ["translate_ir", "translate_named_expr"] @@ -148,7 +153,7 @@ def _( with set_node(visitor, node.input): inp = translate_ir(visitor, n=None) exprs = [translate_named_expr(visitor, n=e) for e in node.expr] - return ir.Select(schema, inp, exprs, node.should_broadcast) + return ir.Select(schema, exprs, node.should_broadcast, inp) @_translate_ir.register @@ -161,11 +166,11 @@ def _( keys = [translate_named_expr(visitor, n=e) for e in node.keys] return ir.GroupBy( schema, - inp, - aggs, keys, + aggs, node.maintain_order, node.options, + inp, ) @@ -182,7 +187,71 @@ def _( with set_node(visitor, node.input_right): inp_right = translate_ir(visitor, n=None) right_on = [translate_named_expr(visitor, n=e) for e in node.right_on] - return ir.Join(schema, inp_left, inp_right, left_on, right_on, node.options) + if (how := node.options[0]) in { + "inner", + "left", + "right", + "full", + "cross", + "semi", + "anti", + }: + return ir.Join(schema, left_on, right_on, node.options, inp_left, inp_right) + else: + how, op1, op2 = how + if how != "ie_join": + raise NotImplementedError( + f"Unsupported join type {how}" + ) # pragma: no cover; asof joins not yet exposed + # No exposure of mixed/conditional joins in pylibcudf yet, so in + # the first instance, implement by doing a cross join followed by + # a filter. + _, join_nulls, zlice, suffix, coalesce = node.options + cross = ir.Join( + schema, + [], + [], + ("cross", join_nulls, None, suffix, coalesce), + inp_left, + inp_right, + ) + dtype = plc.DataType(plc.TypeId.BOOL8) + if op2 is None: + ops = [op1] + else: + ops = [op1, op2] + suffix = cross.options[3] + + # Column references in the right table refer to the post-join + # names, so with suffixes. + def _rename(e: expr.Expr, rec: ExprTransformer) -> expr.Expr: + if isinstance(e, expr.Col) and e.name in inp_left.schema: + return type(e)(e.dtype, f"{e.name}{suffix}") + return reuse_if_unchanged(e, rec) + + mapper = make_recursive(_rename) + right_on = [ + expr.NamedExpr( + f"{old.name}{suffix}" if old.name in inp_left.schema else old.name, new + ) + for new, old in zip( + (mapper(e.value) for e in right_on), right_on, strict=True + ) + ] + mask = functools.reduce( + functools.partial( + expr.BinOp, dtype, plc.binaryop.BinaryOperator.LOGICAL_AND + ), + ( + expr.BinOp(dtype, expr.BinOp._MAPPING[op], left.value, right.value) + for op, left, right in zip(ops, left_on, right_on, strict=True) + ), + ) + filtered = ir.Filter(schema, expr.NamedExpr("mask", mask), cross) + if zlice is not None: + offset, length = zlice + return ir.Slice(schema, offset, length, filtered) + return filtered @_translate_ir.register @@ -192,7 +261,7 @@ def _( with set_node(visitor, node.input): inp = translate_ir(visitor, n=None) exprs = [translate_named_expr(visitor, n=e) for e in node.exprs] - return ir.HStack(schema, inp, exprs, node.should_broadcast) + return ir.HStack(schema, exprs, node.should_broadcast, inp) @_translate_ir.register @@ -202,17 +271,23 @@ def _( with set_node(visitor, node.input): inp = translate_ir(visitor, n=None) exprs = [translate_named_expr(visitor, n=e) for e in node.expr] - return ir.Reduce(schema, inp, exprs) + return ir.Reduce(schema, exprs, inp) @_translate_ir.register def _( node: pl_ir.Distinct, visitor: NodeTraverser, schema: dict[str, plc.DataType] ) -> ir.IR: + (keep, subset, maintain_order, zlice) = node.options + keep = ir.Distinct._KEEP_MAP[keep] + subset = frozenset(subset) if subset is not None else None return ir.Distinct( schema, + keep, + subset, + zlice, + maintain_order, translate_ir(visitor, n=node.input), - node.options, ) @@ -223,14 +298,18 @@ def _( with set_node(visitor, node.input): inp = translate_ir(visitor, n=None) by = [translate_named_expr(visitor, n=e) for e in node.by_column] - return ir.Sort(schema, inp, by, node.sort_options, node.slice) + stable, nulls_last, descending = node.sort_options + order, null_order = sorting.sort_order( + descending, nulls_last=nulls_last, num_keys=len(by) + ) + return ir.Sort(schema, by, order, null_order, stable, node.slice, inp) @_translate_ir.register def _( node: pl_ir.Slice, visitor: NodeTraverser, schema: dict[str, plc.DataType] ) -> ir.IR: - return ir.Slice(schema, translate_ir(visitor, n=node.input), node.offset, node.len) + return ir.Slice(schema, node.offset, node.len, translate_ir(visitor, n=node.input)) @_translate_ir.register @@ -240,7 +319,7 @@ def _( with set_node(visitor, node.input): inp = translate_ir(visitor, n=None) mask = translate_named_expr(visitor, n=node.predicate) - return ir.Filter(schema, inp, mask) + return ir.Filter(schema, mask, inp) @_translate_ir.register @@ -259,10 +338,10 @@ def _( name, *options = node.function return ir.MapFunction( schema, - # TODO: merge_sorted breaks this pattern - translate_ir(visitor, n=node.input), name, options, + # TODO: merge_sorted breaks this pattern + translate_ir(visitor, n=node.input), ) @@ -271,7 +350,7 @@ def _( node: pl_ir.Union, visitor: NodeTraverser, schema: dict[str, plc.DataType] ) -> ir.IR: return ir.Union( - schema, [translate_ir(visitor, n=n) for n in node.inputs], node.options + schema, node.options, *(translate_ir(visitor, n=n) for n in node.inputs) ) @@ -279,7 +358,7 @@ def _( def _( node: pl_ir.HConcat, visitor: NodeTraverser, schema: dict[str, plc.DataType] ) -> ir.IR: - return ir.HConcat(schema, [translate_ir(visitor, n=n) for n in node.inputs]) + return ir.HConcat(schema, *(translate_ir(visitor, n=n) for n in node.inputs)) def translate_ir(visitor: NodeTraverser, *, n: int | None = None) -> ir.IR: @@ -309,8 +388,7 @@ def translate_ir(visitor: NodeTraverser, *, 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). - # Polars 1.7 changes definition of the CSV reader options schema name. - if (version := visitor.version()) >= (3, 0): + if (version := visitor.version()) >= (4, 0): raise NotImplementedError( f"No support for polars IR {version=}" ) # pragma: no cover; no such version for now. diff --git a/python/cudf_polars/cudf_polars/dsl/traversal.py b/python/cudf_polars/cudf_polars/dsl/traversal.py new file mode 100644 index 00000000000..be8338cb9a9 --- /dev/null +++ b/python/cudf_polars/cudf_polars/dsl/traversal.py @@ -0,0 +1,175 @@ +# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. +# SPDX-License-Identifier: Apache-2.0 + +"""Traversal and visitor utilities for nodes.""" + +from __future__ import annotations + +from typing import TYPE_CHECKING, Any, Generic + +from cudf_polars.typing import U_contra, V_co + +if TYPE_CHECKING: + from collections.abc import Callable, Generator, Mapping, MutableMapping + + from cudf_polars.typing import GenericTransformer, NodeT + + +__all__: list[str] = [ + "traversal", + "reuse_if_unchanged", + "make_recursive", + "CachingVisitor", +] + + +def traversal(node: NodeT) -> Generator[NodeT, None, None]: + """ + Pre-order traversal of nodes in an expression. + + Parameters + ---------- + node + Root of expression to traverse. + + Yields + ------ + Unique nodes in the expression, parent before child, children + in-order from left to right. + """ + seen = {node} + lifo = [node] + + while lifo: + node = lifo.pop() + yield node + for child in reversed(node.children): + if child not in seen: + seen.add(child) + lifo.append(child) + + +def reuse_if_unchanged(node: NodeT, fn: GenericTransformer[NodeT, NodeT]) -> NodeT: + """ + Recipe for transforming nodes that returns the old object if unchanged. + + Parameters + ---------- + node + Node to recurse on + fn + Function to transform children + + Notes + ----- + This can be used as a generic "base case" handler when + writing transforms that take nodes and produce new nodes. + + Returns + ------- + Existing node `e` if transformed children are unchanged, otherwise + reconstructed node with new children. + """ + new_children = [fn(c) for c in node.children] + if all(new == old for new, old in zip(new_children, node.children, strict=True)): + return node + return node.reconstruct(new_children) + + +def make_recursive( + fn: Callable[[U_contra, GenericTransformer[U_contra, V_co]], V_co], + *, + state: Mapping[str, Any] | None = None, +) -> GenericTransformer[U_contra, V_co]: + """ + No-op wrapper for recursive visitors. + + Facilitates using visitors that don't need caching but are written + in the same style. + + Parameters + ---------- + fn + Function to transform inputs to outputs. Should take as its + second argument a callable from input to output. + state + Arbitrary *immutable* state that should be accessible to the + visitor through the `state` property. + + Notes + ----- + All transformation functions *must* be free of side-effects. + + Usually, prefer a :class:`CachingVisitor`, but if we know that we + don't need caching in a transformation and then this no-op + approach is slightly cheaper. + + Returns + ------- + Recursive function without caching. + + See Also + -------- + CachingVisitor + """ + + def rec(node: U_contra) -> V_co: + return fn(node, rec) # type: ignore[arg-type] + + rec.state = state if state is not None else {} # type: ignore[attr-defined] + return rec # type: ignore[return-value] + + +class CachingVisitor(Generic[U_contra, V_co]): + """ + Caching wrapper for recursive visitors. + + Facilitates writing visitors where already computed results should + be cached and reused. The cache is managed automatically, and is + tied to the lifetime of the wrapper. + + Parameters + ---------- + fn + Function to transform inputs to outputs. Should take as its + second argument the recursive cache manager. + state + Arbitrary *immutable* state that should be accessible to the + visitor through the `state` property. + + Notes + ----- + All transformation functions *must* be free of side-effects. + + Returns + ------- + Recursive function with caching. + """ + + def __init__( + self, + fn: Callable[[U_contra, GenericTransformer[U_contra, V_co]], V_co], + *, + state: Mapping[str, Any] | None = None, + ) -> None: + self.fn = fn + self.cache: MutableMapping[U_contra, V_co] = {} + self.state = state if state is not None else {} + + def __call__(self, value: U_contra) -> V_co: + """ + Apply the function to a value. + + Parameters + ---------- + value + The value to transform. + + Returns + ------- + A transformed value. + """ + try: + return self.cache[value] + except KeyError: + return self.cache.setdefault(value, self.fn(value, self)) diff --git a/python/cudf_polars/cudf_polars/testing/plugin.py b/python/cudf_polars/cudf_polars/testing/plugin.py index 05b76d76808..a3607159e01 100644 --- a/python/cudf_polars/cudf_polars/testing/plugin.py +++ b/python/cudf_polars/cudf_polars/testing/plugin.py @@ -53,12 +53,34 @@ def pytest_configure(config: pytest.Config): "tests/unit/io/test_lazy_parquet.py::test_parquet_is_in_statistics": "Debug output on stderr doesn't match", "tests/unit/io/test_lazy_parquet.py::test_parquet_statistics": "Debug output on stderr doesn't match", "tests/unit/io/test_lazy_parquet.py::test_parquet_different_schema[False]": "Needs cudf#16394", + "tests/unit/io/test_lazy_parquet.py::test_parquet_schema_arg[False-columns]": "Correctly raises but different error", + "tests/unit/io/test_lazy_parquet.py::test_parquet_schema_arg[False-row_groups]": "Correctly raises but different error", + "tests/unit/io/test_lazy_parquet.py::test_parquet_schema_arg[False-prefiltered]": "Correctly raises but different error", + "tests/unit/io/test_lazy_parquet.py::test_parquet_schema_arg[False-none]": "Correctly raises but different error", "tests/unit/io/test_lazy_parquet.py::test_parquet_schema_mismatch_panic_17067[False]": "Needs cudf#16394", + "tests/unit/io/test_lazy_parquet.py::test_scan_parquet_ignores_dtype_mismatch_for_non_projected_columns_19249[False-False]": "Needs some variant of cudf#16394", + "tests/unit/io/test_lazy_parquet.py::test_scan_parquet_ignores_dtype_mismatch_for_non_projected_columns_19249[True-False]": "Needs some variant of cudf#16394", "tests/unit/io/test_lazy_parquet.py::test_parquet_slice_pushdown_non_zero_offset[False]": "Thrift data not handled correctly/slice pushdown wrong?", "tests/unit/io/test_lazy_parquet.py::test_parquet_unaligned_schema_read[False]": "Incomplete handling of projected reads with mismatching schemas, cudf#16394", "tests/unit/io/test_lazy_parquet.py::test_parquet_unaligned_schema_read_dtype_mismatch[False]": "Different exception raised, but correctly raises an exception", "tests/unit/io/test_lazy_parquet.py::test_parquet_unaligned_schema_read_missing_cols_from_first[False]": "Different exception raised, but correctly raises an exception", "tests/unit/io/test_parquet.py::test_read_parquet_only_loads_selected_columns_15098": "Memory usage won't be correct due to GPU", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-False-none]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-False-none]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-False-prefiltered]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-False-prefiltered]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-False-row_groups]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-False-row_groups]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-False-columns]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-False-columns]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-True-none]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-True-none]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-True-prefiltered]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-True-prefiltered]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-True-row_groups]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-True-row_groups]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection0-True-columns]": "Mismatching column read cudf#16394", + "tests/unit/io/test_parquet.py::test_allow_missing_columns[projection1-True-columns]": "Mismatching column read cudf#16394", "tests/unit/io/test_scan.py::test_scan[single-csv-async]": "Debug output on stderr doesn't match", "tests/unit/io/test_scan.py::test_scan_with_limit[single-csv-async]": "Debug output on stderr doesn't match", "tests/unit/io/test_scan.py::test_scan_with_filter[single-csv-async]": "Debug output on stderr doesn't match", @@ -107,6 +129,14 @@ def pytest_configure(config: pytest.Config): "tests/unit/operations/aggregation/test_aggregations.py::test_sum_empty_and_null_set": "libcudf sums column of all nulls to null, not zero", "tests/unit/operations/aggregation/test_aggregations.py::test_binary_op_agg_context_no_simplify_expr_12423": "groupby-agg of just literals should not produce collect_list", "tests/unit/operations/aggregation/test_aggregations.py::test_nan_inf_aggregation": "treatment of nans and nulls together is different in libcudf and polars in groupby-agg context", + "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func0-func0-none]": "cudf-polars doesn't nullify division by zero", + "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func0-func1-none]": "cudf-polars doesn't nullify division by zero", + "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func0-func2-none]": "cudf-polars doesn't nullify division by zero", + "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func0-func3-none]": "cudf-polars doesn't nullify division by zero", + "tests/unit/operations/arithmetic/test_list_arithmetic.py::test_list_arithmetic_values[func1-func0-none]": "cudf-polars doesn't nullify division by zero", + "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/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", @@ -124,13 +154,6 @@ def pytest_configure(config: pytest.Config): "tests/unit/operations/test_group_by.py::test_group_by_binary_agg_with_literal": "Incorrect broadcasting of literals in groupby-agg", "tests/unit/operations/test_group_by.py::test_aggregated_scalar_elementwise_15602": "Unsupported boolean function/dtype combination in groupby-agg", "tests/unit/operations/test_group_by.py::test_schemas[data1-expr1-expected_select1-expected_gb1]": "Mismatching dtypes, needs cudf#15852", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_by_monday_and_offset_5444": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_label[left-expected0]": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_label[right-expected1]": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_label[datapoint-expected2]": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_rolling_dynamic_sortedness_check": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_validation": "IR needs to expose groupby-dynamic information", - "tests/unit/operations/test_group_by_dynamic.py::test_group_by_dynamic_15225": "IR needs to expose groupby-dynamic information", "tests/unit/operations/test_join.py::test_cross_join_slice_pushdown": "Need to implement slice pushdown for cross joins", "tests/unit/sql/test_cast.py::test_cast_errors[values0-values::uint8-conversion from `f64` to `u64` failed]": "Casting that raises not supported on GPU", "tests/unit/sql/test_cast.py::test_cast_errors[values1-values::uint4-conversion from `i64` to `u32` failed]": "Casting that raises not supported on GPU", @@ -140,6 +163,7 @@ def pytest_configure(config: pytest.Config): "tests/unit/streaming/test_streaming_io.py::test_parquet_eq_statistics": "Debug output on stderr doesn't match", "tests/unit/test_cse.py::test_cse_predicate_self_join": "Debug output on stderr doesn't match", "tests/unit/test_empty.py::test_empty_9137": "Mismatching dtypes, needs cudf#15852", + "tests/unit/test_errors.py::test_error_on_empty_group_by": "Incorrect exception raised", # Maybe flaky, order-dependent? "tests/unit/test_projections.py::test_schema_full_outer_join_projection_pd_13287": "Order-specific result check, query is correct but in different order", "tests/unit/test_queries.py::test_group_by_agg_equals_zero_3535": "libcudf sums all nulls to null, not zero", diff --git a/python/cudf_polars/cudf_polars/typing/__init__.py b/python/cudf_polars/cudf_polars/typing/__init__.py index 240b11bdf59..a27a3395c35 100644 --- a/python/cudf_polars/cudf_polars/typing/__init__.py +++ b/python/cudf_polars/cudf_polars/typing/__init__.py @@ -5,8 +5,8 @@ from __future__ import annotations -from collections.abc import Mapping -from typing import TYPE_CHECKING, Literal, Protocol, Union +from collections.abc import Hashable, Mapping +from typing import TYPE_CHECKING, Any, Literal, Protocol, TypeVar, Union import pylibcudf as plc @@ -18,7 +18,19 @@ import polars as pl -IR: TypeAlias = Union[ + from cudf_polars.dsl import expr, ir, nodebase + +__all__: list[str] = [ + "PolarsIR", + "PolarsExpr", + "NodeTraverser", + "OptimizationArgs", + "GenericTransformer", + "ExprTransformer", + "IRTransformer", +] + +PolarsIR: TypeAlias = Union[ pl_ir.PythonScan, pl_ir.Scan, pl_ir.Cache, @@ -38,7 +50,7 @@ pl_ir.ExtContext, ] -Expr: TypeAlias = Union[ +PolarsExpr: TypeAlias = Union[ pl_expr.Function, pl_expr.Window, pl_expr.Literal, @@ -68,7 +80,7 @@ def set_node(self, n: int) -> None: """Set the current plan node to n.""" ... - def view_current_node(self) -> IR: + def view_current_node(self) -> PolarsIR: """Convert current plan node to python rep.""" ... @@ -80,7 +92,7 @@ def get_dtype(self, n: int) -> pl.DataType: """Get the datatype of the given expression id.""" ... - def view_expression(self, n: int) -> Expr: + def view_expression(self, n: int) -> PolarsExpr: """Convert the given expression to python rep.""" ... @@ -107,3 +119,29 @@ def set_udf( "cluster_with_columns", "no_optimization", ] + + +U_contra = TypeVar("U_contra", bound=Hashable, contravariant=True) +V_co = TypeVar("V_co", covariant=True) +NodeT = TypeVar("NodeT", bound="nodebase.Node[Any]") + + +class GenericTransformer(Protocol[U_contra, V_co]): + """Abstract protocol for recursive visitors.""" + + def __call__(self, __value: U_contra) -> V_co: + """Apply the visitor to the node.""" + ... + + @property + def state(self) -> Mapping[str, Any]: + """Arbitrary immutable state.""" + ... + + +# Quotes to avoid circular import +ExprTransformer: TypeAlias = GenericTransformer["expr.Expr", "expr.Expr"] +"""Protocol for transformation of Expr nodes.""" + +IRTransformer: TypeAlias = GenericTransformer["ir.IR", "ir.IR"] +"""Protocol for transformation of IR nodes.""" diff --git a/python/cudf_polars/docs/overview.md b/python/cudf_polars/docs/overview.md index 7837a275f20..74b2cd4e5de 100644 --- a/python/cudf_polars/docs/overview.md +++ b/python/cudf_polars/docs/overview.md @@ -11,14 +11,17 @@ You will need: environment](https://github.com/rapidsai/cudf/blob/branch-24.12/CONTRIBUTING.md#setting-up-your-build-environment). The combined devcontainer works, or whatever your favourite approach is. -> ![NOTE] These instructions will get simpler as we merge code in. +:::{note} +These instructions will get simpler as we merge code in. +::: ## Installing polars -`cudf-polars` works with polars >= 1.3, as long as the internal IR -version doesn't get a major version bump. So `pip install polars>=1.3` -should work. For development, if we're adding things to the polars -side of things, we will need to build polars from source: +The `cudf-polars` `pyproject.toml` advertises which polars versions it +works with. So for pure `cudf-polars` development, installing as +normal and satisfying the dependencies in the repository is +sufficient. For development, if we're adding things to the polars side +of things, we will need to build polars from source: ```sh git clone https://github.com/pola-rs/polars @@ -36,7 +39,9 @@ pip install --upgrade uv uv pip install --upgrade -r py-polars/requirements-dev.txt ``` -> ![NOTE] plain `pip install` works fine, but `uv` is _much_ faster! +:::{note} +plain `pip install` works fine, but `uv` is _much_ faster! +::: Now we have the necessary machinery to build polars ```sh @@ -83,7 +88,7 @@ representation (IR). Second, an execution phase which executes using our IR. The translation phase receives the a low-level Rust `NodeTraverser` -object which delivers Python representations of the plan nodes (and +object that delivers Python representations of the plan nodes (and expressions) one at a time. During translation, we endeavour to raise `NotImplementedError` for any unsupported functionality. This way, if we can't execute something, we just don't modify the logical plan at @@ -126,7 +131,6 @@ arguments, at the moment, `raise_on_fail` is also supported, which raises, rather than falling back, during translation: ```python - result = q.collect(engine=pl.GPUEngine(raise_on_fail=True)) ``` @@ -144,13 +148,73 @@ changes. We can therefore attempt to detect the IR version appropriately. This should be done during IR translation in `translate.py`. -## Adding a handler for a new plan node +# IR design + +As noted, we translate the polars DSL into our own IR. This is both so +that we can smooth out minor version differences (advertised by +`NodeTraverser` version changes) within `cudf-polars`, and so that we +have the freedom to introduce new IR nodes and rewrite rules as might +be appropriate for GPU execution. + +To that end, we provide facilities for definition of nodes as well as +writing traversals and rewrite rules. The abstract base class `Node` +in `dsl/nodebase.py` defines the interface for implementing new nodes, +and provides many useful default methods. See also the docstrings of +the `Node` class. + +:::{note} +This generic implementation relies on nodes being treated as +*immutable*. Do not implement in-place modification of nodes, bad +things will happen. +::: + +## Defining nodes + +A concrete node type (`cudf-polars` has expression nodes, `Expr`; +and plan nodes, `IR`), should inherit from `Node`. Nodes have +two types of data: + +1. `children`: a tuple (possibly empty) of concrete nodes; +2. non-child: arbitrary data attached to the node that is _not_ a + concrete node. + +The base `Node` class requires that one advertise the names of the +non-child attributes in the `_non_child` class variable. The +constructor of the concrete node should take its arguments in the +order `*_non_child` (ordered as the class variable does) and then +`*children`. For example, the `Sort` node, which sorts a column +generated by an expression, has this definition: + +```python +class Expr(Node): + children: tuple[Expr, ...] + +class Sort(Expr): + _non_child = ("dtype", "options") + children: tuple[Expr] + def __init__(self, dtype, options, column: Expr): + self.dtype = dtype + self.options = options + self.children = (column,) +``` + +By following this pattern, we get an automatic (caching) +implementation of `__hash__` and `__eq__`, as well as a useful +`reconstruct` method that will rebuild the node with new children. + +If you want to control the behaviour of `__hash__` and `__eq__` for a +single node, override (respectively) the `get_hashable` and `is_equal` +methods. + +## Adding new translation rules from the polars IR + +### Plan nodes -Plan node definitions live in `cudf_polars/dsl/ir.py`, these are -`dataclasses` that inherit from the base `IR` node. The evaluation of -a plan node is done by implementing the `evaluate` method. +Plan node definitions live in `cudf_polars/dsl/ir.py`, these all +inherit from the base `IR` node. The evaluation of a plan node is done +by implementing the `evaluate` method. -To translate the plan node, add a case handler in `translate_ir` which +To translate the plan node, add a case handler in `translate_ir` that lives in `cudf_polars/dsl/translate.py`. As well as child nodes that are plans, most plan nodes contain child @@ -163,25 +227,12 @@ translating a `Join` node, the left keys (expressions) should be translated with the left input active (and right keys with right input). To facilitate this, use the `set_node` context manager. -## Adding a handler for a new expression node +### Expression nodes Adding a handle for an expression node is very similar to a plan node. -Expressions are all defined in `cudf_polars/dsl/expr.py` and inherit -from `Expr`. Unlike plan nodes, these are not `dataclasses`, since it -is simpler for us to implement efficient hashing, repr, and equality if we -can write that ourselves. - -Every expression consists of two types of data: -1. child data (other `Expr`s) -2. non-child data (anything other than an `Expr`) -The generic implementations of special methods in the base `Expr` base -class require that the subclasses advertise which arguments to the -constructor are non-child in a `_non_child` class slot. The -constructor should then take arguments: -```python -def __init__(self, *non_child_data: Any, *children: Expr): -``` -Read the docstrings in the `Expr` class for more details. +Expressions are defined in `cudf_polars/dsl/expressions/` and exported +into the `dsl` namespace via `expr.py`. They inherit +from `Expr`. Expressions are evaluated by implementing a `do_evaluate` method that takes a `DataFrame` as context (this provides columns) along with an @@ -198,6 +249,124 @@ To simplify state tracking, all columns should be considered immutable on construction. This matches the "functional" description coming from the logical plan in any case, so is reasonably natural. +## Traversing and transforming nodes + +In addition to representing and evaluating nodes. We also provide +facilities for traversing a tree of nodes and defining transformation +rules in `dsl/traversal.py`. The simplest is `traversal`, a +[pre-order](https://en.wikipedia.org/wiki/Tree_traversal) visit of all +unique nodes in an expression. Use this if you want to know some +specific thing about an expression. For example, to determine if an +expression contains a `Literal` node: + +```python +def has_literal(node: Expr) -> bool: + return any(isinstance(e, Literal) for e in traversal(node)) +``` + +It is often convenient to provide (immutable) state to a visitor, as +well as some facility to perform DAG-aware rewrites (reusing a +transformation for an expression if we have already seen it). We +therefore adopt the following pattern of writing DAG-aware visitors. +Suppose we want a rewrite rule (`rewrite`) between expressions +(`Expr`) and some new type `T`. We define our general transformation +function `rewrite` with type `Expr -> (Expr -> T) -> T`: + +```python +from cudf_polars.typing import GenericTransformer + +@singledispatch +def rewrite(e: Expr, rec: GenericTransformer[Expr, T]) -> T: + ... +``` + +Note in particular that the function to perform the recursion is +passed as the second argument. Rather than defining methods on each +node in turn for a particular rewrite rule, we prefer free functions +and use `functools.singledispatch` to provide dispatching. We now, in +the usual fashion, register handlers for different expression types. +To use this function, we need to be able to provide both the +expression to convert and the recursive function itself. To do this we +must convert our `rewrite` function into something that only takes a +single argument (the expression to rewrite), but carries around +information about how to perform the recursion. To this end, we have +two utilities in `traversal.py`: + +- `make_recursive` and +- `CachingVisitor`. + +These both implement the `GenericTransformer` protocol, and can be +wrapped around a transformation function like `rewrite` to provide a +function `Expr -> T`. They also allow us to attach arbitrary +*immutable* state to our visitor by passing a `state` dictionary. This +dictionary can then be inspected by the concrete transformation +function. `make_recursive` is very simple, and provides no caching of +intermediate results (so any DAGs that are visited will be viewed as +trees). `CachingVisitor` provides the same interface, but maintains a +cache of intermediate results, and reuses them if the same expression +is seen again. + +Finally, for writing transformations that take nodes and deliver new +nodes (e.g. rewrite rules), we have a final utility +`reuse_if_unchanged` that can be used as a base case transformation +for node to node rewrites. It is a depth-first visit that transforms +children but only returns a new node with new children if the rewrite +of children returned new nodes. + +To see how these pieces fit together, let us consider writing a +`rename` function that takes an expression (potentially with +references to columns) along with a mapping defining a renaming +between (some subset of) column names. The goal is to deliver a new +expression with appropriate columns renamed. + +To start, we define the dispatch function +```python +from collections.abc import Mapping +from functools import singledispatch +from cudf_polars.dsl.traversal import ( + CachingVisitor, make_recursive, reuse_if_unchanged +) +from cudf_polars.dsl.expr import Col, Expr +from cudf_polars.typing import ExprTransformer + + +@singledispatch +def _rename(e: Expr, rec: ExprTransformer) -> Expr: + raise NotImplementedError(f"No handler for {type(e)}") +``` +then we register specific handlers, first for columns: +```python +@_rename.register +def _(e: Col, rec: ExprTransformer) -> Expr: + mapping = rec.state["mapping"] # state set on rec + if e.name in mapping: + # If we have a rename, return a new Col reference + # with a new name + return type(e)(e.dtype, mapping[e.name]) + return e +``` +and then for the remaining expressions +```python +_rename.register(Expr)(reuse_if_unchanged) +``` + +:::{note} +In this case, we could have put the generic handler in the `_rename` +function, however, then we would not get a nice error message if we +accidentally sent in an object of the incorrect type. +::: + +Finally we tie everything together with a public function: + +```python +def rename(e: Expr, mapping: Mapping[str, str]) -> Expr: + """Rename column references in an expression.""" + mapper = CachingVisitor(_rename, state={"mapping": mapping}) + # or + # mapper = make_recursive(_rename, state={"mapping": mapping}) + return mapper(e) +``` + # Containers Containers should be constructed as relatively lightweight objects diff --git a/python/cudf_polars/pyproject.toml b/python/cudf_polars/pyproject.toml index 5345fad41a2..2afdab1be4b 100644 --- a/python/cudf_polars/pyproject.toml +++ b/python/cudf_polars/pyproject.toml @@ -19,7 +19,7 @@ authors = [ license = { text = "Apache 2.0" } requires-python = ">=3.10" dependencies = [ - "polars>=1.8,<1.9", + "polars>=1.11,<1.12", "pylibcudf==24.12.*,>=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 = [ @@ -60,7 +60,7 @@ xfail_strict = true [tool.coverage.report] exclude_also = [ "if TYPE_CHECKING:", - "class .*\\bProtocol\\):", + "class .*\\bProtocol(?:\\[[^]]+\\])?\\):", "assert_never\\(" ] # The cudf_polars test suite doesn't exercise the plugin, so we omit diff --git a/python/cudf_polars/tests/dsl/test_expr.py b/python/cudf_polars/tests/dsl/test_expr.py index b7d4672daca..84e33262869 100644 --- a/python/cudf_polars/tests/dsl/test_expr.py +++ b/python/cudf_polars/tests/dsl/test_expr.py @@ -73,3 +73,24 @@ def test_namedexpr_repr_stable(): b2 = expr.NamedExpr("b1", expr.Col(plc.DataType(plc.TypeId.INT8), "a")) assert repr(b1) == repr(b2) + + +def test_equality_cse(): + dt = plc.DataType(plc.TypeId.INT8) + + def make_expr(n1, n2): + a = expr.Col(plc.DataType(plc.TypeId.INT8), n1) + b = expr.Col(plc.DataType(plc.TypeId.INT8), n2) + + return expr.BinOp(dt, plc.binaryop.BinaryOperator.ADD, a, b) + + e1 = make_expr("a", "b") + e2 = make_expr("a", "b") + e3 = make_expr("a", "c") + + assert e1.children is not e2.children + assert e1 == e2 + assert e1.children is e2.children + assert e1 == e2 + assert e1 != e3 + assert e2 != e3 diff --git a/python/cudf_polars/tests/dsl/test_traversal.py b/python/cudf_polars/tests/dsl/test_traversal.py new file mode 100644 index 00000000000..6505a786855 --- /dev/null +++ b/python/cudf_polars/tests/dsl/test_traversal.py @@ -0,0 +1,229 @@ +# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. +# SPDX-License-Identifier: Apache-2.0 + +from __future__ import annotations + +from functools import singledispatch + +import pylibcudf as plc + +import polars as pl +from polars.testing import assert_frame_equal + +from cudf_polars import translate_ir +from cudf_polars.dsl import expr, ir +from cudf_polars.dsl.traversal import ( + CachingVisitor, + make_recursive, + reuse_if_unchanged, + traversal, +) +from cudf_polars.typing import ExprTransformer, IRTransformer + + +def make_expr(dt, n1, n2): + a1 = expr.Col(dt, n1) + a2 = expr.Col(dt, n2) + + return expr.BinOp(dt, plc.binaryop.BinaryOperator.MUL, a1, a2) + + +def test_traversal_unique(): + dt = plc.DataType(plc.TypeId.INT8) + + e1 = make_expr(dt, "a", "a") + unique_exprs = list(traversal(e1)) + + assert len(unique_exprs) == 2 + assert set(unique_exprs) == {expr.Col(dt, "a"), e1} + assert unique_exprs == [e1, expr.Col(dt, "a")] + + e2 = make_expr(dt, "a", "b") + unique_exprs = list(traversal(e2)) + + assert len(unique_exprs) == 3 + assert set(unique_exprs) == {expr.Col(dt, "a"), expr.Col(dt, "b"), e2} + assert unique_exprs == [e2, expr.Col(dt, "a"), expr.Col(dt, "b")] + + e3 = make_expr(dt, "b", "a") + unique_exprs = list(traversal(e3)) + + assert len(unique_exprs) == 3 + assert set(unique_exprs) == {expr.Col(dt, "a"), expr.Col(dt, "b"), e3} + assert unique_exprs == [e3, expr.Col(dt, "b"), expr.Col(dt, "a")] + + +def rename(e, rec): + mapping = rec.state["mapping"] + if isinstance(e, expr.Col) and e.name in mapping: + return type(e)(e.dtype, mapping[e.name]) + return reuse_if_unchanged(e, rec) + + +def test_caching_visitor(): + dt = plc.DataType(plc.TypeId.INT8) + + e1 = make_expr(dt, "a", "b") + + mapper = CachingVisitor(rename, state={"mapping": {"b": "c"}}) + + renamed = mapper(e1) + assert renamed == make_expr(dt, "a", "c") + assert len(mapper.cache) == 3 + + e2 = make_expr(dt, "a", "a") + mapper = CachingVisitor(rename, state={"mapping": {"b": "c"}}) + + renamed = mapper(e2) + assert renamed == make_expr(dt, "a", "a") + assert len(mapper.cache) == 2 + mapper = CachingVisitor(rename, state={"mapping": {"a": "c"}}) + + renamed = mapper(e2) + assert renamed == make_expr(dt, "c", "c") + assert len(mapper.cache) == 2 + + +def test_noop_visitor(): + dt = plc.DataType(plc.TypeId.INT8) + + e1 = make_expr(dt, "a", "b") + + mapper = make_recursive(rename, state={"mapping": {"b": "c"}}) + + renamed = mapper(e1) + assert renamed == make_expr(dt, "a", "c") + + e2 = make_expr(dt, "a", "a") + mapper = make_recursive(rename, state={"mapping": {"b": "c"}}) + + renamed = mapper(e2) + assert renamed == make_expr(dt, "a", "a") + mapper = make_recursive(rename, state={"mapping": {"a": "c"}}) + + renamed = mapper(e2) + assert renamed == make_expr(dt, "c", "c") + + +def test_rewrite_ir_node(): + df = pl.LazyFrame({"a": [1, 2, 1], "b": [1, 3, 4]}) + q = df.group_by("a").agg(pl.col("b").sum()).sort("b") + + orig = translate_ir(q._ldf.visit()) + + new_df = pl.DataFrame({"a": [1, 1, 2], "b": [-1, -2, -4]}) + + def replace_df(node, rec): + if isinstance(node, ir.DataFrameScan): + return ir.DataFrameScan( + node.schema, new_df._df, node.projection, node.predicate + ) + return reuse_if_unchanged(node, rec) + + mapper = CachingVisitor(replace_df) + + new = mapper(orig) + + result = new.evaluate(cache={}).to_polars() + + expect = pl.DataFrame({"a": [2, 1], "b": [-4, -3]}) + + assert_frame_equal(result, expect) + + +def test_rewrite_scan_node(tmp_path): + left = pl.LazyFrame({"a": [1, 2, 3], "b": [1, 3, 4]}) + right = pl.DataFrame({"a": [1, 4, 2], "c": [1, 2, 3]}) + + right.write_parquet(tmp_path / "right.pq") + + right_s = pl.scan_parquet(tmp_path / "right.pq") + + q = left.join(right_s, on="a", how="inner") + + def replace_scan(node, rec): + if isinstance(node, ir.Scan): + return ir.DataFrameScan( + node.schema, right._df, node.with_columns, node.predicate + ) + return reuse_if_unchanged(node, rec) + + mapper = CachingVisitor(replace_scan) + + orig = translate_ir(q._ldf.visit()) + new = mapper(orig) + + result = new.evaluate(cache={}).to_polars() + + expect = q.collect() + + assert_frame_equal(result, expect, check_row_order=False) + + +def test_rewrite_names_and_ops(): + df = pl.LazyFrame({"a": [1, 2, 3], "b": [3, 4, 5], "c": [5, 6, 7], "d": [7, 9, 8]}) + + q = df.select(pl.col("a") - (pl.col("b") + pl.col("c") * 2), pl.col("d")).sort("d") + + # We will replace a -> d, c -> d, and addition with multiplication + expect = ( + df.select( + (pl.col("d") - (pl.col("b") * pl.col("d") * 2)).alias("a"), pl.col("d") + ) + .sort("d") + .collect() + ) + + qir = translate_ir(q._ldf.visit()) + + @singledispatch + def _transform(e: expr.Expr, fn: ExprTransformer) -> expr.Expr: + raise NotImplementedError("Unhandled") + + @_transform.register + def _(e: expr.Col, fn: ExprTransformer): + mapping = fn.state["mapping"] + if e.name in mapping: + return type(e)(e.dtype, mapping[e.name]) + return e + + @_transform.register + def _(e: expr.BinOp, fn: ExprTransformer): + if e.op == plc.binaryop.BinaryOperator.ADD: + return type(e)( + e.dtype, plc.binaryop.BinaryOperator.MUL, *map(fn, e.children) + ) + return reuse_if_unchanged(e, fn) + + _transform.register(expr.Expr)(reuse_if_unchanged) + + @singledispatch + def _rewrite(node: ir.IR, fn: IRTransformer) -> ir.IR: + raise NotImplementedError("Unhandled") + + @_rewrite.register + def _(node: ir.Select, fn: IRTransformer): + expr_mapper = fn.state["expr_mapper"] + return type(node)( + node.schema, + [expr.NamedExpr(e.name, expr_mapper(e.value)) for e in node.exprs], + node.should_broadcast, + fn(node.children[0]), + ) + + _rewrite.register(ir.IR)(reuse_if_unchanged) + + rewriter = CachingVisitor( + _rewrite, + state={ + "expr_mapper": CachingVisitor( + _transform, state={"mapping": {"a": "d", "c": "d"}} + ) + }, + ) + + new_ir = rewriter(qir) + + got = new_ir.evaluate(cache={}).to_polars() + + assert_frame_equal(expect, got) diff --git a/python/cudf_polars/tests/test_config.py b/python/cudf_polars/tests/test_config.py index 3c3986be19b..9900f598e5f 100644 --- a/python/cudf_polars/tests/test_config.py +++ b/python/cudf_polars/tests/test_config.py @@ -10,7 +10,7 @@ import rmm -from cudf_polars.dsl.ir import IR +from cudf_polars.dsl.ir import DataFrameScan from cudf_polars.testing.asserts import ( assert_gpu_result_equal, assert_ir_translation_raises, @@ -18,10 +18,10 @@ def test_polars_verbose_warns(monkeypatch): - def raise_unimplemented(self): + def raise_unimplemented(self, *args): raise NotImplementedError("We don't support this") - monkeypatch.setattr(IR, "__post_init__", raise_unimplemented) + monkeypatch.setattr(DataFrameScan, "__init__", raise_unimplemented) q = pl.LazyFrame({}) # Ensure that things raise assert_ir_translation_raises(q, NotImplementedError) diff --git a/python/cudf_polars/tests/test_join.py b/python/cudf_polars/tests/test_join.py index 7d9ec98db97..501560d15b8 100644 --- a/python/cudf_polars/tests/test_join.py +++ b/python/cudf_polars/tests/test_join.py @@ -2,9 +2,12 @@ # SPDX-License-Identifier: Apache-2.0 from __future__ import annotations +from contextlib import nullcontext + import pytest import polars as pl +from polars.testing import assert_frame_equal from cudf_polars.testing.asserts import ( assert_gpu_result_equal, @@ -22,6 +25,11 @@ def how(request): return request.param +@pytest.fixture(params=[None, (1, 5), (1, None), (0, 2), (0, None)]) +def zlice(request): + return request.param + + @pytest.fixture def left(): return pl.LazyFrame( @@ -37,8 +45,9 @@ def left(): def right(): return pl.LazyFrame( { - "a": [1, 4, 3, 7, None, None], - "c": [2, 3, 4, 5, 6, 7], + "a": [1, 4, 3, 7, None, None, 1], + "c": [2, 3, 4, 5, 6, 7, 8], + "d": [6, None, 7, 8, -1, 2, 4], } ) @@ -70,11 +79,31 @@ def test_coalesce_join(left, right, how, join_nulls, join_expr): query = left.join( right, on=join_expr, how=how, join_nulls=join_nulls, coalesce=True ) - assert_gpu_result_equal(query, check_row_order=False) + assert_gpu_result_equal(query, check_row_order=how == "left") -def test_cross_join(left, right): +def test_left_join_with_slice(left, right, join_nulls, zlice): + q = left.join(right, on="a", how="left", join_nulls=join_nulls, coalesce=True) + ctx = nullcontext() + if zlice is not None: + q_expect = q.collect().slice(*zlice) + q = q.slice(*zlice) + if zlice == (1, 5) or zlice == (0, 2): + # https://github.com/pola-rs/polars/issues/19403 + # https://github.com/pola-rs/polars/issues/19405 + ctx = pytest.raises(AssertionError) + assert_frame_equal( + q_expect, q.collect(engine=pl.GPUEngine(raise_on_fail=True)) + ) + + with ctx: + assert_gpu_result_equal(q) + + +def test_cross_join(left, right, zlice): q = left.join(right, how="cross") + if zlice is not None: + q = q.slice(*zlice) assert_gpu_result_equal(q) @@ -86,3 +115,26 @@ def test_join_literal_key_unsupported(left, right, left_on, right_on): q = left.join(right, left_on=left_on, right_on=right_on, how="inner") assert_ir_translation_raises(q, NotImplementedError) + + +@pytest.mark.parametrize( + "conditions", + [ + [pl.col("a") < pl.col("a_right")], + [pl.col("a_right") <= pl.col("a") * 2], + [pl.col("b") * 2 > pl.col("a_right"), pl.col("a") == pl.col("c_right")], + [pl.col("b") * 2 <= pl.col("a_right"), pl.col("a") < pl.col("c_right")], + [pl.col("b") <= pl.col("a_right") * 7, pl.col("a") < pl.col("d") * 2], + ], +) +def test_join_where(left, right, conditions, zlice): + q = left.join_where(right, *conditions) + + assert_gpu_result_equal(q, check_row_order=False) + + if zlice is not None: + q_len = q.slice(*zlice).select(pl.len()) + # Can't compare result, since row order is not guaranteed and + # therefore we only check the length + + assert_gpu_result_equal(q_len) diff --git a/python/custreamz/custreamz/tests/conftest.py b/python/custreamz/custreamz/tests/conftest.py index 1cda9b71387..c5135bc6414 100644 --- a/python/custreamz/custreamz/tests/conftest.py +++ b/python/custreamz/custreamz/tests/conftest.py @@ -2,6 +2,7 @@ import socket import pytest + from custreamz import kafka diff --git a/python/custreamz/pyproject.toml b/python/custreamz/pyproject.toml index af45f49d9b4..a8ab05a3922 100644 --- a/python/custreamz/pyproject.toml +++ b/python/custreamz/pyproject.toml @@ -65,50 +65,20 @@ include = [ ] exclude = ["*tests*"] -[tool.isort] -line_length = 79 -multi_line_output = 3 -include_trailing_comma = true -force_grid_wrap = 0 -combine_as_imports = true -order_by_type = true -known_dask = [ - "dask", - "distributed", - "dask_cuda", -] -known_rapids = [ - "rmm", - "cudf", - "dask_cudf", -] -known_first_party = [ - "streamz", -] -default_section = "THIRDPARTY" -sections = [ - "FUTURE", - "STDLIB", - "THIRDPARTY", - "DASK", - "RAPIDS", - "FIRSTPARTY", - "LOCALFOLDER", -] -skip = [ - "thirdparty", - ".eggs", - ".git", - ".hg", - ".mypy_cache", - ".tox", - ".venv", - "_build", - "buck-out", - "build", - "dist", - "__init__.py", -] +[tool.ruff] +extend = "../../pyproject.toml" + +[tool.ruff.lint.isort] +combine-as-imports = true +known-first-party = ["streamz"] +section-order = ["future", "standard-library", "third-party", "dask", "rapids", "first-party", "local-folder"] + +[tool.ruff.lint.isort.sections] +dask = ["dask", "distributed", "dask_cuda"] +rapids = ["rmm", "cudf", "dask_cudf"] + +[tool.ruff.lint.per-file-ignores] +"__init__.py" = ["E402", "F401"] [tool.pytest.ini_options] addopts = "--tb=native --strict-config --strict-markers" diff --git a/python/dask_cudf/dask_cudf/__init__.py b/python/dask_cudf/dask_cudf/__init__.py index 04c2ad65b99..f9df22cc436 100644 --- a/python/dask_cudf/dask_cudf/__init__.py +++ b/python/dask_cudf/dask_cudf/__init__.py @@ -7,15 +7,15 @@ # do anything for dask==2024.2.0) config.set({"dataframe.query-planning-warning": False}) -import dask.dataframe as dd -from dask.dataframe import from_delayed +import dask.dataframe as dd # noqa: E402 +from dask.dataframe import from_delayed # noqa: E402 -import cudf +import cudf # noqa: E402 -from . import backends -from ._version import __git_commit__, __version__ -from .core import concat, from_cudf, from_dask_dataframe -from .expr import QUERY_PLANNING_ON +from . import backends # noqa: E402, F401 +from ._version import __git_commit__, __version__ # noqa: E402, F401 +from .core import concat, from_cudf, from_dask_dataframe # noqa: E402 +from .expr import QUERY_PLANNING_ON # noqa: E402 def read_csv(*args, **kwargs): @@ -55,9 +55,9 @@ def inner_func(*args, **kwargs): to_orc = raise_not_implemented_error("to_orc") else: - from .core import DataFrame, Index, Series - from .groupby import groupby_agg - from .io import read_text, to_orc + from .core import DataFrame, Index, Series # noqa: F401 + from .groupby import groupby_agg # noqa: F401 + from .io import read_text, to_orc # noqa: F401 __all__ = [ diff --git a/python/dask_cudf/dask_cudf/expr/__init__.py b/python/dask_cudf/dask_cudf/expr/__init__.py index a76b655ef42..6dadadd5263 100644 --- a/python/dask_cudf/dask_cudf/expr/__init__.py +++ b/python/dask_cudf/dask_cudf/expr/__init__.py @@ -12,8 +12,8 @@ config.set({"dataframe.shuffle.method": "tasks"}) try: - import dask_cudf.expr._collection - import dask_cudf.expr._expr + import dask_cudf.expr._collection # noqa: F401 + import dask_cudf.expr._expr # noqa: F401 except ImportError as err: # Dask *should* raise an error before this. diff --git a/python/dask_cudf/dask_cudf/expr/_expr.py b/python/dask_cudf/dask_cudf/expr/_expr.py index 4a9f4de8b9c..c7cf66fbffd 100644 --- a/python/dask_cudf/dask_cudf/expr/_expr.py +++ b/python/dask_cudf/dask_cudf/expr/_expr.py @@ -12,7 +12,7 @@ ) from dask_expr._reductions import Reduction, Var from dask_expr.io.io import FusedParquetIO -from dask_expr.io.parquet import ReadParquetPyarrowFS +from dask_expr.io.parquet import FragmentWrapper, ReadParquetPyarrowFS from dask.dataframe.core import ( _concat, @@ -302,16 +302,34 @@ def _dataset_info(self): return dataset_info @staticmethod - def _table_to_pandas( - table, - index_name, - *args, - ): + def _table_to_pandas(table, index_name): df = cudf.DataFrame.from_arrow(table) if index_name is not None: df = df.set_index(index_name) return df + def _filtered_task(self, index: int): + columns = self.columns.copy() + index_name = self.index.name + if self.index is not None: + index_name = self.index.name + schema = self._dataset_info["schema"].remove_metadata() + if index_name: + if columns is None: + columns = list(schema.names) + columns.append(index_name) + return ( + self._table_to_pandas, + ( + self._fragment_to_table, + FragmentWrapper(self.fragments[index], filesystem=self.fs), + self.filters, + columns, + schema, + ), + index_name, + ) + def _tune_up(self, parent): if self._fusion_compression_factor >= 1: return diff --git a/python/dask_cudf/dask_cudf/io/__init__.py b/python/dask_cudf/dask_cudf/io/__init__.py index 76bb2ea99b4..0421bd755f4 100644 --- a/python/dask_cudf/dask_cudf/io/__init__.py +++ b/python/dask_cudf/dask_cudf/io/__init__.py @@ -1,11 +1,11 @@ -# Copyright (c) 2018-2022, NVIDIA CORPORATION. +# Copyright (c) 2018-2024, NVIDIA CORPORATION. -from .csv import read_csv -from .json import read_json -from .orc import read_orc, to_orc -from .text import read_text +from .csv import read_csv # noqa: F401 +from .json import read_json # noqa: F401 +from .orc import read_orc, to_orc # noqa: F401 +from .text import read_text # noqa: F401 try: - from .parquet import read_parquet, to_parquet + from .parquet import read_parquet, to_parquet # noqa: F401 except ImportError: pass diff --git a/python/dask_cudf/dask_cudf/io/tests/test_parquet.py b/python/dask_cudf/dask_cudf/io/tests/test_parquet.py index 896c4169f5b..ae5ca480e31 100644 --- a/python/dask_cudf/dask_cudf/io/tests/test_parquet.py +++ b/python/dask_cudf/dask_cudf/io/tests/test_parquet.py @@ -15,7 +15,11 @@ import cudf import dask_cudf -from dask_cudf.tests.utils import skip_dask_expr, xfail_dask_expr +from dask_cudf.tests.utils import ( + require_dask_expr, + skip_dask_expr, + xfail_dask_expr, +) # Check if create_metadata_file is supported by # the current dask.dataframe version @@ -615,3 +619,28 @@ def test_timezone_column(tmpdir): got = dask_cudf.read_parquet(path) expect = cudf.read_parquet(path) dd.assert_eq(got, expect) + + +@require_dask_expr() +@pytest.mark.skipif( + not dask_cudf.backends.PYARROW_GE_15, + reason="Requires pyarrow 15", +) +@pytest.mark.parametrize("min_part_size", ["1B", "1GB"]) +def test_read_parquet_arrow_filesystem(tmpdir, min_part_size): + tmp_path = str(tmpdir) + with dask.config.set( + { + "dataframe.backend": "cudf", + "dataframe.parquet.minimum-partition-size": min_part_size, + } + ): + dd.from_dict( + {"x": range(1000), "y": ["a", "b", "c", "d"] * 250}, + npartitions=10, + ).to_parquet(tmp_path, write_index=False) + df = cudf.read_parquet(tmp_path) + ddf = dask_cudf.read_parquet(tmp_path, filesystem="arrow") + dd.assert_eq(df, ddf, check_index=False) + assert isinstance(ddf._meta, cudf.DataFrame) + assert isinstance(ddf.compute(), cudf.DataFrame) diff --git a/python/dask_cudf/dask_cudf/io/tests/test_s3.py b/python/dask_cudf/dask_cudf/io/tests/test_s3.py index cf8af82e112..90907f6fb99 100644 --- a/python/dask_cudf/dask_cudf/io/tests/test_s3.py +++ b/python/dask_cudf/dask_cudf/io/tests/test_s3.py @@ -11,6 +11,8 @@ from dask.dataframe import assert_eq +import cudf + import dask_cudf from dask_cudf.tests.utils import QUERY_PLANNING_ON @@ -168,6 +170,8 @@ def test_read_parquet_filesystem(s3_base, s3so, pdf, filesystem): filesystem=filesystem, ) assert df.b.sum().compute() == 9 + assert isinstance(df._meta, cudf.DataFrame) + assert isinstance(df.compute(), cudf.DataFrame) def test_read_parquet_filesystem_explicit(s3_base, s3so, pdf): diff --git a/python/dask_cudf/pyproject.toml b/python/dask_cudf/pyproject.toml index fbcd7ae5dfb..862e8f36eaa 100644 --- a/python/dask_cudf/pyproject.toml +++ b/python/dask_cudf/pyproject.toml @@ -69,50 +69,17 @@ version = {file = "dask_cudf/VERSION"} [tool.setuptools.packages.find] exclude = ["*tests*"] -[tool.isort] -line_length = 79 -multi_line_output = 3 -include_trailing_comma = true -force_grid_wrap = 0 -combine_as_imports = true -order_by_type = true +[tool.ruff] +extend = "../../pyproject.toml" -known_dask = [ - "dask", - "distributed", - "dask_cuda", -] -known_rapids = [ - "rmm", - "cudf", -] -known_first_party = [ - "dask_cudf", -] +[tool.ruff.lint.isort] +combine-as-imports = true +known-first-party = ["dask_cudf"] +section-order = ["future", "standard-library", "third-party", "dask", "rapids", "first-party", "local-folder"] -default_section = "THIRDPARTY" -sections = [ - "FUTURE", - "STDLIB", - "THIRDPARTY", - "DASK", - "RAPIDS", - "FIRSTPARTY", - "LOCALFOLDER", -] -skip = [ - "thirdparty", - ".eggs", - ".git", - ".hg", - ".mypy_cache", - ".tox", - ".venv", - "_build", - "buck-out", - "build", - "dist", -] +[tool.ruff.lint.isort.sections] +dask = ["dask", "distributed", "dask_cuda"] +rapids = ["rmm", "cudf"] [tool.pytest.ini_options] addopts = "--tb=native --strict-config --strict-markers" @@ -126,5 +93,8 @@ filterwarnings = [ # https://github.com/dask/partd/blob/main/partd/pandas.py#L198 "ignore:Passing a BlockManager to DataFrame is deprecated and will raise in a future version. Use public APIs instead.:DeprecationWarning", "ignore:String support for `aggregate_files` is experimental. Behavior may change in the future.:FutureWarning:dask", + # Dask now loudly throws warnings: https://github.com/dask/dask/pull/11437 + # When the legacy implementation is removed we can remove this warning and stop running pytests with `DASK_DATAFRAME__QUERY_PLANNING=False` + "ignore:The legacy Dask DataFrame implementation is deprecated and will be removed in a future version.*:FutureWarning", ] xfail_strict = true diff --git a/python/pylibcudf/pylibcudf/__init__.pxd b/python/pylibcudf/pylibcudf/__init__.pxd index aa67b4b1149..9bdfdab97c2 100644 --- a/python/pylibcudf/pylibcudf/__init__.pxd +++ b/python/pylibcudf/pylibcudf/__init__.pxd @@ -13,6 +13,7 @@ from . cimport ( expressions, filling, groupby, + interop, join, json, labeling, @@ -62,6 +63,7 @@ __all__ = [ "filling", "gpumemoryview", "groupby", + "interop", "join", "json", "lists", diff --git a/python/pylibcudf/pylibcudf/interop.pxd b/python/pylibcudf/pylibcudf/interop.pxd new file mode 100644 index 00000000000..2a0a8c15fdd --- /dev/null +++ b/python/pylibcudf/pylibcudf/interop.pxd @@ -0,0 +1,8 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from pylibcudf.table cimport Table + + +cpdef Table from_dlpack(object managed_tensor) + +cpdef object to_dlpack(Table input) diff --git a/python/pylibcudf/pylibcudf/interop.pyx b/python/pylibcudf/pylibcudf/interop.pyx index 642516a1b90..61e812353b7 100644 --- a/python/pylibcudf/pylibcudf/interop.pyx +++ b/python/pylibcudf/pylibcudf/interop.pyx @@ -1,6 +1,11 @@ # Copyright (c) 2023-2024, NVIDIA CORPORATION. -from cpython.pycapsule cimport PyCapsule_GetPointer, PyCapsule_New +from cpython.pycapsule cimport ( + PyCapsule_GetPointer, + PyCapsule_IsValid, + PyCapsule_New, + PyCapsule_SetName, +) from libc.stdlib cimport free from libcpp.memory cimport unique_ptr from libcpp.utility cimport move @@ -16,11 +21,14 @@ from pylibcudf.libcudf.interop cimport ( ArrowArray, ArrowArrayStream, ArrowSchema, + DLManagedTensor, column_metadata, from_arrow_column as cpp_from_arrow_column, from_arrow_stream as cpp_from_arrow_stream, + from_dlpack as cpp_from_dlpack, to_arrow_host_raw, to_arrow_schema_raw, + to_dlpack as cpp_to_dlpack, ) from pylibcudf.libcudf.table.table cimport table @@ -315,3 +323,87 @@ def _to_arrow_scalar(cudf_object, metadata=None): # Note that metadata for scalars is primarily important for preserving # information on nested types since names are otherwise irrelevant. return to_arrow(Column.from_scalar(cudf_object, 1), metadata=metadata)[0] + + +cpdef Table from_dlpack(object managed_tensor): + """ + Convert a DLPack DLTensor into a cudf table. + + For details, see :cpp:func:`cudf::from_dlpack` + + Parameters + ---------- + managed_tensor : PyCapsule + A 1D or 2D column-major (Fortran order) tensor. + + Returns + ------- + Table + Table with a copy of the tensor data. + """ + if not PyCapsule_IsValid(managed_tensor, "dltensor"): + raise ValueError("Invalid PyCapsule object") + cdef unique_ptr[table] c_result + cdef DLManagedTensor* dlpack_tensor = PyCapsule_GetPointer( + managed_tensor, "dltensor" + ) + if dlpack_tensor is NULL: + raise ValueError("PyCapsule object contained a NULL pointer") + PyCapsule_SetName(managed_tensor, "used_dltensor") + + # Note: A copy is always performed when converting the dlpack + # data to a libcudf table. We also delete the dlpack_tensor pointer + # as the pointer is not deleted by libcudf's from_dlpack function. + # TODO: https://github.com/rapidsai/cudf/issues/10874 + # TODO: https://github.com/rapidsai/cudf/issues/10849 + with nogil: + c_result = cpp_from_dlpack(dlpack_tensor) + + cdef Table result = Table.from_libcudf(move(c_result)) + dlpack_tensor.deleter(dlpack_tensor) + return result + + +cpdef object to_dlpack(Table input): + """ + Convert a cudf table into a DLPack DLTensor. + + For details, see :cpp:func:`cudf::to_dlpack` + + Parameters + ---------- + input : Table + A 1D or 2D column-major (Fortran order) tensor. + + Returns + ------- + PyCapsule + 1D or 2D DLPack tensor with a copy of the table data, or nullptr. + """ + for col in input._columns: + if col.null_count(): + raise ValueError( + "Cannot create a DLPack tensor with null values. " + "Input is required to have null count as zero." + ) + cdef DLManagedTensor *dlpack_tensor + + with nogil: + dlpack_tensor = cpp_to_dlpack(input.view()) + + return PyCapsule_New( + dlpack_tensor, + "dltensor", + dlmanaged_tensor_pycapsule_deleter + ) + + +cdef void dlmanaged_tensor_pycapsule_deleter(object pycap_obj) noexcept: + if PyCapsule_IsValid(pycap_obj, "used_dltensor"): + # we do not call a used capsule's deleter + return + cdef DLManagedTensor* dlpack_tensor = PyCapsule_GetPointer( + pycap_obj, "dltensor" + ) + if dlpack_tensor is not NULL: + dlpack_tensor.deleter(dlpack_tensor) diff --git a/python/pylibcudf/pylibcudf/libcudf/interop.pxd b/python/pylibcudf/pylibcudf/libcudf/interop.pxd index 30b97fdec34..b75e9ca7001 100644 --- a/python/pylibcudf/pylibcudf/libcudf/interop.pxd +++ b/python/pylibcudf/pylibcudf/libcudf/interop.pxd @@ -32,11 +32,13 @@ cdef extern from "cudf/interop.hpp" nogil: cdef extern from "cudf/interop.hpp" namespace "cudf" \ nogil: - cdef unique_ptr[table] from_dlpack(const DLManagedTensor* tensor - ) except + + cdef unique_ptr[table] from_dlpack( + const DLManagedTensor* managed_tensor + ) except + - DLManagedTensor* to_dlpack(table_view input_table - ) except + + DLManagedTensor* to_dlpack( + const table_view& input + ) except + cdef cppclass column_metadata: column_metadata() except + diff --git a/python/pylibcudf/pylibcudf/libcudf/nvtext/stemmer.pxd b/python/pylibcudf/pylibcudf/libcudf/nvtext/stemmer.pxd index 673bffa28ae..be3a2d75718 100644 --- a/python/pylibcudf/pylibcudf/libcudf/nvtext/stemmer.pxd +++ b/python/pylibcudf/pylibcudf/libcudf/nvtext/stemmer.pxd @@ -1,6 +1,7 @@ # Copyright (c) 2020-2024, NVIDIA CORPORATION. from libc.stdint cimport int32_t +from libcpp cimport bool from libcpp.memory cimport unique_ptr from pylibcudf.libcudf.column.column cimport column from pylibcudf.libcudf.column.column_view cimport column_view @@ -8,9 +9,9 @@ from pylibcudf.libcudf.types cimport size_type cdef extern from "nvtext/stemmer.hpp" namespace "nvtext" nogil: - ctypedef enum letter_type: - CONSONANT 'nvtext::letter_type::CONSONANT' - VOWEL 'nvtext::letter_type::VOWEL' + cpdef enum class letter_type: + CONSONANT + VOWEL cdef unique_ptr[column] porter_stemmer_measure( const column_view & strings diff --git a/python/pylibcudf/pylibcudf/libcudf/strings/replace_re.pxd b/python/pylibcudf/pylibcudf/libcudf/strings/replace_re.pxd index 40f0e2fa50c..6b0c90d0acc 100644 --- a/python/pylibcudf/pylibcudf/libcudf/strings/replace_re.pxd +++ b/python/pylibcudf/pylibcudf/libcudf/strings/replace_re.pxd @@ -6,6 +6,7 @@ from libcpp.vector cimport vector from pylibcudf.libcudf.column.column cimport column from pylibcudf.libcudf.column.column_view cimport column_view from pylibcudf.libcudf.scalar.scalar cimport string_scalar +from pylibcudf.libcudf.strings.regex_flags cimport regex_flags from pylibcudf.libcudf.strings.regex_program cimport regex_program from pylibcudf.libcudf.table.table cimport table from pylibcudf.libcudf.types cimport size_type @@ -14,17 +15,18 @@ from pylibcudf.libcudf.types cimport size_type cdef extern from "cudf/strings/replace_re.hpp" namespace "cudf::strings" nogil: cdef unique_ptr[column] replace_re( - column_view source_strings, - regex_program, - string_scalar repl, - size_type maxrepl) except + - - cdef unique_ptr[column] replace_with_backrefs( - column_view source_strings, - regex_program, - string repl) except + + column_view input, + regex_program prog, + string_scalar replacement, + size_type max_replace_count) except + cdef unique_ptr[column] replace_re( - column_view source_strings, + column_view input, vector[string] patterns, - column_view repls) except + + column_view replacements, + regex_flags flags) except + + + cdef unique_ptr[column] replace_with_backrefs( + column_view input, + regex_program prog, + string replacement) except + diff --git a/python/pylibcudf/pylibcudf/nvtext/CMakeLists.txt b/python/pylibcudf/pylibcudf/nvtext/CMakeLists.txt index e01ca3fbdd3..d97c0a73267 100644 --- a/python/pylibcudf/pylibcudf/nvtext/CMakeLists.txt +++ b/python/pylibcudf/pylibcudf/nvtext/CMakeLists.txt @@ -13,7 +13,7 @@ # ============================================================================= set(cython_sources edit_distance.pyx generate_ngrams.pyx jaccard.pyx minhash.pyx - ngrams_tokenize.pyx normalize.pyx + ngrams_tokenize.pyx normalize.pyx replace.pyx stemmer.pyx ) set(linked_libraries cudf::cudf) diff --git a/python/pylibcudf/pylibcudf/nvtext/__init__.pxd b/python/pylibcudf/pylibcudf/nvtext/__init__.pxd index 08dbec84090..a658e57018e 100644 --- a/python/pylibcudf/pylibcudf/nvtext/__init__.pxd +++ b/python/pylibcudf/pylibcudf/nvtext/__init__.pxd @@ -7,6 +7,8 @@ from . cimport ( minhash, ngrams_tokenize, normalize, + replace, + stemmer, ) __all__ = [ @@ -16,4 +18,6 @@ __all__ = [ "minhash", "ngrams_tokenize", "normalize", + "replace", + "stemmer", ] diff --git a/python/pylibcudf/pylibcudf/nvtext/__init__.py b/python/pylibcudf/pylibcudf/nvtext/__init__.py index 6dccf3dd9cf..2c1feb089a2 100644 --- a/python/pylibcudf/pylibcudf/nvtext/__init__.py +++ b/python/pylibcudf/pylibcudf/nvtext/__init__.py @@ -7,6 +7,8 @@ minhash, ngrams_tokenize, normalize, + replace, + stemmer, ) __all__ = [ @@ -16,4 +18,6 @@ "minhash", "ngrams_tokenize", "normalize", + "replace", + "stemmer", ] diff --git a/python/pylibcudf/pylibcudf/nvtext/replace.pxd b/python/pylibcudf/pylibcudf/nvtext/replace.pxd new file mode 100644 index 00000000000..624f90e7486 --- /dev/null +++ b/python/pylibcudf/pylibcudf/nvtext/replace.pxd @@ -0,0 +1,20 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from pylibcudf.column cimport Column +from pylibcudf.libcudf.types cimport size_type +from pylibcudf.scalar cimport Scalar + + +cpdef Column replace_tokens( + Column input, + Column targets, + Column replacements, + Scalar delimiter=*, +) + +cpdef Column filter_tokens( + Column input, + size_type min_token_length, + Scalar replacement=*, + Scalar delimiter=* +) diff --git a/python/pylibcudf/pylibcudf/nvtext/replace.pyx b/python/pylibcudf/pylibcudf/nvtext/replace.pyx new file mode 100644 index 00000000000..b65348ce14d --- /dev/null +++ b/python/pylibcudf/pylibcudf/nvtext/replace.pyx @@ -0,0 +1,109 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from cython.operator cimport dereference +from libcpp.memory cimport unique_ptr +from libcpp.utility cimport move +from pylibcudf.column cimport Column +from pylibcudf.libcudf.column.column cimport column +from pylibcudf.libcudf.nvtext.replace cimport ( + filter_tokens as cpp_filter_tokens, + replace_tokens as cpp_replace_tokens, +) +from pylibcudf.libcudf.scalar.scalar cimport string_scalar +from pylibcudf.libcudf.scalar.scalar_factories cimport ( + make_string_scalar as cpp_make_string_scalar, +) +from pylibcudf.libcudf.types cimport size_type +from pylibcudf.scalar cimport Scalar + + +cpdef Column replace_tokens( + Column input, + Column targets, + Column replacements, + Scalar delimiter=None, +): + """ + Replaces specified tokens with corresponding replacement strings. + + For details, see :cpp:func:`replace_tokens` + + Parameters + ---------- + input : Column + Strings column to replace + targets : Column + Strings to compare against tokens found in ``input`` + replacements : Column + Replacement strings for each string in ``targets`` + delimiter : Scalar, optional + Characters used to separate each string into tokens. + The default of empty string will identify tokens using whitespace. + + Returns + ------- + Column + New strings column with replaced strings + """ + cdef unique_ptr[column] c_result + if delimiter is None: + delimiter = Scalar.from_libcudf( + cpp_make_string_scalar("".encode()) + ) + with nogil: + c_result = cpp_replace_tokens( + input.view(), + targets.view(), + replacements.view(), + dereference(delimiter.get()), + ) + return Column.from_libcudf(move(c_result)) + + +cpdef Column filter_tokens( + Column input, + size_type min_token_length, + Scalar replacement=None, + Scalar delimiter=None +): + """ + Removes tokens whose lengths are less than a specified number of characters. + + For details, see :cpp:func:`filter_tokens` + + Parameters + ---------- + input : Column + Strings column to replace + min_token_length : size_type + The minimum number of characters to retain a + token in the output string + replacement : Scalar, optional + Optional replacement string to be used in place of removed tokens + delimiter : Scalar, optional + Characters used to separate each string into tokens. + The default of empty string will identify tokens using whitespace. + Returns + ------- + Column + New strings column of filtered strings + """ + cdef unique_ptr[column] c_result + if delimiter is None: + delimiter = Scalar.from_libcudf( + cpp_make_string_scalar("".encode()) + ) + if replacement is None: + replacement = Scalar.from_libcudf( + cpp_make_string_scalar("".encode()) + ) + + with nogil: + c_result = cpp_filter_tokens( + input.view(), + min_token_length, + dereference(replacement.get()), + dereference(delimiter.get()), + ) + + return Column.from_libcudf(move(c_result)) diff --git a/python/pylibcudf/pylibcudf/nvtext/stemmer.pxd b/python/pylibcudf/pylibcudf/nvtext/stemmer.pxd new file mode 100644 index 00000000000..48762efc01f --- /dev/null +++ b/python/pylibcudf/pylibcudf/nvtext/stemmer.pxd @@ -0,0 +1,14 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from libcpp cimport bool +from pylibcudf.column cimport Column +from pylibcudf.libcudf.nvtext.stemmer cimport letter_type +from pylibcudf.libcudf.types cimport size_type + +ctypedef fused ColumnOrSize: + Column + size_type + +cpdef Column is_letter(Column input, bool check_vowels, ColumnOrSize indices) + +cpdef Column porter_stemmer_measure(Column input) diff --git a/python/pylibcudf/pylibcudf/nvtext/stemmer.pyx b/python/pylibcudf/pylibcudf/nvtext/stemmer.pyx new file mode 100644 index 00000000000..854d1053624 --- /dev/null +++ b/python/pylibcudf/pylibcudf/nvtext/stemmer.pyx @@ -0,0 +1,76 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from libcpp cimport bool +from libcpp.memory cimport unique_ptr +from libcpp.utility cimport move +from pylibcudf.column cimport Column +from pylibcudf.libcudf.column.column cimport column +from pylibcudf.libcudf.nvtext.stemmer cimport ( + is_letter as cpp_is_letter, + letter_type, + porter_stemmer_measure as cpp_porter_stemmer_measure, +) +from pylibcudf.libcudf.types cimport size_type + + +cpdef Column is_letter( + Column input, + bool check_vowels, + ColumnOrSize indices +): + """ + Returns boolean column indicating if the character + or characters at the provided character index or + indices (respectively) are consonants or vowels + + For details, see :cpp:func:`is_letter` + + Parameters + ---------- + input : Column + Input strings + check_vowels : bool + If true, the check is for vowels. Otherwise the check is + for consonants. + indices : Union[Column, size_type] + The character position(s) to check in each string + + Returns + ------- + Column + New boolean column. + """ + cdef unique_ptr[column] c_result + + with nogil: + c_result = cpp_is_letter( + input.view(), + letter_type.VOWEL if check_vowels else letter_type.CONSONANT, + indices if ColumnOrSize is size_type else indices.view() + ) + + return Column.from_libcudf(move(c_result)) + + +cpdef Column porter_stemmer_measure(Column input): + """ + Returns the Porter Stemmer measurements of a strings column. + + For details, see :cpp:func:`porter_stemmer_measure` + + Parameters + ---------- + input : Column + Strings column of words to measure + + Returns + ------- + Column + New column of measure values + """ + cdef unique_ptr[column] c_result + + with nogil: + c_result = cpp_porter_stemmer_measure(input.view()) + + return Column.from_libcudf(move(c_result)) diff --git a/python/pylibcudf/pylibcudf/strings/CMakeLists.txt b/python/pylibcudf/pylibcudf/strings/CMakeLists.txt index 04dd131cd75..5d7fbd24b91 100644 --- a/python/pylibcudf/pylibcudf/strings/CMakeLists.txt +++ b/python/pylibcudf/pylibcudf/strings/CMakeLists.txt @@ -28,6 +28,7 @@ set(cython_sources regex_program.pyx repeat.pyx replace.pyx + replace_re.pyx side_type.pyx slice.pyx strip.pyx diff --git a/python/pylibcudf/pylibcudf/strings/__init__.pxd b/python/pylibcudf/pylibcudf/strings/__init__.pxd index 93c61f3f72c..da1c1c576c0 100644 --- a/python/pylibcudf/pylibcudf/strings/__init__.pxd +++ b/python/pylibcudf/pylibcudf/strings/__init__.pxd @@ -17,6 +17,7 @@ from . cimport ( regex_program, repeat, replace, + replace_re, side_type, slice, split, @@ -42,6 +43,7 @@ __all__ = [ "regex_program", "repeat", "replace", + "replace_re", "slice", "strip", "split", diff --git a/python/pylibcudf/pylibcudf/strings/__init__.py b/python/pylibcudf/pylibcudf/strings/__init__.py index d52b0405f1e..40fa8261905 100644 --- a/python/pylibcudf/pylibcudf/strings/__init__.py +++ b/python/pylibcudf/pylibcudf/strings/__init__.py @@ -17,6 +17,7 @@ regex_program, repeat, replace, + replace_re, side_type, slice, split, @@ -42,6 +43,7 @@ "regex_program", "repeat", "replace", + "replace_re", "slice", "strip", "split", diff --git a/python/pylibcudf/pylibcudf/strings/replace_re.pxd b/python/pylibcudf/pylibcudf/strings/replace_re.pxd new file mode 100644 index 00000000000..e27ccd55f7d --- /dev/null +++ b/python/pylibcudf/pylibcudf/strings/replace_re.pxd @@ -0,0 +1,30 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +from pylibcudf.column cimport Column +from pylibcudf.libcudf.types cimport size_type +from pylibcudf.scalar cimport Scalar +from pylibcudf.strings.regex_flags cimport regex_flags +from pylibcudf.strings.regex_program cimport RegexProgram + +ctypedef fused Replacement: + Column + Scalar + +ctypedef fused Patterns: + RegexProgram + list + + +cpdef Column replace_re( + Column input, + Patterns patterns, + Replacement replacement=*, + size_type max_replace_count=*, + regex_flags flags=* +) + +cpdef Column replace_with_backrefs( + Column input, + RegexProgram prog, + str replacement +) diff --git a/python/pylibcudf/pylibcudf/strings/replace_re.pyx b/python/pylibcudf/pylibcudf/strings/replace_re.pyx new file mode 100644 index 00000000000..ccc33fd4425 --- /dev/null +++ b/python/pylibcudf/pylibcudf/strings/replace_re.pyx @@ -0,0 +1,134 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. +from cython.operator cimport dereference +from libcpp.memory cimport unique_ptr +from libcpp.string cimport string +from libcpp.utility cimport move +from libcpp.vector cimport vector +from pylibcudf.column cimport Column +from pylibcudf.libcudf.column.column cimport column +from pylibcudf.libcudf.scalar.scalar cimport string_scalar +from pylibcudf.libcudf.scalar.scalar_factories cimport ( + make_string_scalar as cpp_make_string_scalar, +) +from pylibcudf.libcudf.strings cimport replace_re as cpp_replace_re +from pylibcudf.libcudf.types cimport size_type +from pylibcudf.scalar cimport Scalar +from pylibcudf.strings.regex_flags cimport regex_flags +from pylibcudf.strings.regex_program cimport RegexProgram + + +cpdef Column replace_re( + Column input, + Patterns patterns, + Replacement replacement=None, + size_type max_replace_count=-1, + regex_flags flags=regex_flags.DEFAULT, +): + """ + For each string, replaces any character sequence matching the given patterns + with the provided replacement. + + For details, see :cpp:func:`cudf::strings::replace_re` + + Parameters + ---------- + input : Column + Strings instance for this operation. + patterns: RegexProgram or list[str] + If RegexProgram, the regex to match to each string. + If list[str], a list of regex strings to search within each string. + replacement : Scalar or Column + If Scalar, the string used to replace the matched sequence in each string. + ``patterns`` must be a RegexProgram. + If Column, the strings used for replacement. + ``patterns`` must be a list[str]. + max_replace_count : int + The maximum number of times to replace the matched pattern + within each string. ``patterns`` must be a RegexProgram. + Default replaces every substring that is matched. + flags : RegexFlags + Regex flags for interpreting special characters in the patterns. + ``patterns`` must be a list[str] + + Returns + ------- + Column + New strings column + """ + cdef unique_ptr[column] c_result + cdef vector[string] c_patterns + + if Patterns is RegexProgram and Replacement is Scalar: + if replacement is None: + replacement = Scalar.from_libcudf( + cpp_make_string_scalar("".encode()) + ) + with nogil: + c_result = move( + cpp_replace_re.replace_re( + input.view(), + patterns.c_obj.get()[0], + dereference((replacement.get())), + max_replace_count + ) + ) + + return Column.from_libcudf(move(c_result)) + elif Patterns is list and Replacement is Column: + c_patterns.reserve(len(patterns)) + for pattern in patterns: + c_patterns.push_back(pattern.encode()) + + with nogil: + c_result = move( + cpp_replace_re.replace_re( + input.view(), + c_patterns, + replacement.view(), + flags, + ) + ) + + return Column.from_libcudf(move(c_result)) + else: + raise TypeError("Must pass either a RegexProgram and a Scalar or a list") + + +cpdef Column replace_with_backrefs( + Column input, + RegexProgram prog, + str replacement +): + """ + For each string, replaces any character sequence matching the given regex + using the replacement template for back-references. + + For details, see :cpp:func:`cudf::strings::replace_with_backrefs` + + Parameters + ---------- + input : Column + Strings instance for this operation. + + prog: RegexProgram + Regex program instance. + + replacement : str + The replacement template for creating the output string. + + Returns + ------- + Column + New strings column. + """ + cdef unique_ptr[column] c_result + cdef string c_replacement = replacement.encode() + + with nogil: + c_result = cpp_replace_re.replace_with_backrefs( + input.view(), + prog.c_obj.get()[0], + c_replacement, + ) + + return Column.from_libcudf(move(c_result)) diff --git a/python/pylibcudf/pylibcudf/tests/common/utils.py b/python/pylibcudf/pylibcudf/tests/common/utils.py index 9f389fa42c4..d95849ef371 100644 --- a/python/pylibcudf/pylibcudf/tests/common/utils.py +++ b/python/pylibcudf/pylibcudf/tests/common/utils.py @@ -7,10 +7,11 @@ import numpy as np import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from pyarrow.orc import write_table as orc_write_table from pyarrow.parquet import write_table as pq_write_table + +import pylibcudf as plc from pylibcudf.io.types import CompressionType diff --git a/python/pylibcudf/pylibcudf/tests/conftest.py b/python/pylibcudf/pylibcudf/tests/conftest.py index fdce6f353ca..a19a8835498 100644 --- a/python/pylibcudf/pylibcudf/tests/conftest.py +++ b/python/pylibcudf/pylibcudf/tests/conftest.py @@ -8,8 +8,9 @@ import numpy as np import pyarrow as pa -import pylibcudf as plc import pytest + +import pylibcudf as plc from pylibcudf.io.types import CompressionType sys.path.insert(0, os.path.join(os.path.dirname(__file__), "common")) diff --git a/python/pylibcudf/pylibcudf/tests/io/test_avro.py b/python/pylibcudf/pylibcudf/tests/io/test_avro.py index 0cd5064a697..3d9d99ffa61 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_avro.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_avro.py @@ -5,10 +5,11 @@ import fastavro import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_table_and_meta_eq +import pylibcudf as plc + avro_dtype_pairs = [ ("boolean", pa.bool_()), ("int", pa.int32()), diff --git a/python/pylibcudf/pylibcudf/tests/io/test_csv.py b/python/pylibcudf/pylibcudf/tests/io/test_csv.py index ab26f23418d..22c83acc47c 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_csv.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_csv.py @@ -5,9 +5,7 @@ import pandas as pd import pyarrow as pa -import pylibcudf as plc import pytest -from pylibcudf.io.types import CompressionType from utils import ( _convert_types, assert_table_and_meta_eq, @@ -15,6 +13,9 @@ write_source_str, ) +import pylibcudf as plc +from pylibcudf.io.types import CompressionType + # Shared kwargs to pass to make_source _COMMON_CSV_SOURCE_KWARGS = { "format": "csv", diff --git a/python/pylibcudf/pylibcudf/tests/io/test_json.py b/python/pylibcudf/pylibcudf/tests/io/test_json.py index 9d976fedf00..453e5ce32a8 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_json.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_json.py @@ -3,9 +3,7 @@ import pandas as pd import pyarrow as pa -import pylibcudf as plc import pytest -from pylibcudf.io.types import CompressionType from utils import ( assert_table_and_meta_eq, make_source, @@ -13,6 +11,9 @@ write_source_str, ) +import pylibcudf as plc +from pylibcudf.io.types import CompressionType + # Shared kwargs to pass to make_source _COMMON_JSON_SOURCE_KWARGS = {"format": "json", "orient": "records"} diff --git a/python/pylibcudf/pylibcudf/tests/io/test_orc.py b/python/pylibcudf/pylibcudf/tests/io/test_orc.py index 42b14b1feff..5ed660ba6cf 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_orc.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_orc.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import _convert_types, assert_table_and_meta_eq, make_source +import pylibcudf as plc + # Shared kwargs to pass to make_source _COMMON_ORC_SOURCE_KWARGS = {"format": "orc"} diff --git a/python/pylibcudf/pylibcudf/tests/io/test_parquet.py b/python/pylibcudf/pylibcudf/tests/io/test_parquet.py index f6e843ccf66..41298601539 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_parquet.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_parquet.py @@ -1,9 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from pyarrow.parquet import read_table +from utils import assert_table_and_meta_eq, make_source + +import pylibcudf as plc from pylibcudf.expressions import ( ASTOperator, ColumnNameReference, @@ -11,7 +13,6 @@ Literal, Operation, ) -from utils import assert_table_and_meta_eq, make_source # Shared kwargs to pass to make_source _COMMON_PARQUET_SOURCE_KWARGS = {"format": "parquet"} diff --git a/python/pylibcudf/pylibcudf/tests/io/test_source_sink_info.py b/python/pylibcudf/pylibcudf/tests/io/test_source_sink_info.py index 747f58ec8cf..0c43c363e55 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_source_sink_info.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_source_sink_info.py @@ -2,9 +2,10 @@ import io -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.fixture(params=[plc.io.SourceInfo, plc.io.SinkInfo]) def io_class(request): diff --git a/python/pylibcudf/pylibcudf/tests/io/test_timezone.py b/python/pylibcudf/pylibcudf/tests/io/test_timezone.py index 76b0424b2af..b3555013927 100644 --- a/python/pylibcudf/pylibcudf/tests/io/test_timezone.py +++ b/python/pylibcudf/pylibcudf/tests/io/test_timezone.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import zoneinfo -import pylibcudf as plc import pytest +import pylibcudf as plc + def test_make_timezone_transition_table(): if len(zoneinfo.TZPATH) == 0: diff --git a/python/pylibcudf/pylibcudf/tests/test_binaryops.py b/python/pylibcudf/pylibcudf/tests/test_binaryops.py index f784cb3c191..bbb08e8b95a 100644 --- a/python/pylibcudf/pylibcudf/tests/test_binaryops.py +++ b/python/pylibcudf/pylibcudf/tests/test_binaryops.py @@ -4,10 +4,11 @@ import numpy as np import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + def idfn(param): ltype, rtype, outtype, plc_op, _ = param diff --git a/python/pylibcudf/pylibcudf/tests/test_column_factories.py b/python/pylibcudf/pylibcudf/tests/test_column_factories.py index 8cedbc6d42f..e317362a76b 100644 --- a/python/pylibcudf/pylibcudf/tests/test_column_factories.py +++ b/python/pylibcudf/pylibcudf/tests/test_column_factories.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import DEFAULT_STRUCT_TESTING_TYPE, assert_column_eq +import pylibcudf as plc + EMPTY_COL_SIZE = 3 NUMERIC_TYPES = [ diff --git a/python/pylibcudf/pylibcudf/tests/test_column_from_device.py b/python/pylibcudf/pylibcudf/tests/test_column_from_device.py index 0e129fdf0ef..24cd6b9e35f 100644 --- a/python/pylibcudf/pylibcudf/tests/test_column_from_device.py +++ b/python/pylibcudf/pylibcudf/tests/test_column_from_device.py @@ -1,12 +1,13 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq import rmm +import pylibcudf as plc + VALID_TYPES = [ pa.int8(), pa.int16(), diff --git a/python/pylibcudf/pylibcudf/tests/test_contiguous_split.py b/python/pylibcudf/pylibcudf/tests/test_contiguous_split.py index 7a5c1664eed..6d8b5993964 100644 --- a/python/pylibcudf/pylibcudf/tests/test_contiguous_split.py +++ b/python/pylibcudf/pylibcudf/tests/test_contiguous_split.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_table_eq +import pylibcudf as plc + param_pyarrow_tables = [ pa.table([]), pa.table({"a": [1, 2, 3], "b": [4, 5, 6], "c": [7, 8, 9]}), diff --git a/python/pylibcudf/pylibcudf/tests/test_copying.py b/python/pylibcudf/pylibcudf/tests/test_copying.py index 628682d0a66..c0a41b96b1a 100644 --- a/python/pylibcudf/pylibcudf/tests/test_copying.py +++ b/python/pylibcudf/pylibcudf/tests/test_copying.py @@ -2,7 +2,6 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import ( DEFAULT_STRUCT_TESTING_TYPE, @@ -16,6 +15,8 @@ metadata_from_arrow_type, ) +import pylibcudf as plc + # TODO: consider moving this to conftest and "pairing" # it with pa_type, so that they don't get out of sync diff --git a/python/pylibcudf/pylibcudf/tests/test_datetime.py b/python/pylibcudf/pylibcudf/tests/test_datetime.py index 75930d59058..a80ab8d9f65 100644 --- a/python/pylibcudf/pylibcudf/tests/test_datetime.py +++ b/python/pylibcudf/pylibcudf/tests/test_datetime.py @@ -4,10 +4,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module", params=["s", "ms", "us", "ns"]) def datetime_column(has_nulls, request): diff --git a/python/pylibcudf/pylibcudf/tests/test_expressions.py b/python/pylibcudf/pylibcudf/tests/test_expressions.py index 5894ef4624c..6eabd6db617 100644 --- a/python/pylibcudf/pylibcudf/tests/test_expressions.py +++ b/python/pylibcudf/pylibcudf/tests/test_expressions.py @@ -1,8 +1,9 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest +import pylibcudf as plc + # We can't really evaluate these expressions, so just make sure # construction works properly diff --git a/python/pylibcudf/pylibcudf/tests/test_interop.py b/python/pylibcudf/pylibcudf/tests/test_interop.py index 01c998f16d4..af80b6e5978 100644 --- a/python/pylibcudf/pylibcudf/tests/test_interop.py +++ b/python/pylibcudf/pylibcudf/tests/test_interop.py @@ -1,8 +1,12 @@ # Copyright (c) 2024, NVIDIA CORPORATION. +import cupy as cp +import numpy as np import pyarrow as pa -import pylibcudf as plc import pytest +from utils import assert_table_eq + +import pylibcudf as plc def test_list_dtype_roundtrip(): @@ -66,3 +70,31 @@ def test_decimal_other(data_type): arrow_type = plc.interop.to_arrow(data_type, precision=precision) assert arrow_type == pa.decimal128(precision, 0) + + +def test_round_trip_dlpack_plc_table(): + expected = pa.table({"a": [1, 2, 3], "b": [5, 6, 7]}) + plc_table = plc.interop.from_arrow(expected) + result = plc.interop.from_dlpack(plc.interop.to_dlpack(plc_table)) + assert_table_eq(expected, result) + + +@pytest.mark.parametrize("array", [np.array, cp.array]) +def test_round_trip_dlpack_array(array): + arr = array([1, 2, 3]) + result = plc.interop.from_dlpack(arr.__dlpack__()) + expected = pa.table({"a": [1, 2, 3]}) + assert_table_eq(expected, result) + + +def test_to_dlpack_error(): + plc_table = plc.interop.from_arrow( + pa.table({"a": [1, None, 3], "b": [5, 6, 7]}) + ) + with pytest.raises(ValueError, match="Cannot create a DLPack tensor"): + plc.interop.from_dlpack(plc.interop.to_dlpack(plc_table)) + + +def test_from_dlpack_error(): + with pytest.raises(ValueError, match="Invalid PyCapsule object"): + plc.interop.from_dlpack(1) diff --git a/python/pylibcudf/pylibcudf/tests/test_join.py b/python/pylibcudf/pylibcudf/tests/test_join.py index 61e02f4d28d..f43a56046a4 100644 --- a/python/pylibcudf/pylibcudf/tests/test_join.py +++ b/python/pylibcudf/pylibcudf/tests/test_join.py @@ -2,9 +2,10 @@ import numpy as np import pyarrow as pa -import pylibcudf as plc from utils import assert_table_eq +import pylibcudf as plc + def test_cross_join(): left = pa.Table.from_arrays([[0, 1, 2], [3, 4, 5]], names=["a", "b"]) diff --git a/python/pylibcudf/pylibcudf/tests/test_json.py b/python/pylibcudf/pylibcudf/tests/test_json.py index 3d2955211f8..486a9524e92 100644 --- a/python/pylibcudf/pylibcudf/tests/test_json.py +++ b/python/pylibcudf/pylibcudf/tests/test_json.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def plc_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_labeling.py b/python/pylibcudf/pylibcudf/tests/test_labeling.py index f7fb7463b50..beacfc63ce5 100644 --- a/python/pylibcudf/pylibcudf/tests/test_labeling.py +++ b/python/pylibcudf/pylibcudf/tests/test_labeling.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.mark.parametrize("left_inclusive", [True, False]) @pytest.mark.parametrize("right_inclusive", [True, False]) diff --git a/python/pylibcudf/pylibcudf/tests/test_lists.py b/python/pylibcudf/pylibcudf/tests/test_lists.py index 2353a6ff8f9..f3ef555f11d 100644 --- a/python/pylibcudf/pylibcudf/tests/test_lists.py +++ b/python/pylibcudf/pylibcudf/tests/test_lists.py @@ -3,10 +3,11 @@ import numpy as np import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture def test_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_null_mask.py b/python/pylibcudf/pylibcudf/tests/test_null_mask.py index 3edcae59edc..cd3da856de2 100644 --- a/python/pylibcudf/pylibcudf/tests/test_null_mask.py +++ b/python/pylibcudf/pylibcudf/tests/test_null_mask.py @@ -1,12 +1,13 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest -from pylibcudf.null_mask import MaskState import rmm +import pylibcudf as plc +from pylibcudf.null_mask import MaskState + @pytest.fixture(params=[False, True]) def nullable(request): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_edit_distance.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_edit_distance.py index 7d93c471cc4..8b14e0db576 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_edit_distance.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_edit_distance.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def edit_distance_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_generate_ngrams.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_generate_ngrams.py index 5cf9874d595..fae4685f81b 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_generate_ngrams.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_generate_ngrams.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def input_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_jaccard.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_jaccard.py index d5a168426b1..05fe7b53c16 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_jaccard.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_jaccard.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def input_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_minhash.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_minhash.py index 4e389a63f90..ead9ee094af 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_minhash.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_minhash.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.fixture(scope="module", params=[pa.uint32(), pa.uint64()]) def minhash_input_data(request): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_ngrams_tokenize.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_ngrams_tokenize.py index 283a009288d..84748b5597e 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_ngrams_tokenize.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_ngrams_tokenize.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def input_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_normalize.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_normalize.py index fe28b83c09a..25b6d1389ec 100644 --- a/python/pylibcudf/pylibcudf/tests/test_nvtext_normalize.py +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_normalize.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def norm_spaces_input_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_replace.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_replace.py new file mode 100644 index 00000000000..65687f31c85 --- /dev/null +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_replace.py @@ -0,0 +1,64 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +import pyarrow as pa +import pytest +from utils import assert_column_eq + +import pylibcudf as plc + + +@pytest.fixture(scope="module") +def input_col(): + arr = ["the quick", "brown fox", "jumps*over the", "lazy dog"] + return pa.array(arr) + + +@pytest.fixture(scope="module") +def targets(): + arr = ["the quick", "brown fox", "jumps*over the", "lazy dog"] + return pa.array(arr) + + +@pytest.mark.parametrize("delim", ["*", None]) +def test_replace_tokens(input_col, targets, delim): + replacements = pa.array(["slow", "cat", "looked", "rat"]) + result = plc.nvtext.replace.replace_tokens( + plc.interop.from_arrow(input_col), + plc.interop.from_arrow(targets), + plc.interop.from_arrow(replacements), + plc.interop.from_arrow(pa.scalar(delim)) if delim else None, + ) + expected = pa.array(["slow", "cat", "jumps*over the", "rat"]) + if not delim: + expected = pa.array( + ["the quick", "brown fox", "jumps*over the", "lazy dog"] + ) + assert_column_eq(result, expected) + + +@pytest.mark.parametrize("min_token_length", [4, 5]) +@pytest.mark.parametrize("replace", ["---", None]) +@pytest.mark.parametrize("delim", ["*", None]) +def test_filter_tokens(input_col, min_token_length, replace, delim): + result = plc.nvtext.replace.filter_tokens( + plc.interop.from_arrow(input_col), + min_token_length, + plc.interop.from_arrow(pa.scalar(replace)) if replace else None, + plc.interop.from_arrow(pa.scalar(delim)) if delim else None, + ) + expected = pa.array( + ["the quick", "brown fox", "jumps*over the", "lazy dog"] + ) + if not delim and not replace and min_token_length == 4: + expected = pa.array([" quick", "brown ", "jumps*over ", "lazy "]) + if not delim and not replace and min_token_length == 5: + expected = pa.array([" quick", "brown ", "jumps*over ", " "]) + if not delim and replace == "---" and min_token_length == 4: + expected = pa.array( + ["--- quick", "brown ---", "jumps*over ---", "lazy ---"] + ) + if not delim and replace == "---" and min_token_length == 5: + expected = pa.array( + ["--- quick", "brown ---", "jumps*over ---", "--- ---"] + ) + assert_column_eq(result, expected) diff --git a/python/pylibcudf/pylibcudf/tests/test_nvtext_stemmer.py b/python/pylibcudf/pylibcudf/tests/test_nvtext_stemmer.py new file mode 100644 index 00000000000..e7f4a971f08 --- /dev/null +++ b/python/pylibcudf/pylibcudf/tests/test_nvtext_stemmer.py @@ -0,0 +1,48 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +import pyarrow as pa +import pytest +from utils import assert_column_eq + +import pylibcudf as plc + + +@pytest.fixture(scope="module") +def input_col(): + arr = ["trouble", "toy", "syzygy"] + return pa.array(arr) + + +@pytest.mark.parametrize("check_vowels", [True, False]) +@pytest.mark.parametrize("indices", [[3, 1, 4], 1]) +def test_is_letter(input_col, check_vowels, indices): + def is_letter(s, i, check): + vowels = "aeiouy" + return (s[i] in vowels) == check + + result = plc.nvtext.stemmer.is_letter( + plc.interop.from_arrow(input_col), + check_vowels, + plc.interop.from_arrow(pa.array(indices)) + if isinstance(indices, list) + else indices, + ) + expected = pa.array( + [ + is_letter( + s, + indices[i] if isinstance(indices, list) else indices, + check_vowels, + ) + for i, s in enumerate(input_col.to_pylist()) + ] + ) + assert_column_eq(result, expected) + + +def test_porter_stemmer_measure(input_col): + result = plc.nvtext.stemmer.porter_stemmer_measure( + plc.interop.from_arrow(input_col), + ) + expected = pa.array([1, 1, 2], type=pa.int32()) + assert_column_eq(result, expected) diff --git a/python/pylibcudf/pylibcudf/tests/test_partitioning.py b/python/pylibcudf/pylibcudf/tests/test_partitioning.py index 444d0089d2c..c55e54cebc6 100644 --- a/python/pylibcudf/pylibcudf/tests/test_partitioning.py +++ b/python/pylibcudf/pylibcudf/tests/test_partitioning.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_table_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def partitioning_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_quantiles.py b/python/pylibcudf/pylibcudf/tests/test_quantiles.py index bac56691306..e4a24fb1c98 100644 --- a/python/pylibcudf/pylibcudf/tests/test_quantiles.py +++ b/python/pylibcudf/pylibcudf/tests/test_quantiles.py @@ -3,10 +3,11 @@ import numpy as np import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq, assert_table_eq +import pylibcudf as plc + # Map pylibcudf interpolation options to pyarrow options interp_mapping = { plc.types.Interpolation.LINEAR: "linear", diff --git a/python/pylibcudf/pylibcudf/tests/test_regex_program.py b/python/pylibcudf/pylibcudf/tests/test_regex_program.py index 777315df538..52598f2c462 100644 --- a/python/pylibcudf/pylibcudf/tests/test_regex_program.py +++ b/python/pylibcudf/pylibcudf/tests/test_regex_program.py @@ -1,8 +1,9 @@ # Copyright (c) 2024, NVIDIA CORPORATION. -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.mark.parametrize("pat", ["(", "*", "\\"]) def test_regex_program_invalid(pat): diff --git a/python/pylibcudf/pylibcudf/tests/test_reshape.py b/python/pylibcudf/pylibcudf/tests/test_reshape.py index 01115bc363a..ef23e23766a 100644 --- a/python/pylibcudf/pylibcudf/tests/test_reshape.py +++ b/python/pylibcudf/pylibcudf/tests/test_reshape.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq, assert_table_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def reshape_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_round.py b/python/pylibcudf/pylibcudf/tests/test_round.py index 0b30316b9a0..2526580bc13 100644 --- a/python/pylibcudf/pylibcudf/tests/test_round.py +++ b/python/pylibcudf/pylibcudf/tests/test_round.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(params=["float32", "float64"]) def column(request, has_nulls): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_attributes.py b/python/pylibcudf/pylibcudf/tests/test_string_attributes.py index a1820def0b1..f461657281a 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_attributes.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_attributes.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture() def str_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_capitalize.py b/python/pylibcudf/pylibcudf/tests/test_string_capitalize.py index 176ccc55b96..3e31c75c38a 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_capitalize.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_capitalize.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def str_data(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_case.py b/python/pylibcudf/pylibcudf/tests/test_string_case.py index 233cc253b14..08ac371fd96 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_case.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_case.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def string_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_char_types.py b/python/pylibcudf/pylibcudf/tests/test_string_char_types.py index bcd030c019e..06b44210d74 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_char_types.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_char_types.py @@ -2,9 +2,10 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_all_characters_of_type(): pa_array = pa.array(["1", "A"]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_combine.py b/python/pylibcudf/pylibcudf/tests/test_string_combine.py index 4a7007a0d6b..eea3ac68e84 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_combine.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_combine.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + def test_concatenate_scalar_seperator(): plc_table = plc.interop.from_arrow( diff --git a/python/pylibcudf/pylibcudf/tests/test_string_contains.py b/python/pylibcudf/pylibcudf/tests/test_string_contains.py index 4e4dd7cbb00..ba9a4a7d3b8 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_contains.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_contains.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def target_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert.py b/python/pylibcudf/pylibcudf/tests/test_string_convert.py index 69f7a0fdd33..3f3f452c4f6 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture( scope="module", diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_booleans.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_booleans.py index 117c59ff1b8..b391d2b290e 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_booleans.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_booleans.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_to_booleans(): pa_array = pa.array(["true", None, "True"]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_datetime.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_datetime.py index f3e84286a36..c9368d858a4 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_datetime.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_datetime.py @@ -3,10 +3,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture def fmt(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_durations.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_durations.py index 6d704309bfd..2d3578e4e71 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_durations.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_durations.py @@ -3,10 +3,11 @@ from datetime import datetime, timedelta import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture( params=[ diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_fixed_point.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_fixed_point.py index b1c4d729604..012e722038e 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_fixed_point.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_fixed_point.py @@ -2,9 +2,10 @@ import decimal import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_to_fixed_point(): typ = pa.decimal128(38, 2) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_floats.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_floats.py index e9918fab559..8ee2b5075af 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_floats.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_floats.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_to_floats(): typ = pa.float32() diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_integers.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_integers.py index 6d1d565af30..01192c2d1f8 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_integers.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_integers.py @@ -1,8 +1,9 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_to_integers(): typ = pa.int8() diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_ipv4.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_ipv4.py index 4dc3e512624..b533809f106 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_ipv4.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_ipv4.py @@ -1,8 +1,9 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_ipv4_to_integers(): arr = pa.array(["123.45.67.890", None]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_lists.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_lists.py index 8591732b39e..737036a4f0f 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_lists.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_lists.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.mark.parametrize("na_rep", [None, pa.scalar("")]) @pytest.mark.parametrize("separators", [None, pa.array([",", "[", "]"])]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_convert_urls.py b/python/pylibcudf/pylibcudf/tests/test_string_convert_urls.py index fee8c3fb8f6..528736798c7 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_convert_urls.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_convert_urls.py @@ -2,9 +2,10 @@ import urllib import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_url_encode(): data = ["/home/nfs", None] diff --git a/python/pylibcudf/pylibcudf/tests/test_string_extract.py b/python/pylibcudf/pylibcudf/tests/test_string_extract.py index 788b86423c4..e70edf4fb33 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_extract.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_extract.py @@ -2,6 +2,7 @@ import pyarrow as pa import pyarrow.compute as pc + import pylibcudf as plc diff --git a/python/pylibcudf/pylibcudf/tests/test_string_find.py b/python/pylibcudf/pylibcudf/tests/test_string_find.py index db3b13a5aae..82ec18832a9 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_find.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_find.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def data_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_find_multiple.py b/python/pylibcudf/pylibcudf/tests/test_string_find_multiple.py index d6b37a388f0..fa9eee3594b 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_find_multiple.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_find_multiple.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_find_multiple(): arr = pa.array(["abc", "def"]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_findall.py b/python/pylibcudf/pylibcudf/tests/test_string_findall.py index debfad92d00..b73d812c898 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_findall.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_findall.py @@ -2,9 +2,10 @@ import re import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_findall(): arr = pa.array(["bunny", "rabbit", "hare", "dog"]) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_padding.py b/python/pylibcudf/pylibcudf/tests/test_string_padding.py index 2ba775d17ae..79498132097 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_padding.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_padding.py @@ -2,6 +2,7 @@ import pyarrow as pa import pyarrow.compute as pc + import pylibcudf as plc diff --git a/python/pylibcudf/pylibcudf/tests/test_string_repeat.py b/python/pylibcudf/pylibcudf/tests/test_string_repeat.py index 18b5d8bf4d0..c06c06be7c6 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_repeat.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_repeat.py @@ -2,9 +2,10 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.mark.parametrize("repeats", [pa.array([2, 2]), 2]) def test_repeat_strings(repeats): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_replace.py b/python/pylibcudf/pylibcudf/tests/test_string_replace.py index 5a9c2007b73..2c7d25133de 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_replace.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_replace.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def data_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_replace_re.py b/python/pylibcudf/pylibcudf/tests/test_string_replace_re.py new file mode 100644 index 00000000000..511f826441a --- /dev/null +++ b/python/pylibcudf/pylibcudf/tests/test_string_replace_re.py @@ -0,0 +1,72 @@ +# Copyright (c) 2024, NVIDIA CORPORATION. + +import pyarrow as pa +import pyarrow.compute as pc +import pytest +from utils import assert_column_eq + +import pylibcudf as plc + + +@pytest.mark.parametrize("max_replace_count", [-1, 1]) +def test_replace_re_regex_program_scalar(max_replace_count): + arr = pa.array(["foo", "fuz", None]) + pat = "f." + repl = "ba" + result = plc.strings.replace_re.replace_re( + plc.interop.from_arrow(arr), + plc.strings.regex_program.RegexProgram.create( + pat, plc.strings.regex_flags.RegexFlags.DEFAULT + ), + plc.interop.from_arrow(pa.scalar(repl)), + max_replace_count=max_replace_count, + ) + expected = pc.replace_substring_regex( + arr, + pat, + repl, + max_replacements=max_replace_count + if max_replace_count != -1 + else None, + ) + assert_column_eq(result, expected) + + +@pytest.mark.parametrize( + "flags", + [ + plc.strings.regex_flags.RegexFlags.DEFAULT, + plc.strings.regex_flags.RegexFlags.DOTALL, + ], +) +def test_replace_re_list_str_columns(flags): + arr = pa.array(["foo", "fuz", None]) + pats = ["oo", "uz"] + repls = ["a", "b"] + result = plc.strings.replace_re.replace_re( + plc.interop.from_arrow(arr), + pats, + plc.interop.from_arrow(pa.array(repls)), + flags=flags, + ) + expected = arr + for pat, repl in zip(pats, repls): + expected = pc.replace_substring_regex( + expected, + pat, + repl, + ) + assert_column_eq(result, expected) + + +def test_replace_with_backrefs(): + arr = pa.array(["Z756", None]) + result = plc.strings.replace_re.replace_with_backrefs( + plc.interop.from_arrow(arr), + plc.strings.regex_program.RegexProgram.create( + "(\\d)(\\d)", plc.strings.regex_flags.RegexFlags.DEFAULT + ), + "V\\2\\1", + ) + expected = pa.array(["ZV576", None]) + assert_column_eq(result, expected) diff --git a/python/pylibcudf/pylibcudf/tests/test_string_slice.py b/python/pylibcudf/pylibcudf/tests/test_string_slice.py index d9ce5591b98..1759f739e31 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_slice.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_slice.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture(scope="module") def pa_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_split_partition.py b/python/pylibcudf/pylibcudf/tests/test_string_split_partition.py index 80cae8d1c6b..4e80f19b814 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_split_partition.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_split_partition.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_table_eq +import pylibcudf as plc + @pytest.fixture def data_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_split_split.py b/python/pylibcudf/pylibcudf/tests/test_string_split_split.py index 2aeffac8209..450b336ce65 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_split_split.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_split_split.py @@ -2,10 +2,11 @@ import pyarrow as pa import pyarrow.compute as pc -import pylibcudf as plc import pytest from utils import assert_column_eq, assert_table_eq +import pylibcudf as plc + @pytest.fixture def data_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_strip.py b/python/pylibcudf/pylibcudf/tests/test_string_strip.py index 005e5e4a405..5869e5f4920 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_strip.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_strip.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + data_strings = [ "AbC", "123abc", diff --git a/python/pylibcudf/pylibcudf/tests/test_string_translate.py b/python/pylibcudf/pylibcudf/tests/test_string_translate.py index 2ae893e69fb..84fd3354ac6 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_translate.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_translate.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from utils import assert_column_eq +import pylibcudf as plc + @pytest.fixture def data_col(): diff --git a/python/pylibcudf/pylibcudf/tests/test_string_wrap.py b/python/pylibcudf/pylibcudf/tests/test_string_wrap.py index a1c820cd586..00442d866e9 100644 --- a/python/pylibcudf/pylibcudf/tests/test_string_wrap.py +++ b/python/pylibcudf/pylibcudf/tests/test_string_wrap.py @@ -2,9 +2,10 @@ import textwrap import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_wrap(): width = 12 diff --git a/python/pylibcudf/pylibcudf/tests/test_table.py b/python/pylibcudf/pylibcudf/tests/test_table.py index e822d6a97a8..ac39ef4c5c9 100644 --- a/python/pylibcudf/pylibcudf/tests/test_table.py +++ b/python/pylibcudf/pylibcudf/tests/test_table.py @@ -1,9 +1,10 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest +import pylibcudf as plc + @pytest.mark.parametrize( "arrow_tbl", diff --git a/python/pylibcudf/pylibcudf/tests/test_transform.py b/python/pylibcudf/pylibcudf/tests/test_transform.py index d5c618f07e4..49802fe64ac 100644 --- a/python/pylibcudf/pylibcudf/tests/test_transform.py +++ b/python/pylibcudf/pylibcudf/tests/test_transform.py @@ -3,9 +3,10 @@ import math import pyarrow as pa -import pylibcudf as plc from utils import assert_column_eq +import pylibcudf as plc + def test_nans_to_nulls(has_nans): if has_nans: diff --git a/python/pylibcudf/pylibcudf/tests/test_transpose.py b/python/pylibcudf/pylibcudf/tests/test_transpose.py index ac11123f680..b0c0bc72ead 100644 --- a/python/pylibcudf/pylibcudf/tests/test_transpose.py +++ b/python/pylibcudf/pylibcudf/tests/test_transpose.py @@ -1,10 +1,11 @@ # Copyright (c) 2024, NVIDIA CORPORATION. import pyarrow as pa -import pylibcudf as plc import pytest from packaging.version import parse +import pylibcudf as plc + @pytest.mark.skipif( parse(pa.__version__) < parse("16.0.0"), diff --git a/python/pylibcudf/pyproject.toml b/python/pylibcudf/pyproject.toml index ea5b3065896..a80c85a1fa8 100644 --- a/python/pylibcudf/pyproject.toml +++ b/python/pylibcudf/pyproject.toml @@ -53,48 +53,20 @@ test = [ Homepage = "https://github.com/rapidsai/cudf" Documentation = "https://docs.rapids.ai/api/cudf/stable/" -[tool.isort] -line_length = 79 -multi_line_output = 3 -include_trailing_comma = true -force_grid_wrap = 0 -combine_as_imports = true -order_by_type = true -known_dask = [ - "dask", - "distributed", - "dask_cuda", -] -known_rapids = [ - "rmm", -] -known_first_party = [ - "cudf", -] -default_section = "THIRDPARTY" -sections = [ - "FUTURE", - "STDLIB", - "THIRDPARTY", - "DASK", - "RAPIDS", - "FIRSTPARTY", - "LOCALFOLDER", -] -skip = [ - "thirdparty", - ".eggs", - ".git", - ".hg", - ".mypy_cache", - ".tox", - ".venv", - "_build", - "buck-out", - "build", - "dist", - "__init__.py", -] +[tool.ruff] +extend = "../../pyproject.toml" + +[tool.ruff.lint.isort] +combine-as-imports = true +known-first-party = ["cudf"] +section-order = ["future", "standard-library", "third-party", "dask", "rapids", "first-party", "local-folder"] + +[tool.ruff.lint.isort.sections] +dask = ["dask", "distributed", "dask_cuda"] +rapids = ["rmm"] + +[tool.ruff.lint.per-file-ignores] +"__init__.py" = ["E402", "F401"] [tool.pytest.ini_options] # --import-mode=importlib because two test_json.py exists and tests directory is not a structured module