Merge branch 'branch-25.02' into fea/extract-pq-bloom-filter-data

.github/workflows/pr.yaml

@@ -328,16 +328,11 @@ jobs:
         run_script: "ci/cudf_pandas_scripts/pandas-tests/diff.sh"
 
   telemetry-summarize:
-    runs-on: ubuntu-latest
+    # This job must use a self-hosted runner to record telemetry traces.
+    runs-on: linux-amd64-cpu4
     needs: pr-builder
     if: ${{ vars.TELEMETRY_ENABLED == 'true' && !cancelled() }}
     continue-on-error: true
     steps:
-      - name: Load stashed telemetry env vars
-        uses: rapidsai/shared-actions/telemetry-dispatch-load-base-env-vars@main
-        with:
-            load_service_name: true
       - name: Telemetry summarize
-        uses: rapidsai/shared-actions/telemetry-dispatch-write-summary@main
-        with:
-          cert_concat: "${{ secrets.OTEL_EXPORTER_OTLP_CA_CERTIFICATE }};${{ secrets.OTEL_EXPORTER_OTLP_CLIENT_CERTIFICATE }};${{ secrets.OTEL_EXPORTER_OTLP_CLIENT_KEY }}"
+        uses: rapidsai/shared-actions/telemetry-dispatch-summarize@main

cpp/CMakeLists.txt

@@ -276,7 +276,7 @@ rapids_cpm_init()
 
 # Not using rapids-cmake since we never want to find, always download.
 CPMAddPackage(
-  NAME rapids_logger GITHUB_REPOSITORY rapidsai/rapids-logger GIT_SHALLOW TRUE GIT_TAG
+  NAME rapids_logger GITHUB_REPOSITORY rapidsai/rapids-logger GIT_SHALLOW FALSE GIT_TAG
   c510947ae9d3a67530cfe3e5eaccb5a3b8ea0e55 VERSION c510947ae9d3a67530cfe3e5eaccb5a3b8ea0e55
 )
 rapids_make_logger(cudf EXPORT_SET cudf-exports)
@@ -917,7 +917,9 @@ if(CUDF_LARGE_STRINGS_DISABLED)
 endif()
 
 # Define logging level
-target_compile_definitions(cudf PRIVATE "CUDF_LOG_ACTIVE_LEVEL=${LIBCUDF_LOGGING_LEVEL}")
+target_compile_definitions(
+  cudf PRIVATE "CUDF_LOG_ACTIVE_LEVEL=CUDF_LOG_LEVEL_${LIBCUDF_LOGGING_LEVEL}"
+)
 
 # Enable remote IO through KvikIO
 target_compile_definitions(cudf PRIVATE $<$<BOOL:${CUDF_KVIKIO_REMOTE_IO}>:CUDF_KVIKIO_REMOTE_IO>)

cpp/cmake/thirdparty/get_nanoarrow.cmake

@@ -14,11 +14,6 @@
 
 # This function finds nanoarrow and sets any additional necessary environment variables.
 function(find_and_configure_nanoarrow)
-  include(${rapids-cmake-dir}/cpm/package_override.cmake)
-
-  set(cudf_patch_dir "${CMAKE_CURRENT_FUNCTION_LIST_DIR}/patches")
-  rapids_cpm_package_override("${cudf_patch_dir}/nanoarrow_override.json")
-
   if(NOT BUILD_SHARED_LIBS)
     set(_exclude_from_all EXCLUDE_FROM_ALL FALSE)
   else()
@@ -31,6 +26,9 @@ function(find_and_configure_nanoarrow)
     nanoarrow 0.6.0.dev
     GLOBAL_TARGETS nanoarrow
     CPM_ARGS
+    GIT_REPOSITORY https://github.com/apache/arrow-nanoarrow.git
+    GIT_TAG 1e2664a70ec14907409cadcceb14d79b9670bcdb
+    GIT_SHALLOW FALSE
     OPTIONS "BUILD_SHARED_LIBS OFF" "NANOARROW_NAMESPACE cudf" ${_exclude_from_all}
   )
   set_target_properties(nanoarrow PROPERTIES POSITION_INDEPENDENT_CODE ON)

cpp/include/cudf/detail/aggregation/aggregation.cuh

@@ -23,6 +23,7 @@
 #include <cudf/detail/utilities/assert.cuh>
 #include <cudf/detail/utilities/device_atomics.cuh>
 #include <cudf/table/table_view.hpp>
+#include <cudf/utilities/span.hpp>
 #include <cudf/utilities/traits.cuh>
 
 #include <rmm/cuda_stream_view.hpp>
@@ -31,7 +32,6 @@
 #include <thrust/fill.h>
 
 #include <type_traits>
-#include <vector>
 
 namespace cudf {
 namespace detail {
@@ -216,12 +216,12 @@ struct identity_initializer {
  * @throw cudf::logic_error if column type is not fixed-width
  *
  * @param table The table of columns to initialize.
- * @param aggs A vector of aggregation operations corresponding to the table
+ * @param aggs A span of aggregation operations corresponding to the table
  * columns. The aggregations determine the identity value for each column.
  * @param stream CUDA stream used for device memory operations and kernel launches.
  */
 void initialize_with_identity(mutable_table_view& table,
-                              std::vector<aggregation::Kind> const& aggs,
+                              host_span<cudf::aggregation::Kind const> aggs,
                               rmm::cuda_stream_view stream);
 
 }  // namespace detail

cpp/src/aggregation/aggregation.cu

@@ -17,15 +17,14 @@
 #include <cudf/detail/aggregation/aggregation.cuh>
 #include <cudf/detail/aggregation/aggregation.hpp>
 #include <cudf/table/table_view.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
 
-#include <vector>
-
 namespace cudf {
 namespace detail {
 void initialize_with_identity(mutable_table_view& table,
-                              std::vector<aggregation::Kind> const& aggs,
+                              host_span<cudf::aggregation::Kind const> aggs,
                               rmm::cuda_stream_view stream)
 {
   // TODO: Initialize all the columns in a single kernel instead of invoking one

cpp/src/groupby/hash/compute_aggregations.cuh

@@ -60,7 +60,7 @@ rmm::device_uvector<cudf::size_type> compute_aggregations(
   rmm::cuda_stream_view stream)
 {
   // flatten the aggs to a table that can be operated on by aggregate_row
-  auto [flattened_values, agg_kinds, aggs] = flatten_single_pass_aggs(requests);
+  auto [flattened_values, agg_kinds, aggs] = flatten_single_pass_aggs(requests, stream);
   auto const d_agg_kinds                   = cudf::detail::make_device_uvector_async(
     agg_kinds, stream, rmm::mr::get_current_device_resource());
 

cpp/src/groupby/hash/compute_global_memory_aggs.cu

@@ -24,7 +24,7 @@ template rmm::device_uvector<cudf::size_type> compute_global_memory_aggs<global_
   bitmask_type const* row_bitmask,
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> const& agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   global_set_t& global_set,
   std::vector<std::unique_ptr<aggregation>>& aggregations,
   cudf::detail::result_cache* sparse_results,

cpp/src/groupby/hash/compute_global_memory_aggs.cuh

@@ -25,6 +25,7 @@
 #include <cudf/groupby.hpp>
 #include <cudf/table/table_device_view.cuh>
 #include <cudf/types.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
 #include <rmm/device_uvector.hpp>
@@ -44,7 +45,7 @@ rmm::device_uvector<cudf::size_type> compute_global_memory_aggs(
   bitmask_type const* row_bitmask,
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> const& agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   SetType& global_set,
   std::vector<std::unique_ptr<aggregation>>& aggregations,
   cudf::detail::result_cache* sparse_results,

cpp/src/groupby/hash/compute_global_memory_aggs.hpp

@@ -19,6 +19,7 @@
 #include <cudf/groupby.hpp>
 #include <cudf/table/table_view.hpp>
 #include <cudf/types.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
 #include <rmm/device_uvector.hpp>
@@ -34,7 +35,7 @@ rmm::device_uvector<cudf::size_type> compute_global_memory_aggs(
   bitmask_type const* row_bitmask,
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> const& agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   SetType& global_set,
   std::vector<std::unique_ptr<aggregation>>& aggregations,
   cudf::detail::result_cache* sparse_results,

cpp/src/groupby/hash/compute_global_memory_aggs_null.cu

@@ -24,7 +24,7 @@ template rmm::device_uvector<cudf::size_type> compute_global_memory_aggs<nullabl
   bitmask_type const* row_bitmask,
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> const& agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   nullable_global_set_t& global_set,
   std::vector<std::unique_ptr<aggregation>>& aggregations,
   cudf::detail::result_cache* sparse_results,

cpp/src/groupby/hash/create_sparse_results_table.cu

@@ -23,6 +23,7 @@
 #include <cudf/detail/aggregation/aggregation.cuh>
 #include <cudf/table/table_view.hpp>
 #include <cudf/types.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
 #include <rmm/device_uvector.hpp>
@@ -48,7 +49,7 @@ void extract_populated_keys(SetType const& key_set,
 template <typename GlobalSetType>
 cudf::table create_sparse_results_table(cudf::table_view const& flattened_values,
                                         cudf::aggregation::Kind const* d_agg_kinds,
-                                        std::vector<cudf::aggregation::Kind> agg_kinds,
+                                        host_span<cudf::aggregation::Kind const> agg_kinds,
                                         bool direct_aggregations,
                                         GlobalSetType const& global_set,
                                         rmm::device_uvector<cudf::size_type>& populated_keys,
@@ -107,7 +108,7 @@ template void extract_populated_keys<nullable_global_set_t>(
 template cudf::table create_sparse_results_table<global_set_t>(
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   bool direct_aggregations,
   global_set_t const& global_set,
   rmm::device_uvector<cudf::size_type>& populated_keys,
@@ -116,7 +117,7 @@ template cudf::table create_sparse_results_table<global_set_t>(
 template cudf::table create_sparse_results_table<nullable_global_set_t>(
   cudf::table_view const& flattened_values,
   cudf::aggregation::Kind const* d_agg_kinds,
-  std::vector<cudf::aggregation::Kind> agg_kinds,
+  host_span<cudf::aggregation::Kind const> agg_kinds,
   bool direct_aggregations,
   nullable_global_set_t const& global_set,
   rmm::device_uvector<cudf::size_type>& populated_keys,

cpp/src/groupby/hash/create_sparse_results_table.hpp

@@ -20,12 +20,11 @@
 #include <cudf/table/table.hpp>
 #include <cudf/table/table_view.hpp>
 #include <cudf/types.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
 #include <rmm/device_uvector.hpp>
 
-#include <vector>
-
 namespace cudf::groupby::detail::hash {
 /**
  * @brief Computes and returns a device vector containing all populated keys in
@@ -47,7 +46,7 @@ void extract_populated_keys(SetType const& key_set,
 template <typename GlobalSetType>
 cudf::table create_sparse_results_table(cudf::table_view const& flattened_values,
                                         cudf::aggregation::Kind const* d_agg_kinds,
-                                        std::vector<cudf::aggregation::Kind> agg_kinds,
+                                        host_span<cudf::aggregation::Kind const> agg_kinds,
                                         bool direct_aggregations,
                                         GlobalSetType const& global_set,
                                         rmm::device_uvector<cudf::size_type>& populated_keys,

cpp/src/groupby/hash/flatten_single_pass_aggs.cpp

@@ -18,6 +18,7 @@
 
 #include <cudf/aggregation.hpp>
 #include <cudf/detail/aggregation/aggregation.hpp>
+#include <cudf/detail/utilities/vector_factories.hpp>
 #include <cudf/dictionary/dictionary_column_view.hpp>
 #include <cudf/groupby.hpp>
 #include <cudf/types.hpp>
@@ -102,12 +103,15 @@ class groupby_simple_aggregations_collector final
 };
 
 // flatten aggs to filter in single pass aggs
-std::tuple<table_view, std::vector<aggregation::Kind>, std::vector<std::unique_ptr<aggregation>>>
-flatten_single_pass_aggs(host_span<aggregation_request const> requests)
+std::tuple<table_view,
+           cudf::detail::host_vector<aggregation::Kind>,
+           std::vector<std::unique_ptr<aggregation>>>
+flatten_single_pass_aggs(host_span<aggregation_request const> requests,
+                         rmm::cuda_stream_view stream)
 {
   std::vector<column_view> columns;
   std::vector<std::unique_ptr<aggregation>> aggs;
-  std::vector<aggregation::Kind> agg_kinds;
+  auto agg_kinds = cudf::detail::make_empty_host_vector<aggregation::Kind>(requests.size(), stream);
 
   for (auto const& request : requests) {
     auto const& agg_v = request.aggregations;

cpp/src/groupby/hash/flatten_single_pass_aggs.hpp

@@ -26,7 +26,10 @@
 namespace cudf::groupby::detail::hash {
 
 // flatten aggs to filter in single pass aggs
-std::tuple<table_view, std::vector<aggregation::Kind>, std::vector<std::unique_ptr<aggregation>>>
-flatten_single_pass_aggs(host_span<aggregation_request const> requests);
+std::tuple<table_view,
+           cudf::detail::host_vector<aggregation::Kind>,
+           std::vector<std::unique_ptr<aggregation>>>
+flatten_single_pass_aggs(host_span<aggregation_request const> requests,
+                         rmm::cuda_stream_view stream);
 
 }  // namespace cudf::groupby::detail::hash

cpp/src/groupby/hash/hash_compound_agg_finalizer.cu

@@ -170,7 +170,8 @@ void hash_compound_agg_finalizer<SetType>::visit(cudf::detail::var_aggregation c
     cudf::detail::target_type(result_type, agg.kind), col.size(), mask_state::ALL_NULL, stream);
   auto var_result_view = mutable_column_device_view::create(var_result->mutable_view(), stream);
   mutable_table_view var_table_view{{var_result->mutable_view()}};
-  cudf::detail::initialize_with_identity(var_table_view, {agg.kind}, stream);
+  cudf::detail::initialize_with_identity(
+    var_table_view, host_span<cudf::aggregation::Kind const>(&agg.kind, 1), stream);
 
   thrust::for_each_n(
     rmm::exec_policy_nosync(stream),

cpp/src/groupby/sort/group_scan_util.cuh

@@ -107,7 +107,10 @@ struct group_scan_functor<K, T, std::enable_if_t<is_group_scan_supported<K, T>()
     if (values.is_empty()) { return result; }
 
     auto result_table = mutable_table_view({*result});
-    cudf::detail::initialize_with_identity(result_table, {K}, stream);
+    // Need an address of the aggregation kind to pass to the span
+    auto const kind = K;
+    cudf::detail::initialize_with_identity(
+      result_table, host_span<aggregation::Kind const>(&kind, 1), stream);
 
     auto result_view = mutable_column_device_view::create(result->mutable_view(), stream);
     auto values_view = column_device_view::create(values, stream);

cpp/tests/streams/replace_test.cpp

@@ -104,9 +104,9 @@ TEST_F(ReplaceTest, NormalizeNansAndZeros)
 
 TEST_F(ReplaceTest, NormalizeNansAndZerosMutable)
 {
-  auto nan          = std::numeric_limits<double>::quiet_NaN();
-  auto input_column = cudf::test::make_type_param_vector<double>({-0.0, 0.0, -nan, nan, nan});
-  cudf::test::fixed_width_column_wrapper<double> input(input_column.begin(), input_column.end());
-  cudf::mutable_column_view mutable_view = cudf::column(input, cudf::test::get_default_stream());
-  cudf::normalize_nans_and_zeros(mutable_view, cudf::test::get_default_stream());
+  auto nan   = std::numeric_limits<double>::quiet_NaN();
+  auto data  = cudf::test::make_type_param_vector<double>({-0.0, 0.0, -nan, nan, nan});
+  auto input = cudf::test::fixed_width_column_wrapper<double>(data.begin(), data.end()).release();
+  auto view  = input->mutable_view();
+  cudf::normalize_nans_and_zeros(view, cudf::test::get_default_stream());
 }

python/cudf/cudf/_lib/CMakeLists.txt

@@ -12,9 +12,7 @@
 # the License.
 # =============================================================================
 
-set(cython_sources column.pyx groupby.pyx interop.pyx scalar.pyx strings_udf.pyx types.pyx
-                   utils.pyx
-)
+set(cython_sources column.pyx groupby.pyx scalar.pyx strings_udf.pyx types.pyx utils.pyx)
 set(linked_libraries cudf::cudf)
 
 rapids_cython_create_modules(
@@ -24,8 +22,3 @@ rapids_cython_create_modules(
 )
 
 target_link_libraries(strings_udf PUBLIC cudf_strings_udf)
-target_include_directories(interop PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
-
-include(${rapids-cmake-dir}/export/find_package_root.cmake)
-include(../../../../cpp/cmake/thirdparty/get_nanoarrow.cmake)
-target_link_libraries(interop PUBLIC nanoarrow)

python/cudf/cudf/_lib/__init__.py

@@ -3,7 +3,6 @@
 
 from . import (
     groupby,
-    interop,
     strings_udf,
 )