Merge branch 'branch-21.10' of https://github.com/rapidsai/cudf into …

…8660

ci/gpu/build.sh

@@ -80,7 +80,7 @@ gpuci_mamba_retry install -y \
                   "rapids-notebook-env=$MINOR_VERSION.*" \
                   "dask-cuda=${MINOR_VERSION}" \
                   "rmm=$MINOR_VERSION.*" \
-                  "ucx-py=0.21.*"
+                  "ucx-py=0.22.*"
 
 # https://docs.rapids.ai/maintainers/depmgmt/
 # gpuci_mamba_retry remove --force rapids-build-env rapids-notebook-env

ci/gpu/java.sh

@@ -80,7 +80,7 @@ gpuci_conda_retry install -y \
                   "rapids-notebook-env=$MINOR_VERSION.*" \
                   "dask-cuda=${MINOR_VERSION}" \
                   "rmm=$MINOR_VERSION.*" \
-                  "ucx-py=0.21.*" \
+                  "ucx-py=0.22.*" \
                   "openjdk=8.*" \
                   "maven"
 

conda/recipes/libcudf/meta.yaml

@@ -53,8 +53,7 @@ test:
     - test -f $PREFIX/include/cudf/aggregation.hpp
     - test -f $PREFIX/include/cudf/ast/detail/expression_parser.hpp
     - test -f $PREFIX/include/cudf/ast/detail/operators.hpp
-    - test -f $PREFIX/include/cudf/ast/nodes.hpp
-    - test -f $PREFIX/include/cudf/ast/operators.hpp
+    - test -f $PREFIX/include/cudf/ast/expressions.hpp
     - test -f $PREFIX/include/cudf/binaryop.hpp
     - test -f $PREFIX/include/cudf/labeling/label_bins.hpp
     - test -f $PREFIX/include/cudf/column/column_factories.hpp
@@ -119,19 +118,25 @@ test:
     - test -f $PREFIX/include/cudf/hashing.hpp
     - test -f $PREFIX/include/cudf/interop.hpp
     - test -f $PREFIX/include/cudf/io/avro.hpp
+    - test -f $PREFIX/include/cudf/io/csv.hpp
     - test -f $PREFIX/include/cudf/io/data_sink.hpp
     - test -f $PREFIX/include/cudf/io/datasource.hpp
-    - test -f $PREFIX/include/cudf/io/orc_metadata.hpp
-    - test -f $PREFIX/include/cudf/io/csv.hpp
     - test -f $PREFIX/include/cudf/io/detail/avro.hpp
     - test -f $PREFIX/include/cudf/io/detail/csv.hpp
     - test -f $PREFIX/include/cudf/io/detail/json.hpp
     - test -f $PREFIX/include/cudf/io/detail/orc.hpp
     - test -f $PREFIX/include/cudf/io/detail/parquet.hpp
     - test -f $PREFIX/include/cudf/io/detail/utils.hpp
     - test -f $PREFIX/include/cudf/io/json.hpp
+    - test -f $PREFIX/include/cudf/io/orc_metadata.hpp
     - test -f $PREFIX/include/cudf/io/orc.hpp
     - test -f $PREFIX/include/cudf/io/parquet.hpp
+    - test -f $PREFIX/include/cudf/io/text/data_chunk_source_factories.hpp
+    - test -f $PREFIX/include/cudf/io/text/data_chunk_source.hpp
+    - test -f $PREFIX/include/cudf/io/text/detail/multistate.hpp
+    - test -f $PREFIX/include/cudf/io/text/detail/tile_state.hpp
+    - test -f $PREFIX/include/cudf/io/text/detail/trie.hpp
+    - test -f $PREFIX/include/cudf/io/text/multibyte_split.hpp
     - test -f $PREFIX/include/cudf/io/types.hpp
     - test -f $PREFIX/include/cudf/ipc.hpp
     - test -f $PREFIX/include/cudf/join.hpp

cpp/CMakeLists.txt

@@ -28,6 +28,17 @@ elseif(CMAKE_CUDA_ARCHITECTURES STREQUAL "")
   set(CUDF_BUILD_FOR_DETECTED_ARCHS TRUE)
 endif()
 
+file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-21.10/RAPIDS.cmake
+    ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
+include(${CMAKE_BINARY_DIR}/RAPIDS.cmake)
+
+include(rapids-cmake)
+include(rapids-cpm)
+include(rapids-cuda)
+include(rapids-export)
+include(rapids-find)
+
+
 project(CUDF VERSION 21.10.00 LANGUAGES C CXX)
 
 # Needed because GoogleBenchmark changes the state of FindThreads.cmake,
@@ -138,6 +149,9 @@ include(cmake/thirdparty/CUDF_GetArrow.cmake)
 include(cmake/thirdparty/CUDF_GetDLPack.cmake)
 # find libcu++
 include(cmake/thirdparty/CUDF_GetLibcudacxx.cmake)
+# find cuCollections
+# Should come after including thrust and libcudacxx
+include(cmake/thirdparty/CUDF_GetcuCollections.cmake)
 # find or install GoogleTest
 include(cmake/thirdparty/CUDF_GetGTest.cmake)
 # preprocess jitify-able kernels
@@ -153,6 +167,7 @@ add_library(cudf
     src/aggregation/aggregation.cu
     src/aggregation/result_cache.cpp
     src/ast/expression_parser.cpp
+    src/ast/expressions.cpp
     src/binaryop/binaryop.cpp
     src/binaryop/compiled/binary_ops.cu
     src/binaryop/compiled/Add.cu
@@ -284,14 +299,15 @@ add_library(cudf
     src/io/orc/writer_impl.cu
     src/io/parquet/compact_protocol_writer.cpp
     src/io/parquet/page_data.cu
-    src/io/parquet/page_dict.cu
+    src/io/parquet/chunk_dict.cu
     src/io/parquet/page_enc.cu
     src/io/parquet/page_hdr.cu
     src/io/parquet/parquet.cpp
     src/io/parquet/reader_impl.cu
     src/io/parquet/writer_impl.cu
     src/io/statistics/orc_column_statistics.cu
     src/io/statistics/parquet_column_statistics.cu
+    src/io/text/multibyte_split.cu
     src/io/utilities/column_buffer.cpp
     src/io/utilities/data_sink.cpp
     src/io/utilities/datasource.cpp
@@ -353,8 +369,9 @@ add_library(cudf
     src/reshape/interleave_columns.cu
     src/reshape/tile.cu
     src/rolling/grouped_rolling.cu
-    src/rolling/rolling.cu
     src/rolling/range_window_bounds.cpp
+    src/rolling/rolling.cu
+    src/rolling/rolling_collect_list.cu
     src/round/round.cu
     src/scalar/scalar.cpp
     src/scalar/scalar_factories.cpp
@@ -526,7 +543,8 @@ target_link_libraries(cudf
            PUBLIC ZLIB::ZLIB
                   ${ARROW_LIBRARIES}
                   cudf::Thrust
-                  rmm::rmm)
+                  rmm::rmm
+           PRIVATE cuco::cuco)
 
 if(CUDA_STATIC_RUNTIME)
     # Tell CMake what CUDA language runtime to use

cpp/benchmarks/CMakeLists.txt

@@ -245,3 +245,8 @@ ConfigureBench(STRINGS_BENCH
 # - json benchmark -------------------------------------------------------------------
 ConfigureBench(JSON_BENCH
   string/json_benchmark.cpp)
+
+###################################################################################################
+# - io benchmark ---------------------------------------------------------------------
+ConfigureBench(MULTIBYTE_SPLIT_BENCHMARK
+  io/text/multibyte_split_benchmark.cpp)

cpp/benchmarks/ast/transform_benchmark.cpp

@@ -95,22 +95,22 @@ static void BM_ast_transform(benchmark::State& state)
   // Note that a std::list is required here because of its guarantees against reference invalidation
   // when items are added or removed. References to items in a std::vector are not safe if the
   // vector must re-allocate.
-  auto expressions = std::list<cudf::ast::expression>();
+  auto expressions = std::list<cudf::ast::operation>();
 
   // Construct tree that chains additions like (((a + b) + c) + d)
   auto const op = cudf::ast::ast_operator::ADD;
   if (reuse_columns) {
-    expressions.push_back(cudf::ast::expression(op, column_refs.at(0), column_refs.at(0)));
+    expressions.push_back(cudf::ast::operation(op, column_refs.at(0), column_refs.at(0)));
     for (cudf::size_type i = 0; i < tree_levels - 1; i++) {
-      expressions.push_back(cudf::ast::expression(op, expressions.back(), column_refs.at(0)));
+      expressions.push_back(cudf::ast::operation(op, expressions.back(), column_refs.at(0)));
     }
   } else {
-    expressions.push_back(cudf::ast::expression(op, column_refs.at(0), column_refs.at(1)));
+    expressions.push_back(cudf::ast::operation(op, column_refs.at(0), column_refs.at(1)));
     std::transform(std::next(column_refs.cbegin(), 2),
                    column_refs.cend(),
                    std::back_inserter(expressions),
                    [&](auto const& column_ref) {
-                     return cudf::ast::expression(op, expressions.back(), column_ref);
+                     return cudf::ast::operation(op, expressions.back(), column_ref);
                    });
   }
 

cpp/benchmarks/binaryop/binaryop_benchmark.cpp

@@ -74,14 +74,14 @@ static void BM_binaryop_transform(benchmark::State& state)
     auto const op         = cudf::binary_operator::ADD;
     auto result_data_type = cudf::data_type(cudf::type_to_id<key_type>());
     if (reuse_columns) {
-      auto result = cudf::binary_operation(columns.at(0), columns.at(0), op, result_data_type);
+      auto result = cudf::jit::binary_operation(columns.at(0), columns.at(0), op, result_data_type);
       for (cudf::size_type i = 0; i < tree_levels - 1; i++) {
-        result = cudf::binary_operation(result->view(), columns.at(0), op, result_data_type);
+        result = cudf::jit::binary_operation(result->view(), columns.at(0), op, result_data_type);
       }
     } else {
-      auto result = cudf::binary_operation(columns.at(0), columns.at(1), op, result_data_type);
+      auto result = cudf::jit::binary_operation(columns.at(0), columns.at(1), op, result_data_type);
       std::for_each(std::next(columns.cbegin(), 2), columns.cend(), [&](auto const& col) {
-        result = cudf::binary_operation(result->view(), col, op, result_data_type);
+        result = cudf::jit::binary_operation(result->view(), col, op, result_data_type);
       });
     }
   }

cpp/benchmarks/binaryop/compiled_binaryop_benchmark.cpp

@@ -41,11 +41,11 @@ void BM_compiled_binaryop(benchmark::State& state, cudf::binary_operator binop)
   auto output_dtype = cudf::data_type(cudf::type_to_id<TypeOut>());
 
   // Call once for hot cache.
-  cudf::experimental::binary_operation(lhs, rhs, binop, output_dtype);
+  cudf::binary_operation(lhs, rhs, binop, output_dtype);
 
   for (auto _ : state) {
     cuda_event_timer timer(state, true);
-    cudf::experimental::binary_operation(lhs, rhs, binop, output_dtype);
+    cudf::binary_operation(lhs, rhs, binop, output_dtype);
   }
 }
 

cpp/benchmarks/binaryop/jit_binaryop_benchmark.cpp

@@ -41,11 +41,11 @@ void BM_binaryop(benchmark::State& state, cudf::binary_operator binop)
   auto output_dtype = cudf::data_type(cudf::type_to_id<TypeOut>());
 
   // Call once for hot cache.
-  cudf::binary_operation(lhs, rhs, binop, output_dtype);
+  cudf::jit::binary_operation(lhs, rhs, binop, output_dtype);
 
   for (auto _ : state) {
     cuda_event_timer timer(state, true);
-    cudf::binary_operation(lhs, rhs, binop, output_dtype);
+    cudf::jit::binary_operation(lhs, rhs, binop, output_dtype);
   }
 }
 

cpp/benchmarks/groupby/group_nth_benchmark.cu

@@ -63,7 +63,8 @@ void BM_pre_sorted_nth(benchmark::State& state)
   std::vector<cudf::groupby::aggregation_request> requests;
   requests.emplace_back(cudf::groupby::aggregation_request());
   requests[0].values = vals;
-  requests[0].aggregations.push_back(cudf::make_nth_element_aggregation(-1));
+  requests[0].aggregations.push_back(
+    cudf::make_nth_element_aggregation<cudf::groupby_aggregation>(-1));
 
   for (auto _ : state) {
     cuda_event_timer timer(state, true);

cpp/benchmarks/groupby/group_sum_benchmark.cu

@@ -58,7 +58,7 @@ void BM_basic_sum(benchmark::State& state)
   std::vector<cudf::groupby::aggregation_request> requests;
   requests.emplace_back(cudf::groupby::aggregation_request());
   requests[0].values = vals;
-  requests[0].aggregations.push_back(cudf::make_sum_aggregation());
+  requests[0].aggregations.push_back(cudf::make_sum_aggregation<cudf::groupby_aggregation>());
 
   for (auto _ : state) {
     cuda_event_timer timer(state, true);
@@ -97,7 +97,7 @@ void BM_pre_sorted_sum(benchmark::State& state)
   std::vector<cudf::groupby::aggregation_request> requests;
   requests.emplace_back(cudf::groupby::aggregation_request());
   requests[0].values = vals;
-  requests[0].aggregations.push_back(cudf::make_sum_aggregation());
+  requests[0].aggregations.push_back(cudf::make_sum_aggregation<cudf::groupby_aggregation>());
 
   for (auto _ : state) {
     cuda_event_timer timer(state, true);

cpp/benchmarks/io/cuio_benchmark_common.hpp

@@ -33,6 +33,8 @@ using cudf::io::io_type;
   benchmark(name##_buffer_output, type_or_group, static_cast<uint32_t>(io_type::HOST_BUFFER)); \
   benchmark(name##_void_output, type_or_group, static_cast<uint32_t>(io_type::VOID));
 
+std::string random_file_in_dir(std::string const& dir_path);
+
 /**
  * @brief Class to create a coupled `source_info` and `sink_info` of given type.
  */