Merge remote-tracking branch 'origin/branch-21.10' into refactor/remo…

…ve_copy_construct
rapidsai · Aug 8, 2021 · 2fdaccf · 2fdaccf
2 parents aecfb6b + a4eabf0
commit 2fdaccf
Show file tree

Hide file tree

Showing 59 changed files with 2,197 additions and 1,967 deletions.
diff --git a/conda/recipes/libcudf/meta.yaml b/conda/recipes/libcudf/meta.yaml
@@ -51,7 +51,6 @@ test:
     - test -f $PREFIX/lib/libcudf.so
     - test -f $PREFIX/lib/libcudftestutil.a
     - test -f $PREFIX/include/cudf/aggregation.hpp
-    - test -f $PREFIX/include/cudf/ast/transform.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
@@ -102,6 +101,7 @@ test:
     - test -f $PREFIX/include/cudf/detail/utilities/integer_utils.hpp
     - test -f $PREFIX/include/cudf/detail/utilities/int_fastdiv.h
     - test -f $PREFIX/include/cudf/detail/utilities/vector_factories.hpp
+    - test -f $PREFIX/include/cudf/detail/utilities/visitor_overload.hpp
     - test -f $PREFIX/include/cudf/dictionary/detail/concatenate.hpp
     - test -f $PREFIX/include/cudf/dictionary/detail/encode.hpp
     - test -f $PREFIX/include/cudf/dictionary/detail/merge.hpp

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
@@ -152,7 +152,6 @@ add_library(cudf
     src/aggregation/aggregation.cu
     src/aggregation/result_cache.cpp
     src/ast/expression_parser.cpp
-    src/ast/transform.cu
     src/binaryop/binaryop.cpp
     src/binaryop/compiled/binary_ops.cu
     src/binaryop/compiled/Add.cu
@@ -305,6 +304,7 @@ add_library(cudf
     src/join/cross_join.cu
     src/join/hash_join.cu
     src/join/join.cu
+    src/join/join_utils.cu
     src/join/semi_join.cu
     src/lists/contains.cu
     src/lists/combine/concatenate_list_elements.cu
@@ -436,6 +436,7 @@ add_library(cudf
     src/text/subword/wordpiece_tokenizer.cu
     src/text/tokenize.cu
     src/transform/bools_to_mask.cu
+    src/transform/compute_column.cu
     src/transform/encode.cu
     src/transform/mask_to_bools.cu
     src/transform/nans_to_nulls.cu

diff --git a/cpp/benchmarks/ast/transform_benchmark.cpp b/cpp/benchmarks/ast/transform_benchmark.cpp
@@ -14,10 +14,10 @@
  * limitations under the License.
  */
 
-#include <cudf/ast/transform.hpp>
 #include <cudf/column/column_factories.hpp>
 #include <cudf/table/table.hpp>
 #include <cudf/table/table_view.hpp>
+#include <cudf/transform.hpp>
 #include <cudf/types.hpp>
 #include <cudf/utilities/error.hpp>
 
@@ -119,7 +119,7 @@ static void BM_ast_transform(benchmark::State& state)
   // Execute benchmark
   for (auto _ : state) {
     cuda_event_timer raii(state, true);  // flush_l2_cache = true, stream = 0
-    cudf::ast::compute_column(table, expression_tree_root);
+    cudf::compute_column(table, expression_tree_root);
   }
 
   // Use the number of bytes read from global memory

diff --git a/cpp/include/cudf/ast/detail/expression_evaluator.cuh b/cpp/include/cudf/ast/detail/expression_evaluator.cuh
@@ -191,13 +191,13 @@ struct mutable_column_expression_result
 };
 
 /**
- * @brief Despite to a binary operator based on a single data type.
+ * @brief Dispatch to a binary operator based on a single data type.
  *
  * This functor is a dispatcher for binary operations that assumes that both
- * operands to a binary operation are of the same type. This assumption is
- * encoded in the one non-deducible template parameter LHS, the type of the
- * left-hand operand, which is then used as the template parameter for both the
- * left and right operands to the binary operator f.
+ * operands are of the same type. This assumption is encoded in the
+ * non-deducible template parameter LHS, the type of the left-hand operand,
+ * which is then used as the template parameter for both the left and right
+ * operands to the binary operator f.
  */
 struct single_dispatch_binary_operator {
   /**

diff --git a/cpp/include/cudf/ast/detail/expression_parser.hpp b/cpp/include/cudf/ast/detail/expression_parser.hpp
@@ -30,7 +30,7 @@ namespace ast {
 namespace detail {
 
 /**
- * @brief Enum defining data reference types used by a node.
+ * @brief Node data reference types.
  *
  * This enum is device-specific. For instance, intermediate data references are generated by the
  * linearization process but cannot be explicitly created by the user.

diff --git a/cpp/include/cudf/ast/detail/transform.cuh b/cpp/include/cudf/ast/detail/transform.cuh
diff --git a/cpp/include/cudf/ast/transform.hpp b/cpp/include/cudf/ast/transform.hpp
diff --git a/cpp/include/cudf/datetime.hpp b/cpp/include/cudf/datetime.hpp
@@ -206,6 +206,21 @@ std::unique_ptr<cudf::column> is_leap_year(
   cudf::column_view const& column,
   rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
 
+/**
+ * @brief Extract the number of days in the month
+ *
+ * output[i] contains the number of days in the month of date `column[i]`
+ * output[i] is null if `column[i]` is null
+ *
+ * @throw cudf::logic_error if input column datatype is not a TIMESTAMP
+ *
+ * @param cudf::column_view of the input datetime values
+ * @return cudf::column of datatype INT16 of days in month of the corresponding date
+ */
+std::unique_ptr<cudf::column> days_in_month(
+  cudf::column_view const& column,
+  rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
+
 /**
  * @brief  Returns the quarter of the date
  *

diff --git a/cpp/include/cudf/detail/transform.hpp b/cpp/include/cudf/detail/transform.hpp
@@ -16,6 +16,7 @@
 
 #pragma once
 
+#include <cudf/ast/nodes.hpp>
 #include <cudf/transform.hpp>
 
 #include <rmm/cuda_stream_view.hpp>
@@ -35,6 +36,17 @@ std::unique_ptr<column> transform(
   rmm::cuda_stream_view stream        = rmm::cuda_stream_default,
   rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
 
+/**
+ * @copydoc cudf::compute_column
+ *
+ * @param stream CUDA stream used for device memory operations and kernel launches.
+ */
+std::unique_ptr<column> compute_column(
+  table_view const table,
+  ast::expression const& expr,
+  rmm::cuda_stream_view stream        = rmm::cuda_stream_default,
+  rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
+
 /**
  * @copydoc cudf::nans_to_nulls
  *

diff --git a/cpp/include/cudf/detail/utilities/visitor_overload.hpp b/cpp/include/cudf/detail/utilities/visitor_overload.hpp
@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2021, 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
+
+namespace cudf::detail {
+
+/**
+ * @brief Helper class to support inline-overloading for all of a variant's alternative types
+ */
+template <class... Ts>
+struct visitor_overload : Ts... {
+  using Ts::operator()...;
+};
+template <class... Ts>
+visitor_overload(Ts...) -> visitor_overload<Ts...>;
+
+}  // namespace cudf::detail