Added ast tree to simplify expression lifetime management

cpp/include/cudf/ast/detail/expression_parser.hpp

@@ -21,6 +21,7 @@
 #include <cudf/table/table_view.hpp>
 #include <cudf/types.hpp>
 #include <cudf/utilities/memory_resource.hpp>
+#include <cudf/utilities/span.hpp>
 
 #include <thrust/scan.h>
 
@@ -300,7 +301,7 @@ class expression_parser {
    * @return The indices of the operands stored in the data references.
    */
   std::vector<cudf::size_type> visit_operands(
-    std::vector<std::reference_wrapper<expression const>> operands);
+    cudf::host_span<std::reference_wrapper<cudf::ast::expression const> const> operands);
 
   /**
    * @brief Add a data reference to the internal list.

cpp/include/cudf/ast/expressions.hpp

@@ -22,6 +22,8 @@
 #include <cudf/utilities/error.hpp>
 
 #include <cstdint>
+#include <memory>
+#include <vector>
 
 namespace CUDF_EXPORT cudf {
 namespace ast {
@@ -478,7 +480,7 @@ class operation : public expression {
    *
    * @return Vector of operands
    */
-  [[nodiscard]] std::vector<std::reference_wrapper<expression const>> get_operands() const
+  [[nodiscard]] std::vector<std::reference_wrapper<expression const>> const& get_operands() const
   {
     return operands;
   }
@@ -506,8 +508,8 @@ class operation : public expression {
   };
 
  private:
-  ast_operator const op;
-  std::vector<std::reference_wrapper<expression const>> const operands;
+  ast_operator op;
+  std::vector<std::reference_wrapper<expression const>> operands;
 };
 
 /**
@@ -552,6 +554,84 @@ class column_name_reference : public expression {
   std::string column_name;
 };
 
+/**
+ * @brief An AST expression tree. it owns and contains multiple dependent expressions. All the
+ * expressions are destroyed once the tree is destructed.
+ */
+class tree {
+ public:
+  /**
+   * @brief construct an empty ast tree
+   */
+  tree()                  = default;
+  tree(tree&&)            = default;
+  tree& operator=(tree&&) = default;
+  ~tree()                 = default;
+
+  // the tree is not copyable
+  tree(tree const&)            = delete;
+  tree& operator=(tree const&) = delete;
+
+  /**
+   * @brief Add an expression to the AST tree
+   * @param args Arguments to use to construct the ast expression
+   * @returns a reference to the added expression
+   */
+  template <typename Expr, typename... Args>
+  expression const& emplace(Args&&... args)
+  {
+    static_assert(std::is_base_of_v<expression, Expr>);
+    return *expressions.emplace_back(std::make_unique<Expr>(std::forward<Args>(args)...));
+  }
+
+  /**
+   * @brief Add an expression to the AST tree
+   * @param expr AST expression to be added
+   * @returns a reference to the added expression
+   */
+  template <typename Expr>
+  expression const& push(Expr expr)
+  {
+    return emplace<Expr>(std::move(expr));
+  }
+
+  /**
+   * @brief get the first expression in the tree
+   */
+  expression const& front() const { return *expressions.front(); }
+
+  /**
+   * @brief get the last expression in the tree
+   */
+  expression const& back() const { return *expressions.back(); }
+
+  /**
+   * @brief get the number of expressions added to the tree
+   */
+  size_t size() const { return expressions.size(); }
+
+  /**
+   * @brief get the expression at a checked index in the tree
+   * @returns the expression at the specified index
+   */
+  expression const& at(size_t index) { return *expressions.at(index); }
+
+  /**
+   * @brief get the expression at an unchecked index in the tree
+   * @returns the expression at the specified index
+   */
+  expression const& operator[](size_t index) const { return *expressions[index]; }
+
+  /**
+   * @brief get an immutable span to the expressions in the tree
+   * @returns all expressions added to the tree
+   */
+  std::vector<std::unique_ptr<expression>> const& get_expressions() const { return expressions; }
+
+ private:
+  std::vector<std::unique_ptr<expression>> expressions;
+};
+
 /** @} */  // end of group
 }  // namespace ast
 

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.
@@ -210,7 +210,7 @@ cudf::data_type expression_parser::output_type() const
 }
 
 std::vector<cudf::size_type> expression_parser::visit_operands(
-  std::vector<std::reference_wrapper<expression const>> operands)
+  cudf::host_span<std::reference_wrapper<expression const> const> operands)
 {
   auto operand_data_reference_indices = std::vector<cudf::size_type>();
   for (auto const& operand : operands) {

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.
@@ -23,36 +23,41 @@
 #include <cudf/types.hpp>
 #include <cudf/utilities/error.hpp>
 
+#include <stdexcept>
+
 namespace cudf {
 namespace ast {
 
-operation::operation(ast_operator op, expression const& input) : op(op), operands({input})
+operation::operation(ast_operator op, expression const& input) : op{op}, operands{input}
 {
-  if (cudf::ast::detail::ast_operator_arity(op) != 1) {
-    CUDF_FAIL("The provided operator is not a unary operator.");
-  }
+  CUDF_EXPECTS(cudf::ast::detail::ast_operator_arity(op) == 1,
+               "The provided operator is not a unary operator.",
+               std::invalid_argument);
 }
 
 operation::operation(ast_operator op, expression const& left, expression const& right)
-  : op(op), operands({left, right})
+  : op{op}, operands{left, right}
 {
-  if (cudf::ast::detail::ast_operator_arity(op) != 2) {
-    CUDF_FAIL("The provided operator is not a binary operator.");
-  }
+  CUDF_EXPECTS(cudf::ast::detail::ast_operator_arity(op) == 2,
+               "The provided operator is not a binary operator.",
+               std::invalid_argument);
 }
 
 cudf::size_type literal::accept(detail::expression_parser& visitor) const
 {
   return visitor.visit(*this);
 }
+
 cudf::size_type column_reference::accept(detail::expression_parser& visitor) const
 {
   return visitor.visit(*this);
 }
+
 cudf::size_type operation::accept(detail::expression_parser& visitor) const
 {
   return visitor.visit(*this);
 }
+
 cudf::size_type column_name_reference::accept(detail::expression_parser& visitor) const
 {
   return visitor.visit(*this);
@@ -63,16 +68,19 @@ auto literal::accept(detail::expression_transformer& visitor) const
 {
   return visitor.visit(*this);
 }
+
 auto column_reference::accept(detail::expression_transformer& visitor) const
   -> decltype(visitor.visit(*this))
 {
   return visitor.visit(*this);
 }
+
 auto operation::accept(detail::expression_transformer& visitor) const
   -> decltype(visitor.visit(*this))
 {
   return visitor.visit(*this);
 }
+
 auto column_name_reference::accept(detail::expression_transformer& visitor) const
   -> decltype(visitor.visit(*this))
 {

cpp/src/io/parquet/predicate_pushdown.cpp

@@ -26,6 +26,7 @@
 #include <cudf/utilities/default_stream.hpp>
 #include <cudf/utilities/error.hpp>
 #include <cudf/utilities/memory_resource.hpp>
+#include <cudf/utilities/span.hpp>
 #include <cudf/utilities/traits.hpp>
 #include <cudf/utilities/type_dispatcher.hpp>
 
@@ -374,7 +375,7 @@ class stats_expression_converter : public ast::detail::expression_transformer {
 
  private:
   std::vector<std::reference_wrapper<ast::expression const>> visit_operands(
-    std::vector<std::reference_wrapper<ast::expression const>> operands)
+    cudf::host_span<std::reference_wrapper<ast::expression const> const> operands)
   {
     std::vector<std::reference_wrapper<ast::expression const>> transformed_operands;
     for (auto const& operand : operands) {
@@ -554,7 +555,7 @@ std::reference_wrapper<ast::expression const> named_to_reference_converter::visi
 
 std::vector<std::reference_wrapper<ast::expression const>>
 named_to_reference_converter::visit_operands(
-  std::vector<std::reference_wrapper<ast::expression const>> operands)
+  cudf::host_span<std::reference_wrapper<ast::expression const> const> operands)
 {
   std::vector<std::reference_wrapper<ast::expression const>> transformed_operands;
   for (auto const& operand : operands) {
@@ -624,7 +625,7 @@ class names_from_expression : public ast::detail::expression_transformer {
   }
 
  private:
-  void visit_operands(std::vector<std::reference_wrapper<ast::expression const>> operands)
+  void visit_operands(cudf::host_span<std::reference_wrapper<ast::expression const> const> operands)
   {
     for (auto const& operand : operands) {
       operand.get().accept(*this);

cpp/src/io/parquet/reader_impl_helpers.hpp

@@ -425,7 +425,7 @@ class named_to_reference_converter : public ast::detail::expression_transformer
 
  private:
   std::vector<std::reference_wrapper<ast::expression const>> visit_operands(
-    std::vector<std::reference_wrapper<ast::expression const>> operands);
+    cudf::host_span<std::reference_wrapper<ast::expression const> const> operands);
 
   std::unordered_map<std::string, size_type> column_name_to_index;
   std::optional<std::reference_wrapper<ast::expression const>> _stats_expr;