duckdb 0.8.2-dev2700.0 → 0.8.2-dev2842.0

package/src/duckdb/src/include/duckdb/optimizer/join_order/join_order_optimizer.hpp

@@ -10,6 +10,7 @@
 
 #include "duckdb/common/unordered_map.hpp"
 #include "duckdb/common/unordered_set.hpp"
+#include "duckdb/optimizer/join_order/query_graph_manager.hpp"
 #include "duckdb/optimizer/join_order/join_relation.hpp"
 #include "duckdb/optimizer/join_order/cardinality_estimator.hpp"
 #include "duckdb/optimizer/join_order/query_graph.hpp"
@@ -22,22 +23,13 @@
 
 namespace duckdb {
 
-struct GenerateJoinRelation {
-	GenerateJoinRelation(JoinRelationSet &set, unique_ptr<LogicalOperator> op_p) : set(set), op(std::move(op_p)) {
-	}
-
-	JoinRelationSet &set;
-	unique_ptr<LogicalOperator> op;
-};
-
 class JoinOrderOptimizer {
 public:
-	explicit JoinOrderOptimizer(ClientContext &context)
-	    : context(context), cardinality_estimator(context), full_plan_found(false), must_update_full_plan(false) {
+	explicit JoinOrderOptimizer(ClientContext &context) : context(context), query_graph_manager(context) {
 	}
 
 	//! Perform join reordering inside a plan
-	unique_ptr<LogicalOperator> Optimize(unique_ptr<LogicalOperator> plan);
+	unique_ptr<LogicalOperator> Optimize(unique_ptr<LogicalOperator> plan, optional_ptr<RelationStats> stats = nullptr);
 
 	unique_ptr<JoinNode> CreateJoinTree(JoinRelationSet &set,
 	                                    const vector<reference<NeighborInfo>> &possible_connections, JoinNode &left,
@@ -45,16 +37,10 @@ public:
 
 private:
 	ClientContext &context;
-	//! The total amount of join pairs that have been considered
-	idx_t pairs = 0;
-	//! Set of all relations considered in the join optimizer
-	vector<unique_ptr<SingleJoinRelation>> relations;
-	//! A mapping of base table index -> index into relations array (relation number)
-	unordered_map<idx_t, idx_t> relation_mapping;
-	//! A structure holding all the created JoinRelationSet objects
-	JoinRelationSetManager set_manager;
-	//! The set of edges used in the join optimizer
-	QueryGraph query_graph;
+
+	//! manages the query graph, relations, and edges between relations
+	QueryGraphManager query_graph_manager;
+
 	//! The optimal join plan found for the specific JoinRelationSet*
 	unordered_map<JoinRelationSet *, unique_ptr<JoinNode>> plans;
 

package/src/duckdb/src/include/duckdb/optimizer/join_order/join_relation.hpp

@@ -11,19 +11,8 @@
 #include "duckdb/common/common.hpp"
 #include "duckdb/common/unordered_map.hpp"
 #include "duckdb/common/unordered_set.hpp"
-#include "duckdb/common/optional_ptr.hpp"
 
 namespace duckdb {
-class LogicalOperator;
-
-//! Represents a single relation and any metadata accompanying that relation
-struct SingleJoinRelation {
-	LogicalOperator &op;
-	optional_ptr<LogicalOperator> parent;
-
-	SingleJoinRelation(LogicalOperator &op, optional_ptr<LogicalOperator> parent) : op(op), parent(parent) {
-	}
-};
 
 //! Set of relations, used in the join graph.
 struct JoinRelationSet {

package/src/duckdb/src/include/duckdb/optimizer/join_order/plan_enumerator.hpp

@@ -0,0 +1,89 @@
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/optimizer/join_order/plan_enumerator.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include "duckdb/common/unordered_map.hpp"
+#include "duckdb/common/unordered_set.hpp"
+#include "duckdb/optimizer/join_order/join_relation.hpp"
+#include "duckdb/optimizer/join_order/cardinality_estimator.hpp"
+#include "duckdb/optimizer/join_order/query_graph.hpp"
+#include "duckdb/optimizer/join_order/join_node.hpp"
+#include "duckdb/optimizer/join_order/cost_model.hpp"
+#include "duckdb/parser/expression_map.hpp"
+#include "duckdb/common/reference_map.hpp"
+#include "duckdb/planner/logical_operator.hpp"
+#include "duckdb/planner/logical_operator_visitor.hpp"
+
+#include <functional>
+
+namespace duckdb {
+
+class QueryGraphManager;
+
+class PlanEnumerator {
+public:
+	explicit PlanEnumerator(QueryGraphManager &query_graph_manager, CostModel &cost_model,
+	                        const QueryGraphEdges &query_graph)
+	    : query_graph(query_graph), query_graph_manager(query_graph_manager), cost_model(cost_model),
+	      full_plan_found(false), must_update_full_plan(false) {
+	}
+
+	//! Perform the join order solving
+	unique_ptr<JoinNode> SolveJoinOrder();
+	void InitLeafPlans();
+
+	static unique_ptr<LogicalOperator> BuildSideProbeSideSwaps(unique_ptr<LogicalOperator> plan);
+
+private:
+	QueryGraphEdges const &query_graph;
+	//! The total amount of join pairs that have been considered
+	idx_t pairs = 0;
+	//! The set of edges used in the join optimizer
+	QueryGraphManager &query_graph_manager;
+	//! Cost model to evaluate cost of joins
+	CostModel &cost_model;
+	//! A map to store the optimal join plan found for a specific JoinRelationSet*
+	reference_map_t<JoinRelationSet, unique_ptr<JoinNode>> plans;
+
+	bool full_plan_found;
+	bool must_update_full_plan;
+	unordered_set<string> join_nodes_in_full_plan;
+
+	unique_ptr<JoinNode> CreateJoinTree(JoinRelationSet &set,
+	                                    const vector<reference<NeighborInfo>> &possible_connections, JoinNode &left,
+	                                    JoinNode &right);
+
+	//! Emit a pair as a potential join candidate. Returns the best plan found for the (left, right) connection (either
+	//! the newly created plan, or an existing plan)
+	JoinNode &EmitPair(JoinRelationSet &left, JoinRelationSet &right, const vector<reference<NeighborInfo>> &info);
+	//! Tries to emit a potential join candidate pair. Returns false if too many pairs have already been emitted,
+	//! cancelling the dynamic programming step.
+	bool TryEmitPair(JoinRelationSet &left, JoinRelationSet &right, const vector<reference<NeighborInfo>> &info);
+
+	bool EnumerateCmpRecursive(JoinRelationSet &left, JoinRelationSet &right, unordered_set<idx_t> &exclusion_set);
+	//! Emit a relation set node
+	bool EmitCSG(JoinRelationSet &node);
+	//! Enumerate the possible connected subgraphs that can be joined together in the join graph
+	bool EnumerateCSGRecursive(JoinRelationSet &node, unordered_set<idx_t> &exclusion_set);
+	//! Generate cross product edges inside the side
+	void GenerateCrossProducts();
+
+	//! Solve the join order exactly using dynamic programming. Returns true if it was completed successfully (i.e. did
+	//! not time-out)
+	bool SolveJoinOrderExactly();
+	//! Solve the join order approximately using a greedy algorithm
+	void SolveJoinOrderApproximately();
+
+	void UpdateDPTree(JoinNode &new_plan);
+
+	void UpdateJoinNodesInFullPlan(JoinNode &node);
+	bool NodeInFullPlan(JoinNode &node);
+};
+
+} // namespace duckdb

package/src/duckdb/src/include/duckdb/optimizer/join_order/query_graph.hpp

@@ -9,47 +9,32 @@
 #pragma once
 
 #include "duckdb/common/common.hpp"
+#include "duckdb/common/optional_ptr.hpp"
+#include "duckdb/optimizer/join_order/join_relation.hpp"
+#include "duckdb/optimizer/join_order/join_node.hpp"
+#include "duckdb/optimizer/join_order/relation_manager.hpp"
 #include "duckdb/common/pair.hpp"
 #include "duckdb/common/unordered_map.hpp"
 #include "duckdb/common/unordered_set.hpp"
-#include "duckdb/optimizer/join_order/join_relation.hpp"
 #include "duckdb/common/vector.hpp"
 #include "duckdb/planner/column_binding.hpp"
-#include "duckdb/common/optional_ptr.hpp"
 
 #include <functional>
 
 namespace duckdb {
-class Expression;
-class LogicalOperator;
-
-struct FilterInfo {
-	FilterInfo(JoinRelationSet &set, idx_t filter_index) : set(set), filter_index(filter_index) {
-	}
-
-	JoinRelationSet &set;
-	idx_t filter_index;
-	optional_ptr<JoinRelationSet> left_set;
-	optional_ptr<JoinRelationSet> right_set;
-	ColumnBinding left_binding;
-	ColumnBinding right_binding;
-};
 
-struct FilterNode {
-	vector<reference<FilterInfo>> filters;
-	unordered_map<idx_t, unique_ptr<FilterNode>> children;
-};
+struct FilterInfo;
 
 struct NeighborInfo {
-	NeighborInfo(JoinRelationSet &neighbor) : neighbor(neighbor) {
+	NeighborInfo(optional_ptr<JoinRelationSet> neighbor) : neighbor(neighbor) {
 	}
 
-	JoinRelationSet &neighbor;
-	vector<reference<FilterInfo>> filters;
+	optional_ptr<JoinRelationSet> neighbor;
+	vector<optional_ptr<FilterInfo>> filters;
 };
 
 //! The QueryGraph contains edges between relations and allows edges to be created/queried
-class QueryGraph {
+class QueryGraphEdges {
 public:
 	//! Contains a node with info about neighboring relations and child edge infos
 	struct QueryEdge {
@@ -61,22 +46,23 @@ public:
 	string ToString() const;
 	void Print();
 
-	//! Create an edge in the edge_set
-	void CreateEdge(JoinRelationSet &left, JoinRelationSet &right, optional_ptr<FilterInfo> info);
 	//! Returns a connection if there is an edge that connects these two sets, or nullptr otherwise
-	vector<reference<NeighborInfo>> GetConnections(JoinRelationSet &node, JoinRelationSet &other);
+	const vector<reference<NeighborInfo>> GetConnections(JoinRelationSet &node, JoinRelationSet &other) const;
 	//! Enumerate the neighbors of a specific node that do not belong to any of the exclusion_set. Note that if a
 	//! neighbor has multiple nodes, this function will return the lowest entry in that set.
-	vector<idx_t> GetNeighbors(JoinRelationSet &node, unordered_set<idx_t> &exclusion_set);
+	const vector<idx_t> GetNeighbors(JoinRelationSet &node, unordered_set<idx_t> &exclusion_set) const;
+
 	//! Enumerate all neighbors of a given JoinRelationSet node
-	void EnumerateNeighbors(JoinRelationSet &node, const std::function<bool(NeighborInfo &)> &callback);
+	void EnumerateNeighbors(JoinRelationSet &node, const std::function<bool(NeighborInfo &)> &callback) const;
+	//! Create an edge in the edge_set
+	void CreateEdge(JoinRelationSet &left, JoinRelationSet &right, optional_ptr<FilterInfo> info);
 
 private:
 	//! Get the QueryEdge of a specific node
-	QueryEdge &GetQueryEdge(JoinRelationSet &left);
+	optional_ptr<QueryEdge> GetQueryEdge(JoinRelationSet &left);
 
 	void EnumerateNeighborsDFS(JoinRelationSet &node, reference<QueryEdge> info, idx_t index,
-	                           const std::function<bool(NeighborInfo &)> &callback);
+	                           const std::function<bool(NeighborInfo &)> &callback) const;
 
 	QueryEdge root;
 };

package/src/duckdb/src/include/duckdb/optimizer/join_order/query_graph_manager.hpp

@@ -0,0 +1,113 @@
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/optimizer/join_order/query_graph_manager.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include "duckdb/common/common.hpp"
+#include "duckdb/common/optional_ptr.hpp"
+#include "duckdb/common/pair.hpp"
+#include "duckdb/common/unordered_map.hpp"
+#include "duckdb/common/unordered_set.hpp"
+#include "duckdb/common/vector.hpp"
+#include "duckdb/optimizer/join_order/join_node.hpp"
+#include "duckdb/optimizer/join_order/join_relation.hpp"
+#include "duckdb/optimizer/join_order/query_graph.hpp"
+#include "duckdb/optimizer/join_order/relation_manager.hpp"
+#include "duckdb/planner/column_binding.hpp"
+#include "duckdb/planner/logical_operator.hpp"
+
+#include <functional>
+
+namespace duckdb {
+
+struct GenerateJoinRelation {
+	GenerateJoinRelation(optional_ptr<JoinRelationSet> set, unique_ptr<LogicalOperator> op_p)
+	    : set(set), op(std::move(op_p)) {
+	}
+
+	optional_ptr<JoinRelationSet> set;
+	unique_ptr<LogicalOperator> op;
+};
+
+//! Filter info struct that is used by the cardinality estimator to set the initial cardinality
+//! but is also eventually transformed into a query edge.
+struct FilterInfo {
+	FilterInfo(unique_ptr<Expression> filter, JoinRelationSet &set, idx_t filter_index)
+	    : filter(std::move(filter)), set(set), filter_index(filter_index) {
+	}
+
+	unique_ptr<Expression> filter;
+	JoinRelationSet &set;
+	idx_t filter_index;
+	optional_ptr<JoinRelationSet> left_set;
+	optional_ptr<JoinRelationSet> right_set;
+	ColumnBinding left_binding;
+	ColumnBinding right_binding;
+};
+
+//! The QueryGraphManager manages the process of extracting the reorderable and nonreorderable operations
+//! from the logical plan and creating the intermediate structures needed by the plan enumerator.
+//! When the plan enumerator finishes, the Query Graph Manger can then recreate the logical plan.
+class QueryGraphManager {
+public:
+	QueryGraphManager(ClientContext &context) : relation_manager(context), context(context) {
+	}
+
+	//! manage relations and the logical operators they represent
+	RelationManager relation_manager;
+
+	//! A structure holding all the created JoinRelationSet objects
+	JoinRelationSetManager set_manager;
+
+	ClientContext &context;
+
+	//! Extract the join relations, optimizing non-reoderable relations when encountered
+	bool Build(LogicalOperator &op);
+
+	//! Reconstruct the logical plan using the plan found by the plan enumerator
+	unique_ptr<LogicalOperator> Reconstruct(unique_ptr<LogicalOperator> plan, JoinNode &node);
+
+	//! Get a reference to the QueryGraphEdges structure that stores edges between
+	//! nodes and hypernodes.
+	const QueryGraphEdges &GetQueryGraphEdges() const;
+
+	//! Get a list of the join filters in the join plan than eventually are
+	//! transformed into the query graph edges
+	const vector<unique_ptr<FilterInfo>> &GetFilterBindings() const;
+
+	//! Plan enumerator may not find a full plan and therefore will need to create cross
+	//! products to create edges.
+	void CreateQueryGraphCrossProduct(JoinRelationSet &left, JoinRelationSet &right);
+
+	//! after join order optimization, we perform build side probe side optimizations.
+	//! (Basically we put lower expected cardinality columns on the build side, and larger
+	//! tables on the probe side)
+	unique_ptr<LogicalOperator> LeftRightOptimizations(unique_ptr<LogicalOperator> op);
+
+private:
+	vector<reference<LogicalOperator>> filter_operators;
+
+	//! Filter information including the column_bindings that join filters
+	//! used by the cardinality estimator to estimate distinct counts
+	vector<unique_ptr<FilterInfo>> filters_and_bindings;
+
+	QueryGraphEdges query_graph;
+
+	void GetColumnBinding(Expression &expression, ColumnBinding &binding);
+
+	bool ExtractBindings(Expression &expression, unordered_set<idx_t> &bindings);
+	bool LeftCardLessThanRight(LogicalOperator &op);
+
+	void CreateHyperGraphEdges();
+
+	GenerateJoinRelation GenerateJoins(vector<unique_ptr<LogicalOperator>> &extracted_relations, JoinNode &node);
+
+	unique_ptr<LogicalOperator> RewritePlan(unique_ptr<LogicalOperator> plan, JoinNode &node);
+};
+
+} // namespace duckdb

package/src/duckdb/src/include/duckdb/optimizer/join_order/relation_manager.hpp

@@ -0,0 +1,73 @@
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/optimizer/join_order/relation_manager.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include "duckdb/common/unordered_map.hpp"
+#include "duckdb/common/unordered_set.hpp"
+#include "duckdb/optimizer/join_order/join_relation.hpp"
+#include "duckdb/optimizer/join_order/cardinality_estimator.hpp"
+#include "duckdb/optimizer/join_order/relation_statistics_helper.hpp"
+#include "duckdb/optimizer/join_order/join_node.hpp"
+#include "duckdb/parser/expression_map.hpp"
+#include "duckdb/planner/logical_operator.hpp"
+#include "duckdb/planner/logical_operator_visitor.hpp"
+
+namespace duckdb {
+
+struct FilterInfo;
+
+//! Represents a single relation and any metadata accompanying that relation
+struct SingleJoinRelation {
+	LogicalOperator &op;
+	optional_ptr<LogicalOperator> parent;
+	RelationStats stats;
+
+	SingleJoinRelation(LogicalOperator &op, optional_ptr<LogicalOperator> parent) : op(op), parent(parent) {
+	}
+	SingleJoinRelation(LogicalOperator &op, optional_ptr<LogicalOperator> parent, RelationStats stats)
+	    : op(op), parent(parent), stats(stats) {
+	}
+};
+
+class RelationManager {
+public:
+	explicit RelationManager(ClientContext &context) : context(context) {
+	}
+
+	idx_t NumRelations();
+
+	bool ExtractJoinRelations(LogicalOperator &input_op, vector<reference<LogicalOperator>> &filter_operators,
+	                          optional_ptr<LogicalOperator> parent = nullptr);
+
+	//! for each join filter in the logical plan op, extract the relations that are referred to on
+	//! both sides of the join filter, along with the tables & indexes.
+	vector<unique_ptr<FilterInfo>> ExtractEdges(LogicalOperator &op,
+	                                            vector<reference<LogicalOperator>> &filter_operators,
+	                                            JoinRelationSetManager &set_manager);
+
+	//! Extract the set of relations referred to inside an expression
+	bool ExtractBindings(Expression &expression, unordered_set<idx_t> &bindings);
+	void AddRelation(LogicalOperator &op, optional_ptr<LogicalOperator> parent, const RelationStats &stats);
+
+	void AddAggregateRelation(LogicalOperator &op, optional_ptr<LogicalOperator> parent, const RelationStats &stats);
+	vector<unique_ptr<SingleJoinRelation>> GetRelations();
+
+	const vector<RelationStats> GetRelationStats();
+	//! A mapping of base table index -> index into relations array (relation number)
+	unordered_map<idx_t, idx_t> relation_mapping;
+
+	void PrintRelationStats();
+
+private:
+	ClientContext &context;
+	//! Set of all relations considered in the join optimizer
+	vector<unique_ptr<SingleJoinRelation>> relations;
+};
+
+} // namespace duckdb

package/src/duckdb/src/include/duckdb/optimizer/join_order/relation_statistics_helper.hpp

@@ -0,0 +1,73 @@
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/optimizer/join_order/statistics_extractor.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+#pragma once
+
+#include "duckdb/planner/filter/conjunction_filter.hpp"
+#include "duckdb/planner/logical_operator.hpp"
+
+namespace duckdb {
+
+class CardinalityEstimator;
+
+struct DistinctCount {
+	idx_t distinct_count;
+	bool from_hll;
+};
+
+struct ExpressionBinding {
+	bool found_expression = false;
+	ColumnBinding child_binding;
+	bool expression_is_constant = false;
+};
+
+struct RelationStats {
+	// column_id -> estimated distinct count for column
+	vector<DistinctCount> column_distinct_count;
+	idx_t cardinality;
+	double filter_strength = 1;
+	bool stats_initialized = false;
+
+	// for debug, column names and tables
+	vector<string> column_names;
+	string table_name;
+
+	RelationStats() : cardinality(1), filter_strength(1), stats_initialized(false) {
+	}
+};
+
+class RelationStatisticsHelper {
+public:
+	static constexpr double DEFAULT_SELECTIVITY = 0.2;
+
+public:
+	static idx_t InspectConjunctionAND(idx_t cardinality, idx_t column_index, ConjunctionAndFilter &filter,
+	                                   BaseStatistics &base_stats);
+	//	static idx_t InspectConjunctionOR(idx_t cardinality, idx_t column_index, ConjunctionOrFilter &filter,
+	//	                                  BaseStatistics &base_stats);
+	//! Extract Statistics from a LogicalGet.
+	static RelationStats ExtractGetStats(LogicalGet &get, ClientContext &context);
+	static RelationStats ExtractDelimGetStats(LogicalDelimGet &delim_get, ClientContext &context);
+	//! Create the statistics for a projection using the statistics of the operator that sits underneath the
+	//! projection. Then also create statistics for any extra columns the projection creates.
+	static RelationStats ExtractDummyScanStats(LogicalDummyScan &dummy_scan, ClientContext &context);
+	static RelationStats ExtractExpressionGetStats(LogicalExpressionGet &expression_get, ClientContext &context);
+	//! All relation extractors for blocking relations
+	static RelationStats ExtractProjectionStats(LogicalProjection &proj, RelationStats &child_stats);
+	static RelationStats ExtractAggregationStats(LogicalAggregate &aggr, RelationStats &child_stats);
+	static RelationStats ExtractWindowStats(LogicalWindow &window, RelationStats &child_stats);
+	//! Called after reordering a query plan with potentially 2+ relations.
+	static RelationStats CombineStatsOfReorderableOperator(vector<ColumnBinding> &bindings,
+	                                                       vector<RelationStats> relation_stats);
+	//! Called after reordering a query plan with potentially 2+ relations.
+	static RelationStats CombineStatsOfNonReorderableOperator(LogicalOperator &op, vector<RelationStats> child_stats);
+	static void CopyRelationStats(RelationStats &to, const RelationStats &from);
+
+private:
+};
+
+} // namespace duckdb

package/src/duckdb/src/include/duckdb/parallel/task_scheduler.hpp

@@ -39,7 +39,7 @@ class TaskScheduler {
 	constexpr static int64_t TASK_TIMEOUT_USECS = 5000;
 
 public:
-	TaskScheduler(DatabaseInstance &db);
+	explicit TaskScheduler(DatabaseInstance &db);
 	~TaskScheduler();
 
 	DUCKDB_API static TaskScheduler &GetScheduler(ClientContext &context);
@@ -67,6 +67,9 @@ public:
 	//! Send signals to n threads, signalling for them to wake up and attempt to execute a task
 	void Signal(idx_t n);
 
+	//! Yield to other threads
+	void YieldThread();
+
 	//! Set the allocator flush threshold
 	void SetAllocatorFlushTreshold(idx_t threshold);
 

package/src/duckdb/src/include/duckdb/parser/group_by_node.hpp

@@ -22,6 +22,17 @@ public:
 	vector<unique_ptr<ParsedExpression>> group_expressions;
 	//! The different grouping sets as they map to the group expressions
 	vector<GroupingSet> grouping_sets;
+
+public:
+	GroupByNode Copy() {
+		GroupByNode node;
+		node.group_expressions.reserve(group_expressions.size());
+		for (auto &expr : group_expressions) {
+			node.group_expressions.push_back(expr->Copy());
+		}
+		node.grouping_sets = grouping_sets;
+		return node;
+	}
 };
 
 } // namespace duckdb

package/src/duckdb/src/include/duckdb/parser/parser.hpp

@@ -22,6 +22,8 @@ struct PGList;
 
 namespace duckdb {
 
+class GroupByNode;
+
 //! The parser is responsible for parsing the query and converting it into a set
 //! of parsed statements. The parsed statements can then be converted into a
 //! plan and executed.
@@ -50,6 +52,8 @@ public:
 	//! Parses a list of expressions (i.e. the list found in a SELECT clause)
 	DUCKDB_API static vector<unique_ptr<ParsedExpression>> ParseExpressionList(const string &select_list,
 	                                                                           ParserOptions options = ParserOptions());
+	//! Parses a list of GROUP BY expressions
+	static GroupByNode ParseGroupByList(const string &group_by, ParserOptions options = ParserOptions());
 	//! Parses a list as found in an ORDER BY expression (i.e. including optional ASCENDING/DESCENDING modifiers)
 	static vector<OrderByNode> ParseOrderList(const string &select_list, ParserOptions options = ParserOptions());
 	//! Parses an update list (i.e. the list found in the SET clause of an UPDATE statement)

package/src/duckdb/src/include/duckdb/planner/logical_operator.hpp

@@ -49,8 +49,6 @@ public:
 	idx_t estimated_cardinality;
 	bool has_estimated_cardinality;
 
-	unique_ptr<EstimatedProperties> estimated_props;
-
 public:
 	virtual vector<ColumnBinding> GetColumnBindings();
 	static vector<ColumnBinding> GenerateColumnBindings(idx_t table_idx, idx_t column_count);

package/src/duckdb/src/include/duckdb.h

@@ -308,7 +308,8 @@ typedef enum { DuckDBSuccess = 0, DuckDBError = 1 } duckdb_state;
 typedef enum {
 	DUCKDB_PENDING_RESULT_READY = 0,
 	DUCKDB_PENDING_RESULT_NOT_READY = 1,
-	DUCKDB_PENDING_ERROR = 2
+	DUCKDB_PENDING_ERROR = 2,
+	DUCKDB_PENDING_NO_TASKS_AVAILABLE = 3
 } duckdb_pending_state;
 
 //===--------------------------------------------------------------------===//
@@ -1260,6 +1261,15 @@ Otherwise, all remaining tasks must be executed first.
 */
 DUCKDB_API duckdb_state duckdb_execute_pending(duckdb_pending_result pending_result, duckdb_result *out_result);
 
+/*!
+Returns whether a duckdb_pending_state is finished executing. For example if `pending_state` is
+DUCKDB_PENDING_RESULT_READY, this function will return true.
+
+* pending_state: The pending state on which to decide whether to finish execution.
+* returns: Boolean indicating pending execution should be considered finished.
+*/
+DUCKDB_API bool duckdb_pending_execution_is_finished(duckdb_pending_state pending_state);
+
 //===--------------------------------------------------------------------===//
 // Value Interface
 //===--------------------------------------------------------------------===//

package/src/duckdb/src/main/capi/pending-c.cpp

@@ -92,6 +92,8 @@ duckdb_pending_state duckdb_pending_execute_task(duckdb_pending_result pending_r
 	switch (return_value) {
 	case PendingExecutionResult::RESULT_READY:
 		return DUCKDB_PENDING_RESULT_READY;
+	case PendingExecutionResult::NO_TASKS_AVAILABLE:
+		return DUCKDB_PENDING_NO_TASKS_AVAILABLE;
 	case PendingExecutionResult::RESULT_NOT_READY:
 		return DUCKDB_PENDING_RESULT_NOT_READY;
 	default:
@@ -99,6 +101,21 @@ duckdb_pending_state duckdb_pending_execute_task(duckdb_pending_result pending_r
 	}
 }
 
+bool duckdb_pending_execution_is_finished(duckdb_pending_state pending_state) {
+	switch (pending_state) {
+	case DUCKDB_PENDING_RESULT_READY:
+		return PendingQueryResult::IsFinished(PendingExecutionResult::RESULT_READY);
+	case DUCKDB_PENDING_NO_TASKS_AVAILABLE:
+		return PendingQueryResult::IsFinished(PendingExecutionResult::NO_TASKS_AVAILABLE);
+	case DUCKDB_PENDING_RESULT_NOT_READY:
+		return PendingQueryResult::IsFinished(PendingExecutionResult::RESULT_NOT_READY);
+	case DUCKDB_PENDING_ERROR:
+		return PendingQueryResult::IsFinished(PendingExecutionResult::EXECUTION_ERROR);
+	default:
+		return PendingQueryResult::IsFinished(PendingExecutionResult::EXECUTION_ERROR);
+	}
+}
+
 duckdb_state duckdb_execute_pending(duckdb_pending_result pending_result, duckdb_result *out_result) {
 	if (!pending_result || !out_result) {
 		return DuckDBError;

package/src/duckdb/src/main/pending_query_result.cpp

@@ -55,7 +55,8 @@ PendingExecutionResult PendingQueryResult::ExecuteTaskInternal(ClientContextLock
 
 unique_ptr<QueryResult> PendingQueryResult::ExecuteInternal(ClientContextLock &lock) {
 	CheckExecutableInternal(lock);
-	while (ExecuteTaskInternal(lock) == PendingExecutionResult::RESULT_NOT_READY) {
+	// Busy wait while execution is not finished
+	while (!IsFinished(ExecuteTaskInternal(lock))) {
 	}
 	if (HasError()) {
 		return make_uniq<MaterializedQueryResult>(error);
@@ -74,4 +75,11 @@ void PendingQueryResult::Close() {
 	context.reset();
 }
 
+bool PendingQueryResult::IsFinished(PendingExecutionResult result) {
+	if (result == PendingExecutionResult::RESULT_READY || result == PendingExecutionResult::EXECUTION_ERROR) {
+		return true;
+	}
+	return false;
+}
+
 } // namespace duckdb