duckdb 0.7.2-dev0.0 → 0.7.2-dev1034.0

package/src/duckdb/src/function/scalar/list/contains_or_position.cpp

@@ -5,156 +5,14 @@
 
 namespace duckdb {
 
-struct ContainsFunctor {
-	static inline bool Initialize() {
-		return false;
-	}
-	static inline bool UpdateResultEntries(idx_t child_idx) {
-		return true;
-	}
-};
-
-struct PositionFunctor {
-	static inline int32_t Initialize() {
-		return 0;
-	}
-	static inline int32_t UpdateResultEntries(idx_t child_idx) {
-		return child_idx + 1;
-	}
-};
-
-template <class CHILD_TYPE, class RETURN_TYPE, class OP>
-static void TemplatedContainsOrPosition(DataChunk &args, ExpressionState &state, Vector &result,
-                                        bool is_nested = false) {
-	D_ASSERT(args.ColumnCount() == 2);
-	auto count = args.size();
-	Vector &list = args.data[0];
-	Vector &value_vector = args.data[1];
-
-	// Create a result vector of type RETURN_TYPE
-	result.SetVectorType(VectorType::FLAT_VECTOR);
-	auto result_entries = FlatVector::GetData<RETURN_TYPE>(result);
-	auto &result_validity = FlatVector::Validity(result);
-
-	if (list.GetType().id() == LogicalTypeId::SQLNULL) {
-		result_validity.SetInvalid(0);
-		return;
-	}
-
-	auto list_size = ListVector::GetListSize(list);
-	auto &child_vector = ListVector::GetEntry(list);
-
-	UnifiedVectorFormat child_data;
-	child_vector.ToUnifiedFormat(list_size, child_data);
-
-	UnifiedVectorFormat list_data;
-	list.ToUnifiedFormat(count, list_data);
-	auto list_entries = (list_entry_t *)list_data.data;
-
-	UnifiedVectorFormat value_data;
-	value_vector.ToUnifiedFormat(count, value_data);
-
-	// not required for a comparison of nested types
-	auto child_value = (CHILD_TYPE *)child_data.data;
-	auto values = (CHILD_TYPE *)value_data.data;
-
-	for (idx_t i = 0; i < count; i++) {
-		auto list_index = list_data.sel->get_index(i);
-		auto value_index = value_data.sel->get_index(i);
-
-		if (!list_data.validity.RowIsValid(list_index) || !value_data.validity.RowIsValid(value_index)) {
-			result_validity.SetInvalid(i);
-			continue;
-		}
-
-		const auto &list_entry = list_entries[list_index];
-
-		result_entries[i] = OP::Initialize();
-		for (idx_t child_idx = 0; child_idx < list_entry.length; child_idx++) {
-
-			auto child_value_idx = child_data.sel->get_index(list_entry.offset + child_idx);
-			if (!child_data.validity.RowIsValid(child_value_idx)) {
-				continue;
-			}
-
-			if (!is_nested) {
-				if (Equals::Operation(child_value[child_value_idx], values[value_index])) {
-					result_entries[i] = OP::UpdateResultEntries(child_idx);
-					break; // Found value in list, no need to look further
-				}
-			} else {
-				// FIXME: using Value is less efficient than modifying the vector comparison code
-				// to more efficiently compare nested types
-				auto lvalue = child_vector.GetValue(child_value_idx);
-				auto rvalue = value_vector.GetValue(value_index);
-				if (Value::NotDistinctFrom(lvalue, rvalue)) {
-					result_entries[i] = OP::UpdateResultEntries(child_idx);
-					break; // Found value in list, no need to look further
-				}
-			}
-		}
-	}
-
-	if (args.AllConstant()) {
-		result.SetVectorType(VectorType::CONSTANT_VECTOR);
-	}
-}
-
-template <class T, class OP>
-static void ListContainsOrPosition(DataChunk &args, ExpressionState &state, Vector &result) {
-	switch (args.data[1].GetType().InternalType()) {
-	case PhysicalType::BOOL:
-	case PhysicalType::INT8:
-		TemplatedContainsOrPosition<int8_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::INT16:
-		TemplatedContainsOrPosition<int16_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::INT32:
-		TemplatedContainsOrPosition<int32_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::INT64:
-		TemplatedContainsOrPosition<int64_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::INT128:
-		TemplatedContainsOrPosition<hugeint_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::UINT8:
-		TemplatedContainsOrPosition<uint8_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::UINT16:
-		TemplatedContainsOrPosition<uint16_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::UINT32:
-		TemplatedContainsOrPosition<uint32_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::UINT64:
-		TemplatedContainsOrPosition<uint64_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::FLOAT:
-		TemplatedContainsOrPosition<float, T, OP>(args, state, result);
-		break;
-	case PhysicalType::DOUBLE:
-		TemplatedContainsOrPosition<double, T, OP>(args, state, result);
-		break;
-	case PhysicalType::VARCHAR:
-		TemplatedContainsOrPosition<string_t, T, OP>(args, state, result);
-		break;
-	case PhysicalType::STRUCT:
-	case PhysicalType::LIST:
-		TemplatedContainsOrPosition<int8_t, T, OP>(args, state, result, true);
-		break;
-	default:
-		throw NotImplementedException("This function has not been implemented for this type");
-	}
-}
-
 static void ListContainsFunction(DataChunk &args, ExpressionState &state, Vector &result) {
-	return ListContainsOrPosition<bool, ContainsFunctor>(args, state, result);
+	(void)state;
+	return ListContainsOrPosition<bool, ContainsFunctor, ListArgFunctor>(args, result);
 }
 
 static void ListPositionFunction(DataChunk &args, ExpressionState &state, Vector &result) {
-	return ListContainsOrPosition<int32_t, PositionFunctor>(args, state, result);
+	(void)state;
+	return ListContainsOrPosition<int32_t, PositionFunctor, ListArgFunctor>(args, result);
 }
 
 template <LogicalTypeId RETURN_TYPE>

package/src/duckdb/src/function/scalar/list/flatten.cpp

@@ -1,7 +1,7 @@
 #include "duckdb/common/types/data_chunk.hpp"
 #include "duckdb/function/scalar/nested_functions.hpp"
 #include "duckdb/planner/expression/bound_function_expression.hpp"
-#include "duckdb/storage/statistics/list_statistics.hpp"
+#include "duckdb/storage/statistics/list_stats.hpp"
 
 namespace duckdb {
 
@@ -121,17 +121,10 @@ static unique_ptr<FunctionData> ListFlattenBind(ClientContext &context, ScalarFu
 
 static unique_ptr<BaseStatistics> ListFlattenStats(ClientContext &context, FunctionStatisticsInput &input) {
 	auto &child_stats = input.child_stats;
-	if (!child_stats[0]) {
-		return nullptr;
-	}
-	auto &list_stats = (ListStatistics &)*child_stats[0];
-	if (!list_stats.child_stats || list_stats.child_stats->type == LogicalTypeId::SQLNULL) {
-		return nullptr;
-	}
-
-	auto child_copy = list_stats.child_stats->Copy();
-	child_copy->validity_stats = make_unique<ValidityStatistics>(true);
-	return child_copy;
+	auto &list_child_stats = ListStats::GetChildStats(child_stats[0]);
+	auto child_copy = list_child_stats.Copy();
+	child_copy.Set(StatsInfo::CAN_HAVE_NULL_VALUES);
+	return child_copy.ToUnique();
 }
 
 void ListFlattenFun::RegisterFunction(BuiltinFunctions &set) {

package/src/duckdb/src/function/scalar/list/list_aggregates.cpp

@@ -507,7 +507,7 @@ ScalarFunction ListUniqueFun::GetFunction() {
 }
 
 void ListAggregateFun::RegisterFunction(BuiltinFunctions &set) {
-	set.AddFunction({"list_aggregate", "array_aggregate", "list_aggr", "array_aggr"}, GetFunction());
+	set.AddFunction({"list_aggregate", "array_aggregate", "list_aggr", "array_aggr", "aggregate"}, GetFunction());
 }
 
 void ListDistinctFun::RegisterFunction(BuiltinFunctions &set) {

package/src/duckdb/src/function/scalar/list/list_concat.cpp

@@ -3,8 +3,6 @@
 #include "duckdb/planner/expression/bound_function_expression.hpp"
 #include "duckdb/planner/expression/bound_parameter_expression.hpp"
 #include "duckdb/planner/expression_binder.hpp"
-#include "duckdb/storage/statistics/list_statistics.hpp"
-#include "duckdb/storage/statistics/validity_statistics.hpp"
 
 namespace duckdb {
 
@@ -84,9 +82,10 @@ static unique_ptr<FunctionData> ListConcatBind(ClientContext &context, ScalarFun
 		throw ParameterNotResolvedException();
 	} else if (lhs.id() == LogicalTypeId::SQLNULL || rhs.id() == LogicalTypeId::SQLNULL) {
 		// we mimic postgres behaviour: list_concat(NULL, my_list) = my_list
-		bound_function.arguments[0] = lhs;
-		bound_function.arguments[1] = rhs;
-		bound_function.return_type = rhs.id() == LogicalTypeId::SQLNULL ? lhs : rhs;
+		auto return_type = rhs.id() == LogicalTypeId::SQLNULL ? lhs : rhs;
+		bound_function.arguments[0] = return_type;
+		bound_function.arguments[1] = return_type;
+		bound_function.return_type = return_type;
 	} else {
 		D_ASSERT(lhs.id() == LogicalTypeId::LIST);
 		D_ASSERT(rhs.id() == LogicalTypeId::LIST);
@@ -96,7 +95,7 @@ static unique_ptr<FunctionData> ListConcatBind(ClientContext &context, ScalarFun
 		for (const auto &argument : arguments) {
 			child_type = LogicalType::MaxLogicalType(child_type, ListType::GetChildType(argument->return_type));
 		}
-		auto list_type = LogicalType::LIST(std::move(child_type));
+		auto list_type = LogicalType::LIST(child_type);
 
 		bound_function.arguments[0] = list_type;
 		bound_function.arguments[1] = list_type;
@@ -108,14 +107,11 @@ static unique_ptr<FunctionData> ListConcatBind(ClientContext &context, ScalarFun
 static unique_ptr<BaseStatistics> ListConcatStats(ClientContext &context, FunctionStatisticsInput &input) {
 	auto &child_stats = input.child_stats;
 	D_ASSERT(child_stats.size() == 2);
-	if (!child_stats[0] || !child_stats[1]) {
-		return nullptr;
-	}
 
-	auto &left_stats = (ListStatistics &)*child_stats[0];
-	auto &right_stats = (ListStatistics &)*child_stats[1];
+	auto &left_stats = child_stats[0];
+	auto &right_stats = child_stats[1];
 
-	auto stats = left_stats.Copy();
+	auto stats = left_stats.ToUnique();
 	stats->Merge(right_stats);
 
 	return stats;

package/src/duckdb/src/function/scalar/list/list_extract.cpp

@@ -7,8 +7,7 @@
 #include "duckdb/function/scalar/string_functions.hpp"
 #include "duckdb/parser/expression/bound_expression.hpp"
 #include "duckdb/planner/expression/bound_function_expression.hpp"
-#include "duckdb/storage/statistics/list_statistics.hpp"
-#include "duckdb/storage/statistics/validity_statistics.hpp"
+#include "duckdb/storage/statistics/list_stats.hpp"
 
 namespace duckdb {
 
@@ -212,17 +211,11 @@ static unique_ptr<FunctionData> ListExtractBind(ClientContext &context, ScalarFu
 
 static unique_ptr<BaseStatistics> ListExtractStats(ClientContext &context, FunctionStatisticsInput &input) {
 	auto &child_stats = input.child_stats;
-	if (!child_stats[0]) {
-		return nullptr;
-	}
-	auto &list_stats = (ListStatistics &)*child_stats[0];
-	if (!list_stats.child_stats) {
-		return nullptr;
-	}
-	auto child_copy = list_stats.child_stats->Copy();
+	auto &list_child_stats = ListStats::GetChildStats(child_stats[0]);
+	auto child_copy = list_child_stats.Copy();
 	// list_extract always pushes a NULL, since if the offset is out of range for a list it inserts a null
-	child_copy->validity_stats = make_unique<ValidityStatistics>(true);
-	return child_copy;
+	child_copy.Set(StatsInfo::CAN_HAVE_NULL_VALUES);
+	return child_copy.ToUnique();
 }
 
 void ListExtractFun::RegisterFunction(BuiltinFunctions &set) {

package/src/duckdb/src/function/scalar/list/list_lambdas.cpp

@@ -124,7 +124,7 @@ static void ExecuteExpression(vector<LogicalType> &types, vector<LogicalType> &r
 	// set the other vectors
 	vector<Vector> slices;
 	for (idx_t col_idx = 0; col_idx < args.ColumnCount() - 1; col_idx++) {
-		slices.emplace_back(Vector(args.data[col_idx + 1], sel_vectors[col_idx], elem_cnt));
+		slices.emplace_back(args.data[col_idx + 1], sel_vectors[col_idx], elem_cnt);
 		slices[col_idx].Flatten(elem_cnt);
 		input_chunk.data[col_idx + 2].Reference(slices[col_idx]);
 	}
@@ -195,10 +195,10 @@ static void ListLambdaFunction(DataChunk &args, ExpressionState &state, Vector &
 
 	// skip the list column
 	for (idx_t i = 1; i < args.ColumnCount(); i++) {
-		columns.emplace_back(UnifiedVectorFormat());
+		columns.emplace_back();
 		args.data[i].ToUnifiedFormat(count, columns[i - 1]);
 		indexes.push_back(0);
-		sel_vectors.emplace_back(SelectionVector(STANDARD_VECTOR_SIZE));
+		sel_vectors.emplace_back(STANDARD_VECTOR_SIZE);
 		types.push_back(args.data[i].GetType());
 	}
 
@@ -392,6 +392,8 @@ void ListTransformFun::RegisterFunction(BuiltinFunctions &set) {
 	set.AddFunction(fun);
 	fun.name = "array_apply";
 	set.AddFunction(fun);
+	fun.name = "apply";
+	set.AddFunction(fun);
 }
 
 void ListFilterFun::RegisterFunction(BuiltinFunctions &set) {
@@ -405,6 +407,8 @@ void ListFilterFun::RegisterFunction(BuiltinFunctions &set) {
 
 	fun.name = "array_filter";
 	set.AddFunction(fun);
+	fun.name = "filter";
+	set.AddFunction(fun);
 }
 
 } // namespace duckdb

package/src/duckdb/src/function/scalar/list/list_value.cpp

@@ -4,7 +4,7 @@
 #include "duckdb/function/scalar/nested_functions.hpp"
 #include "duckdb/common/types/data_chunk.hpp"
 #include "duckdb/common/pair.hpp"
-#include "duckdb/storage/statistics/list_statistics.hpp"
+#include "duckdb/storage/statistics/list_stats.hpp"
 #include "duckdb/planner/expression_binder.hpp"
 
 namespace duckdb {
@@ -42,23 +42,19 @@ static unique_ptr<FunctionData> ListValueBind(ClientContext &context, ScalarFunc
 
 	// this is more for completeness reasons
 	bound_function.varargs = child_type;
-	bound_function.return_type = LogicalType::LIST(std::move(child_type));
+	bound_function.return_type = LogicalType::LIST(child_type);
 	return make_unique<VariableReturnBindData>(bound_function.return_type);
 }
 
 unique_ptr<BaseStatistics> ListValueStats(ClientContext &context, FunctionStatisticsInput &input) {
 	auto &child_stats = input.child_stats;
 	auto &expr = input.expr;
-	auto list_stats = make_unique<ListStatistics>(expr.return_type);
+	auto list_stats = ListStats::CreateEmpty(expr.return_type);
+	auto &list_child_stats = ListStats::GetChildStats(list_stats);
 	for (idx_t i = 0; i < child_stats.size(); i++) {
-		if (child_stats[i]) {
-			list_stats->child_stats->Merge(*child_stats[i]);
-		} else {
-			list_stats->child_stats.reset();
-			return std::move(list_stats);
-		}
+		list_child_stats.Merge(child_stats[i]);
 	}
-	return std::move(list_stats);
+	return list_stats.ToUnique();
 }
 
 void ListValueFun::RegisterFunction(BuiltinFunctions &set) {

package/src/duckdb/src/function/scalar/map/map.cpp

@@ -98,6 +98,52 @@ static void AlignVectorToReference(const Vector &original, const Vector &referen
 	result.Reference(expanded_const);
 }
 
+static bool ListEntriesEqual(Vector &keys, Vector &values, idx_t count) {
+	auto key_count = ListVector::GetListSize(keys);
+	auto value_count = ListVector::GetListSize(values);
+	bool same_vector_type = keys.GetVectorType() == values.GetVectorType();
+
+	D_ASSERT(keys.GetType().id() == LogicalTypeId::LIST);
+	D_ASSERT(values.GetType().id() == LogicalTypeId::LIST);
+
+	UnifiedVectorFormat keys_data;
+	UnifiedVectorFormat values_data;
+
+	keys.ToUnifiedFormat(count, keys_data);
+	values.ToUnifiedFormat(count, values_data);
+
+	auto keys_entries = (list_entry_t *)keys_data.data;
+	auto values_entries = (list_entry_t *)values_data.data;
+
+	if (same_vector_type) {
+		const auto key_data = keys_data.data;
+		const auto value_data = values_data.data;
+
+		if (keys.GetVectorType() == VectorType::CONSTANT_VECTOR) {
+			D_ASSERT(values.GetVectorType() == VectorType::CONSTANT_VECTOR);
+			// Only need to compare one entry in this case
+			return memcmp(key_data, value_data, sizeof(list_entry_t)) == 0;
+		}
+
+		// Fast path if the vector types are equal, can just check if the entries are the same
+		if (key_count != value_count) {
+			return false;
+		}
+		return memcmp(key_data, value_data, count * sizeof(list_entry_t)) == 0;
+	}
+
+	// Compare the list_entries one by one
+	for (idx_t i = 0; i < count; i++) {
+		auto keys_idx = keys_data.sel->get_index(i);
+		auto values_idx = values_data.sel->get_index(i);
+
+		if (keys_entries[keys_idx] != values_entries[values_idx]) {
+			return false;
+		}
+	}
+	return true;
+}
+
 static void MapFunction(DataChunk &args, ExpressionState &state, Vector &result) {
 	D_ASSERT(result.GetType().id() == LogicalTypeId::MAP);
 
@@ -132,7 +178,7 @@ static void MapFunction(DataChunk &args, ExpressionState &state, Vector &result)
 	} else if (values_are_const && !keys_are_const) {
 		AlignVectorToReference(args.data[1], args.data[0], args.size(), value_vector);
 	} else {
-		if (key_count != value_count || memcmp(key_data, value_data, args.size() * sizeof(list_entry_t)) != 0) {
+		if (!ListEntriesEqual(args.data[0], args.data[1], args.size())) {
 			throw InvalidInputException("Error in MAP creation: key list and value list do not align. i.e. different "
 			                            "size or incompatible structure");
 		}

package/src/duckdb/src/function/scalar/map/map_entries.cpp

@@ -0,0 +1,61 @@
+#include "duckdb/planner/expression/bound_function_expression.hpp"
+#include "duckdb/common/string_util.hpp"
+#include "duckdb/parser/expression/bound_expression.hpp"
+#include "duckdb/function/scalar/nested_functions.hpp"
+#include "duckdb/common/types/data_chunk.hpp"
+#include "duckdb/common/pair.hpp"
+
+namespace duckdb {
+
+// Reverse of map_from_entries
+static void MapEntriesFunction(DataChunk &args, ExpressionState &state, Vector &result) {
+	idx_t count = args.size();
+
+	result.Reinterpret(args.data[0]);
+
+	if (args.AllConstant()) {
+		result.SetVectorType(VectorType::CONSTANT_VECTOR);
+	}
+	result.Verify(count);
+}
+
+static unique_ptr<FunctionData> MapEntriesBind(ClientContext &context, ScalarFunction &bound_function,
+                                               vector<unique_ptr<Expression>> &arguments) {
+	child_list_t<LogicalType> child_types;
+
+	if (arguments.size() != 1) {
+		throw InvalidInputException("Too many arguments provided, only expecting a single map");
+	}
+	auto &map = arguments[0]->return_type;
+
+	if (map.id() == LogicalTypeId::UNKNOWN) {
+		// Prepared statement
+		bound_function.arguments.emplace_back(LogicalTypeId::UNKNOWN);
+		bound_function.return_type = LogicalType(LogicalTypeId::SQLNULL);
+		return nullptr;
+	}
+
+	if (map.id() != LogicalTypeId::MAP) {
+		throw InvalidInputException("The provided argument is not a map");
+	}
+	auto &key_type = MapType::KeyType(map);
+	auto &value_type = MapType::ValueType(map);
+
+	child_types.push_back(make_pair("k", key_type));
+	child_types.push_back(make_pair("v", value_type));
+
+	auto row_type = LogicalType::STRUCT(child_types);
+
+	bound_function.return_type = LogicalType::LIST(row_type);
+	return make_unique<VariableReturnBindData>(bound_function.return_type);
+}
+
+void MapEntriesFun::RegisterFunction(BuiltinFunctions &set) {
+	//! the arguments and return types are actually set in the binder function
+	ScalarFunction fun("map_entries", {}, LogicalTypeId::LIST, MapEntriesFunction, MapEntriesBind);
+	fun.null_handling = FunctionNullHandling::DEFAULT_NULL_HANDLING;
+	fun.varargs = LogicalType::ANY;
+	set.AddFunction(fun);
+}
+
+} // namespace duckdb

package/src/duckdb/src/function/scalar/map/map_extract.cpp

@@ -5,19 +5,55 @@
 #include "duckdb/common/types/data_chunk.hpp"
 
 namespace duckdb {
-void FillResult(Value &values, Vector &result, idx_t row) {
-	//! First Initialize List Vector
-	idx_t current_offset = ListVector::GetListSize(result);
-	//! Push Values to List Vector
-	auto &list_values = ListValue::GetChildren(values);
-	for (idx_t i = 0; i < list_values.size(); i++) {
-		ListVector::PushBack(result, list_values[i]);
+void FillResult(Vector &map, Vector &offsets, Vector &result, idx_t count) {
+	UnifiedVectorFormat map_data;
+	map.ToUnifiedFormat(count, map_data);
+
+	UnifiedVectorFormat offset_data;
+	offsets.ToUnifiedFormat(count, offset_data);
+
+	UnifiedVectorFormat result_data;
+	result.ToUnifiedFormat(count, result_data);
+
+	auto entry_count = ListVector::GetListSize(map);
+	auto &values_entries = MapVector::GetValues(map);
+	UnifiedVectorFormat values_entry_data;
+	// Note: this vector can have a different size than the map
+	values_entries.ToUnifiedFormat(entry_count, values_entry_data);
+
+	for (idx_t row = 0; row < count; row++) {
+		idx_t offset_idx = offset_data.sel->get_index(row);
+		auto offset = ((int32_t *)offset_data.data)[offset_idx];
+
+		// Get the current size of the list, for the offset
+		idx_t current_offset = ListVector::GetListSize(result);
+		if (!offset_data.validity.RowIsValid(offset_idx) || !offset) {
+			// Set the entry data for this result row
+			idx_t result_index = result_data.sel->get_index(row);
+			auto &entry = ((list_entry_t *)result_data.data)[result_index];
+			entry.length = 0;
+			entry.offset = current_offset;
+			continue;
+		}
+		// All list indices start at 1, reduce by 1 to get the actual index
+		offset--;
+
+		// Get the 'values' list entry corresponding to the offset
+		idx_t value_index = map_data.sel->get_index(row);
+		auto &value_list_entry = ((list_entry_t *)map_data.data)[value_index];
+
+		// Add the values to the result
+		idx_t list_offset = value_list_entry.offset + offset;
+		// All keys are unique, only one will ever match
+		idx_t length = 1;
+		ListVector::Append(result, values_entries, length + list_offset, list_offset);
+
+		// Set the entry data for this result row
+		idx_t result_index = result_data.sel->get_index(row);
+		auto &entry = ((list_entry_t *)result_data.data)[result_index];
+		entry.length = length;
+		entry.offset = current_offset;
 	}
-
-	//! now set the pointer
-	auto &entry = ListVector::GetData(result)[row];
-	entry.length = list_values.size();
-	entry.offset = current_offset;
 }
 
 static void MapExtractFunction(DataChunk &args, ExpressionState &state, Vector &result) {
@@ -25,6 +61,8 @@ static void MapExtractFunction(DataChunk &args, ExpressionState &state, Vector &
 	D_ASSERT(args.data[0].GetType().id() == LogicalTypeId::MAP);
 	result.SetVectorType(VectorType::FLAT_VECTOR);
 
+	idx_t tuple_count = args.size();
+	// Optimization: because keys are not allowed to be NULL, we can early-out
 	if (args.data[1].GetType().id() == LogicalTypeId::SQLNULL) {
 		//! We don't need to look through the map if the 'key' to look for is NULL
 		ListVector::SetListSize(result, 0);
@@ -32,7 +70,7 @@ static void MapExtractFunction(DataChunk &args, ExpressionState &state, Vector &
 		auto list_data = ConstantVector::GetData<list_entry_t>(result);
 		list_data->offset = 0;
 		list_data->length = 0;
-		result.Verify(args.size());
+		result.Verify(tuple_count);
 		return;
 	}
 
@@ -41,26 +79,30 @@ static void MapExtractFunction(DataChunk &args, ExpressionState &state, Vector &
 
 	UnifiedVectorFormat map_data;
 
-	auto &map_keys = MapVector::GetKeys(map);
-	auto &map_values = MapVector::GetValues(map);
+	// Create the chunk we'll feed to ListPosition
+	DataChunk list_position_chunk;
+	vector<LogicalType> chunk_types;
+	chunk_types.reserve(2);
+	chunk_types.push_back(map.GetType());
+	chunk_types.push_back(key.GetType());
+	list_position_chunk.InitializeEmpty(chunk_types.begin(), chunk_types.end());
 
-	map.ToUnifiedFormat(args.size(), map_data);
+	// Populate it with the map keys list and the key vector
+	list_position_chunk.data[0].Reference(map);
+	list_position_chunk.data[1].Reference(key);
+	list_position_chunk.SetCardinality(tuple_count);
 
-	for (idx_t row = 0; row < args.size(); row++) {
-		idx_t row_index = map_data.sel->get_index(row);
-		auto key_value = key.GetValue(row);
+	Vector position_vector(LogicalType::LIST(LogicalType::INTEGER), tuple_count);
+	// We can pass around state as it's not used by ListPositionFunction anyways
+	ListContainsOrPosition<int32_t, PositionFunctor, MapKeyArgFunctor>(list_position_chunk, position_vector);
 
-		list_entry_t entry = ListVector::GetData(map)[row_index];
-		auto offsets = MapVector::Search(map_keys, args.size(), key_value, entry);
-		auto values = FlatVector::GetValuesFromOffsets(map_values, offsets);
-		FillResult(values, result, row);
-	}
+	FillResult(map, position_vector, result, tuple_count);
 
-	if (args.size() == 1) {
+	if (tuple_count == 1) {
 		result.SetVectorType(VectorType::CONSTANT_VECTOR);
 	}
 
-	result.Verify(args.size());
+	result.Verify(tuple_count);
 }
 
 static unique_ptr<FunctionData> MapExtractBind(ClientContext &context, ScalarFunction &bound_function,