duckdb 0.7.2-dev12.0 → 0.7.2-dev1244.0

package/src/duckdb/src/include/duckdb/common/types.hpp

@@ -17,12 +17,28 @@
 
 namespace duckdb {
 
+class FormatSerializer;
+class FormatDeserializer;
 class Serializer;
 class Deserializer;
 class Value;
 class TypeCatalogEntry;
 class Vector;
 class ClientContext;
+class FieldWriter;
+
+//! Extra Type Info Type
+enum class ExtraTypeInfoType : uint8_t {
+	INVALID_TYPE_INFO = 0,
+	GENERIC_TYPE_INFO = 1,
+	DECIMAL_TYPE_INFO = 2,
+	STRING_TYPE_INFO = 3,
+	LIST_TYPE_INFO = 4,
+	STRUCT_TYPE_INFO = 5,
+	ENUM_TYPE_INFO = 6,
+	USER_TYPE_INFO = 7,
+	AGGREGATE_STATE_TYPE_INFO = 8
+};
 
 struct hugeint_t {
 public:
@@ -32,7 +48,7 @@ public:
 public:
 	DUCKDB_API hugeint_t() = default;
 	DUCKDB_API hugeint_t(int64_t value); // NOLINT: Allow implicit conversion from `int64_t`
-	DUCKDB_API constexpr hugeint_t(int64_t upper, uint64_t lower): lower(lower), upper(upper) {};
+	DUCKDB_API constexpr hugeint_t(int64_t upper, uint64_t lower): lower(lower), upper(upper) {}
 	DUCKDB_API constexpr hugeint_t(const hugeint_t &rhs) = default;
 	DUCKDB_API constexpr hugeint_t(hugeint_t &&rhs) = default;
 	DUCKDB_API hugeint_t &operator=(const hugeint_t &rhs) = default;
@@ -94,6 +110,12 @@ struct list_entry_t {
 	list_entry_t() = default;
 	list_entry_t(uint64_t offset, uint64_t length) : offset(offset), length(length) {
 	}
+	inline constexpr bool operator != (const list_entry_t &other) const {
+		return !(*this == other);
+	}
+	inline constexpr bool operator == (const list_entry_t &other) const {
+		return offset == other.offset && length == other.length;
+	}
 
 	uint64_t offset;
 	uint64_t length;
@@ -261,7 +283,6 @@ enum class LogicalTypeId : uint8_t {
 
 	HUGEINT = 50,
 	POINTER = 51,
-	// HASH = 52, // deprecated, uses UBIGINT instead
 	VALIDITY = 53,
 	UUID = 54,
 
@@ -275,8 +296,8 @@ enum class LogicalTypeId : uint8_t {
 	UNION = 107
 };
 
-struct ExtraTypeInfo;
 
+struct ExtraTypeInfo;
 
 struct aggregate_state_t;
 
@@ -298,6 +319,11 @@ struct LogicalType {
 	inline const ExtraTypeInfo *AuxInfo() const {
 		return type_info_.get();
 	}
+
+	inline shared_ptr<ExtraTypeInfo> GetAuxInfoShrPtr() const {
+		return type_info_;
+	}
+
 	inline void CopyAuxInfo(const LogicalType& other) {
 		type_info_ = other.type_info_;
 	}
@@ -314,7 +340,7 @@ struct LogicalType {
 	inline LogicalType& operator=(LogicalType&& other) noexcept {
 		id_ = other.id_;
 		physical_type_ = other.physical_type_;
-		type_info_ = std::move(other.type_info_);
+		std::swap(type_info_, other.type_info_);
 		return *this;
 	}
 
@@ -325,9 +351,16 @@ struct LogicalType {
 
 	//! Serializes a LogicalType to a stand-alone binary blob
 	DUCKDB_API void Serialize(Serializer &serializer) const;
+
+	DUCKDB_API void SerializeEnumType(Serializer &serializer) const;
+
 	//! Deserializes a blob back into an LogicalType
 	DUCKDB_API static LogicalType Deserialize(Deserializer &source);
 
+	DUCKDB_API void FormatSerialize(FormatSerializer &serializer) const;
+	DUCKDB_API static LogicalType FormatDeserialize(FormatDeserializer &deserializer);
+
+
 	DUCKDB_API static bool TypeIsTimestamp(LogicalTypeId id) {
 		return (id == LogicalTypeId::TIMESTAMP ||
 				id == LogicalTypeId::TIMESTAMP_MS ||
@@ -350,6 +383,8 @@ struct LogicalType {
 	DUCKDB_API static void SetCatalog(LogicalType &type, TypeCatalogEntry* catalog_entry);
 	DUCKDB_API static TypeCatalogEntry* GetCatalog(const LogicalType &type);
 
+	DUCKDB_API static ExtraTypeInfoType GetExtraTypeInfoType(const ExtraTypeInfo &type);
+
 	//! Gets the decimal properties of a numeric type. Fails if the type is not numeric.
 	DUCKDB_API bool GetDecimalProperties(uint8_t &width, uint8_t &scale) const;
 
@@ -404,10 +439,10 @@ public:
 	// explicitly allowing these functions to be capitalized to be in-line with the remaining functions
 	DUCKDB_API static LogicalType DECIMAL(int width, int scale);                 // NOLINT
 	DUCKDB_API static LogicalType VARCHAR_COLLATION(string collation);           // NOLINT
-	DUCKDB_API static LogicalType LIST( LogicalType child);                       // NOLINT
-	DUCKDB_API static LogicalType STRUCT( child_list_t<LogicalType> children);    // NOLINT
+	DUCKDB_API static LogicalType LIST(const LogicalType &child);                       // NOLINT
+	DUCKDB_API static LogicalType STRUCT(const child_list_t<LogicalType> &children);    // NOLINT
 	DUCKDB_API static LogicalType AGGREGATE_STATE(aggregate_state_t state_type);    // NOLINT
-	DUCKDB_API static LogicalType MAP(LogicalType child);				// NOLINT
+	DUCKDB_API static LogicalType MAP(const LogicalType &child);				// NOLINT
 	DUCKDB_API static LogicalType MAP( child_list_t<LogicalType> children);       // NOLINT
 	DUCKDB_API static LogicalType MAP(LogicalType key, LogicalType value); // NOLINT
 	DUCKDB_API static LogicalType UNION( child_list_t<LogicalType> members);     // NOLINT
@@ -442,12 +477,14 @@ struct UserType{
 struct EnumType{
 	DUCKDB_API static const string &GetTypeName(const LogicalType &type);
 	DUCKDB_API static int64_t GetPos(const LogicalType &type, const string_t& key);
-	DUCKDB_API static Vector &GetValuesInsertOrder(const LogicalType &type);
+	DUCKDB_API static const Vector &GetValuesInsertOrder(const LogicalType &type);
 	DUCKDB_API static idx_t GetSize(const LogicalType &type);
 	DUCKDB_API static const string GetValue(const Value &val);
 	DUCKDB_API static void SetCatalog(LogicalType &type, TypeCatalogEntry* catalog_entry);
 	DUCKDB_API static TypeCatalogEntry* GetCatalog(const LogicalType &type);
+	DUCKDB_API static string GetSchemaName(const LogicalType &type);
 	DUCKDB_API static PhysicalType GetPhysicalType(const LogicalType &type);
+	DUCKDB_API static void Serialize(FieldWriter& writer, const ExtraTypeInfo& type_info, bool serialize_internals);
 };
 
 struct StructType {

package/src/duckdb/src/include/duckdb/common/vector_operations/unary_executor.hpp

@@ -209,8 +209,8 @@ public:
 
 	template <class INPUT_TYPE, class RESULT_TYPE, class OP>
 	static void ExecuteString(Vector &input, Vector &result, idx_t count) {
-		UnaryExecutor::GenericExecute<string_t, string_t, UnaryStringOperator<OP>>(input, result, count,
-		                                                                           (void *)&result);
+		UnaryExecutor::GenericExecute<INPUT_TYPE, RESULT_TYPE, UnaryStringOperator<OP>>(input, result, count,
+		                                                                                (void *)&result);
 	}
 };
 

package/src/duckdb/src/include/duckdb/execution/aggregate_hashtable.hpp

@@ -62,6 +62,20 @@ struct AggregateHTScanState {
 	idx_t scan_position = 0;
 };
 
+struct AggregateHTAppendState {
+	AggregateHTAppendState();
+
+	Vector ht_offsets;
+	Vector hash_salts;
+	SelectionVector group_compare_vector;
+	SelectionVector no_match_vector;
+	SelectionVector empty_vector;
+	SelectionVector new_groups;
+	Vector addresses;
+	unique_ptr<UnifiedVectorFormat[]> group_data;
+	DataChunk group_chunk;
+};
+
 class GroupedAggregateHashTable : public BaseAggregateHashTable {
 public:
 	//! The hash table load factor, when a resize is triggered
@@ -71,10 +85,12 @@ public:
 public:
 	GroupedAggregateHashTable(ClientContext &context, Allocator &allocator, vector<LogicalType> group_types,
 	                          vector<LogicalType> payload_types, const vector<BoundAggregateExpression *> &aggregates,
-	                          HtEntryType entry_type = HtEntryType::HT_WIDTH_64);
+	                          HtEntryType entry_type = HtEntryType::HT_WIDTH_64,
+	                          idx_t initial_capacity = InitialCapacity());
 	GroupedAggregateHashTable(ClientContext &context, Allocator &allocator, vector<LogicalType> group_types,
 	                          vector<LogicalType> payload_types, vector<AggregateObject> aggregates,
-	                          HtEntryType entry_type = HtEntryType::HT_WIDTH_64);
+	                          HtEntryType entry_type = HtEntryType::HT_WIDTH_64,
+	                          idx_t initial_capacity = InitialCapacity());
 	GroupedAggregateHashTable(ClientContext &context, Allocator &allocator, vector<LogicalType> group_types);
 	~GroupedAggregateHashTable() override;
 
@@ -85,9 +101,10 @@ public:
 	//! Add the given data to the HT, computing the aggregates grouped by the
 	//! data in the group chunk. When resize = true, aggregates will not be
 	//! computed but instead just assigned.
-	idx_t AddChunk(DataChunk &groups, DataChunk &payload, const vector<idx_t> &filter);
-	idx_t AddChunk(DataChunk &groups, Vector &group_hashes, DataChunk &payload, const vector<idx_t> &filter);
-	idx_t AddChunk(DataChunk &groups, DataChunk &payload, AggregateType filter);
+	idx_t AddChunk(AggregateHTAppendState &state, DataChunk &groups, DataChunk &payload, const vector<idx_t> &filter);
+	idx_t AddChunk(AggregateHTAppendState &state, DataChunk &groups, Vector &group_hashes, DataChunk &payload,
+	               const vector<idx_t> &filter);
+	idx_t AddChunk(AggregateHTAppendState &state, DataChunk &groups, DataChunk &payload, AggregateType filter);
 
 	//! Scan the HT starting from the scan_position until the result and group
 	//! chunks are filled. scan_position will be updated by this function.
@@ -100,19 +117,26 @@ public:
 	//! Finds or creates groups in the hashtable using the specified group keys. The addresses vector will be filled
 	//! with pointers to the groups in the hash table, and the new_groups selection vector will point to the newly
 	//! created groups. The return value is the amount of newly created groups.
-	idx_t FindOrCreateGroups(DataChunk &groups, Vector &group_hashes, Vector &addresses_out,
+	idx_t FindOrCreateGroups(AggregateHTAppendState &state, DataChunk &groups, Vector &group_hashes,
+	                         Vector &addresses_out, SelectionVector &new_groups_out);
+	idx_t FindOrCreateGroups(AggregateHTAppendState &state, DataChunk &groups, Vector &addresses_out,
 	                         SelectionVector &new_groups_out);
-	idx_t FindOrCreateGroups(DataChunk &groups, Vector &addresses_out, SelectionVector &new_groups_out);
-	void FindOrCreateGroups(DataChunk &groups, Vector &addresses_out);
+	void FindOrCreateGroups(AggregateHTAppendState &state, DataChunk &groups, Vector &addresses_out);
 
 	//! Executes the filter(if any) and update the aggregates
 	void Combine(GroupedAggregateHashTable &other);
 
+	static idx_t InitialCapacity();
 	idx_t Size() {
 		return entries;
 	}
+	idx_t Capacity() {
+		return capacity;
+	}
 
+	idx_t ResizeThreshold();
 	idx_t MaxCapacity();
+	static idx_t GetMaxCapacity(HtEntryType entry_type, idx_t tuple_size);
 
 	void Partition(vector<GroupedAggregateHashTable *> &partition_hts, hash_t mask, idx_t shift);
 
@@ -137,8 +161,7 @@ private:
 	//! The hashes of the HT
 	BufferHandle hashes_hdl;
 	data_ptr_t hashes_hdl_ptr;
-	data_ptr_t hashes_end_ptr; // of hashes
-	idx_t hash_offset;         // Offset into the layout of the hash column
+	idx_t hash_offset; // Offset into the layout of the hash column
 
 	hash_t hash_prefix_shift;
 	idx_t payload_page_offset;
@@ -146,16 +169,8 @@ private:
 	//! Bitmask for getting relevant bits from the hashes to determine the position
 	hash_t bitmask;
 
-	vector<unique_ptr<GroupedAggregateHashTable>> distinct_hashes;
-
 	bool is_finalized;
 
-	// some stuff from FindOrCreateGroupsInternal() to avoid allocation there
-	Vector ht_offsets;
-	Vector hash_salts;
-	SelectionVector group_compare_vector;
-	SelectionVector no_match_vector;
-	SelectionVector empty_vector;
 	vector<ExpressionType> predicates;
 
 private:
@@ -175,8 +190,8 @@ private:
 	template <class ENTRY>
 	void Resize(idx_t size);
 	template <class ENTRY>
-	idx_t FindOrCreateGroupsInternal(DataChunk &groups, Vector &group_hashes, Vector &addresses,
-	                                 SelectionVector &new_groups);
+	idx_t FindOrCreateGroupsInternal(AggregateHTAppendState &state, DataChunk &groups, Vector &group_hashes,
+	                                 Vector &addresses, SelectionVector &new_groups);
 
 	template <class FUNC = std::function<void(idx_t, idx_t, data_ptr_t)>>
 	void PayloadApply(FUNC fun);

package/src/duckdb/src/include/duckdb/execution/index/art/art.hpp

@@ -27,18 +27,7 @@
 namespace duckdb {
 
 class ConflictManager;
-
-struct ARTIndexScanState : public IndexScanState {
-
-	//! Scan predicates (single predicate scan or range scan)
-	Value values[2];
-	//! Expressions of the scan predicates
-	ExpressionType expressions[2];
-	bool checked = false;
-	//! All scanned row IDs
-	vector<row_t> result_ids;
-	Iterator iterator;
-};
+struct ARTIndexScanState;
 
 enum class VerifyExistenceType : uint8_t {
 	APPEND = 0,    // appends to a table
@@ -72,7 +61,7 @@ public:
 	          vector<row_t> &result_ids) override;
 
 	//! Called when data is appended to the index. The lock obtained from InitializeLock must be held
-	bool Append(IndexLock &lock, DataChunk &entries, Vector &row_identifiers) override;
+	PreservedError Append(IndexLock &lock, DataChunk &entries, Vector &row_identifiers) override;
 	//! Verify that data can be appended to the index without a constraint violation
 	void VerifyAppend(DataChunk &chunk) override;
 	//! Verify that data can be appended to the index without a constraint violation using the conflict manager
@@ -80,7 +69,7 @@ public:
 	//! Delete a chunk of entries from the index. The lock obtained from InitializeLock must be held
 	void Delete(IndexLock &lock, DataChunk &entries, Vector &row_identifiers) override;
 	//! Insert a chunk of entries into the index
-	bool Insert(IndexLock &lock, DataChunk &data, Vector &row_ids) override;
+	PreservedError Insert(IndexLock &lock, DataChunk &data, Vector &row_ids) override;
 
 	//! Construct an ART from a vector of sorted keys
 	bool ConstructFromSorted(idx_t count, vector<Key> &keys, Vector &row_identifiers);

package/src/duckdb/src/include/duckdb/execution/operator/aggregate/physical_perfecthash_aggregate.hpp

@@ -20,7 +20,7 @@ class PhysicalPerfectHashAggregate : public PhysicalOperator {
 public:
 	PhysicalPerfectHashAggregate(ClientContext &context, vector<LogicalType> types,
 	                             vector<unique_ptr<Expression>> aggregates, vector<unique_ptr<Expression>> groups,
-	                             vector<unique_ptr<BaseStatistics>> group_stats, vector<idx_t> required_bits,
+	                             const vector<unique_ptr<BaseStatistics>> &group_stats, vector<idx_t> required_bits,
 	                             idx_t estimated_cardinality);
 
 	//! The groups

package/src/duckdb/src/include/duckdb/execution/operator/join/physical_cross_product.hpp

@@ -60,10 +60,12 @@ public:
 
 	OperatorResultType Execute(DataChunk &input, DataChunk &output);
 
+	// returns if the left side is scanned as a constant vector
 	bool ScanLHS() {
 		return scan_input_chunk;
 	}
 
+	// returns the position in the chunk of chunk scanned as a constant input vector
 	idx_t PositionInChunk() {
 		return position_in_chunk;
 	}

package/src/duckdb/src/include/duckdb/execution/operator/persistent/csv_file_handle.hpp

@@ -87,6 +87,7 @@ public:
 			// we have data left to read from the file
 			// read directly into the buffer
 			auto bytes_read = file_handle->Read((char *)buffer + result_offset, nr_bytes - result_offset);
+			file_size = file_handle->GetFileSize();
 			read_position += bytes_read;
 			if (reset_enabled) {
 				// if reset caching is enabled, we need to cache the bytes that we have read

package/src/duckdb/src/include/duckdb/execution/operator/persistent/csv_reader_options.hpp

@@ -14,6 +14,7 @@
 #include "duckdb/common/types/value.hpp"
 #include "duckdb/common/field_writer.hpp"
 #include "duckdb/common/case_insensitive_map.hpp"
+#include "duckdb/common/types.hpp"
 
 namespace duckdb {
 
@@ -77,12 +78,19 @@ struct BufferedCSVReaderOptions {
 	vector<LogicalType> sql_type_list;
 	//! User-defined name list
 	vector<string> name_list;
+	//! Types considered as candidates for auto detection ordered by descending specificity (~ from high to low)
+	vector<LogicalType> auto_type_candidates = {LogicalType::VARCHAR, LogicalType::TIMESTAMP, LogicalType::DATE,
+	                                            LogicalType::TIME,    LogicalType::DOUBLE,    LogicalType::BIGINT,
+	                                            LogicalType::BOOLEAN, LogicalType::SQLNULL};
+
 	//===--------------------------------------------------------------------===//
 	// ReadCSVOptions
 	//===--------------------------------------------------------------------===//
 
 	//! How many leading rows to skip
 	idx_t skip_rows = 0;
+	//! Whether or not the skip_rows is set by the user
+	bool skip_rows_set = false;
 	//! Maximum CSV line size: specified because if we reach this amount, we likely have wrong delimiters (default: 2MB)
 	//! note that this is the guaranteed line length that will succeed, longer lines may be accepted if slightly above
 	idx_t maximum_line_size = 2097152;
@@ -110,6 +118,8 @@ struct BufferedCSVReaderOptions {
 	idx_t buffer_size = CSVBuffer::INITIAL_BUFFER_SIZE_COLOSSAL;
 	//! Decimal separator when reading as numeric
 	string decimal_separator = ".";
+	//! Whether or not to pad rows that do not have enough columns with NULL values
+	bool null_padding = true;
 
 	//===--------------------------------------------------------------------===//
 	// WriteCSVOptions

package/src/duckdb/src/include/duckdb/execution/operator/projection/physical_projection.hpp

@@ -30,6 +30,11 @@ public:
 	}
 
 	string ParamsToString() const override;
+
+	static unique_ptr<PhysicalOperator>
+	CreateJoinProjection(vector<LogicalType> proj_types, const vector<LogicalType> &lhs_types,
+	                     const vector<LogicalType> &rhs_types, const vector<idx_t> &left_projection_map,
+	                     const vector<idx_t> &right_projection_map, const idx_t estimated_cardinality);
 };
 
 } // namespace duckdb

package/src/duckdb/src/include/duckdb/execution/partitionable_hashtable.hpp

@@ -54,12 +54,16 @@ private:
 	vector<idx_t> sel_vector_sizes;
 	DataChunk group_subset, payload_subset;
 	Vector hashes, hashes_subset;
+	AggregateHTAppendState append_state;
 
 	HashTableList unpartitioned_hts;
-	unordered_map<hash_t, HashTableList> radix_partitioned_hts;
+	vector<HashTableList> radix_partitioned_hts;
+	idx_t tuple_size;
 
 private:
 	idx_t ListAddChunk(HashTableList &list, DataChunk &groups, Vector &group_hashes, DataChunk &payload,
 	                   const vector<idx_t> &filter);
+	//! Returns the HT entry size used for intermediate hash tables
+	HtEntryType GetHTEntrySize();
 };
 } // namespace duckdb

package/src/duckdb/src/include/duckdb/execution/physical_plan_generator.hpp

@@ -48,6 +48,7 @@ protected:
 
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalAggregate &op);
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalAnyJoin &op);
+	unique_ptr<PhysicalOperator> CreatePlan(LogicalAsOfJoin &op);
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalColumnDataGet &op);
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalComparisonJoin &op);
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalCreate &op);
@@ -88,9 +89,6 @@ protected:
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalRecursiveCTE &op);
 	unique_ptr<PhysicalOperator> CreatePlan(LogicalCTERef &op);
 
-	unique_ptr<PhysicalOperator> CreateDistinctOn(unique_ptr<PhysicalOperator> child,
-	                                              vector<unique_ptr<Expression>> distinct_targets);
-
 	unique_ptr<PhysicalOperator> ExtractAggregateExpressions(unique_ptr<PhysicalOperator> child,
 	                                                         vector<unique_ptr<Expression>> &expressions,
 	                                                         vector<unique_ptr<Expression>> &groups);

package/src/duckdb/src/include/duckdb/execution/window_segment_tree.hpp

@@ -15,6 +15,60 @@
 
 namespace duckdb {
 
+class WindowAggregateState {
+public:
+	WindowAggregateState(AggregateFunction &aggregate, FunctionData *bind_info, const LogicalType &result_type_p);
+	virtual ~WindowAggregateState();
+
+	virtual void Sink(DataChunk &payload_chunk, SelectionVector *filter_sel, idx_t filtered);
+	virtual void Finalize();
+	virtual void Compute(Vector &result, idx_t rid, idx_t start, idx_t end);
+
+protected:
+	void AggregateInit();
+	void AggegateFinal(Vector &result, idx_t rid);
+
+	//! The aggregate that the window function is computed over
+	AggregateFunction aggregate;
+	//! The bind info of the aggregate
+	FunctionData *bind_info;
+	//! The result type of the window function
+	LogicalType result_type;
+
+	//! Data pointer that contains a single state, used for intermediate window segment aggregation
+	vector<data_t> state;
+	//! Reused result state container for the window functions
+	Vector statev;
+	//! A vector of pointers to "state", used for intermediate window segment aggregation
+	Vector statep;
+	//! Input data chunk, used for intermediate window segment aggregation
+	DataChunk inputs;
+};
+
+class WindowConstantAggregate : public WindowAggregateState {
+public:
+	static bool IsConstantAggregate(const BoundWindowExpression &wexpr);
+
+	WindowConstantAggregate(AggregateFunction &aggregate, FunctionData *bind_info, const LogicalType &result_type_p,
+	                        const ValidityMask &partition_mask, const idx_t count);
+	~WindowConstantAggregate() override {
+	}
+
+	void Sink(DataChunk &payload_chunk, SelectionVector *filter_sel, idx_t filtered) override;
+	void Finalize() override;
+	void Compute(Vector &result, idx_t rid, idx_t start, idx_t end) override;
+
+private:
+	//! Partition starts
+	vector<idx_t> partition_offsets;
+	//! Aggregate results
+	unique_ptr<Vector> results;
+	//! The current result partition being built/read
+	idx_t partition;
+	//! The current input row being built/read
+	idx_t row;
+};
+
 class WindowSegmentTree {
 public:
 	using FrameBounds = std::pair<idx_t, idx_t>;

package/src/duckdb/src/include/duckdb/function/aggregate/distributive_functions.hpp

@@ -27,6 +27,11 @@ struct BitXorFun {
 	static void RegisterFunction(BuiltinFunctions &set);
 };
 
+struct BitStringAggFun {
+	static void GetBitStringAggregate(const LogicalType &type, AggregateFunctionSet &bitstring_agg);
+	static void RegisterFunction(BuiltinFunctions &set);
+};
+
 struct CountStarFun {
 	static AggregateFunction GetFunction();
 

package/src/duckdb/src/include/duckdb/function/aggregate_function.hpp

@@ -17,8 +17,9 @@
 
 namespace duckdb {
 
-//! This allows us to use the & operator to check if the type is contained in the set
 enum class AggregateType : uint8_t { NON_DISTINCT = 1, DISTINCT = 2 };
+//! Whether or not the input order influences the result of the aggregate
+enum class AggregateOrderDependent : uint8_t { ORDER_DEPENDENT = 1, NOT_ORDER_DEPENDENT = 2 };
 
 class BoundAggregateExpression;
 
@@ -29,6 +30,17 @@ struct AggregateInputData {
 	Allocator &allocator;
 };
 
+struct AggregateStatisticsInput {
+	AggregateStatisticsInput(FunctionData *bind_data_p, vector<BaseStatistics> &child_stats_p,
+	                         NodeStatistics *node_stats_p)
+	    : bind_data(bind_data_p), child_stats(child_stats_p), node_stats(node_stats_p) {
+	}
+
+	FunctionData *bind_data;
+	vector<BaseStatistics> &child_stats;
+	NodeStatistics *node_stats;
+};
+
 //! The type used for sizing hashed aggregate function states
 typedef idx_t (*aggregate_size_t)();
 //! The type used for initializing hashed aggregate function states
@@ -43,9 +55,7 @@ typedef void (*aggregate_finalize_t)(Vector &state, AggregateInputData &aggr_inp
                                      idx_t offset);
 //! The type used for propagating statistics in aggregate functions (optional)
 typedef unique_ptr<BaseStatistics> (*aggregate_statistics_t)(ClientContext &context, BoundAggregateExpression &expr,
-                                                             FunctionData *bind_data,
-                                                             vector<unique_ptr<BaseStatistics>> &child_stats,
-                                                             NodeStatistics *node_stats);
+                                                             AggregateStatisticsInput &input);
 //! Binds the scalar function and creates the function data
 typedef unique_ptr<FunctionData> (*bind_aggregate_function_t)(ClientContext &context, AggregateFunction &function,
                                                               vector<unique_ptr<Expression>> &arguments);
@@ -83,7 +93,7 @@ public:
 	                         LogicalType(LogicalTypeId::INVALID), null_handling),
 	      state_size(state_size), initialize(initialize), update(update), combine(combine), finalize(finalize),
 	      simple_update(simple_update), window(window), bind(bind), destructor(destructor), statistics(statistics),
-	      serialize(serialize), deserialize(deserialize) {
+	      serialize(serialize), deserialize(deserialize), order_dependent(AggregateOrderDependent::ORDER_DEPENDENT) {
 	}
 
 	DUCKDB_API
@@ -98,7 +108,7 @@ public:
 	                         LogicalType(LogicalTypeId::INVALID)),
 	      state_size(state_size), initialize(initialize), update(update), combine(combine), finalize(finalize),
 	      simple_update(simple_update), window(window), bind(bind), destructor(destructor), statistics(statistics),
-	      serialize(serialize), deserialize(deserialize) {
+	      serialize(serialize), deserialize(deserialize), order_dependent(AggregateOrderDependent::ORDER_DEPENDENT) {
 	}
 
 	DUCKDB_API AggregateFunction(const vector<LogicalType> &arguments, const LogicalType &return_type,
@@ -151,6 +161,8 @@ public:
 
 	aggregate_serialize_t serialize;
 	aggregate_deserialize_t deserialize;
+	//! Whether or not the aggregate is order dependent
+	AggregateOrderDependent order_dependent;
 
 	DUCKDB_API bool operator==(const AggregateFunction &rhs) const {
 		return state_size == rhs.state_size && initialize == rhs.initialize && update == rhs.update &&

package/src/duckdb/src/include/duckdb/function/cast/bound_cast_data.hpp

@@ -0,0 +1,84 @@
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/function/cast/bound_cast_data.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include "duckdb/function/cast/default_casts.hpp"
+
+namespace duckdb {
+
+struct ListBoundCastData : public BoundCastData {
+	explicit ListBoundCastData(BoundCastInfo child_cast) : child_cast_info(std::move(child_cast)) {
+	}
+
+	BoundCastInfo child_cast_info;
+	static unique_ptr<BoundCastData> BindListToListCast(BindCastInput &input, const LogicalType &source,
+	                                                    const LogicalType &target);
+	static unique_ptr<FunctionLocalState> InitListLocalState(CastLocalStateParameters &parameters);
+
+public:
+	unique_ptr<BoundCastData> Copy() const override {
+		return make_unique<ListBoundCastData>(child_cast_info.Copy());
+	}
+};
+
+struct ListCast {
+	static bool ListToListCast(Vector &source, Vector &result, idx_t count, CastParameters &parameters);
+};
+
+struct StructBoundCastData : public BoundCastData {
+	StructBoundCastData(vector<BoundCastInfo> child_casts, LogicalType target_p)
+	    : child_cast_info(std::move(child_casts)), target(std::move(target_p)) {
+	}
+
+	vector<BoundCastInfo> child_cast_info;
+	LogicalType target;
+
+	static unique_ptr<BoundCastData> BindStructToStructCast(BindCastInput &input, const LogicalType &source,
+	                                                        const LogicalType &target);
+	static unique_ptr<FunctionLocalState> InitStructCastLocalState(CastLocalStateParameters &parameters);
+
+public:
+	unique_ptr<BoundCastData> Copy() const override {
+		vector<BoundCastInfo> copy_info;
+		for (auto &info : child_cast_info) {
+			copy_info.push_back(info.Copy());
+		}
+		return make_unique<StructBoundCastData>(std::move(copy_info), target);
+	}
+};
+
+struct StructCastLocalState : public FunctionLocalState {
+public:
+	vector<unique_ptr<FunctionLocalState>> local_states;
+};
+
+struct MapBoundCastData : public BoundCastData {
+	MapBoundCastData(BoundCastInfo key_cast, BoundCastInfo value_cast)
+	    : key_cast(std::move(key_cast)), value_cast(std::move(value_cast)) {
+	}
+
+	BoundCastInfo key_cast;
+	BoundCastInfo value_cast;
+
+	static unique_ptr<BoundCastData> BindMapToMapCast(BindCastInput &input, const LogicalType &source,
+	                                                  const LogicalType &target);
+
+public:
+	unique_ptr<BoundCastData> Copy() const override {
+		return make_unique<MapBoundCastData>(key_cast.Copy(), value_cast.Copy());
+	}
+};
+
+struct MapCastLocalState : public FunctionLocalState {
+public:
+	unique_ptr<FunctionLocalState> key_state;
+	unique_ptr<FunctionLocalState> value_state;
+};
+
+} // namespace duckdb