duckdb 0.5.1-dev97.0 → 0.5.1

package/src/duckdb.cpp

@@ -3763,6 +3763,19 @@ idx_t TableCatalogEntry::StandardColumnCount() const {
 	return count;
 }
 
+unique_ptr<BaseStatistics> TableCatalogEntry::GetStatistics(ClientContext &context, column_t column_id) {
+	if (column_id == COLUMN_IDENTIFIER_ROW_ID) {
+		return nullptr;
+	}
+	if (column_id >= columns.size()) {
+		throw InternalException("TableCatalogEntry::GetStatistics column_id out of range");
+	}
+	if (columns[column_id].Generated()) {
+		return nullptr;
+	}
+	return storage->GetStatistics(context, columns[column_id].StorageOid());
+}
+
 unique_ptr<CatalogEntry> TableCatalogEntry::AlterEntry(ClientContext &context, AlterInfo *info) {
 	D_ASSERT(!internal);
 	if (info->type != AlterType::ALTER_TABLE) {
@@ -3830,6 +3843,9 @@ static void RenameExpression(ParsedExpression &expr, RenameColumnInfo &info) {
 
 unique_ptr<CatalogEntry> TableCatalogEntry::RenameColumn(ClientContext &context, RenameColumnInfo &info) {
 	auto rename_idx = GetColumnIndex(info.old_name);
+	if (rename_idx == COLUMN_IDENTIFIER_ROW_ID) {
+		throw CatalogException("Cannot rename rowid column");
+	}
 	auto create_info = make_unique<CreateTableInfo>(schema->name, name);
 	create_info->temporary = temporary;
 	for (idx_t i = 0; i < columns.size(); i++) {
@@ -3932,6 +3948,9 @@ unique_ptr<CatalogEntry> TableCatalogEntry::AddColumn(ClientContext &context, Ad
 unique_ptr<CatalogEntry> TableCatalogEntry::RemoveColumn(ClientContext &context, RemoveColumnInfo &info) {
 	auto removed_index = GetColumnIndex(info.removed_column, info.if_column_exists);
 	if (removed_index == DConstants::INVALID_INDEX) {
+		if (!info.if_column_exists) {
+			throw CatalogException("Cannot drop column: rowid column cannot be dropped");
+		}
 		return nullptr;
 	}
 
@@ -4038,7 +4057,7 @@ unique_ptr<CatalogEntry> TableCatalogEntry::RemoveColumn(ClientContext &context,
 		return make_unique<TableCatalogEntry>(catalog, schema, (BoundCreateTableInfo *)bound_create_info.get(),
 		                                      storage);
 	}
-	auto new_storage = make_shared<DataTable>(context, *storage, removed_index);
+	auto new_storage = make_shared<DataTable>(context, *storage, columns[removed_index].StorageOid());
 	return make_unique<TableCatalogEntry>(catalog, schema, (BoundCreateTableInfo *)bound_create_info.get(),
 	                                      new_storage);
 }
@@ -4046,13 +4065,18 @@ unique_ptr<CatalogEntry> TableCatalogEntry::RemoveColumn(ClientContext &context,
 unique_ptr<CatalogEntry> TableCatalogEntry::SetDefault(ClientContext &context, SetDefaultInfo &info) {
 	auto create_info = make_unique<CreateTableInfo>(schema->name, name);
 	auto default_idx = GetColumnIndex(info.column_name);
+	if (default_idx == COLUMN_IDENTIFIER_ROW_ID) {
+		throw CatalogException("Cannot SET DEFAULT for rowid column");
+	}
 
 	// Copy all the columns, changing the value of the one that was specified by 'column_name'
 	for (idx_t i = 0; i < columns.size(); i++) {
 		auto copy = columns[i].Copy();
 		if (default_idx == i) {
 			// set the default value of this column
-			D_ASSERT(!copy.Generated()); // Shouldnt reach here - DEFAULT value isn't supported for Generated Columns
+			if (copy.Generated()) {
+				throw BinderException("Cannot SET DEFAULT for generated column \"%s\"", columns[i].Name());
+			}
 			copy.SetDefaultValue(info.expression ? info.expression->Copy() : nullptr);
 		}
 		create_info->columns.push_back(move(copy));
@@ -4077,6 +4101,9 @@ unique_ptr<CatalogEntry> TableCatalogEntry::SetNotNull(ClientContext &context, S
 	}
 
 	idx_t not_null_idx = GetColumnIndex(info.column_name);
+	if (columns[not_null_idx].Generated()) {
+		throw BinderException("Unsupported constraint for generated column!");
+	}
 	bool has_not_null = false;
 	for (idx_t i = 0; i < constraints.size(); i++) {
 		auto constraint = constraints[i]->Copy();
@@ -4100,8 +4127,9 @@ unique_ptr<CatalogEntry> TableCatalogEntry::SetNotNull(ClientContext &context, S
 		                                      storage);
 	}
 
-	// Return with new storage info
-	auto new_storage = make_shared<DataTable>(context, *storage, make_unique<NotNullConstraint>(not_null_idx));
+	// Return with new storage info. Note that we need the bound column index here.
+	auto new_storage = make_shared<DataTable>(context, *storage,
+	                                          make_unique<BoundNotNullConstraint>(columns[not_null_idx].StorageOid()));
 	return make_unique<TableCatalogEntry>(catalog, schema, (BoundCreateTableInfo *)bound_create_info.get(),
 	                                      new_storage);
 }
@@ -4207,12 +4235,19 @@ unique_ptr<CatalogEntry> TableCatalogEntry::ChangeColumnType(ClientContext &cont
 	auto expression = info.expression->Copy();
 	auto bound_expression = expr_binder.Bind(expression);
 	auto bound_create_info = binder->BindCreateTableInfo(move(create_info));
+	vector<column_t> storage_oids;
 	if (bound_columns.empty()) {
-		bound_columns.push_back(COLUMN_IDENTIFIER_ROW_ID);
+		storage_oids.push_back(COLUMN_IDENTIFIER_ROW_ID);
+	}
+	// transform to storage_oid
+	else {
+		for (idx_t i = 0; i < bound_columns.size(); i++) {
+			storage_oids.push_back(columns[bound_columns[i]].StorageOid());
+		}
 	}
 
-	auto new_storage =
-	    make_shared<DataTable>(context, *storage, change_idx, info.target_type, move(bound_columns), *bound_expression);
+	auto new_storage = make_shared<DataTable>(context, *storage, columns[change_idx].StorageOid(), info.target_type,
+	                                          move(storage_oids), *bound_expression);
 	auto result =
 	    make_unique<TableCatalogEntry>(catalog, schema, (BoundCreateTableInfo *)bound_create_info.get(), new_storage);
 	return move(result);
@@ -4460,7 +4495,7 @@ void TableCatalogEntry::CommitAlter(AlterInfo &info) {
 		}
 	}
 	D_ASSERT(removed_index != DConstants::INVALID_INDEX);
-	storage->CommitDropColumn(removed_index);
+	storage->CommitDropColumn(columns[removed_index].StorageOid());
 }
 
 void TableCatalogEntry::CommitDrop() {
@@ -5030,11 +5065,13 @@ bool CatalogSet::AlterEntry(ClientContext &context, const string &name, AlterInf
 				throw CatalogException(rename_err_msg, original_name, value->name);
 			}
 		}
+	}
+
+	if (value->name != original_name) {
+		// Do PutMapping and DeleteMapping after dependency check
 		PutMapping(context, value->name, entry_index);
 		DeleteMapping(context, original_name);
 	}
-	//! Check the dependency manager to verify that there are no conflicting dependencies with this alter
-	catalog.dependency_manager->AlterObject(context, entry, value.get());
 
 	value->timestamp = transaction.transaction_id;
 	value->child = move(entries[entry_index]);
@@ -5046,10 +5083,18 @@ bool CatalogSet::AlterEntry(ClientContext &context, const string &name, AlterInf
 	alter_info->Serialize(serializer);
 	BinaryData serialized_alter = serializer.GetData();
 
+	auto new_entry = value.get();
+
 	// push the old entry in the undo buffer for this transaction
 	transaction.PushCatalogEntry(value->child.get(), serialized_alter.data.get(), serialized_alter.size);
 	entries[entry_index] = move(value);
 
+	// Check the dependency manager to verify that there are no conflicting dependencies with this alter
+	// Note that we do this AFTER the new entry has been entirely set up in the catalog set
+	// that is because in case the alter fails because of a dependency conflict, we need to be able to cleanly roll back
+	// to the old entry.
+	catalog.dependency_manager->AlterObject(context, entry, new_entry);
+
 	return true;
 }
 
@@ -6602,7 +6647,7 @@ static void GetBitPosition(idx_t row_idx, idx_t &current_byte, uint8_t &current_
 }
 
 static void UnsetBit(uint8_t *data, idx_t current_byte, uint8_t current_bit) {
-	data[current_byte] &= ~(1 << current_bit);
+	data[current_byte] &= ~((uint64_t)1 << current_bit);
 }
 
 static void NextBit(idx_t &current_byte, uint8_t &current_bit) {
@@ -9546,6 +9591,8 @@ void Exception::ThrowAsTypeWithMessage(ExceptionType type, const string &message
 		throw ParameterNotAllowedException(message);
 	case ExceptionType::PARAMETER_NOT_RESOLVED:
 		throw ParameterNotResolvedException();
+	case ExceptionType::FATAL:
+		throw FatalException(message);
 	default:
 		throw Exception(type, message);
 	}
@@ -16811,9 +16858,15 @@ string FileSystem::ConvertSeparators(const string &path) {
 }
 
 string FileSystem::ExtractBaseName(const string &path) {
+	if (path.empty()) {
+		return string();
+	}
 	auto normalized_path = ConvertSeparators(path);
 	auto sep = PathSeparator();
-	auto vec = StringUtil::Split(StringUtil::Split(normalized_path, sep).back(), ".");
+	auto splits = StringUtil::Split(normalized_path, sep);
+	D_ASSERT(!splits.empty());
+	auto vec = StringUtil::Split(splits.back(), ".");
+	D_ASSERT(!vec.empty());
 	return vec[0];
 }
 
@@ -18888,7 +18941,8 @@ namespace duckdb {
 
 static unordered_map<column_t, string> GetKnownColumnValues(string &filename,
                                                             unordered_map<string, column_t> &column_map,
-                                                            bool filename_col, bool hive_partition_cols) {
+                                                            duckdb_re2::RE2 &compiled_regex, bool filename_col,
+                                                            bool hive_partition_cols) {
 	unordered_map<column_t, string> result;
 
 	if (filename_col) {
@@ -18899,7 +18953,7 @@ static unordered_map<column_t, string> GetKnownColumnValues(string &filename,
 	}
 
 	if (hive_partition_cols) {
-		auto partitions = HivePartitioning::Parse(filename);
+		auto partitions = HivePartitioning::Parse(filename, compiled_regex);
 		for (auto &partition : partitions) {
 			auto lookup_column_id = column_map.find(partition.first);
 			if (lookup_column_id != column_map.end()) {
@@ -18937,10 +18991,10 @@ static void ConvertKnownColRefToConstants(unique_ptr<Expression> &expr,
 // 	- s3://bucket/var1=value1/bla/bla/var2=value2
 //  - http(s)://domain(:port)/lala/kasdl/var1=value1/?not-a-var=not-a-value
 //  - folder/folder/folder/../var1=value1/etc/.//var2=value2
-std::map<string, string> HivePartitioning::Parse(string &filename) {
-	std::map<string, string> result;
+const string HivePartitioning::REGEX_STRING = "[\\/\\\\]([^\\/\\?\\\\]+)=([^\\/\\n\\?\\\\]+)";
 
-	string regex = "[\\/\\\\]([^\\/\\?\\\\]+)=([^\\/\\n\\?\\\\]+)";
+std::map<string, string> HivePartitioning::Parse(string &filename, duckdb_re2::RE2 &regex) {
+	std::map<string, string> result;
 	duckdb_re2::StringPiece input(filename); // Wrap a StringPiece around it
 
 	string var;
@@ -18951,6 +19005,11 @@ std::map<string, string> HivePartitioning::Parse(string &filename) {
 	return result;
 }
 
+std::map<string, string> HivePartitioning::Parse(string &filename) {
+	duckdb_re2::RE2 regex(REGEX_STRING);
+	return Parse(filename, regex);
+}
+
 // TODO: this can still be improved by removing the parts of filter expressions that are true for all remaining files.
 //		 currently, only expressions that cannot be evaluated during pushdown are removed.
 void HivePartitioning::ApplyFiltersToFileList(vector<string> &files, vector<unique_ptr<Expression>> &filters,
@@ -18958,6 +19017,7 @@ void HivePartitioning::ApplyFiltersToFileList(vector<string> &files, vector<uniq
                                               bool hive_enabled, bool filename_enabled) {
 	vector<string> pruned_files;
 	vector<unique_ptr<Expression>> pruned_filters;
+	duckdb_re2::RE2 regex(REGEX_STRING);
 
 	if ((!filename_enabled && !hive_enabled) || filters.empty()) {
 		return;
@@ -18966,7 +19026,7 @@ void HivePartitioning::ApplyFiltersToFileList(vector<string> &files, vector<uniq
 	for (idx_t i = 0; i < files.size(); i++) {
 		auto &file = files[i];
 		bool should_prune_file = false;
-		auto known_values = GetKnownColumnValues(file, column_map, filename_enabled, hive_enabled);
+		auto known_values = GetKnownColumnValues(file, column_map, regex, filename_enabled, hive_enabled);
 
 		FilterCombiner combiner;
 		for (auto &filter : filters) {
@@ -19198,6 +19258,8 @@ private:
 	//! Set the file pointer of a file handle to a specified location. Reads and writes will happen from this location
 	void SetFilePointer(FileHandle &handle, idx_t location);
 	idx_t GetFilePointer(FileHandle &handle);
+
+	vector<string> FetchFileWithoutGlob(const string &path, FileOpener *opener, bool absolute_path);
 };
 
 } // namespace duckdb
@@ -20079,6 +20141,26 @@ static void GlobFiles(FileSystem &fs, const string &path, const string &glob, bo
 	});
 }
 
+vector<string> LocalFileSystem::FetchFileWithoutGlob(const string &path, FileOpener *opener, bool absolute_path) {
+	vector<string> result;
+	if (FileExists(path) || IsPipe(path)) {
+		result.push_back(path);
+	} else if (!absolute_path) {
+		Value value;
+		if (opener->TryGetCurrentSetting("file_search_path", value)) {
+			auto search_paths_str = value.ToString();
+			std::vector<std::string> search_paths = StringUtil::Split(search_paths_str, ',');
+			for (const auto &search_path : search_paths) {
+				auto joined_path = JoinPath(search_path, path);
+				if (FileExists(joined_path) || IsPipe(joined_path)) {
+					result.push_back(joined_path);
+				}
+			}
+		}
+	}
+	return result;
+}
+
 vector<string> LocalFileSystem::Glob(const string &path, FileOpener *opener) {
 	if (path.empty()) {
 		return vector<string>();
@@ -20125,23 +20207,7 @@ vector<string> LocalFileSystem::Glob(const string &path, FileOpener *opener) {
 	// Check if the path has a glob at all
 	if (!HasGlob(path)) {
 		// no glob: return only the file (if it exists or is a pipe)
-		vector<string> result;
-		if (FileExists(path) || IsPipe(path)) {
-			result.push_back(path);
-		} else if (!absolute_path) {
-			Value value;
-			if (opener->TryGetCurrentSetting("file_search_path", value)) {
-				auto search_paths_str = value.ToString();
-				std::vector<std::string> search_paths = StringUtil::Split(search_paths_str, ',');
-				for (const auto &search_path : search_paths) {
-					auto joined_path = JoinPath(search_path, path);
-					if (FileExists(joined_path) || IsPipe(joined_path)) {
-						result.push_back(joined_path);
-					}
-				}
-			}
-		}
-		return result;
+		return FetchFileWithoutGlob(path, opener, absolute_path);
 	}
 	vector<string> previous_directories;
 	if (absolute_path) {
@@ -20175,7 +20241,12 @@ vector<string> LocalFileSystem::Glob(const string &path, FileOpener *opener) {
 				}
 			}
 		}
-		if (is_last_chunk || result.empty()) {
+		if (result.empty()) {
+			// no result found that matches the glob
+			// last ditch effort: search the path as a string literal
+			return FetchFileWithoutGlob(path, opener, absolute_path);
+		}
+		if (is_last_chunk) {
 			return result;
 		}
 		previous_directories = move(result);
@@ -22524,14 +22595,16 @@ struct IntervalToStringCast {
 			if (micros < 0) {
 				// negative time: append negative sign
 				buffer[length++] = '-';
+			} else {
 				micros = -micros;
 			}
-			int64_t hour = micros / Interval::MICROS_PER_HOUR;
-			micros -= hour * Interval::MICROS_PER_HOUR;
-			int64_t min = micros / Interval::MICROS_PER_MINUTE;
-			micros -= min * Interval::MICROS_PER_MINUTE;
-			int64_t sec = micros / Interval::MICROS_PER_SEC;
-			micros -= sec * Interval::MICROS_PER_SEC;
+			int64_t hour = -(micros / Interval::MICROS_PER_HOUR);
+			micros += hour * Interval::MICROS_PER_HOUR;
+			int64_t min = -(micros / Interval::MICROS_PER_MINUTE);
+			micros += min * Interval::MICROS_PER_MINUTE;
+			int64_t sec = -(micros / Interval::MICROS_PER_SEC);
+			micros += sec * Interval::MICROS_PER_SEC;
+			micros = -micros;
 
 			if (hour < 10) {
 				buffer[length++] = '0';
@@ -28654,7 +28727,7 @@ template <idx_t radix_bits>
 struct RadixPartitioningConstants {
 public:
 	static constexpr const idx_t NUM_RADIX_BITS = radix_bits;
-	static constexpr const idx_t NUM_PARTITIONS = 1 << NUM_RADIX_BITS;
+	static constexpr const idx_t NUM_PARTITIONS = (idx_t)1 << NUM_RADIX_BITS;
 	static constexpr const idx_t TMP_BUF_SIZE = 8;
 
 public:
@@ -28672,7 +28745,7 @@ private:
 struct RadixPartitioning {
 public:
 	static idx_t NumberOfPartitions(idx_t radix_bits) {
-		return 1 << radix_bits;
+		return (idx_t)1 << radix_bits;
 	}
 
 	//! Partition the data in block_collection/string_heap to multiple partitions
@@ -39531,7 +39604,7 @@ public:
 namespace duckdb {
 
 enum class UnicodeType { INVALID, ASCII, UNICODE };
-enum class UnicodeInvalidReason { BYTE_MISMATCH, NULL_BYTE };
+enum class UnicodeInvalidReason { BYTE_MISMATCH, NULL_BYTE, INVALID_UNICODE };
 
 class Utf8Proc {
 public:
@@ -50360,6 +50433,24 @@ void StringVector::AddHeapReference(Vector &vector, Vector &other) {
 	StringVector::AddBuffer(vector, other.auxiliary);
 }
 
+Vector &MapVector::GetKeys(Vector &vector) {
+	auto &entries = StructVector::GetEntries(vector);
+	D_ASSERT(entries.size() == 2);
+	return *entries[0];
+}
+Vector &MapVector::GetValues(Vector &vector) {
+	auto &entries = StructVector::GetEntries(vector);
+	D_ASSERT(entries.size() == 2);
+	return *entries[1];
+}
+
+const Vector &MapVector::GetKeys(const Vector &vector) {
+	return GetKeys((Vector &)vector);
+}
+const Vector &MapVector::GetValues(const Vector &vector) {
+	return GetValues((Vector &)vector);
+}
+
 vector<unique_ptr<Vector>> &StructVector::GetEntries(Vector &vector) {
 	D_ASSERT(vector.GetType().id() == LogicalTypeId::STRUCT || vector.GetType().id() == LogicalTypeId::MAP);
 	if (vector.GetVectorType() == VectorType::DICTIONARY_VECTOR) {
@@ -63217,6 +63308,16 @@ bool DistinctAggregateData::IsDistinct(idx_t index) const {
 
 
 
+//===----------------------------------------------------------------------===//
+//                         DuckDB
+//
+// duckdb/parallel/base_pipeline_event.hpp
+//
+//
+//===----------------------------------------------------------------------===//
+
+
+
 //===----------------------------------------------------------------------===//
 //                         DuckDB
 //
@@ -63290,6 +63391,22 @@ protected:
 
 
 
+namespace duckdb {
+
+//! A BasePipelineEvent is used as the basis of any event that belongs to a specific pipeline
+class BasePipelineEvent : public Event {
+public:
+	BasePipelineEvent(shared_ptr<Pipeline> pipeline);
+	BasePipelineEvent(Pipeline &pipeline);
+
+	//! The pipeline that this event belongs to
+	shared_ptr<Pipeline> pipeline;
+};
+
+} // namespace duckdb
+
+
+
 namespace duckdb {
 
 PhysicalHashAggregate::PhysicalHashAggregate(ClientContext &context, vector<LogicalType> types,
@@ -63446,16 +63563,15 @@ void PhysicalHashAggregate::Combine(ExecutionContext &context, GlobalSinkState &
 	}
 }
 
-class HashAggregateFinalizeEvent : public Event {
+class HashAggregateFinalizeEvent : public BasePipelineEvent {
 public:
 	HashAggregateFinalizeEvent(const PhysicalHashAggregate &op_p, HashAggregateGlobalState &gstate_p,
 	                           Pipeline *pipeline_p)
-	    : Event(pipeline_p->executor), op(op_p), gstate(gstate_p), pipeline(pipeline_p) {
+	    : BasePipelineEvent(*pipeline_p), op(op_p), gstate(gstate_p) {
 	}
 
 	const PhysicalHashAggregate &op;
 	HashAggregateGlobalState &gstate;
-	Pipeline *pipeline;
 
 public:
 	void Schedule() override {
@@ -64717,15 +64833,14 @@ private:
 };
 
 // TODO: Create tasks and run these in parallel instead of doing this all in Schedule, single threaded
-class DistinctAggregateFinalizeEvent : public Event {
+class DistinctAggregateFinalizeEvent : public BasePipelineEvent {
 public:
 	DistinctAggregateFinalizeEvent(const PhysicalUngroupedAggregate &op_p, UngroupedAggregateGlobalState &gstate_p,
-	                               Pipeline *pipeline_p, ClientContext &context)
-	    : Event(pipeline_p->executor), op(op_p), gstate(gstate_p), pipeline(pipeline_p), context(context) {
+	                               Pipeline &pipeline_p, ClientContext &context)
+	    : BasePipelineEvent(pipeline_p), op(op_p), gstate(gstate_p), context(context) {
 	}
 	const PhysicalUngroupedAggregate &op;
 	UngroupedAggregateGlobalState &gstate;
-	Pipeline *pipeline;
 	ClientContext &context;
 
 public:
@@ -64738,16 +64853,15 @@ public:
 	}
 };
 
-class DistinctCombineFinalizeEvent : public Event {
+class DistinctCombineFinalizeEvent : public BasePipelineEvent {
 public:
 	DistinctCombineFinalizeEvent(const PhysicalUngroupedAggregate &op_p, UngroupedAggregateGlobalState &gstate_p,
-	                             Pipeline *pipeline_p, ClientContext &client)
-	    : Event(pipeline_p->executor), op(op_p), gstate(gstate_p), pipeline(pipeline_p), client(client) {
+	                             Pipeline &pipeline_p, ClientContext &client)
+	    : BasePipelineEvent(pipeline_p), op(op_p), gstate(gstate_p), client(client) {
 	}
 
 	const PhysicalUngroupedAggregate &op;
 	UngroupedAggregateGlobalState &gstate;
-	Pipeline *pipeline;
 	ClientContext &client;
 
 public:
@@ -64763,7 +64877,7 @@ public:
 		SetTasks(move(tasks));
 
 		//! Now that all tables are combined, it's time to do the distinct aggregations
-		auto new_event = make_shared<DistinctAggregateFinalizeEvent>(op, gstate, pipeline, client);
+		auto new_event = make_shared<DistinctAggregateFinalizeEvent>(op, gstate, *pipeline, client);
 		this->InsertEvent(move(new_event));
 	}
 };
@@ -64792,12 +64906,12 @@ SinkFinalizeType PhysicalUngroupedAggregate::FinalizeDistinct(Pipeline &pipeline
 		}
 	}
 	if (any_partitioned) {
-		auto new_event = make_shared<DistinctCombineFinalizeEvent>(*this, gstate, &pipeline, context);
+		auto new_event = make_shared<DistinctCombineFinalizeEvent>(*this, gstate, pipeline, context);
 		event.InsertEvent(move(new_event));
 	} else {
 		//! Hashtables aren't partitioned, they dont need to be joined first
 		//! So we can compute the aggregate already
-		auto new_event = make_shared<DistinctAggregateFinalizeEvent>(*this, gstate, &pipeline, context);
+		auto new_event = make_shared<DistinctAggregateFinalizeEvent>(*this, gstate, pipeline, context);
 		event.InsertEvent(move(new_event));
 	}
 	return SinkFinalizeType::READY;
@@ -66543,19 +66657,18 @@ private:
 	WindowGlobalHashGroup &hash_group;
 };
 
-class WindowMergeEvent : public Event {
+class WindowMergeEvent : public BasePipelineEvent {
 public:
 	WindowMergeEvent(WindowGlobalSinkState &gstate_p, Pipeline &pipeline_p, WindowGlobalHashGroup &hash_group_p)
-	    : Event(pipeline_p.executor), gstate(gstate_p), pipeline(pipeline_p), hash_group(hash_group_p) {
+	    : BasePipelineEvent(pipeline_p), gstate(gstate_p), hash_group(hash_group_p) {
 	}
 
 	WindowGlobalSinkState &gstate;
-	Pipeline &pipeline;
 	WindowGlobalHashGroup &hash_group;
 
 public:
 	void Schedule() override {
-		auto &context = pipeline.GetClientContext();
+		auto &context = pipeline->GetClientContext();
 
 		// Schedule tasks equal to the number of threads, which will each merge multiple partitions
 		auto &ts = TaskScheduler::GetScheduler(context);
@@ -66570,7 +66683,7 @@ public:
 
 	void FinishEvent() override {
 		hash_group.global_sort->CompleteMergeRound(true);
-		CreateMergeTasks(pipeline, *this, gstate, hash_group);
+		CreateMergeTasks(*pipeline, *this, gstate, hash_group);
 	}
 
 	static void CreateMergeTasks(Pipeline &pipeline, Event &event, WindowGlobalSinkState &state,
@@ -67979,6 +68092,11 @@ public:
 
 private:
 	static const vector<string> PathComponents();
+	//! For tagged releases we use the tag, else we use the git commit hash
+	static const string GetVersionDirectoryName();
+	//! Version tags occur with and without 'v', tag in extension path is always with 'v'
+	static const string NormalizeVersionTag(const string &version_tag);
+	static bool IsRelease(const string &version_tag);
 
 private:
 	static ExtensionLoadResult LoadExtensionInternal(DuckDB &db, const std::string &extension, bool initial_load);
@@ -70905,18 +71023,17 @@ private:
 	bool parallel;
 };
 
-class HashJoinFinalizeEvent : public Event {
+class HashJoinFinalizeEvent : public BasePipelineEvent {
 public:
 	HashJoinFinalizeEvent(Pipeline &pipeline_p, HashJoinGlobalSinkState &sink)
-	    : Event(pipeline_p.executor), pipeline(pipeline_p), sink(sink) {
+	    : BasePipelineEvent(pipeline_p), sink(sink) {
 	}
 
-	Pipeline &pipeline;
 	HashJoinGlobalSinkState &sink;
 
 public:
 	void Schedule() override {
-		auto &context = pipeline.GetClientContext();
+		auto &context = pipeline->GetClientContext();
 		auto parallel_construct_count =
 		    context.config.verify_parallelism ? STANDARD_VECTOR_SIZE : PARALLEL_CONSTRUCT_COUNT;
 
@@ -70983,20 +71100,19 @@ private:
 	JoinHashTable &local_ht;
 };
 
-class HashJoinPartitionEvent : public Event {
+class HashJoinPartitionEvent : public BasePipelineEvent {
 public:
 	HashJoinPartitionEvent(Pipeline &pipeline_p, HashJoinGlobalSinkState &sink,
 	                       vector<unique_ptr<JoinHashTable>> &local_hts)
-	    : Event(pipeline_p.executor), pipeline(pipeline_p), sink(sink), local_hts(local_hts) {
+	    : BasePipelineEvent(pipeline_p), sink(sink), local_hts(local_hts) {
 	}
 
-	Pipeline &pipeline;
 	HashJoinGlobalSinkState &sink;
 	vector<unique_ptr<JoinHashTable>> &local_hts;
 
 public:
 	void Schedule() override {
-		auto &context = pipeline.GetClientContext();
+		auto &context = pipeline->GetClientContext();
 		vector<unique_ptr<Task>> partition_tasks;
 		partition_tasks.reserve(local_hts.size());
 		for (auto &local_ht : local_hts) {
@@ -71009,7 +71125,7 @@ public:
 	void FinishEvent() override {
 		local_hts.clear();
 		sink.hash_table->PrepareExternalFinalize();
-		sink.ScheduleFinalize(pipeline, *this);
+		sink.ScheduleFinalize(*pipeline, *this);
 	}
 };
 
@@ -74713,21 +74829,20 @@ private:
 	GlobalSortedTable &table;
 };
 
-class RangeJoinMergeEvent : public Event {
+class RangeJoinMergeEvent : public BasePipelineEvent {
 public:
 	using GlobalSortedTable = PhysicalRangeJoin::GlobalSortedTable;
 
 public:
 	RangeJoinMergeEvent(GlobalSortedTable &table_p, Pipeline &pipeline_p)
-	    : Event(pipeline_p.executor), table(table_p), pipeline(pipeline_p) {
+	    : BasePipelineEvent(pipeline_p), table(table_p) {
 	}
 
 	GlobalSortedTable &table;
-	Pipeline &pipeline;
 
 public:
 	void Schedule() override {
-		auto &context = pipeline.GetClientContext();
+		auto &context = pipeline->GetClientContext();
 
 		// Schedule tasks equal to the number of threads, which will each merge multiple partitions
 		auto &ts = TaskScheduler::GetScheduler(context);
@@ -74746,7 +74861,7 @@ public:
 		global_sort_state.CompleteMergeRound(true);
 		if (global_sort_state.sorted_blocks.size() > 1) {
 			// Multiple blocks remaining: Schedule the next round
-			table.ScheduleMergeTasks(pipeline, *this);
+			table.ScheduleMergeTasks(*pipeline, *this);
 		}
 	}
 };
@@ -75134,18 +75249,17 @@ private:
 	OrderGlobalState &state;
 };
 
-class OrderMergeEvent : public Event {
+class OrderMergeEvent : public BasePipelineEvent {
 public:
 	OrderMergeEvent(OrderGlobalState &gstate_p, Pipeline &pipeline_p)
-	    : Event(pipeline_p.executor), gstate(gstate_p), pipeline(pipeline_p) {
+	    : BasePipelineEvent(pipeline_p), gstate(gstate_p) {
 	}
 
 	OrderGlobalState &gstate;
-	Pipeline &pipeline;
 
 public:
 	void Schedule() override {
-		auto &context = pipeline.GetClientContext();
+		auto &context = pipeline->GetClientContext();
 
 		// Schedule tasks equal to the number of threads, which will each merge multiple partitions
 		auto &ts = TaskScheduler::GetScheduler(context);
@@ -75164,7 +75278,7 @@ public:
 		global_sort_state.CompleteMergeRound();
 		if (global_sort_state.sorted_blocks.size() > 1) {
 			// Multiple blocks remaining: Schedule the next round
-			PhysicalOrder::ScheduleMergeTasks(pipeline, *this, gstate);
+			PhysicalOrder::ScheduleMergeTasks(*pipeline, *this, gstate);
 		}
 	}
 };
@@ -80064,10 +80178,17 @@ void PhysicalCreateIndex::GetData(ExecutionContext &context, DataChunk &chunk, G
 		return;
 	}
 
+	// convert virtual column ids to storage column ids
+	vector<column_t> storage_ids;
+	for (auto &column_id : column_ids) {
+		D_ASSERT(column_id < table.columns.size());
+		storage_ids.push_back(table.columns[column_id].StorageOid());
+	}
+
 	unique_ptr<Index> index;
 	switch (info->index_type) {
 	case IndexType::ART: {
-		index = make_unique<ART>(column_ids, unbound_expressions, info->constraint_type, *context.client.db);
+		index = make_unique<ART>(storage_ids, unbound_expressions, info->constraint_type, *context.client.db);
 		break;
 	}
 	default:
@@ -80372,11 +80493,10 @@ unique_ptr<GlobalSinkState> PhysicalCreateTableAs::GetGlobalSinkState(ClientCont
 SinkResultType PhysicalCreateTableAs::Sink(ExecutionContext &context, GlobalSinkState &state, LocalSinkState &lstate_p,
                                            DataChunk &input) const {
 	auto &sink = (CreateTableAsGlobalState &)state;
-	if (sink.table) {
-		lock_guard<mutex> client_guard(sink.append_lock);
-		sink.table->storage->Append(*sink.table, context.client, input);
-		sink.inserted_count += input.size();
-	}
+	D_ASSERT(sink.table);
+	lock_guard<mutex> client_guard(sink.append_lock);
+	sink.table->storage->Append(*sink.table, context.client, input);
+	sink.inserted_count += input.size();
 	return SinkResultType::NEED_MORE_INPUT;
 }
 
@@ -80786,6 +80906,7 @@ void PhysicalRecursiveCTE::ExecuteRecursivePipelines(ExecutionContext &context)
 void PhysicalRecursiveCTE::BuildPipelines(Executor &executor, Pipeline &current, PipelineBuildState &state) {
 	op_state.reset();
 	sink_state.reset();
+	pipelines.clear();
 
 	// recursive CTE
 	state.SetPipelineSource(current, this);
@@ -81085,7 +81206,7 @@ PerfectAggregateHashTable::PerfectAggregateHashTable(Allocator &allocator, Buffe
 		total_required_bits += group_bits;
 	}
 	// the total amount of groups we allocate space for is 2^required_bits
-	total_groups = 1 << total_required_bits;
+	total_groups = (uint64_t)1 << total_required_bits;
 	// we don't need to store the groups in a perfect hash table, since the group keys can be deduced by their location
 	grouping_columns = group_types_p.size();
 	layout.Initialize(move(aggregate_objects_p));
@@ -81269,7 +81390,7 @@ static void ReconstructGroupVectorTemplated(uint32_t group_values[], Value &min,
 static void ReconstructGroupVector(uint32_t group_values[], Value &min, idx_t required_bits, idx_t shift,
                                    idx_t entry_count, Vector &result) {
 	// construct the mask for this entry
-	idx_t mask = (1 << required_bits) - 1;
+	idx_t mask = ((uint64_t)1 << required_bits) - 1;
 	switch (result.GetType().InternalType()) {
 	case PhysicalType::INT8:
 		ReconstructGroupVectorTemplated<int8_t>(group_values, min, mask, shift, entry_count, result);
@@ -85516,7 +85637,7 @@ void RadixPartitionedHashTable::SetGroupingValues() {
 		for (idx_t i = 0; i < grouping.size(); i++) {
 			if (grouping_set.find(grouping[i]) == grouping_set.end()) {
 				// we don't group on this value!
-				grouping_value += 1 << (grouping.size() - (i + 1));
+				grouping_value += (int64_t)1 << (grouping.size() - (i + 1));
 			}
 		}
 		grouping_values.push_back(Value::BIGINT(grouping_value));
@@ -91074,7 +91195,21 @@ struct ModeIncluded {
 	const idx_t bias;
 };
 
-template <typename KEY_TYPE>
+struct ModeAssignmentStandard {
+	template <class INPUT_TYPE, class RESULT_TYPE>
+	static RESULT_TYPE Assign(Vector &result, INPUT_TYPE input) {
+		return RESULT_TYPE(input);
+	}
+};
+
+struct ModeAssignmentString {
+	template <class INPUT_TYPE, class RESULT_TYPE>
+	static RESULT_TYPE Assign(Vector &result, INPUT_TYPE input) {
+		return StringVector::AddString(result, input);
+	}
+};
+
+template <typename KEY_TYPE, typename ASSIGN_OP>
 struct ModeFunction {
 	template <class STATE>
 	static void Initialize(STATE *state) {
@@ -91187,7 +91322,7 @@ struct ModeFunction {
 		}
 
 		if (state->valid) {
-			rdata[rid] = RESULT_TYPE(*state->mode);
+			rdata[rid] = ASSIGN_OP::template Assign<INPUT_TYPE, RESULT_TYPE>(result, *state->mode);
 		} else {
 			rmask.Set(rid, false);
 		}
@@ -91203,10 +91338,10 @@ struct ModeFunction {
 	}
 };
 
-template <typename INPUT_TYPE, typename KEY_TYPE>
+template <typename INPUT_TYPE, typename KEY_TYPE, typename ASSIGN_OP = ModeAssignmentStandard>
 AggregateFunction GetTypedModeFunction(const LogicalType &type) {
 	using STATE = ModeState<KEY_TYPE>;
-	using OP = ModeFunction<KEY_TYPE>;
+	using OP = ModeFunction<KEY_TYPE, ASSIGN_OP>;
 	auto func = AggregateFunction::UnaryAggregateDestructor<STATE, INPUT_TYPE, INPUT_TYPE, OP>(type, type);
 	func.window = AggregateFunction::UnaryWindow<STATE, INPUT_TYPE, INPUT_TYPE, OP>;
 	return func;
@@ -91242,7 +91377,7 @@ AggregateFunction GetModeAggregate(const LogicalType &type) {
 		return GetTypedModeFunction<interval_t, interval_t>(type);
 
 	case PhysicalType::VARCHAR:
-		return GetTypedModeFunction<string_t, string>(type);
+		return GetTypedModeFunction<string_t, string, ModeAssignmentString>(type);
 
 	default:
 		throw NotImplementedException("Unimplemented mode aggregate");
@@ -105407,16 +105542,21 @@ static void MapExtractFunction(DataChunk &args, ExpressionState &state, Vector &
 	auto &map = args.data[0];
 	auto &key = args.data[1];
 
-	UnifiedVectorFormat offset_data;
+	UnifiedVectorFormat map_keys_data;
+	UnifiedVectorFormat key_data;
 
-	auto &children = StructVector::GetEntries(map);
+	auto &map_keys = MapVector::GetKeys(map);
+	auto &map_values = MapVector::GetValues(map);
+
+	map_keys.ToUnifiedFormat(args.size(), map_keys_data);
+	key.ToUnifiedFormat(args.size(), key_data);
 
-	children[0]->ToUnifiedFormat(args.size(), offset_data);
 	for (idx_t row = 0; row < args.size(); row++) {
-		idx_t row_index = offset_data.sel->get_index(row);
-		auto key_value = key.GetValue(row_index);
-		auto offsets = ListVector::Search(*children[0], key_value, offset_data.sel->get_index(row));
-		auto values = ListVector::GetValuesFromOffsets(*children[1], offsets);
+		idx_t row_index = map_keys_data.sel->get_index(row);
+		idx_t key_index = key_data.sel->get_index(row);
+		auto key_value = key.GetValue(key_index);
+		auto offsets = ListVector::Search(map_keys, key_value, row_index);
+		auto values = ListVector::GetValuesFromOffsets(map_values, offsets);
 		FillResult(values, result, row);
 	}
 
@@ -108311,6 +108451,24 @@ interval_t DivideOperator::Operation(interval_t left, int64_t right) {
 	return left;
 }
 
+struct BinaryNumericDivideWrapper {
+	template <class FUNC, class OP, class LEFT_TYPE, class RIGHT_TYPE, class RESULT_TYPE>
+	static inline RESULT_TYPE Operation(FUNC fun, LEFT_TYPE left, RIGHT_TYPE right, ValidityMask &mask, idx_t idx) {
+		if (left == NumericLimits<LEFT_TYPE>::Minimum() && right == -1) {
+			throw OutOfRangeException("Overflow in division of %d / %d", left, right);
+		} else if (right == 0) {
+			mask.SetInvalid(idx);
+			return left;
+		} else {
+			return OP::template Operation<LEFT_TYPE, RIGHT_TYPE, RESULT_TYPE>(left, right);
+		}
+	}
+
+	static bool AddsNulls() {
+		return true;
+	}
+};
+
 struct BinaryZeroIsNullWrapper {
 	template <class FUNC, class OP, class LEFT_TYPE, class RIGHT_TYPE, class RESULT_TYPE>
 	static inline RESULT_TYPE Operation(FUNC fun, LEFT_TYPE left, RIGHT_TYPE right, ValidityMask &mask, idx_t idx) {
@@ -108352,13 +108510,13 @@ template <class OP>
 static scalar_function_t GetBinaryFunctionIgnoreZero(const LogicalType &type) {
 	switch (type.id()) {
 	case LogicalTypeId::TINYINT:
-		return BinaryScalarFunctionIgnoreZero<int8_t, int8_t, int8_t, OP>;
+		return BinaryScalarFunctionIgnoreZero<int8_t, int8_t, int8_t, OP, BinaryNumericDivideWrapper>;
 	case LogicalTypeId::SMALLINT:
-		return BinaryScalarFunctionIgnoreZero<int16_t, int16_t, int16_t, OP>;
+		return BinaryScalarFunctionIgnoreZero<int16_t, int16_t, int16_t, OP, BinaryNumericDivideWrapper>;
 	case LogicalTypeId::INTEGER:
-		return BinaryScalarFunctionIgnoreZero<int32_t, int32_t, int32_t, OP>;
+		return BinaryScalarFunctionIgnoreZero<int32_t, int32_t, int32_t, OP, BinaryNumericDivideWrapper>;
 	case LogicalTypeId::BIGINT:
-		return BinaryScalarFunctionIgnoreZero<int64_t, int64_t, int64_t, OP>;
+		return BinaryScalarFunctionIgnoreZero<int64_t, int64_t, int64_t, OP, BinaryNumericDivideWrapper>;
 	case LogicalTypeId::UTINYINT:
 		return BinaryScalarFunctionIgnoreZero<uint8_t, uint8_t, uint8_t, OP>;
 	case LogicalTypeId::USMALLINT:
@@ -114806,11 +114964,22 @@ static void CurrentSchemaFunction(DataChunk &input, ExpressionState &state, Vect
 
 // current_schemas
 static void CurrentSchemasFunction(DataChunk &input, ExpressionState &state, Vector &result) {
+	if (!input.AllConstant()) {
+		throw NotImplementedException("current_schemas requires a constant input");
+	}
+	if (ConstantVector::IsNull(input.data[0])) {
+		result.SetVectorType(VectorType::CONSTANT_VECTOR);
+		ConstantVector::SetNull(result, true);
+		return;
+	}
+	auto implicit_schemas = *ConstantVector::GetData<bool>(input.data[0]);
 	vector<Value> schema_list;
-	vector<string> search_path = ClientData::Get(SystemBindData::GetFrom(state).context).catalog_search_path->Get();
+	auto &catalog_search_path = ClientData::Get(SystemBindData::GetFrom(state).context).catalog_search_path;
+	vector<string> search_path = implicit_schemas ? catalog_search_path->Get() : catalog_search_path->GetSetPaths();
 	std::transform(search_path.begin(), search_path.end(), std::back_inserter(schema_list),
 	               [](const string &s) -> Value { return Value(s); });
-	auto val = Value::LIST(schema_list);
+
+	auto val = Value::LIST(LogicalType::VARCHAR, schema_list);
 	result.Reference(val);
 }
 
@@ -115109,8 +115278,8 @@ struct ArrowScanLocalState : public LocalTableFunctionState {
 struct ArrowScanGlobalState : public GlobalTableFunctionState {
 	unique_ptr<ArrowArrayStreamWrapper> stream;
 	mutex main_mutex;
-	bool ready = false;
 	idx_t max_threads = 1;
+	bool done = false;
 
 	idx_t MaxThreads() const override {
 		return max_threads;
@@ -115398,6 +115567,9 @@ idx_t ArrowTableFunction::ArrowScanMaxThreads(ClientContext &context, const Func
 bool ArrowScanParallelStateNext(ClientContext &context, const FunctionData *bind_data_p, ArrowScanLocalState &state,
                                 ArrowScanGlobalState &parallel_state) {
 	lock_guard<mutex> parallel_lock(parallel_state.main_mutex);
+	if (parallel_state.done) {
+		return false;
+	}
 	state.chunk_offset = 0;
 
 	auto current_chunk = parallel_state.stream->GetNextChunk();
@@ -115407,6 +115579,7 @@ bool ArrowScanParallelStateNext(ClientContext &context, const FunctionData *bind
 	state.chunk = move(current_chunk);
 	//! have we run out of chunks? we are done
 	if (!state.chunk->arrow_array.release) {
+		parallel_state.done = true;
 		return false;
 	}
 	return true;
@@ -117808,6 +117981,7 @@ static void ReadCSVAddNamedParameters(TableFunction &table_function) {
 	table_function.named_parameters["skip"] = LogicalType::BIGINT;
 	table_function.named_parameters["max_line_size"] = LogicalType::VARCHAR;
 	table_function.named_parameters["maximum_line_size"] = LogicalType::VARCHAR;
+	table_function.named_parameters["ignore_errors"] = LogicalType::BOOLEAN;
 }
 
 double CSVReaderProgress(ClientContext &context, const FunctionData *bind_data_p,
@@ -121638,8 +121812,7 @@ static unique_ptr<BaseStatistics> TableScanStatistics(ClientContext &context, co
 		// we don't emit any statistics for tables that have outstanding transaction-local data
 		return nullptr;
 	}
-	auto storage_idx = GetStorageIndex(*bind_data.table, column_id);
-	return bind_data.table->storage->GetStatistics(context, storage_idx);
+	return bind_data.table->GetStatistics(context, column_id);
 }
 
 static void TableScanFunc(ClientContext &context, TableFunctionInput &data_p, DataChunk &output) {
@@ -123211,7 +123384,7 @@ bool duckdb_validity_row_is_valid(uint64_t *validity, idx_t row) {
 	}
 	idx_t entry_idx = row / 64;
 	idx_t idx_in_entry = row % 64;
-	return validity[entry_idx] & (1 << idx_in_entry);
+	return validity[entry_idx] & ((idx_t)1 << idx_in_entry);
 }
 
 void duckdb_validity_set_row_validity(uint64_t *validity, idx_t row, bool valid) {
@@ -123228,7 +123401,7 @@ void duckdb_validity_set_row_invalid(uint64_t *validity, idx_t row) {
 	}
 	idx_t entry_idx = row / 64;
 	idx_t idx_in_entry = row % 64;
-	validity[entry_idx] &= ~(1 << idx_in_entry);
+	validity[entry_idx] &= ~((uint64_t)1 << idx_in_entry);
 }
 
 void duckdb_validity_set_row_valid(uint64_t *validity, idx_t row) {
@@ -123237,7 +123410,7 @@ void duckdb_validity_set_row_valid(uint64_t *validity, idx_t row) {
 	}
 	idx_t entry_idx = row / 64;
 	idx_t idx_in_entry = row % 64;
-	validity[entry_idx] |= 1 << idx_in_entry;
+	validity[entry_idx] |= (uint64_t)1 << idx_in_entry;
 }
 
 
@@ -126420,6 +126593,11 @@ PendingExecutionResult ClientContext::ExecuteTaskInternal(ClientContextLock &loc
 			query_progress = active_query->progress_bar->GetCurrentPercentage();
 		}
 		return result;
+	} catch (FatalException &ex) {
+		// fatal exceptions invalidate the entire database
+		result.SetError(PreservedError(ex));
+		auto &db = DatabaseInstance::GetDatabase(*this);
+		db.Invalidate();
 	} catch (const Exception &ex) {
 		result.SetError(PreservedError(ex));
 	} catch (std::exception &ex) {
@@ -126639,9 +126817,19 @@ unique_ptr<PendingQueryResult> ClientContext::PendingStatementOrPreparedStatemen
 		case StatementType::INSERT_STATEMENT:
 		case StatementType::DELETE_STATEMENT:
 		case StatementType::UPDATE_STATEMENT: {
-			auto sql = statement->ToString();
 			Parser parser;
-			parser.ParseQuery(sql);
+			PreservedError error;
+			try {
+				parser.ParseQuery(statement->ToString());
+			} catch (const Exception &ex) {
+				error = PreservedError(ex);
+			} catch (std::exception &ex) {
+				error = PreservedError(ex);
+			}
+			if (error) {
+				// error in verifying query
+				return make_unique<PendingQueryResult>(error);
+			}
 			statement = move(parser.statements[0]);
 			break;
 		}
@@ -137289,8 +137477,27 @@ namespace duckdb {
 //===--------------------------------------------------------------------===//
 // Install Extension
 //===--------------------------------------------------------------------===//
+const string ExtensionHelper::NormalizeVersionTag(const string &version_tag) {
+	if (version_tag.length() > 0 && version_tag[0] != 'v') {
+		return "v" + version_tag;
+	}
+	return version_tag;
+}
+
+bool ExtensionHelper::IsRelease(const string &version_tag) {
+	return !StringUtil::Contains(version_tag, "-dev");
+}
+
+const string ExtensionHelper::GetVersionDirectoryName() {
+	if (IsRelease(DuckDB::LibraryVersion())) {
+		return NormalizeVersionTag(DuckDB::LibraryVersion());
+	} else {
+		return DuckDB::SourceID();
+	}
+}
+
 const vector<string> ExtensionHelper::PathComponents() {
-	return vector<string> {".duckdb", "extensions", DuckDB::SourceID(), DuckDB::Platform()};
+	return vector<string> {".duckdb", "extensions", GetVersionDirectoryName(), DuckDB::Platform()};
 }
 
 string ExtensionHelper::ExtensionDirectory(ClientContext &context) {
@@ -137363,7 +137570,7 @@ void ExtensionHelper::InstallExtension(ClientContext &context, const string &ext
 		extension_name = "";
 	}
 
-	auto url = StringUtil::Replace(url_template, "${REVISION}", DuckDB::SourceID());
+	auto url = StringUtil::Replace(url_template, "${REVISION}", GetVersionDirectoryName());
 	url = StringUtil::Replace(url, "${PLATFORM}", DuckDB::Platform());
 	url = StringUtil::Replace(url, "${NAME}", extension_name);
 
@@ -142541,9 +142748,7 @@ void CardinalityEstimator::UpdateTotalDomains(JoinNode *node, LogicalOperator *o
 			// Get HLL stats here
 			auto actual_binding = relation_column_to_original_column[key];
 
-			// sometimes base stats is null (test_709.test) returns null for base stats while
-			// there is still a catalog table. Anybody know anything about this?
-			auto base_stats = catalog_table->storage->GetStatistics(context, actual_binding.column_index);
+			auto base_stats = catalog_table->GetStatistics(context, actual_binding.column_index);
 			if (base_stats) {
 				count = base_stats->GetDistinctCount();
 			}
@@ -143239,6 +143444,7 @@ private:
 
 
 
+
 namespace duckdb {
 
 class DeliminatorPlanUpdater : LogicalOperatorVisitor {
@@ -143266,7 +143472,15 @@ void DeliminatorPlanUpdater::VisitOperator(LogicalOperator &op) {
 			    cond.comparison != ExpressionType::COMPARE_NOT_DISTINCT_FROM) {
 				continue;
 			}
-			auto &colref = (BoundColumnRefExpression &)*cond.right;
+			Expression *rhs = cond.right.get();
+			while (rhs->type == ExpressionType::OPERATOR_CAST) {
+				auto &cast = (BoundCastExpression &)*rhs;
+				rhs = cast.child.get();
+			}
+			if (rhs->type != ExpressionType::BOUND_COLUMN_REF) {
+				throw InternalException("Erorr in deliminator: expected a bound column reference");
+			}
+			auto &colref = (BoundColumnRefExpression &)*rhs;
 			if (projection_map.find(colref.binding) != projection_map.end()) {
 				// value on the right is a projection of removed DelimGet
 				for (idx_t i = 0; i < decs->size(); i++) {
@@ -144414,7 +144628,10 @@ FilterResult FilterCombiner::AddBoundComparisonFilter(Expression *expr) {
 		auto node = GetNode(left_is_scalar ? comparison.right.get() : comparison.left.get());
 		idx_t equivalence_set = GetEquivalenceSet(node);
 		auto scalar = left_is_scalar ? comparison.left.get() : comparison.right.get();
-		auto constant_value = ExpressionExecutor::EvaluateScalar(*scalar);
+		Value constant_value;
+		if (!ExpressionExecutor::TryEvaluateScalar(*scalar, constant_value)) {
+			return FilterResult::UNSATISFIABLE;
+		}
 		if (constant_value.IsNull()) {
 			// comparisons with null are always null (i.e. will never result in rows)
 			return FilterResult::UNSATISFIABLE;
@@ -144495,7 +144712,11 @@ FilterResult FilterCombiner::AddFilter(Expression *expr) {
 	}
 	if (expr->IsFoldable()) {
 		// scalar condition, evaluate it
-		auto result = ExpressionExecutor::EvaluateScalar(*expr).CastAs(LogicalType::BOOLEAN);
+		Value result;
+		if (!ExpressionExecutor::TryEvaluateScalar(*expr, result)) {
+			return FilterResult::UNSUPPORTED;
+		}
+		result = result.CastAs(LogicalType::BOOLEAN);
 		// check if the filter passes
 		if (result.IsNull() || !BooleanValue::Get(result)) {
 			// the filter does not pass the scalar test, create an empty result
@@ -144519,7 +144740,10 @@ FilterResult FilterCombiner::AddFilter(Expression *expr) {
 
 			if (lower_is_scalar) {
 				auto scalar = comparison.lower.get();
-				auto constant_value = ExpressionExecutor::EvaluateScalar(*scalar);
+				Value constant_value;
+				if (!ExpressionExecutor::TryEvaluateScalar(*scalar, constant_value)) {
+					return FilterResult::UNSUPPORTED;
+				}
 
 				// create the ExpressionValueInformation
 				ExpressionValueInformation info;
@@ -144552,7 +144776,10 @@ FilterResult FilterCombiner::AddFilter(Expression *expr) {
 
 			if (upper_is_scalar) {
 				auto scalar = comparison.upper.get();
-				auto constant_value = ExpressionExecutor::EvaluateScalar(*scalar);
+				Value constant_value;
+				if (!ExpressionExecutor::TryEvaluateScalar(*scalar, constant_value)) {
+					return FilterResult::UNSUPPORTED;
+				}
 
 				// create the ExpressionValueInformation
 				ExpressionValueInformation info;
@@ -145464,7 +145691,6 @@ unique_ptr<Expression> InClauseRewriter::VisitReplace(BoundOperatorExpression &e
 	// IN clause with many children: try to generate a mark join that replaces this IN expression
 	// we can only do this if the expressions in the expression list are scalar
 	for (idx_t i = 1; i < expr.children.size(); i++) {
-		D_ASSERT(expr.children[i]->return_type == in_type);
 		if (!expr.children[i]->IsFoldable()) {
 			// non-scalar expression
 			all_scalar = false;
@@ -147903,21 +148129,35 @@ unique_ptr<LogicalOperator> FilterPushdown::PushdownAggregate(unique_ptr<Logical
 	FilterPushdown child_pushdown(optimizer);
 	for (idx_t i = 0; i < filters.size(); i++) {
 		auto &f = *filters[i];
-		// check if any aggregate or GROUPING functions are in the set
-		if (f.bindings.find(aggr.aggregate_index) == f.bindings.end() &&
-		    f.bindings.find(aggr.groupings_index) == f.bindings.end()) {
-			// no aggregate! we can push this down
-			// rewrite any group bindings within the filter
-			f.filter = ReplaceGroupBindings(aggr, move(f.filter));
-			// add the filter to the child node
-			if (child_pushdown.AddFilter(move(f.filter)) == FilterResult::UNSATISFIABLE) {
-				// filter statically evaluates to false, strip tree
-				return make_unique<LogicalEmptyResult>(move(op));
+		if (f.bindings.find(aggr.aggregate_index) != f.bindings.end()) {
+			// filter on aggregate: cannot pushdown
+			continue;
+		}
+		if (f.bindings.find(aggr.groupings_index) != f.bindings.end()) {
+			// filter on GROUPINGS function: cannot pushdown
+			continue;
+		}
+		// if there are any empty grouping sets, we cannot push down filters
+		bool has_empty_grouping_sets = false;
+		for (auto &grp : aggr.grouping_sets) {
+			if (grp.empty()) {
+				has_empty_grouping_sets = true;
 			}
-			// erase the filter from here
-			filters.erase(filters.begin() + i);
-			i--;
 		}
+		if (has_empty_grouping_sets) {
+			continue;
+		}
+		// no aggregate! we can push this down
+		// rewrite any group bindings within the filter
+		f.filter = ReplaceGroupBindings(aggr, move(f.filter));
+		// add the filter to the child node
+		if (child_pushdown.AddFilter(move(f.filter)) == FilterResult::UNSATISFIABLE) {
+			// filter statically evaluates to false, strip tree
+			return make_unique<LogicalEmptyResult>(move(op));
+		}
+		// erase the filter from here
+		filters.erase(filters.begin() + i);
+		i--;
 	}
 	child_pushdown.GenerateFilters();
 
@@ -152623,6 +152863,19 @@ unique_ptr<LogicalOperator> TopN::Optimize(unique_ptr<LogicalOperator> op) {
 } // namespace duckdb
 
 
+namespace duckdb {
+
+BasePipelineEvent::BasePipelineEvent(shared_ptr<Pipeline> pipeline_p)
+    : Event(pipeline_p->executor), pipeline(move(pipeline_p)) {
+}
+
+BasePipelineEvent::BasePipelineEvent(Pipeline &pipeline_p)
+    : Event(pipeline_p.executor), pipeline(pipeline_p.shared_from_this()) {
+}
+
+} // namespace duckdb
+
+
 
 
 
@@ -152742,16 +152995,13 @@ public:
 
 
 
-
 namespace duckdb {
 
-class PipelineEvent : public Event {
+//! A PipelineEvent is responsible for scheduling a pipeline
+class PipelineEvent : public BasePipelineEvent {
 public:
 	PipelineEvent(shared_ptr<Pipeline> pipeline);
 
-	//! The pipeline that this event belongs to
-	shared_ptr<Pipeline> pipeline;
-
 public:
 	void Schedule() override;
 	void FinishEvent() override;
@@ -152879,17 +153129,13 @@ private:
 
 
 
-
 namespace duckdb {
 class Executor;
 
-class PipelineFinishEvent : public Event {
+class PipelineFinishEvent : public BasePipelineEvent {
 public:
 	PipelineFinishEvent(shared_ptr<Pipeline> pipeline);
 
-	//! The pipeline that this event belongs to
-	shared_ptr<Pipeline> pipeline;
-
 public:
 	void Schedule() override;
 	void FinishEvent() override;
@@ -152916,6 +153162,9 @@ Executor &Executor::Get(ClientContext &context) {
 
 void Executor::AddEvent(shared_ptr<Event> event) {
 	lock_guard<mutex> elock(executor_lock);
+	if (cancelled) {
+		return;
+	}
 	events.push_back(move(event));
 }
 
@@ -153219,6 +153468,7 @@ void Executor::CancelTasks() {
 	vector<weak_ptr<Pipeline>> weak_references;
 	{
 		lock_guard<mutex> elock(executor_lock);
+		cancelled = true;
 		weak_references.reserve(pipelines.size());
 		for (auto &pipeline : pipelines) {
 			weak_references.push_back(weak_ptr<Pipeline>(pipeline));
@@ -153295,10 +153545,10 @@ PendingExecutionResult Executor::ExecuteTask() {
 	lock_guard<mutex> elock(executor_lock);
 	pipelines.clear();
 	NextExecutor();
-	if (!exceptions.empty()) { // LCOV_EXCL_START
+	if (HasError()) { // LCOV_EXCL_START
 		// an exception has occurred executing one of the pipelines
 		execution_result = PendingExecutionResult::EXECUTION_ERROR;
-		ThrowExceptionInternal();
+		ThrowException();
 	} // LCOV_EXCL_STOP
 	execution_result = PendingExecutionResult::RESULT_READY;
 	return execution_result;
@@ -153307,6 +153557,7 @@ PendingExecutionResult Executor::ExecuteTask() {
 void Executor::Reset() {
 	lock_guard<mutex> elock(executor_lock);
 	physical_plan = nullptr;
+	cancelled = false;
 	owned_plan.reset();
 	root_executor.reset();
 	root_pipelines.clear();
@@ -153343,7 +153594,7 @@ vector<LogicalType> Executor::GetTypes() {
 }
 
 void Executor::PushError(PreservedError exception) {
-	lock_guard<mutex> elock(executor_lock);
+	lock_guard<mutex> elock(error_lock);
 	// interrupt execution of any other pipelines that belong to this executor
 	context.interrupted = true;
 	// push the exception onto the stack
@@ -153351,20 +153602,16 @@ void Executor::PushError(PreservedError exception) {
 }
 
 bool Executor::HasError() {
-	lock_guard<mutex> elock(executor_lock);
+	lock_guard<mutex> elock(error_lock);
 	return !exceptions.empty();
 }
 
 void Executor::ThrowException() {
-	lock_guard<mutex> elock(executor_lock);
-	ThrowExceptionInternal();
-}
-
-void Executor::ThrowExceptionInternal() { // LCOV_EXCL_START
+	lock_guard<mutex> elock(error_lock);
 	D_ASSERT(!exceptions.empty());
 	auto &entry = exceptions[0];
 	entry.Throw();
-} // LCOV_EXCL_STOP
+}
 
 void Executor::Flush(ThreadContext &tcontext) {
 	profiler->Flush(tcontext.profiler);
@@ -153629,6 +153876,9 @@ void Pipeline::Ready() {
 }
 
 void Pipeline::Finalize(Event &event) {
+	if (executor.HasError()) {
+		return;
+	}
 	D_ASSERT(ready);
 	try {
 		auto sink_state = sink->Finalize(*this, event, executor.context, *sink->sink_state);
@@ -153739,16 +153989,25 @@ void PipelineCompleteEvent::FinalizeFinish() {
 } // namespace duckdb
 
 
+
 namespace duckdb {
 
-PipelineEvent::PipelineEvent(shared_ptr<Pipeline> pipeline_p)
-    : Event(pipeline_p->executor), pipeline(move(pipeline_p)) {
+PipelineEvent::PipelineEvent(shared_ptr<Pipeline> pipeline_p) : BasePipelineEvent(move(pipeline_p)) {
 }
 
 void PipelineEvent::Schedule() {
 	auto event = shared_from_this();
-	pipeline->Schedule(event);
-	D_ASSERT(total_tasks > 0);
+	auto &executor = pipeline->executor;
+	try {
+		pipeline->Schedule(event);
+		D_ASSERT(total_tasks > 0);
+	} catch (Exception &ex) {
+		executor.PushError(PreservedError(ex));
+	} catch (std::exception &ex) {
+		executor.PushError(PreservedError(ex));
+	} catch (...) { // LCOV_EXCL_START
+		executor.PushError(PreservedError("Unknown exception in Finalize!"));
+	} // LCOV_EXCL_STOP
 }
 
 void PipelineEvent::FinishEvent() {
@@ -154131,8 +154390,7 @@ void PipelineExecutor::EndOperator(PhysicalOperator *op, DataChunk *chunk) {
 
 namespace duckdb {
 
-PipelineFinishEvent::PipelineFinishEvent(shared_ptr<Pipeline> pipeline_p)
-    : Event(pipeline_p->executor), pipeline(move(pipeline_p)) {
+PipelineFinishEvent::PipelineFinishEvent(shared_ptr<Pipeline> pipeline_p) : BasePipelineEvent(move(pipeline_p)) {
 }
 
 void PipelineFinishEvent::Schedule() {
@@ -167667,7 +167925,7 @@ string QueryNode::ResultModifiersToString() const {
 		} else if (modifier.type == ResultModifierType::LIMIT_PERCENT_MODIFIER) {
 			auto &limit_p_modifier = (LimitPercentModifier &)modifier;
 			if (limit_p_modifier.limit) {
-				result += " LIMIT " + limit_p_modifier.limit->ToString() + " %";
+				result += " LIMIT (" + limit_p_modifier.limit->ToString() + ") %";
 			}
 			if (limit_p_modifier.offset) {
 				result += " OFFSET " + limit_p_modifier.offset->ToString();
@@ -175139,6 +175397,8 @@ BindResult SelectBinder::BindAggregate(FunctionExpression &aggr, AggregateFuncti
 			// we didn't bind columns, try again in children
 			return BindResult(error);
 		}
+	} else if (depth > 0 && !aggregate_binder.HasBoundColumns()) {
+		return BindResult("Aggregate with only constant parameters has to be bound in the root subquery");
 	}
 	if (!filter_error.empty()) {
 		return BindResult(filter_error);
@@ -175146,8 +175406,9 @@ BindResult SelectBinder::BindAggregate(FunctionExpression &aggr, AggregateFuncti
 
 	if (aggr.filter) {
 		auto &child = (BoundExpression &)*aggr.filter;
-		bound_filter = move(child.expr);
+		bound_filter = BoundCastExpression::AddCastToType(move(child.expr), LogicalType::BOOLEAN);
 	}
+
 	// all children bound successfully
 	// extract the children and types
 	vector<LogicalType> types;
@@ -176300,7 +176561,7 @@ BindResult ExpressionBinder::BindMacro(FunctionExpression &function, ScalarMacro
 	string error =
 	    MacroFunction::ValidateArguments(*macro_func->function, macro_func->name, function, positionals, defaults);
 	if (!error.empty()) {
-		return BindResult(binder.FormatError(*expr->get(), error));
+		throw BinderException(binder.FormatError(*expr->get(), error));
 	}
 
 	// create a MacroBinding to bind this macro's parameters to its arguments
@@ -177323,10 +177584,13 @@ public:
 public:
 	unique_ptr<Expression> Bind(unique_ptr<ParsedExpression> expr);
 
-	idx_t MaxCount() {
+	idx_t MaxCount() const {
 		return max_count;
 	}
 
+	bool HasExtraList() const {
+		return extra_list;
+	}
 	unique_ptr<Expression> CreateExtraReference(unique_ptr<ParsedExpression> expr);
 
 private:
@@ -177368,6 +177632,9 @@ unique_ptr<Expression> Binder::BindDelimiter(ClientContext &context, OrderBinder
                                              Value &delimiter_value) {
 	auto new_binder = Binder::CreateBinder(context, this, true);
 	if (delimiter->HasSubquery()) {
+		if (!order_binder.HasExtraList()) {
+			throw BinderException("Subquery in LIMIT/OFFSET not supported in set operation");
+		}
 		return order_binder.CreateExtraReference(move(delimiter));
 	}
 	ExpressionBinder expr_binder(*new_binder, context);
@@ -177378,6 +177645,8 @@ unique_ptr<Expression> Binder::BindDelimiter(ClientContext &context, OrderBinder
 		delimiter_value = ExpressionExecutor::EvaluateScalar(*expr).CastAs(type);
 		return nullptr;
 	}
+	// move any correlated columns to this binder
+	MoveCorrelatedExpressions(*new_binder);
 	return expr;
 }
 
@@ -179981,11 +180250,13 @@ unique_ptr<BoundCreateTableInfo> Binder::BindCreateTableInfo(unique_ptr<CreateIn
 		BindDefaultValues(base.columns, result->bound_defaults);
 	}
 
+	idx_t regular_column_count = 0;
 	// bind collations to detect any unsupported collation errors
 	for (auto &column : base.columns) {
 		if (column.Generated()) {
 			continue;
 		}
+		regular_column_count++;
 		if (column.Type().id() == LogicalTypeId::VARCHAR) {
 			ExpressionBinder::TestCollation(context, StringType::GetCollation(column.Type()));
 		}
@@ -179997,6 +180268,9 @@ unique_ptr<BoundCreateTableInfo> Binder::BindCreateTableInfo(unique_ptr<CreateIn
 			result->dependencies.insert(type_dependency);
 		}
 	}
+	if (regular_column_count == 0) {
+		throw BinderException("Creating a table without physical (non-generated) columns is not supported");
+	}
 	properties.allow_stream_result = false;
 	return result;
 }
@@ -180424,6 +180698,13 @@ BoundStatement Binder::Bind(ExportStatement &stmt) {
 		info->schema = table->schema->name;
 		info->table = table->name;
 
+		// We can not export generated columns
+		for (auto &col : table->columns) {
+			if (!col.Generated()) {
+				info->select_list.push_back(col.GetName());
+			}
+		}
+
 		exported_data.table_name = info->table;
 		exported_data.schema_name = info->schema;
 		exported_data.file_path = info->file_path;
@@ -180669,7 +180950,10 @@ BoundStatement Binder::Bind(InsertStatement &stmt) {
 	}
 
 	// parse select statement and add to logical plan
-	auto root_select = Bind(*stmt.select_statement);
+	auto select_binder = Binder::CreateBinder(context, this);
+	auto root_select = select_binder->Bind(*stmt.select_statement);
+	MoveCorrelatedExpressions(*select_binder);
+
 	CheckInsertColumnCountMismatch(expected_columns, root_select.types.size(), !stmt.columns.empty(),
 	                               table->name.c_str());
 
@@ -181951,6 +182235,18 @@ string Binder::RetrieveUsingBinding(Binder &current_binder, UsingColumnSet *curr
 	return binding;
 }
 
+static vector<string> RemoveDuplicateUsingColumns(const vector<string> &using_columns) {
+	vector<string> result;
+	case_insensitive_set_t handled_columns;
+	for (auto &using_column : using_columns) {
+		if (handled_columns.find(using_column) == handled_columns.end()) {
+			handled_columns.insert(using_column);
+			result.push_back(using_column);
+		}
+	}
+	return result;
+}
+
 unique_ptr<BoundTableRef> Binder::Bind(JoinRef &ref) {
 	auto result = make_unique<BoundJoinRef>();
 	result->left_binder = Binder::CreateBinder(context, this);
@@ -182020,6 +182316,8 @@ unique_ptr<BoundTableRef> Binder::Bind(JoinRef &ref) {
 		D_ASSERT(!result->condition);
 		extra_using_columns = ref.using_columns;
 	}
+	extra_using_columns = RemoveDuplicateUsingColumns(extra_using_columns);
+
 	if (!extra_using_columns.empty()) {
 		vector<UsingColumnSet *> left_using_bindings;
 		vector<UsingColumnSet *> right_using_bindings;
@@ -182465,7 +182763,7 @@ unique_ptr<LogicalOperator> Binder::CreatePlan(BoundEmptyTableRef &ref) {
 namespace duckdb {
 
 unique_ptr<LogicalOperator> Binder::CreatePlan(BoundExpressionListRef &ref) {
-	auto root = make_unique_base<LogicalOperator, LogicalDummyScan>(0);
+	auto root = make_unique_base<LogicalOperator, LogicalDummyScan>(GenerateTableIndex());
 	// values list, first plan any subqueries in the list
 	for (auto &expr_list : ref.values) {
 		for (auto &expr : expr_list) {
@@ -185018,7 +185316,7 @@ BindResult ConstantBinder::BindExpression(unique_ptr<ParsedExpression> *expr_ptr
 	case ExpressionClass::COLUMN_REF:
 		return BindResult(clause + " cannot contain column names");
 	case ExpressionClass::SUBQUERY:
-		return BindResult(clause + " cannot contain subqueries");
+		throw BinderException(clause + " cannot contain subqueries");
 	case ExpressionClass::DEFAULT:
 		return BindResult(clause + " cannot contain DEFAULT clause");
 	case ExpressionClass::WINDOW:
@@ -185278,6 +185576,9 @@ unique_ptr<Expression> OrderBinder::CreateProjectionReference(ParsedExpression &
 }
 
 unique_ptr<Expression> OrderBinder::CreateExtraReference(unique_ptr<ParsedExpression> expr) {
+	if (!extra_list) {
+		throw InternalException("CreateExtraReference called without extra_list");
+	}
 	auto result = CreateProjectionReference(*expr, extra_list->size());
 	extra_list->push_back(move(expr));
 	return result;
@@ -189404,6 +189705,9 @@ unique_ptr<LogicalOperator> FlattenDependentJoins::PushDownDependentJoinInternal
 	case LogicalOperatorType::LOGICAL_ORDER_BY:
 		plan->children[0] = PushDownDependentJoin(move(plan->children[0]));
 		return plan;
+	case LogicalOperatorType::LOGICAL_RECURSIVE_CTE: {
+		throw ParserException("Recursive CTEs not supported in correlated subquery");
+	}
 	default:
 		throw InternalException("Logical operator type \"%s\" for dependent join", LogicalOperatorToString(plan->type));
 	}
@@ -191530,7 +191834,7 @@ void CheckpointManager::CreateCheckpoint() {
 	wal->Flush();
 
 	if (config.options.checkpoint_abort == CheckpointAbort::DEBUG_ABORT_BEFORE_HEADER) {
-		throw IOException("Checkpoint aborted before header write because of PRAGMA checkpoint_abort flag");
+		throw FatalException("Checkpoint aborted before header write because of PRAGMA checkpoint_abort flag");
 	}
 
 	// finally write the updated header
@@ -191539,7 +191843,7 @@ void CheckpointManager::CreateCheckpoint() {
 	block_manager.WriteHeader(header);
 
 	if (config.options.checkpoint_abort == CheckpointAbort::DEBUG_ABORT_BEFORE_TRUNCATE) {
-		throw IOException("Checkpoint aborted before truncate because of PRAGMA checkpoint_abort flag");
+		throw FatalException("Checkpoint aborted before truncate because of PRAGMA checkpoint_abort flag");
 	}
 
 	// truncate the WAL
@@ -197090,7 +197394,7 @@ DataTable::DataTable(ClientContext &context, DataTable &parent, idx_t removed_co
 }
 
 // Alter column to add new constraint
-DataTable::DataTable(ClientContext &context, DataTable &parent, unique_ptr<Constraint> constraint)
+DataTable::DataTable(ClientContext &context, DataTable &parent, unique_ptr<BoundConstraint> constraint)
     : info(parent.info), db(parent.db), total_rows(parent.total_rows.load()), row_groups(parent.row_groups),
       is_root(true) {
 
@@ -197265,7 +197569,7 @@ void DataTable::InitializeParallelScan(ClientContext &context, ParallelTableScan
 
 bool DataTable::NextParallelScan(ClientContext &context, ParallelTableScanState &state, TableScanState &scan_state,
                                  const vector<column_t> &column_ids) {
-	while (state.current_row_group) {
+	while (state.current_row_group && state.current_row_group->count > 0) {
 		idx_t vector_index;
 		idx_t max_row;
 		if (ClientConfig::GetConfig(context).verify_parallelism) {
@@ -197279,13 +197583,8 @@ bool DataTable::NextParallelScan(ClientContext &context, ParallelTableScanState
 			max_row = state.current_row_group->start + state.current_row_group->count;
 		}
 		max_row = MinValue<idx_t>(max_row, state.max_row);
-		bool need_to_scan;
-		if (state.current_row_group->count == 0) {
-			need_to_scan = false;
-		} else {
-			need_to_scan = InitializeScanInRowGroup(scan_state, column_ids, scan_state.table_filters,
-			                                        state.current_row_group, vector_index, max_row);
-		}
+		bool need_to_scan = InitializeScanInRowGroup(scan_state, column_ids, scan_state.table_filters,
+		                                             state.current_row_group, vector_index, max_row);
 		if (ClientConfig::GetConfig(context).verify_parallelism) {
 			state.vector_index++;
 			if (state.vector_index * STANDARD_VECTOR_SIZE >= state.current_row_group->count) {
@@ -197544,14 +197843,15 @@ static void VerifyDeleteForeignKeyConstraint(const BoundForeignKeyConstraint &bf
 	VerifyForeignKeyConstraint(bfk, context, chunk, false);
 }
 
-void DataTable::VerifyNewConstraint(ClientContext &context, DataTable &parent, const Constraint *constraint) {
+void DataTable::VerifyNewConstraint(ClientContext &context, DataTable &parent, const BoundConstraint *constraint) {
 	if (constraint->type != ConstraintType::NOT_NULL) {
 		throw NotImplementedException("FIXME: ALTER COLUMN with such constraint is not supported yet");
 	}
 	// scan the original table, check if there's any null value
-	auto &not_null_constraint = (NotNullConstraint &)*constraint;
+	auto &not_null_constraint = (BoundNotNullConstraint &)*constraint;
 	auto &transaction = Transaction::GetTransaction(context);
 	vector<LogicalType> scan_types;
+	D_ASSERT(not_null_constraint.index < parent.column_definitions.size());
 	scan_types.push_back(parent.column_definitions[not_null_constraint.index].Type());
 	DataChunk scan_chunk;
 	auto &allocator = Allocator::Get(context);
@@ -198308,6 +198608,9 @@ unique_ptr<BaseStatistics> DataTable::GetStatistics(ClientContext &context, colu
 		return nullptr;
 	}
 	lock_guard<mutex> stats_guard(stats_lock);
+	if (column_id >= column_stats.size()) {
+		throw InternalException("Call to GetStatistics is out of range");
+	}
 	return column_stats[column_id]->stats->Copy();
 }
 
@@ -199596,7 +199899,7 @@ void SingleFileBlockManager::WriteHeader(DatabaseHeader header) {
 
 	auto &config = DBConfig::GetConfig(db);
 	if (config.options.checkpoint_abort == CheckpointAbort::DEBUG_ABORT_AFTER_FREE_LIST_WRITE) {
-		throw IOException("Checkpoint aborted after free list write because of PRAGMA checkpoint_abort flag");
+		throw FatalException("Checkpoint aborted after free list write because of PRAGMA checkpoint_abort flag");
 	}
 
 	if (!use_direct_io) {
@@ -201122,6 +201425,7 @@ idx_t ChunkVectorInfo::Delete(Transaction &transaction, row_t rows[], idx_t coun
 		}
 		// after verifying that there are no conflicts we mark the tuple as deleted
 		deleted[rows[i]] = transaction.transaction_id;
+		rows[deleted_tuples] = rows[i];
 		deleted_tuples++;
 	}
 	return deleted_tuples;
@@ -201449,6 +201753,8 @@ public:
 	idx_t ScanCommitted(idx_t vector_index, ColumnScanState &state, Vector &result, bool allow_updates) override;
 	idx_t ScanCount(ColumnScanState &state, Vector &result, idx_t count) override;
 
+	void Skip(ColumnScanState &state, idx_t count = STANDARD_VECTOR_SIZE) override;
+
 	void InitializeAppend(ColumnAppendState &state) override;
 	void Append(BaseStatistics &stats, ColumnAppendState &state, Vector &vector, idx_t count) override;
 	void RevertAppend(row_t start_row) override;
@@ -204028,9 +204334,15 @@ void VersionDeleteState::Flush() {
 		return;
 	}
 	// delete in the current info
-	delete_count += current_info->Delete(transaction, rows, count);
-	// now push the delete into the undo buffer
-	transaction.PushDelete(table, current_info, rows, count, base_row + chunk_row);
+	// it is possible for delete statements to delete the same tuple multiple times when combined with a USING clause
+	// in the current_info->Delete, we check which tuples are actually deleted (excluding duplicate deletions)
+	// this is returned in the actual_delete_count
+	auto actual_delete_count = current_info->Delete(transaction, rows, count);
+	delete_count += actual_delete_count;
+	if (actual_delete_count > 0) {
+		// now push the delete into the undo buffer, but only if any deletes were actually performed
+		transaction.PushDelete(table, current_info, rows, actual_delete_count, base_row + chunk_row);
+	}
 	count = 0;
 }
 
@@ -204407,6 +204719,15 @@ idx_t StructColumnData::ScanCount(ColumnScanState &state, Vector &result, idx_t
 	return scan_count;
 }
 
+void StructColumnData::Skip(ColumnScanState &state, idx_t count) {
+	validity.Skip(state.child_states[0], count);
+
+	// skip inside the sub-columns
+	for (idx_t child_idx = 0; child_idx < sub_columns.size(); child_idx++) {
+		sub_columns[child_idx]->Skip(state.child_states[child_idx + 1], count);
+	}
+}
+
 void StructColumnData::InitializeAppend(ColumnAppendState &state) {
 	ColumnAppendState validity_append;
 	validity.InitializeAppend(validity_append);
@@ -206866,6 +207187,7 @@ void CleanupState::CleanupUpdate(UpdateInfo *info) {
 
 void CleanupState::CleanupDelete(DeleteInfo *info) {
 	auto version_table = info->table;
+	D_ASSERT(version_table->info->cardinality >= info->count);
 	version_table->info->cardinality -= info->count;
 	if (version_table->info->indexes.Empty()) {
 		// this table has no indexes: no cleanup to be done
@@ -260291,49 +260613,84 @@ static void AssignInvalidUTF8Reason(UnicodeInvalidReason *invalid_reason, size_t
 	}
 }
 
-UnicodeType Utf8Proc::Analyze(const char *s, size_t len, UnicodeInvalidReason *invalid_reason, size_t *invalid_pos) {
-	UnicodeType type = UnicodeType::ASCII;
-	char c;
-	for (size_t i = 0; i < len; i++) {
-		c = s[i];
-		if (c == '\0') {
-			AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::NULL_BYTE);
-			return UnicodeType::INVALID;
-		}
-		// 1 Byte / ASCII
-		if ((c & 0x80) == 0) {
-			continue;
-		}
-		type = UnicodeType::UNICODE;
-		if ((s[++i] & 0xC0) != 0x80) {
-			AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::BYTE_MISMATCH);
-			return UnicodeType::INVALID;
-		}
-		if ((c & 0xE0) == 0xC0) {
-			continue;
-		}
-		if ((s[++i] & 0xC0) != 0x80) {
-			AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::BYTE_MISMATCH);
-			return UnicodeType::INVALID;
-		}
-		if ((c & 0xF0) == 0xE0) {
-			continue;
-		}
-		if ((s[++i] & 0xC0) != 0x80) {
+template <const int nextra_bytes, const int mask>
+static inline UnicodeType
+UTF8ExtraByteLoop(const int first_pos_seq, int utf8char, size_t& i,
+				  const char *s, const size_t len, UnicodeInvalidReason *invalid_reason, size_t *invalid_pos) {
+	if ((len - i) < (nextra_bytes + 1)) {
+		/* incomplete byte sequence */
+		AssignInvalidUTF8Reason(invalid_reason, invalid_pos, first_pos_seq, UnicodeInvalidReason::BYTE_MISMATCH);
+		return UnicodeType::INVALID;
+	}
+	for (size_t j = 0 ; j < nextra_bytes; j++) {
+		int c = (int) s[++i];
+		/* now validate the extra bytes */
+		if ((c & 0xC0) != 0x80) {
+			/* extra byte is not in the format 10xxxxxx */
 			AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::BYTE_MISMATCH);
 			return UnicodeType::INVALID;
 		}
-		if ((c & 0xF8) == 0xF0) {
-			continue;
-		}
-		AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::BYTE_MISMATCH);
+		utf8char = (utf8char << 6) | (c & 0x3F);
+	}
+	if ((utf8char & mask) == 0) {
+		/* invalid UTF-8 codepoint, not shortest possible */
+		AssignInvalidUTF8Reason(invalid_reason, invalid_pos, first_pos_seq, UnicodeInvalidReason::INVALID_UNICODE);
+		return UnicodeType::INVALID;
+	}
+	if (utf8char > 0x10FFFF) {
+		/* value not representable by Unicode */
+		AssignInvalidUTF8Reason(invalid_reason, invalid_pos, first_pos_seq, UnicodeInvalidReason::INVALID_UNICODE);
 		return UnicodeType::INVALID;
 	}
+	if ((utf8char & 0x1FFF800) == 0xD800) {
+		/* Unicode characters from U+D800 to U+DFFF are surrogate characters used by UTF-16 which are invalid in UTF-8 */
+		AssignInvalidUTF8Reason(invalid_reason, invalid_pos, first_pos_seq, UnicodeInvalidReason::INVALID_UNICODE);
+		return UnicodeType::INVALID;
+	}
+	return UnicodeType::UNICODE;
+}
 
+UnicodeType Utf8Proc::Analyze(const char *s, size_t len, UnicodeInvalidReason *invalid_reason, size_t *invalid_pos) {
+	UnicodeType type = UnicodeType::ASCII;
+
+	for (size_t i = 0; i < len; i++) {
+		int c = (int) s[i];
+
+		if ((c & 0x80) == 0) {
+			/* 1 byte sequence */
+			if (c == '\0') {
+				/* NULL byte not allowed */
+				AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::NULL_BYTE);
+				return UnicodeType::INVALID;
+			}
+		} else {
+			int first_pos_seq = i;
+
+			if ((c & 0xE0) == 0xC0) {
+				/* 2 byte sequence */
+				int utf8char = c & 0x1F;
+				type = UTF8ExtraByteLoop<1, 0x000780>(first_pos_seq, utf8char, i, s, len, invalid_reason, invalid_pos);
+			} else if ((c & 0xF0) == 0xE0) {
+				/* 3 byte sequence */
+				int utf8char = c & 0x0F;
+				type = UTF8ExtraByteLoop<2, 0x00F800>(first_pos_seq, utf8char, i, s, len, invalid_reason, invalid_pos);
+			} else if ((c & 0xF8) == 0xF0) {
+				/* 4 byte sequence */
+				int utf8char = c & 0x07;
+				type = UTF8ExtraByteLoop<3, 0x1F0000>(first_pos_seq, utf8char, i, s, len, invalid_reason, invalid_pos);
+			} else {
+				/* invalid UTF-8 start byte */
+				AssignInvalidUTF8Reason(invalid_reason, invalid_pos, i, UnicodeInvalidReason::BYTE_MISMATCH);
+				return UnicodeType::INVALID;
+			}
+			if (type == UnicodeType::INVALID) {
+				return type;
+			}
+		}
+	}
 	return type;
 }
 
-
 char* Utf8Proc::Normalize(const char *s, size_t len) {
 	assert(s);
 	assert(Utf8Proc::Analyze(s, len) != UnicodeType::INVALID);