#pragma once #include #include #include #include #include #include #include #include #include namespace DB { class HashedDictionary final : public IDictionary { public: HashedDictionary(const std::string & name, const DictionaryStructure & dict_struct, DictionarySourcePtr source_ptr, const DictionaryLifetime dict_lifetime, bool require_nonempty); HashedDictionary(const HashedDictionary & other); std::exception_ptr getCreationException() const override { return creation_exception; } std::string getName() const override { return name; } std::string getTypeName() const override { return "Hashed"; } size_t getBytesAllocated() const override { return bytes_allocated; } size_t getQueryCount() const override { return query_count.load(std::memory_order_relaxed); } double getHitRate() const override { return 1.0; } size_t getElementCount() const override { return element_count; } double getLoadFactor() const override { return static_cast(element_count) / bucket_count; } bool isCached() const override { return false; } DictionaryPtr clone() const override { return std::make_unique(*this); } const IDictionarySource * getSource() const override { return source_ptr.get(); } const DictionaryLifetime & getLifetime() const override { return dict_lifetime; } const DictionaryStructure & getStructure() const override { return dict_struct; } std::chrono::time_point getCreationTime() const override { return creation_time; } bool isInjective(const std::string & attribute_name) const override { return dict_struct.attributes[&getAttribute(attribute_name) - attributes.data()].injective; } bool hasHierarchy() const override { return hierarchical_attribute; } void toParent(const PaddedPODArray & ids, PaddedPODArray & out) const override; #define DECLARE(TYPE)\ void get##TYPE(const std::string & attribute_name, const PaddedPODArray & ids, PaddedPODArray & out) const; DECLARE(UInt8) DECLARE(UInt16) DECLARE(UInt32) DECLARE(UInt64) DECLARE(Int8) DECLARE(Int16) DECLARE(Int32) DECLARE(Int64) DECLARE(Float32) DECLARE(Float64) #undef DECLARE void getString(const std::string & attribute_name, const PaddedPODArray & ids, ColumnString * out) const; #define DECLARE(TYPE)\ void get##TYPE(\ const std::string & attribute_name, const PaddedPODArray & ids, const PaddedPODArray & def,\ PaddedPODArray & out) const; DECLARE(UInt8) DECLARE(UInt16) DECLARE(UInt32) DECLARE(UInt64) DECLARE(Int8) DECLARE(Int16) DECLARE(Int32) DECLARE(Int64) DECLARE(Float32) DECLARE(Float64) #undef DECLARE void getString( const std::string & attribute_name, const PaddedPODArray & ids, const ColumnString * const def, ColumnString * const out) const; #define DECLARE(TYPE)\ void get##TYPE(\ const std::string & attribute_name, const PaddedPODArray & ids, const TYPE & def, PaddedPODArray & out) const; DECLARE(UInt8) DECLARE(UInt16) DECLARE(UInt32) DECLARE(UInt64) DECLARE(Int8) DECLARE(Int16) DECLARE(Int32) DECLARE(Int64) DECLARE(Float32) DECLARE(Float64) #undef DECLARE void getString( const std::string & attribute_name, const PaddedPODArray & ids, const String & def, ColumnString * const out) const; void has(const PaddedPODArray & ids, PaddedPODArray & out) const override; void isInVectorVector(const PaddedPODArray & child_ids, const PaddedPODArray & ancestor_ids, PaddedPODArray & out) const override; void isInVectorConstant(const PaddedPODArray & child_ids, const Key ancestor_id, PaddedPODArray & out) const override; void isInConstantVector(const Key child_id, const PaddedPODArray & ancestor_ids, PaddedPODArray & out) const override; BlockInputStreamPtr getBlockInputStream(const Names & column_names, size_t max_block_size) const override; private: template using CollectionType = HashMap; template using CollectionPtrType = std::unique_ptr>; struct Attribute final { AttributeUnderlyingType type; std::tuple< UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, Float32, Float64, String> null_values; std::tuple< CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType, CollectionPtrType> maps; std::unique_ptr string_arena; }; void createAttributes(); void loadData(); template void addAttributeSize(const Attribute & attribute); void calculateBytesAllocated(); template void createAttributeImpl(Attribute & attribute, const Field & null_value); Attribute createAttributeWithType(const AttributeUnderlyingType type, const Field & null_value); template void getItemsNumber( const Attribute & attribute, const PaddedPODArray & ids, ValueSetter && set_value, DefaultGetter && get_default) const; template void getItemsImpl( const Attribute & attribute, const PaddedPODArray & ids, ValueSetter && set_value, DefaultGetter && get_default) const; template void setAttributeValueImpl(Attribute & attribute, const Key id, const T value); void setAttributeValue(Attribute & attribute, const Key id, const Field & value); const Attribute & getAttribute(const std::string & attribute_name) const; template void has(const Attribute & attribute, const PaddedPODArray & ids, PaddedPODArray & out) const; template PaddedPODArray getIds(const Attribute & attribute) const; PaddedPODArray getIds() const; template void isInImpl( const ChildType & child_ids, const AncestorType & ancestor_ids, PaddedPODArray & out) const; const std::string name; const DictionaryStructure dict_struct; const DictionarySourcePtr source_ptr; const DictionaryLifetime dict_lifetime; const bool require_nonempty; std::map attribute_index_by_name; std::vector attributes; const Attribute * hierarchical_attribute = nullptr; size_t bytes_allocated = 0; size_t element_count = 0; size_t bucket_count = 0; mutable std::atomic query_count{0}; std::chrono::time_point creation_time; std::exception_ptr creation_exception; }; }