Class ArrowResultSetConverter

class ArrowResultSetConverter

Public Functions

ArrowResultSetConverter(const std::shared_ptr<ResultSet> &results, const std::shared_ptr<Data_Namespace::DataMgr> data_mgr, const ExecutorDeviceType device_type, const int32_t device_id, const std::vector<std::string> &col_names, const int32_t first_n, const ArrowTransport transport_method)

ArrowResultSetConverter(const std::shared_ptr<ResultSet> &results, const std::shared_ptr<Data_Namespace::DataMgr> data_mgr, const ExecutorDeviceType device_type, const int32_t device_id, const std::vector<std::string> &col_names, const int32_t first_n, const ArrowTransport transport_method, const size_t min_result_size_for_bulk_dictionary_fetch, const double max_dictionary_to_result_size_ratio_for_bulk_dictionary_fetch)

ArrowResult getArrowResult() const

Serialize an Arrow result to IPC memory. Users are responsible for freeing all CPU IPC buffers using deallocateArrowResultBuffer. GPU buffers will become owned by the caller upon deserialization, and will be automatically freed when they go out of scope.

ArrowResultSetConverter(const std::shared_ptr<ResultSet> &results, const std::vector<std::string> &col_names, const int32_t first_n)

ArrowResultSetConverter(const std::shared_ptr<ResultSet> &results, const std::vector<std::string> &col_names, const int32_t first_n, const size_t min_result_size_for_bulk_dictionary_fetch, const double max_dictionary_to_result_size_ratio_for_bulk_dictionary_fetch)

std::shared_ptr<arrow::RecordBatch> convertToArrow() const

Public Static Attributes

constexpr size_t default_min_result_size_for_bulk_dictionary_fetch = {10000UL}

constexpr double default_max_dictionary_to_result_size_ratio_for_bulk_dictionary_fetch = {0.1}

Private Functions

std::shared_ptr<arrow::RecordBatch> getArrowBatch(const std::shared_ptr<arrow::Schema> &schema) const

std::shared_ptr<arrow::Field> makeField(const std::string name, const SQLTypeInfo &target_type) const

ArrowResultSetConverter::SerializedArrowOutput getSerializedArrowOutput(arrow::ipc::DictionaryFieldMapper *mapper) const

void initializeColumnBuilder(ColumnBuilder &column_builder, const SQLTypeInfo &col_type, const size_t result_col_idx, const std::shared_ptr<arrow::Field> &field) const

void append(ColumnBuilder &column_builder, const ValueArray &values, const std::shared_ptr<std::vector<bool>> &is_valid) const

std::shared_ptr<arrow::Array> finishColumnBuilder(ColumnBuilder &column_builder) const

Private Members

std::shared_ptr<ResultSet> results_

std::shared_ptr<Data_Namespace::DataMgr> data_mgr_ = nullptr

ExecutorDeviceType device_type_ = ExecutorDeviceType::GPU

int32_t device_id_ = 0

std::vector<std::string> col_names_

int32_t top_n_

ArrowTransport transport_method_

const size_t min_result_size_for_bulk_dictionary_fetch_

const double max_dictionary_to_result_size_ratio_for_bulk_dictionary_fetch_

Friends

friend ArrowResultSetConverter::ArrowResultSet

struct ColumnBuilder

Public Types

using StrId = int32_t

using ArrowStrId = int32_t

Public Members

std::shared_ptr<arrow::Field> field

std::unique_ptr<arrow::ArrayBuilder> builder

SQLTypeInfo col_type

SQLTypes physical_type

ArrowStringRemapMode string_remap_mode = {ArrowStringRemapMode::INVALID}

std::unordered_map<StrId, ArrowStrId> string_remapping

struct SerializedArrowOutput

Public Members

std::shared_ptr<arrow::Buffer> schema

std::shared_ptr<arrow::Buffer> records