Bool support for columnshard (#26865)

Co-authored-by: Matveev Sergei <[email protected]>

Author: xyliganSereja

ydb/core/formats/arrow/arrow_batch_builder.cpp

@@ -29,6 +29,14 @@ arrow::Status AppendCell(arrow::NumericBuilder<T>& builder, const TCell& cell) {
     return builder.Append(cell.AsValue<ui8>());
 }
 
+[[maybe_unused]] arrow::Status AppendCell(arrow::UInt8Builder& builder, const TCell& cell) {
+    if (cell.IsNull()) {
+        return builder.AppendNull();
+    }
+
+    return builder.Append(cell.AsValue<ui8>());
+}
+
 [[maybe_unused]] arrow::Status AppendCell(arrow::BinaryBuilder& builder, const TCell& cell) {
     if (cell.IsNull()) {
         return builder.AppendNull();

ydb/core/formats/arrow/arrow_helpers.cpp

@@ -35,6 +35,11 @@ std::shared_ptr<arrow::DataType> CreateEmptyArrowImpl() {
     return std::make_shared<TType>();
 }
 
+template <>
+std::shared_ptr<arrow::DataType> CreateEmptyArrowImpl<arrow::BooleanType>() {
+    return arrow::uint8();
+}
+
 template <>
 std::shared_ptr<arrow::DataType> CreateEmptyArrowImpl<arrow::Decimal128Type>() {
     return arrow::fixed_size_binary(NScheme::FSB_SIZE);
@@ -62,6 +67,7 @@ arrow::Result<std::shared_ptr<arrow::DataType>> GetArrowType(NScheme::TTypeInfo
         result = CreateEmptyArrowImpl<TType>();
         return true;
     });
+
     if (success) {
         return result;
     }
@@ -72,6 +78,8 @@ arrow::Result<std::shared_ptr<arrow::DataType>> GetArrowType(NScheme::TTypeInfo
 arrow::Result<std::shared_ptr<arrow::DataType>> GetCSVArrowType(NScheme::TTypeInfo typeId) {
     std::shared_ptr<arrow::DataType> result;
     switch (typeId.GetTypeId()) {
+        case NScheme::NTypeIds::Bool:
+            return std::make_shared<arrow::UInt8Type>();
         case NScheme::NTypeIds::Datetime:
         case NScheme::NTypeIds::Datetime64:
             return std::make_shared<arrow::TimestampType>(arrow::TimeUnit::SECOND);

ydb/core/formats/arrow/switch/switch_type.h

@@ -25,7 +25,7 @@ template <typename TFunc>
 [[nodiscard]] bool SwitchYqlTypeToArrowType(const NScheme::TTypeInfo& typeInfo, TFunc&& callback) {
     switch (typeInfo.GetTypeId()) {
         case NScheme::NTypeIds::Bool:
-            return callback(TTypeWrapper<arrow::BooleanType>());
+            return callback(TTypeWrapper<arrow::UInt8Type>());
         case NScheme::NTypeIds::Int8:
             return callback(TTypeWrapper<arrow::Int8Type>());
         case NScheme::NTypeIds::Uint8:

ydb/core/formats/arrow/ut/ut_arrow.cpp

@@ -117,7 +117,7 @@ struct TDataRow {
 
     static std::shared_ptr<arrow::Schema> MakeArrowSchema() {
         std::vector<std::shared_ptr<arrow::Field>> fields = {
-            arrow::field("bool", arrow::boolean()),
+            arrow::field("bool", arrow::uint8()),
             arrow::field("i8", arrow::int8()),
             arrow::field("i16", arrow::int16()),
             arrow::field("i32", arrow::int32()),

ydb/core/grpc_services/rpc_load_rows.cpp

@@ -31,9 +31,6 @@ namespace {
 // TODO: no mapping for DATE, DATETIME, TZ_*, YSON, JSON, UUID, JSON_DOCUMENT, DYNUMBER
 bool ConvertArrowToYdbPrimitive(const arrow::DataType& type, Ydb::Type& toType, const NScheme::TTypeInfo* tableColumnType = nullptr) {
     switch (type.id()) {
-        case arrow::Type::BOOL:
-            toType.set_type_id(Ydb::Type::BOOL);
-            return true;
         case arrow::Type::UINT8:
             toType.set_type_id(Ydb::Type::UINT8);
             return true;
@@ -86,6 +83,7 @@ bool ConvertArrowToYdbPrimitive(const arrow::DataType& type, Ydb::Type& toType,
 
             break;
         }
+        case arrow::Type::BOOL:
         case arrow::Type::NA:
         case arrow::Type::HALF_FLOAT:
         case arrow::Type::DATE32:

ydb/core/io_formats/arrow/scheme/scheme.cpp

@@ -24,6 +24,8 @@ arrow::Result<TArrowCSV> TArrowCSVScheme::Create(const TVector<std::pair<TString
         if (type.GetTypeId() == NScheme::NTypeIds::Decimal) {
             columnInfo.Precision = type.GetDecimalType().GetPrecision();
             columnInfo.Scale = type.GetDecimalType().GetScale();
+        } else if (type.GetTypeId() == NScheme::NTypeIds::Bool) {
+            columnInfo.IsBool = true;
         }
 
         convertedColumns.emplace_back(columnInfo);

ydb/core/kqp/ut/common/columnshard.cpp

@@ -344,7 +344,7 @@ namespace NKqp {
     std::shared_ptr<arrow::Field> TTestHelper::TColumnTableBase::BuildField(const TString name, const NScheme::TTypeInfo& typeInfo, bool nullable) const {
         switch (typeInfo.GetTypeId()) {
         case NScheme::NTypeIds::Bool:
-            return arrow::field(name, arrow::boolean(), nullable);
+            return arrow::field(name, arrow::uint8(), nullable);
         case NScheme::NTypeIds::Int8:
             return arrow::field(name, arrow::int8(), nullable);
         case NScheme::NTypeIds::Int16:

ydb/core/kqp/ut/olap/bool_test_enums.h

@@ -9,8 +9,8 @@ enum EQueryMode {
 };
 
 enum ETableKind {
-    COLUMN_SHARD,
-    DATA_SHARD
+    COLUMNSHARD,
+    DATASHARD
 };
 
 enum ELoadKind {