dont use insert table totally (#10088)

Author: ivanmorozov333

ydb/core/formats/arrow/reader/merger.cpp

@@ -154,13 +154,16 @@ void TMergePartialStream::DrainCurrentPosition(TRecordBatchBuilder* builder, std
     Y_ABORT_UNLESS(SortHeap.Size());
     Y_ABORT_UNLESS(!SortHeap.Current().IsControlPoint());
     if (!SortHeap.Current().IsDeleted()) {
+//        AFL_ERROR(NKikimrServices::TX_COLUMNSHARD)("key_add", SortHeap.Current().GetKeyColumns().DebugJson().GetStringRobust());
         if (builder) {
             builder->AddRecord(SortHeap.Current().GetKeyColumns());
         }
         if (resultScanData && resultPosition) {
             *resultScanData = SortHeap.Current().GetKeyColumns().GetSorting();
             *resultPosition = SortHeap.Current().GetKeyColumns().GetPosition();
         }
+    } else {
+//        AFL_ERROR(NKikimrServices::TX_COLUMNSHARD)("key_skip", SortHeap.Current().GetKeyColumns().DebugJson().GetStringRobust());
     }
     CheckSequenceInDebug(SortHeap.Current().GetKeyColumns());
     const ui64 startPosition = SortHeap.Current().GetKeyColumns().GetPosition();
@@ -169,6 +172,7 @@ void TMergePartialStream::DrainCurrentPosition(TRecordBatchBuilder* builder, std
     bool isFirst = true;
     while (SortHeap.Size() && (isFirst || SortHeap.Current().GetKeyColumns().Compare(*startSorting, startPosition) == std::partial_ordering::equivalent)) {
         if (!isFirst) {
+//            AFL_ERROR(NKikimrServices::TX_COLUMNSHARD)("key_skip1", SortHeap.Current().GetKeyColumns().DebugJson().GetStringRobust());
             auto& anotherIterator = SortHeap.Current();
             if (PossibleSameVersionFlag) {
                 AFL_VERIFY(anotherIterator.GetVersionColumns().Compare(*startVersion, startPosition) != std::partial_ordering::greater)

ydb/core/formats/arrow/reader/position.h

@@ -451,6 +451,8 @@ class TIntervalPositions {
 private:
     std::vector<TIntervalPosition> Positions;
 public:
+    using const_iterator = std::vector<TIntervalPosition>::const_iterator;
+
     bool IsEmpty() const {
         return Positions.empty();
     }
@@ -459,6 +461,16 @@ class TIntervalPositions {
         return Positions.begin();
     }
 
+    TString DebugString() const {
+        TStringBuilder sb;
+        sb << "[";
+        for (auto&& p : Positions) {
+            sb << p.DebugJson().GetStringRobust() << ";";
+        }
+        sb << "]";
+        return sb;
+    }
+
     std::vector<TIntervalPosition>::const_iterator end() const {
         return Positions.end();
     }
@@ -662,6 +674,35 @@ class TRWSortableBatchPosition: public TSortableBatchPosition, public TMoveOnly
         return SplitByBorders(batch, columnNames, it);
     }
 
+    class TIntervalPointsIterator {
+    private:
+        typename TIntervalPositions::const_iterator Current;
+        typename TIntervalPositions::const_iterator End;
+
+    public:
+        TIntervalPointsIterator(const TIntervalPositions& container)
+            : Current(container.begin())
+            , End(container.end()) {
+        }
+
+        bool IsValid() const {
+            return Current != End;
+        }
+
+        void Next() {
+            ++Current;
+        }
+
+        const auto& CurrentPosition() const {
+            return Current->GetPosition();
+        }
+    };
+
+    static std::vector<std::shared_ptr<arrow::RecordBatch>> SplitByBordersInIntervalPositions(
+        const std::shared_ptr<arrow::RecordBatch>& batch, const std::vector<std::string>& columnNames, const TIntervalPositions& container) {
+        TIntervalPointsIterator it(container);
+        return SplitByBorders(batch, columnNames, it);
+    }
 };
 
 }

ydb/core/formats/arrow/serializer/native.cpp

@@ -99,7 +99,7 @@ TString TNativeSerializer::DoSerializePayload(const std::shared_ptr<arrow::Recor
     // Write prepared payload into the resultant string. No extra allocation will be made.
     TStatusValidator::Validate(arrow::ipc::WriteIpcPayload(payload, Options, &out, &metadata_length));
     Y_ABORT_UNLESS(out.GetPosition() == str.size());
-    Y_DEBUG_ABORT_UNLESS(Deserialize(str, batch->schema()).ok());
+    AFL_VERIFY_DEBUG(Deserialize(str, batch->schema()).ok());
     AFL_DEBUG(NKikimrServices::ARROW_HELPER)("event", "serialize")("size", str.size())("columns", batch->schema()->num_fields());
     return str;
 }

ydb/core/formats/arrow/serializer/native.h

@@ -62,6 +62,18 @@ class TNativeSerializer: public ISerializer {
     virtual void DoSerializeToProto(NKikimrSchemeOp::TOlapColumn::TSerializer& proto) const override;
 
 public:
+    static std::shared_ptr<ISerializer> GetUncompressed() {
+        static std::shared_ptr<ISerializer> result =
+            std::make_shared<NArrow::NSerialization::TNativeSerializer>(arrow::Compression::UNCOMPRESSED);
+        return result;
+    }
+
+    static std::shared_ptr<ISerializer> GetFast() {
+        static std::shared_ptr<ISerializer> result =
+            std::make_shared<NArrow::NSerialization::TNativeSerializer>(arrow::Compression::LZ4_FRAME);
+        return result;
+    }
+
     virtual TString GetClassName() const override {
         return GetClassNameStatic();
     }

ydb/core/kqp/ut/common/kqp_ut_common.h

@@ -98,6 +98,8 @@ struct TKikimrSettings: public TTestFeatureFlagsHolder<TKikimrSettings> {
         exchangerSettings->SetMaxDelayMs(10);
         AppConfig.MutableColumnShardConfig()->SetDisabledOnSchemeShard(false);
         FeatureFlags.SetEnableSparsedColumns(true);
+        FeatureFlags.SetEnableImmediateWritingOnBulkUpsert(true);
+        FeatureFlags.SetEnableWritePortionsOnInsert(true);
         FeatureFlags.SetEnableParameterizedDecimal(true);
         FeatureFlags.SetEnableTopicAutopartitioningForCDC(true);
     }

ydb/core/kqp/ut/olap/aggregations_ut.cpp

@@ -95,7 +95,7 @@ Y_UNIT_TEST_SUITE(KqpOlapAggregations) {
             WriteTestData(kikimr, "/Root/olapStore/olapTable", 20000, 2000000, 7000);
             WriteTestData(kikimr, "/Root/olapStore/olapTable", 30000, 1000000, 11000);
         }
-        while (csController->GetInsertFinishedCounter().Val() == 0) {
+        while (csController->GetCompactionFinishedCounter().Val() == 0) {
             Cout << "Wait indexation..." << Endl;
             Sleep(TDuration::Seconds(2));
         }
@@ -374,7 +374,7 @@ Y_UNIT_TEST_SUITE(KqpOlapAggregations) {
             .AddExpectedPlanOptions("KqpOlapFilter")
 #if SSA_RUNTIME_VERSION >= 2U
             .AddExpectedPlanOptions("TKqpOlapAgg")
-            .MutableLimitChecker().SetExpectedResultCount(1)
+            .MutableLimitChecker().SetExpectedResultCount(2)
 #else
             .AddExpectedPlanOptions("Condense")
 #endif
@@ -417,7 +417,7 @@ Y_UNIT_TEST_SUITE(KqpOlapAggregations) {
             .AddExpectedPlanOptions("KqpOlapFilter")
 #if SSA_RUNTIME_VERSION >= 2U
             .AddExpectedPlanOptions("TKqpOlapAgg")
-            .MutableLimitChecker().SetExpectedResultCount(1)
+            .MutableLimitChecker().SetExpectedResultCount(2)
 #else
             .AddExpectedPlanOptions("CombineCore")
             .AddExpectedPlanOptions("KqpOlapFilter")