fix segfault in case normalization different versions (#7239)

3 files changed, 13 insertions, 2 deletions

diff --git a/ydb/core/formats/arrow/common/container.cpp b/ydb/core/formats/arrow/common/container.cpp
index 7e8209dfad5..e910a6837a1 100644
--- a/ydb/core/formats/arrow/common/container.cpp
+++ b/ydb/core/formats/arrow/common/container.cpp
@@ -123,6 +123,11 @@ TGeneralContainer::TGeneralContainer(const std::shared_ptr<arrow::RecordBatch>&
     Initialize();
 }
 
+TGeneralContainer::TGeneralContainer(const ui32 recordsCount)
+    : RecordsCount(recordsCount)
+    , Schema(std::make_shared<NModifier::TSchema>()) {
+}
+
 std::shared_ptr<NKikimr::NArrow::NAccessor::IChunkedArray> TGeneralContainer::GetAccessorByNameVerified(const std::string& fieldId) const {
     auto result = GetAccessorByNameOptional(fieldId);
     AFL_VERIFY(result)("event", "cannot_find_accessor_in_general_container")("field_id", fieldId)("schema", Schema->ToString());
diff --git a/ydb/core/formats/arrow/common/container.h b/ydb/core/formats/arrow/common/container.h
index 90d686ce4e4..823f04821d5 100644
--- a/ydb/core/formats/arrow/common/container.h
+++ b/ydb/core/formats/arrow/common/container.h
@@ -29,6 +29,13 @@ private:
     std::vector<std::shared_ptr<NAccessor::IChunkedArray>> Columns;
     void Initialize();
 public:
+    TGeneralContainer(const ui32 recordsCount);
+
+    ui32 GetRecordsCountVerified() const {
+        AFL_VERIFY(RecordsCount);
+        return *RecordsCount;
+    }
+
     TString DebugString() const;
 
     [[nodiscard]] TConclusionStatus SyncSchemaTo(const std::shared_ptr<arrow::Schema>& schema,
@@ -67,7 +74,6 @@ public:
 
     [[nodiscard]] TConclusionStatus AddField(const std::shared_ptr<arrow::Field>& f, const std::shared_ptr<arrow::ChunkedArray>& data);
 
-    TGeneralContainer() = default;
     TGeneralContainer(const std::shared_ptr<arrow::Table>& table);
     TGeneralContainer(const std::shared_ptr<arrow::RecordBatch>& table);
     TGeneralContainer(const std::shared_ptr<arrow::Schema>& schema, std::vector<std::shared_ptr<NAccessor::IChunkedArray>>&& columns);
diff --git a/ydb/core/tx/columnshard/engines/scheme/versions/abstract_scheme.cpp b/ydb/core/tx/columnshard/engines/scheme/versions/abstract_scheme.cpp
index c6efc7a5567..3eeca97bde2 100644
--- a/ydb/core/tx/columnshard/engines/scheme/versions/abstract_scheme.cpp
+++ b/ydb/core/tx/columnshard/engines/scheme/versions/abstract_scheme.cpp
@@ -35,7 +35,7 @@ TConclusion<std::shared_ptr<NArrow::TGeneralContainer>> ISnapshotSchema::Normali
     AFL_VERIFY(dataSchema.GetSnapshot() < GetSnapshot());
     const std::shared_ptr<arrow::Schema>& resultArrowSchema = GetSchema();
 
-    std::shared_ptr<NArrow::TGeneralContainer> result = std::make_shared<NArrow::TGeneralContainer>();
+    std::shared_ptr<NArrow::TGeneralContainer> result = std::make_shared<NArrow::TGeneralContainer>(batch->GetRecordsCountVerified());
     for (size_t i = 0; i < resultArrowSchema->fields().size(); ++i) {
         auto& resultField = resultArrowSchema->fields()[i];
         auto columnId = GetIndexInfo().GetColumnId(resultField->name());