Gix visitor

Co-authored-by: Tyler Neely <tyler.neely@memgraph.io>

src/common/types.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -16,4 +16,6 @@
 namespace memgraph::common {
 enum class SchemaType : uint8_t { BOOL, INT, STRING, DATE, LOCALTIME, LOCALDATETIME, DURATION };
 
+enum class SchemaConfigParams : uint8_t { REPLICATION_FACTOR, SPLIT_THRESHOLD };
+
 }  // namespace memgraph::common

src/coordinator/hybrid_logical_clock.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -23,17 +23,17 @@ namespace memgraph::coordinator {
 using Time = memgraph::io::Time;
 
 /// Hybrid-logical clock
-struct Hlc {
-  uint64_t logical_id = 0;
+struct Hlc final {
+  uint64_t logical_id{0};
   Time coordinator_wall_clock = Time::min();
 
   auto operator<=>(const Hlc &other) const { return logical_id <=> other.logical_id; }
 
-  bool operator==(const Hlc &other) const = default;
-  bool operator<(const Hlc &other) const = default;
-  bool operator==(const uint64_t other) const { return logical_id == other; }
-  bool operator<(const uint64_t other) const { return logical_id < other; }
-  bool operator>=(const uint64_t other) const { return logical_id >= other; }
+  bool operator==(const Hlc &other) const noexcept = default;
+  bool operator<(const Hlc &other) const noexcept = default;
+  bool operator==(const uint64_t other) const noexcept { return logical_id == other; }
+  bool operator<(const uint64_t other) const noexcept { return logical_id < other; }
+  bool operator>=(const uint64_t other) const noexcept { return logical_id >= other; }
 
   friend std::ostream &operator<<(std::ostream &in, const Hlc &hlc) {
     auto wall_clock = std::chrono::system_clock::to_time_t(hlc.coordinator_wall_clock);

src/coordinator/shard_map.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -128,9 +128,6 @@ struct LabelSpace {
   // Maps between the smallest primary key stored in the shard and the shard
   std::map<PrimaryKey, ShardMetadata> shards;
   size_t replication_factor;
-  // TODO
-  // Stub value. Should be replaced once the shard-split logic is in place.
-  int64_t split_threshold{10000};
 
   friend std::ostream &operator<<(std::ostream &in, const LabelSpace &label_space) {
     using utils::print_helpers::operator<<;

src/expr/ast/cypher_main_visitor.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -33,6 +33,7 @@
 
 #include <boost/preprocessor/cat.hpp>
 
+#include "common/types.hpp"
 #include "expr/ast.hpp"
 #include "expr/ast/ast_visitor.hpp"
 #include "expr/exceptions.hpp"
@@ -2941,6 +2942,40 @@ class CypherMainVisitor : public antlropencypher::MemgraphCypherBaseVisitor {
     return schema_property_map;
   }
 
+  inline memgraph::common::SchemaConfigParams SchemaConfigKeyToEnum(std::string_view val) {
+    if (val == "replication_factor") {
+      return memgraph::common::SchemaConfigParams::REPLICATION_FACTOR;
+    } else if (val == "split_threshold") {
+      return memgraph::common::SchemaConfigParams::SPLIT_THRESHOLD;
+    }
+    throw memgraph::expr::SemanticException("Schema configuration parameter not recognized!");
+  }
+
+  /**
+   * @return Schema*
+   */
+  antlrcpp::Any visitSchemaConfigKeyValuePair(MemgraphCypher::SchemaConfigKeyValuePairContext *ctx) override {
+    MG_ASSERT(ctx->literal().size() == 2);
+    const auto key_value =
+        SchemaConfigKeyToEnum(utils::ToLowerCase(std::any_cast<std::string>(ctx->literal(0)->accept(this))));
+    // TODO(jbajic) Introduce variable values here
+    const auto value = std::any_cast<int64_t>(ctx->literal(1)->accept(this));
+    return std::pair<common::SchemaConfigParams, int64_t>{key_value, value};
+  }
+
+  /**
+   * @return Schema*
+   */
+  antlrcpp::Any visitSchemaConfiguration(MemgraphCypher::SchemaConfigurationContext *ctx) override {
+    std::unordered_map<common::SchemaConfigParams, int64_t> map;
+    for (auto *key_value_pair : ctx->schemaConfigKeyValuePair()) {
+      // If the queries are cached, then only the stripped query is parsed, so the actual keys cannot be determined
+      // here. That means duplicates cannot be checked.
+      map.insert(std::any_cast<std::pair<common::SchemaConfigParams, int64_t>>(key_value_pair->accept(this)));
+    }
+    return map;
+  }
+
   /**
    * @return Schema*
    */
@@ -2981,7 +3016,11 @@ class CypherMainVisitor : public antlropencypher::MemgraphCypherBaseVisitor {
     schema_query->label_ = AddLabel(std::any_cast<std::string>(ctx->labelName()->accept(this)));
     schema_query->schema_type_map_ =
         std::any_cast<std::vector<std::pair<PropertyIx, common::SchemaType>>>(ctx->schemaPropertyMap()->accept(this));
-    query_ = schema_query;
+    if (ctx->schemaConfiguration()) {
+      schema_query->schema_config_map_ = std::any_cast<std::unordered_map<common::SchemaConfigParams, int64_t>>(
+          ctx->schemaConfiguration()->accept(this));
+      query_ = schema_query;
+    }
     return schema_query;
   }
 

src/expr/interpret/frame.hpp

@@ -34,7 +34,6 @@ class Frame {
   const TypedValue &at(const Symbol &symbol) const { return elems_.at(symbol.position()); }
 
   auto &elems() { return elems_; }
-  const auto &elems() const { return elems_; }
 
   utils::MemoryResource *GetMemoryResource() const { return elems_.get_allocator().GetMemoryResource(); }
 

src/io/local_transport/local_system.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source

src/io/simulator/simulator.hpp

@@ -41,7 +41,7 @@ class Simulator {
   Io<SimulatorTransport> Register(Address address) {
     std::uniform_int_distribution<uint64_t> seed_distrib;
     uint64_t seed = seed_distrib(rng_);
-    return Io{SimulatorTransport(simulator_handle_, address, seed), address};
+    return Io{SimulatorTransport{simulator_handle_, address, seed}, address};
   }
 
   void IncrementServerCountAndWaitForQuiescentState(Address address) {
@@ -50,12 +50,8 @@ class Simulator {
 
   SimulatorStats Stats() { return simulator_handle_->Stats(); }
 
-  std::shared_ptr<SimulatorHandle> GetSimulatorHandle() const { return simulator_handle_; }
-
   std::function<bool()> GetSimulatorTickClosure() {
-    std::function<bool()> tick_closure = [handle_copy = simulator_handle_] {
-      return handle_copy->MaybeTickSimulator();
-    };
+    std::function<bool()> tick_closure = [handle_copy = simulator_handle_] { return handle_copy->MaybeTickSimulator(); };
     return tick_closure;
   }
 };

src/io/simulator/simulator_transport.hpp

@@ -26,7 +26,7 @@ using memgraph::io::Time;
 
 class SimulatorTransport {
   std::shared_ptr<SimulatorHandle> simulator_handle_;
-  Address address_;
+  const Address address_;
   std::mt19937 rng_;
 
  public:
@@ -36,9 +36,7 @@ class SimulatorTransport {
   template <Message RequestT, Message ResponseT>
   ResponseFuture<ResponseT> Request(Address to_address, Address from_address, RequestT request,
                                     std::function<void()> notification, Duration timeout) {
-    std::function<bool()> tick_simulator = [handle_copy = simulator_handle_] {
-      return handle_copy->MaybeTickSimulator();
-    };
+    std::function<bool()> tick_simulator = [handle_copy = simulator_handle_] { return handle_copy->MaybeTickSimulator(); };
 
     return simulator_handle_->template SubmitRequest<RequestT, ResponseT>(
         to_address, from_address, std::move(request), timeout, std::move(tick_simulator), std::move(notification));

src/memgraph.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -640,8 +640,6 @@ int main(int argc, char **argv) {
   memgraph::machine_manager::MachineManager<memgraph::io::local_transport::LocalTransport> mm{io, config, coordinator};
   std::jthread mm_thread([&mm] { mm.Run(); });
 
-  auto rr_factory = std::make_unique<memgraph::query::v2::LocalRequestRouterFactory>(io);
-
   memgraph::query::v2::InterpreterContext interpreter_context{
       (memgraph::storage::v3::Shard *)(nullptr),
       {.query = {.allow_load_csv = FLAGS_allow_load_csv},
@@ -652,7 +650,7 @@ int main(int argc, char **argv) {
        .stream_transaction_conflict_retries = FLAGS_stream_transaction_conflict_retries,
        .stream_transaction_retry_interval = std::chrono::milliseconds(FLAGS_stream_transaction_retry_interval)},
       FLAGS_data_directory,
-      std::move(rr_factory),
+      std::move(io),
       mm.CoordinatorAddress()};
 
   SessionData session_data{&interpreter_context};

src/parser/opencypher/grammar/MemgraphCypher.g4

@@ -394,6 +394,10 @@ propertyKeyTypePair : propertyKeyName propertyType ;
 
 schemaPropertyMap : '(' propertyKeyTypePair ( ',' propertyKeyTypePair )* ')' ;
 
-createSchema : CREATE SCHEMA ON ':' labelName schemaPropertyMap ;
+schemaConfigKeyValuePair : literal '=' literal ;
+
+schemaConfiguration : ( schemaConfigKeyValuePair ( ',' schemaConfigKeyValuePair )* )? ;
+
+createSchema : CREATE SCHEMA ON ':' labelName schemaPropertyMap ( CONFIG schemaConfiguration ) ? ;
 
 dropSchema : DROP SCHEMA ON ':' labelName ;

src/query/v2/frontend/ast/ast.lcp

@@ -147,6 +147,14 @@ cpp<#
     }
     cpp<#)
 
+(defun clone-map (source dest)
+    #>cpp
+    ${dest}.reserve(${source}.size());
+    for (const auto &[config_key, config_val]: ${source}) {
+      ${dest}.emplace(config_key, config_val);
+    }
+    cpp<#)
+
 ;; The following index structs serve as a decoupling point of AST from
 ;; concrete database types. All the names are collected in AstStorage, and can
 ;; be indexed through these instances. This means that we can create a vector
@@ -2698,10 +2706,15 @@ cpp<#
                    #>cpp
                    ${dest} = storage->GetLabelIx(${source}.name);
                    cpp<#))
+
     (schema_type_map "std::vector<std::pair<PropertyIx, common::SchemaType>>"
       :slk-save #'slk-save-property-map
       :slk-load #'slk-load-property-map
       :clone #'clone-schema-property-vector
+      :scope :public)
+
+   (schema_config_map "std::unordered_map<common::SchemaConfigParams, int64_t>"
+      :clone #'clone-map
       :scope :public))
 
   (:public

src/query/v2/interpreter.cpp

@@ -22,6 +22,7 @@
 #include <memory>
 #include <optional>
 
+#include "common/types.hpp"
 #include "coordinator/coordinator_client.hpp"
 #include "expr/ast/ast_visitor.hpp"
 #include "io/local_transport/local_system.hpp"
@@ -64,11 +65,7 @@
 #include "utils/tsc.hpp"
 #include "utils/variant_helpers.hpp"
 
-// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
-DEFINE_bool(use_multi_frame, false, "Whether to use MultiFrame or not");
-
 namespace EventCounter {
-
 extern Event ReadQuery;
 extern Event WriteQuery;
 extern Event ReadWriteQuery;
@@ -78,7 +75,6 @@ extern const Event LabelPropertyIndexCreated;
 
 extern const Event StreamsCreated;
 extern const Event TriggersCreated;
-
 }  // namespace EventCounter
 
 namespace memgraph::query::v2 {
@@ -547,7 +543,7 @@ Callback HandleSchemaQuery(SchemaQuery *schema_query, InterpreterContext *interp
         std::vector<std::vector<TypedValue>> results;
         results.reserve(schemas_info.schemas.size());
 
-        for (const auto &[label_id, schema_types] : schemas_info.schemas) {
+        for ([[maybe_unused]] const auto &[label_id, schema_types, config] : schemas_info.schemas) {
           std::vector<TypedValue> schema_info_row;
           schema_info_row.reserve(3);
 
@@ -575,7 +571,7 @@ Callback HandleSchemaQuery(SchemaQuery *schema_query, InterpreterContext *interp
         const auto *schema = db->GetSchema(label);
         std::vector<std::vector<TypedValue>> results;
         if (schema) {
-          for (const auto &schema_property : schema->second) {
+          for (const auto &schema_property : schema->properties) {
             std::vector<TypedValue> schema_info_row;
             schema_info_row.reserve(2);
             schema_info_row.emplace_back(db->PropertyToName(schema_property.property_id));
@@ -594,7 +590,8 @@ Callback HandleSchemaQuery(SchemaQuery *schema_query, InterpreterContext *interp
         throw SyntaxException("One or more types have to be defined in schema definition.");
       }
       callback.fn = [interpreter_context, primary_label = schema_query->label_,
-                     schema_type_map = std::move(schema_type_map)]() {
+                     schema_type_map = std::move(schema_type_map),
+                     config_map = std::move(schema_query->schema_config_map_)]() {
         auto *db = interpreter_context->db;
         const auto label = interpreter_context->NameToLabelId(primary_label.name);
         std::vector<storage::v3::SchemaProperty> schemas_types;
@@ -603,8 +600,30 @@ Callback HandleSchemaQuery(SchemaQuery *schema_query, InterpreterContext *interp
           auto property_id = interpreter_context->NameToPropertyId(schema_type.first.name);
           schemas_types.push_back({property_id, schema_type.second});
         }
-        if (!db->CreateSchema(label, schemas_types)) {
-          throw QueryException(fmt::format("Schema on label :{} already exists!", primary_label.name));
+        auto schema_config = std::invoke([&config_map]() -> std::optional<storage::v3::Schema::SchemaConfiguration> {
+          if (!config_map.empty()) {
+            return std::nullopt;
+          }
+          storage::v3::Schema::SchemaConfiguration config;
+          if (const auto replication_factor_it = config_map.find(common::SchemaConfigParams::REPLICATION_FACTOR);
+              replication_factor_it != config_map.end()) {
+            config.replication_factor = replication_factor_it->second;
+          }
+          if (const auto split_threshold_it = config_map.find(common::SchemaConfigParams::SPLIT_THRESHOLD);
+              split_threshold_it != config_map.end()) {
+            config.split_threshold = split_threshold_it->second;
+          }
+          return config;
+        });
+        // TODO FIx this late
+        if (schema_config) {
+          if (!db->CreateSchema(label, schemas_types, *schema_config)) {
+            throw QueryException(fmt::format("Schema on label :{} already exists!", primary_label.name));
+          }
+        } else {
+          if (!db->CreateSchema(label, schemas_types)) {
+            throw QueryException(fmt::format("Schema on label :{} already exists!", primary_label.name));
+          }
         }
         return std::vector<std::vector<TypedValue>>{};
       };
@@ -693,7 +712,7 @@ PullPlan::PullPlan(const std::shared_ptr<CachedPlan> plan, const Parameters &par
     : plan_(plan),
       cursor_(plan->plan().MakeCursor(execution_memory)),
       frame_(plan->symbol_table().max_position(), execution_memory),
-      multi_frame_(plan->symbol_table().max_position(), FLAGS_default_multi_frame_size, execution_memory),
+      multi_frame_(plan->symbol_table().max_position(), kNumberOfFramesInMultiframe, execution_memory),
       memory_limit_(memory_limit) {
   ctx_.db_accessor = dba;
   ctx_.symbol_table = plan->symbol_table();
@@ -709,6 +728,7 @@ PullPlan::PullPlan(const std::shared_ptr<CachedPlan> plan, const Parameters &par
   ctx_.request_router = request_router;
   ctx_.edge_ids_alloc = &interpreter_context->edge_ids_alloc;
 }
+
 std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::PullMultiple(AnyStream *stream, std::optional<int> n,
                                                                         const std::vector<Symbol> &output_symbols,
                                                                         std::map<std::string, TypedValue> *summary) {
@@ -736,7 +756,10 @@ std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::PullMultiple(AnyStrea
   }
 
   // Returns true if a result was pulled.
-  const auto pull_result = [&]() -> bool { return cursor_->PullMultiple(multi_frame_, ctx_); };
+  const auto pull_result = [&]() -> bool {
+    cursor_->PullMultiple(multi_frame_, ctx_);
+    return !multi_frame_.HasInvalidFrame();
+  };
 
   const auto stream_values = [&output_symbols, &stream](const Frame &frame) {
     // TODO: The streamed values should also probably use the above memory.
@@ -756,14 +779,13 @@ std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::PullMultiple(AnyStrea
   int i = 0;
   if (has_unsent_results_ && !output_symbols.empty()) {
     // stream unsent results from previous pull
-    for (auto &frame : multi_frame_.GetValidFramesConsumer()) {
+
+    auto iterator_for_valid_frame_only = multi_frame_.GetValidFramesReader();
+    for (const auto &frame : iterator_for_valid_frame_only) {
       stream_values(frame);
-      frame.MakeInvalid();
       ++i;
-      if (i == n) {
-        break;
-      }
     }
+    multi_frame_.MakeAllFramesInvalid();
   }
 
   for (; !n || i < n;) {
@@ -772,17 +794,13 @@ std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::PullMultiple(AnyStrea
     }
 
     if (!output_symbols.empty()) {
-      for (auto &frame : multi_frame_.GetValidFramesConsumer()) {
+      auto iterator_for_valid_frame_only = multi_frame_.GetValidFramesReader();
+      for (const auto &frame : iterator_for_valid_frame_only) {
         stream_values(frame);
-        frame.MakeInvalid();
         ++i;
-        if (i == n) {
-          break;
-        }
       }
-    } else {
-      multi_frame_.MakeAllFramesInvalid();
     }
+    multi_frame_.MakeAllFramesInvalid();
   }
 
   // If we finished because we streamed the requested n results,
@@ -817,7 +835,8 @@ std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::PullMultiple(AnyStrea
 std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::Pull(AnyStream *stream, std::optional<int> n,
                                                                 const std::vector<Symbol> &output_symbols,
                                                                 std::map<std::string, TypedValue> *summary) {
-  if (FLAGS_use_multi_frame) {
+  auto should_pull_multiple = false;  // TODO on the long term, we will only use PullMultiple
+  if (should_pull_multiple) {
     return PullMultiple(stream, n, output_symbols, summary);
   }
   // Set up temporary memory for a single Pull. Initial memory comes from the
@@ -911,24 +930,34 @@ using RWType = plan::ReadWriteTypeChecker::RWType;
 
 InterpreterContext::InterpreterContext(storage::v3::Shard *db, const InterpreterConfig config,
                                        const std::filesystem::path & /*data_directory*/,
-                                       std::unique_ptr<RequestRouterFactory> request_router_factory,
+                                       io::Io<io::local_transport::LocalTransport> io,
                                        coordinator::Address coordinator_addr)
-    : db(db),
-      config(config),
-      coordinator_address{coordinator_addr},
-      request_router_factory_{std::move(request_router_factory)} {}
+    : db(db), config(config), io{std::move(io)}, coordinator_address{coordinator_addr} {}
 
 Interpreter::Interpreter(InterpreterContext *interpreter_context) : interpreter_context_(interpreter_context) {
   MG_ASSERT(interpreter_context_, "Interpreter context must not be NULL");
 
-  request_router_ =
-      interpreter_context_->request_router_factory_->CreateRequestRouter(interpreter_context_->coordinator_address);
+  // TODO(tyler) make this deterministic so that it can be tested.
+  auto random_uuid = boost::uuids::uuid{boost::uuids::random_generator()()};
+  auto query_io = interpreter_context_->io.ForkLocal(random_uuid);
 
+  request_router_ = std::make_unique<RequestRouter<io::local_transport::LocalTransport>>(
+      coordinator::CoordinatorClient<io::local_transport::LocalTransport>(
+          query_io, interpreter_context_->coordinator_address, std::vector{interpreter_context_->coordinator_address}),
+      std::move(query_io));
   // Get edge ids
-  const auto edge_ids_alloc_min_max_pair =
-      request_router_->AllocateInitialEdgeIds(interpreter_context_->coordinator_address);
-  if (edge_ids_alloc_min_max_pair) {
-    interpreter_context_->edge_ids_alloc = {edge_ids_alloc_min_max_pair->first, edge_ids_alloc_min_max_pair->second};
+  coordinator::CoordinatorWriteRequests requests{coordinator::AllocateEdgeIdBatchRequest{.batch_size = 1000000}};
+  io::rsm::WriteRequest<coordinator::CoordinatorWriteRequests> ww;
+  ww.operation = requests;
+  auto resp = interpreter_context_->io
+                  .Request<io::rsm::WriteRequest<coordinator::CoordinatorWriteRequests>,
+                           io::rsm::WriteResponse<coordinator::CoordinatorWriteResponses>>(
+                      interpreter_context_->coordinator_address, ww)
+                  .Wait();
+  if (resp.HasValue()) {
+    const auto alloc_edge_id_reps =
+        std::get<coordinator::AllocateEdgeIdBatchResponse>(resp.GetValue().message.write_return);
+    interpreter_context_->edge_ids_alloc = {alloc_edge_id_reps.low, alloc_edge_id_reps.high};
   }
 }
 

src/query/v2/interpreter.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -16,6 +16,7 @@
 
 #include "coordinator/coordinator.hpp"
 #include "coordinator/coordinator_client.hpp"
+#include "io/local_transport/local_transport.hpp"
 #include "io/transport.hpp"
 #include "query/v2/auth_checker.hpp"
 #include "query/v2/bindings/cypher_main_visitor.hpp"
@@ -171,8 +172,7 @@ struct PreparedQuery {
 struct InterpreterContext {
   explicit InterpreterContext(storage::v3::Shard *db, InterpreterConfig config,
                               const std::filesystem::path &data_directory,
-                              std::unique_ptr<RequestRouterFactory> request_router_factory,
-                              coordinator::Address coordinator_addr);
+                              io::Io<io::local_transport::LocalTransport> io, coordinator::Address coordinator_addr);
 
   storage::v3::Shard *db;
 
@@ -188,24 +188,26 @@ struct InterpreterContext {
   const InterpreterConfig config;
   IdAllocator edge_ids_alloc;
 
+  // TODO (antaljanosbenjamin) Figure out an abstraction for io::Io to make it possible to construct an interpreter
+  // context with a simulator transport without templatizing it.
+  io::Io<io::local_transport::LocalTransport> io;
   coordinator::Address coordinator_address;
-  std::unique_ptr<RequestRouterFactory> request_router_factory_;
 
   storage::v3::LabelId NameToLabelId(std::string_view label_name) {
-    return storage::v3::LabelId::FromUint(query_id_mapper_.NameToId(label_name));
+    return storage::v3::LabelId::FromUint(query_id_mapper.NameToId(label_name));
   }
 
   storage::v3::PropertyId NameToPropertyId(std::string_view property_name) {
-    return storage::v3::PropertyId::FromUint(query_id_mapper_.NameToId(property_name));
+    return storage::v3::PropertyId::FromUint(query_id_mapper.NameToId(property_name));
   }
 
   storage::v3::EdgeTypeId NameToEdgeTypeId(std::string_view edge_type_name) {
-    return storage::v3::EdgeTypeId::FromUint(query_id_mapper_.NameToId(edge_type_name));
+    return storage::v3::EdgeTypeId::FromUint(query_id_mapper.NameToId(edge_type_name));
   }
 
  private:
   // TODO Replace with local map of labels, properties and edge type ids
-  storage::v3::NameIdMapper query_id_mapper_;
+  storage::v3::NameIdMapper query_id_mapper;
 };
 
 /// Function that is used to tell all active interpreters that they should stop
@@ -295,15 +297,12 @@ class Interpreter final {
   void Abort();
 
   const RequestRouterInterface *GetRequestRouter() const { return request_router_.get(); }
-  void InstallSimulatorTicker(std::function<bool()> &&tick_simulator) {
-    request_router_->InstallSimulatorTicker(tick_simulator);
-  }
 
  private:
   struct QueryExecution {
+    std::optional<PreparedQuery> prepared_query;
     utils::MonotonicBufferResource execution_memory{kExecutionMemoryBlockSize};
     utils::ResourceWithOutOfMemoryException execution_memory_with_exception{&execution_memory};
-    std::optional<PreparedQuery> prepared_query;
 
     std::map<std::string, TypedValue> summary;
     std::vector<Notification> notifications;

src/query/v2/multiframe.cpp

@@ -17,9 +17,6 @@
 #include "query/v2/bindings/frame.hpp"
 #include "utils/pmr/vector.hpp"
 
-// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
-DEFINE_uint64(default_multi_frame_size, 100, "Default size of MultiFrame");
-
 namespace memgraph::query::v2 {
 
 static_assert(std::forward_iterator<ValidFramesReader::Iterator>);
@@ -57,17 +54,15 @@ bool MultiFrame::HasInvalidFrame() const noexcept {
 
 // NOLINTNEXTLINE (bugprone-exception-escape)
 void MultiFrame::DefragmentValidFrames() noexcept {
-  static constexpr auto kIsValid = [](const FrameWithValidity &frame) { return frame.IsValid(); };
-  static constexpr auto kIsInvalid = [](const FrameWithValidity &frame) { return !frame.IsValid(); };
-  auto first_invalid_frame = std::find_if(frames_.begin(), frames_.end(), kIsInvalid);
-  auto following_first_valid = std::find_if(first_invalid_frame, frames_.end(), kIsValid);
-  while (first_invalid_frame != frames_.end() && following_first_valid != frames_.end()) {
-    std::swap(*first_invalid_frame, *following_first_valid);
-    first_invalid_frame++;
-    first_invalid_frame = std::find_if(first_invalid_frame, frames_.end(), kIsInvalid);
-    following_first_valid++;
-    following_first_valid = std::find_if(following_first_valid, frames_.end(), kIsValid);
-  }
+  /*
+  from: https://en.cppreference.com/w/cpp/algorithm/remove
+  "Removing is done by shifting (by means of copy assignment (until C++11)move assignment (since C++11)) the elements
+  in the range in such a way that the elements that are not to be removed appear in the beginning of the range.
+  Relative order of the elements that remain is preserved and the physical size of the container is unchanged."
+  */
+
+  // NOLINTNEXTLINE (bugprone-unused-return-value)
+  std::remove_if(frames_.begin(), frames_.end(), [](auto &frame) { return !frame.IsValid(); });
 }
 
 ValidFramesReader MultiFrame::GetValidFramesReader() { return ValidFramesReader{*this}; }

src/query/v2/multiframe.hpp

@@ -13,14 +13,10 @@
 
 #include <iterator>
 
-#include <gflags/gflags.h>
-
 #include "query/v2/bindings/frame.hpp"
 
-// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
-DECLARE_uint64(default_multi_frame_size);
-
 namespace memgraph::query::v2 {
+constexpr uint64_t kNumberOfFramesInMultiframe = 1000;  // TODO have it configurable
 
 class ValidFramesConsumer;
 class ValidFramesModifier;
@@ -37,7 +33,6 @@ class MultiFrame {
   MultiFrame(size_t size_of_frame, size_t number_of_frames, utils::MemoryResource *execution_memory);
   ~MultiFrame() = default;
 
-  // Assigning and moving the MultiFrame is not allowed if any accessor from the above ones are alive.
   MultiFrame(const MultiFrame &other);
   MultiFrame(MultiFrame &&other) noexcept;
   MultiFrame &operator=(const MultiFrame &other) = delete;
@@ -102,9 +97,9 @@ class ValidFramesReader {
 
   ~ValidFramesReader() = default;
   ValidFramesReader(const ValidFramesReader &other) = delete;
-  ValidFramesReader(ValidFramesReader &&other) noexcept = default;
+  ValidFramesReader(ValidFramesReader &&other) noexcept = delete;
   ValidFramesReader &operator=(const ValidFramesReader &other) = delete;
-  ValidFramesReader &operator=(ValidFramesReader &&other) noexcept = default;
+  ValidFramesReader &operator=(ValidFramesReader &&other) noexcept = delete;
 
   struct Iterator {
     using iterator_category = std::forward_iterator_tag;
@@ -152,9 +147,9 @@ class ValidFramesModifier {
 
   ~ValidFramesModifier() = default;
   ValidFramesModifier(const ValidFramesModifier &other) = delete;
-  ValidFramesModifier(ValidFramesModifier &&other) noexcept = default;
+  ValidFramesModifier(ValidFramesModifier &&other) noexcept = delete;
   ValidFramesModifier &operator=(const ValidFramesModifier &other) = delete;
-  ValidFramesModifier &operator=(ValidFramesModifier &&other) noexcept = default;
+  ValidFramesModifier &operator=(ValidFramesModifier &&other) noexcept = delete;
 
   struct Iterator {
     using iterator_category = std::forward_iterator_tag;
@@ -207,9 +202,9 @@ class ValidFramesConsumer {
 
   ~ValidFramesConsumer() noexcept;
   ValidFramesConsumer(const ValidFramesConsumer &other) = delete;
-  ValidFramesConsumer(ValidFramesConsumer &&other) noexcept = default;
+  ValidFramesConsumer(ValidFramesConsumer &&other) noexcept = delete;
   ValidFramesConsumer &operator=(const ValidFramesConsumer &other) = delete;
-  ValidFramesConsumer &operator=(ValidFramesConsumer &&other) noexcept = default;
+  ValidFramesConsumer &operator=(ValidFramesConsumer &&other) noexcept = delete;
 
   struct Iterator {
     using iterator_category = std::forward_iterator_tag;
@@ -261,9 +256,9 @@ class InvalidFramesPopulator {
   ~InvalidFramesPopulator() = default;
 
   InvalidFramesPopulator(const InvalidFramesPopulator &other) = delete;
-  InvalidFramesPopulator(InvalidFramesPopulator &&other) noexcept = default;
+  InvalidFramesPopulator(InvalidFramesPopulator &&other) noexcept = delete;
   InvalidFramesPopulator &operator=(const InvalidFramesPopulator &other) = delete;
-  InvalidFramesPopulator &operator=(InvalidFramesPopulator &&other) noexcept = default;
+  InvalidFramesPopulator &operator=(InvalidFramesPopulator &&other) noexcept = delete;
 
   struct Iterator {
     using iterator_category = std::forward_iterator_tag;

src/query/v2/plan/cost_estimator.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source

src/query/v2/plan/operator.lcp

@@ -72,21 +72,7 @@ class Cursor {
   /// @throws QueryRuntimeException if something went wrong with execution
   virtual bool Pull(Frame &, ExecutionContext &) = 0;
 
-  /// Run an iteration of a @c LogicalOperator with MultiFrame.
-  ///
-  /// Since operators may be chained, the iteration may pull results from
-  /// multiple operators.
-  ///
-  /// @param MultiFrame May be read from or written to while performing the
-  ///     iteration.
-  /// @param ExecutionContext Used to get the position of symbols in frame and
-  ///     other information.
-  /// @return True if the operator was able to populate at least one Frame on the MultiFrame,
-  ///     thus if an operator returns true, that means there is at least one valid Frame in the
-  ///     MultiFrame.
-  ///
-  /// @throws QueryRuntimeException if something went wrong with execution
-  virtual bool PullMultiple(MultiFrame &, ExecutionContext &) {MG_ASSERT(false, "PullMultipleIsNotImplemented"); return false; }
+  virtual void PullMultiple(MultiFrame &, ExecutionContext &) { LOG_FATAL("PullMultipleIsNotImplemented"); }
 
   /// Resets the Cursor to its initial state.
   virtual void Reset() = 0;
@@ -349,7 +335,7 @@ and false on every following Pull.")
    class OnceCursor : public Cursor {
     public:
      OnceCursor() {}
-     bool PullMultiple(MultiFrame &, ExecutionContext &) override;
+     void PullMultiple(MultiFrame &, ExecutionContext &) override;
      bool Pull(Frame &, ExecutionContext &) override;
      void Shutdown() override;
      void Reset() override;
@@ -1176,7 +1162,6 @@ a boolean value.")
     public:
      FilterCursor(const Filter &, utils::MemoryResource *);
      bool Pull(Frame &, ExecutionContext &) override;
-     bool PullMultiple(MultiFrame &, ExecutionContext &) override;
      void Shutdown() override;
      void Reset() override;
 
@@ -1228,7 +1213,7 @@ RETURN clause) the Produce's pull succeeds exactly once.")
     public:
      ProduceCursor(const Produce &, utils::MemoryResource *);
      bool Pull(Frame &, ExecutionContext &) override;
-     bool PullMultiple(MultiFrame &, ExecutionContext &) override;
+     void PullMultiple(MultiFrame &, ExecutionContext &) override;
      void Shutdown() override;
      void Reset() override;
 
@@ -1276,7 +1261,6 @@ Has a flag for using DETACH DELETE when deleting vertices.")
     public:
      DeleteCursor(const Delete &, utils::MemoryResource *);
      bool Pull(Frame &, ExecutionContext &) override;
-     bool PullMultiple(MultiFrame &, ExecutionContext &) override;
      void Shutdown() override;
      void Reset() override;
 
@@ -1570,7 +1554,6 @@ edge lists).")
      EdgeUniquenessFilterCursor(const EdgeUniquenessFilter &,
                                 utils::MemoryResource *);
      bool Pull(Frame &, ExecutionContext &) override;
-     bool PullMultiple(MultiFrame &, ExecutionContext &) override;
      void Shutdown() override;
      void Reset() override;
 

src/query/v2/plan/rewrite/index_lookup.hpp

@@ -17,7 +17,6 @@
 #pragma once
 
 #include <algorithm>
-#include <limits>
 #include <memory>
 #include <optional>
 #include <unordered_map>
@@ -562,12 +561,8 @@ class IndexLookupRewriter final : public HierarchicalLogicalOperatorVisitor {
   // `max_vertex_count` controls, whether no operator should be created if the
   // vertex count in the best index exceeds this number. In such a case,
   // `nullptr` is returned and `input` is not chained.
-  // std::unique_ptr<ScanAll> GenScanByIndex(const ScanAll &scan, const std::optional<int64_t> &max_vertex_count =
-  // std::nullopt) {
-  std::unique_ptr<ScanAll> GenScanByIndex(const ScanAll &scan, std::optional<int64_t> max_vertex_count = std::nullopt) {
-    // debug (gvolfing)
-    max_vertex_count = std::numeric_limits<int64_t>::max();
-
+  std::unique_ptr<ScanAll> GenScanByIndex(const ScanAll &scan,
+                                          const std::optional<int64_t> &max_vertex_count = std::nullopt) {
     const auto &input = scan.input();
     const auto &node_symbol = scan.output_symbol_;
     const auto &view = scan.view_;
@@ -602,9 +597,6 @@ class IndexLookupRewriter final : public HierarchicalLogicalOperatorVisitor {
                      [](const auto &schema_elem) { return schema_elem.property_id; });
 
       for (const auto &property_filter : property_filters) {
-        if (property_filter.property_filter->type_ != PropertyFilter::Type::EQUAL) {
-          continue;
-        }
         const auto &property_id = db_->NameToProperty(property_filter.property_filter->property_.name);
         if (std::find(schema_properties.begin(), schema_properties.end(), property_id) != schema_properties.end()) {
           pk_temp.emplace_back(std::make_pair(property_filter.expression, property_filter));

src/query/v2/plan/vertex_count_cache.hpp

@@ -38,15 +38,12 @@ class VertexCountCache {
   auto NameToProperty(const std::string &name) { return request_router_->NameToProperty(name); }
   auto NameToEdgeType(const std::string &name) { return request_router_->NameToEdgeType(name); }
 
-  int64_t VerticesCount() { return request_router_->GetApproximateVertexCount(); }
+  int64_t VerticesCount() { return 1; }
 
-  int64_t VerticesCount(storage::v3::LabelId label) { return request_router_->GetApproximateVertexCount(label); }
+  int64_t VerticesCount(storage::v3::LabelId /*label*/) { return 1; }
 
-  int64_t VerticesCount(storage::v3::LabelId label, storage::v3::PropertyId property) {
-    return request_router_->GetApproximateVertexCount(label, property);
-  }
+  int64_t VerticesCount(storage::v3::LabelId /*label*/, storage::v3::PropertyId /*property*/) { return 1; }
 
-  // TODO(gvolfing) check if we actually use these overloads...
   int64_t VerticesCount(storage::v3::LabelId /*label*/, storage::v3::PropertyId /*property*/,
                         const storage::v3::PropertyValue & /*value*/) {
     return 1;

src/query/v2/request_router.hpp

@@ -24,18 +24,14 @@
 #include <stdexcept>
 #include <thread>
 #include <unordered_map>
-#include <variant>
 #include <vector>
 
-#include <boost/uuid/uuid.hpp>
-
 #include "coordinator/coordinator.hpp"
 #include "coordinator/coordinator_client.hpp"
 #include "coordinator/coordinator_rsm.hpp"
 #include "coordinator/shard_map.hpp"
 #include "io/address.hpp"
 #include "io/errors.hpp"
-#include "io/local_transport/local_transport.hpp"
 #include "io/notifier.hpp"
 #include "io/rsm/raft.hpp"
 #include "io/rsm/rsm_client.hpp"
@@ -117,15 +113,8 @@ class RequestRouterInterface {
   virtual std::optional<storage::v3::EdgeTypeId> MaybeNameToEdgeType(const std::string &name) const = 0;
   virtual std::optional<storage::v3::LabelId> MaybeNameToLabel(const std::string &name) const = 0;
   virtual bool IsPrimaryLabel(storage::v3::LabelId label) const = 0;
-  virtual bool IsPrimaryProperty(storage::v3::LabelId primary_label, storage::v3::PropertyId property) const = 0;
-
-  virtual std::optional<std::pair<uint64_t, uint64_t>> AllocateInitialEdgeIds(io::Address coordinator_address) = 0;
-  virtual void InstallSimulatorTicker(std::function<bool()> tick_simulator) = 0;
+  virtual bool IsPrimaryKey(storage::v3::LabelId primary_label, storage::v3::PropertyId property) const = 0;
   virtual const std::vector<coordinator::SchemaProperty> &GetSchemaForLabel(storage::v3::LabelId label) const = 0;
-
-  virtual int64_t GetApproximateVertexCount() const = 0;
-  virtual int64_t GetApproximateVertexCount(storage::v3::LabelId label) const = 0;
-  virtual int64_t GetApproximateVertexCount(storage::v3::LabelId label, storage::v3::PropertyId property) const = 0;
 };
 
 // TODO(kostasrim)rename this class template
@@ -150,7 +139,7 @@ class RequestRouter : public RequestRouterInterface {
 
   ~RequestRouter() override {}
 
-  void InstallSimulatorTicker(std::function<bool()> tick_simulator) override {
+  void InstallSimulatorTicker(std::function<bool()> tick_simulator) {
     notifier_.InstallSimulatorTicker(tick_simulator);
   }
 
@@ -235,7 +224,7 @@ class RequestRouter : public RequestRouterInterface {
     return edge_types_.IdToName(id.AsUint());
   }
 
-  bool IsPrimaryProperty(storage::v3::LabelId primary_label, storage::v3::PropertyId property) const override {
+  bool IsPrimaryKey(storage::v3::LabelId primary_label, storage::v3::PropertyId property) const override {
     const auto schema_it = shards_map_.schemas.find(primary_label);
     MG_ASSERT(schema_it != shards_map_.schemas.end(), "Invalid primary label id: {}", primary_label.AsUint());
 
@@ -419,30 +408,6 @@ class RequestRouter : public RequestRouterInterface {
     return shards_map_.GetLabelId(name);
   }
 
-  int64_t GetApproximateVertexCount() const override {
-    int64_t vertex_count = 0;
-
-    for (const auto &label_space : shards_map_.label_spaces) {
-      const auto split_threshold = label_space.second.split_threshold;
-      const auto shard_count = static_cast<int64_t>(label_space.second.shards.size());
-      vertex_count += split_threshold * shard_count;
-    }
-
-    return vertex_count;
-  }
-
-  int64_t GetApproximateVertexCount(storage::v3::LabelId label) const override {
-    const auto &label_space = shards_map_.label_spaces.at(label);
-    return label_space.split_threshold * label_space.shards.size();
-  }
-
-  int64_t GetApproximateVertexCount(storage::v3::LabelId label, storage::v3::PropertyId /*property*/) const override {
-    // TODO(gvolfing)
-    // Once we have reliable metadata to approximate the
-    // vertex count -based on properties- rework this function.
-    return GetApproximateVertexCount(label);
-  }
-
  private:
   std::vector<ShardRequestState<msgs::CreateVerticesRequest>> RequestsForCreateVertices(
       const std::vector<msgs::NewVertex> &new_vertices) {
@@ -750,23 +715,6 @@ class RequestRouter : public RequestRouterInterface {
     edge_types_.StoreMapping(std::move(id_to_name));
   }
 
-  std::optional<std::pair<uint64_t, uint64_t>> AllocateInitialEdgeIds(io::Address coordinator_address) override {
-    coordinator::CoordinatorWriteRequests requests{coordinator::AllocateEdgeIdBatchRequest{.batch_size = 1000000}};
-
-    io::rsm::WriteRequest<coordinator::CoordinatorWriteRequests> ww;
-    ww.operation = requests;
-    auto resp =
-        io_.template Request<io::rsm::WriteRequest<coordinator::CoordinatorWriteRequests>,
-                             io::rsm::WriteResponse<coordinator::CoordinatorWriteResponses>>(coordinator_address, ww)
-            .Wait();
-    if (resp.HasValue()) {
-      const auto alloc_edge_id_reps =
-          std::get<coordinator::AllocateEdgeIdBatchResponse>(resp.GetValue().message.write_return);
-      return std::make_pair(alloc_edge_id_reps.low, alloc_edge_id_reps.high);
-    }
-    return {};
-  }
-
   ShardMap shards_map_;
   storage::v3::NameIdMapper properties_;
   storage::v3::NameIdMapper edge_types_;
@@ -778,66 +726,4 @@ class RequestRouter : public RequestRouterInterface {
   io::Notifier notifier_ = {};
   // TODO(kostasrim) Add batch prefetching
 };
-
-class RequestRouterFactory {
- public:
-  RequestRouterFactory() = default;
-  RequestRouterFactory(const RequestRouterFactory &) = delete;
-  RequestRouterFactory &operator=(const RequestRouterFactory &) = delete;
-  RequestRouterFactory(RequestRouterFactory &&) = delete;
-  RequestRouterFactory &operator=(RequestRouterFactory &&) = delete;
-
-  virtual ~RequestRouterFactory() = default;
-
-  virtual std::unique_ptr<RequestRouterInterface> CreateRequestRouter(
-      const coordinator::Address &coordinator_address) const = 0;
-};
-
-class LocalRequestRouterFactory : public RequestRouterFactory {
-  using LocalTransportIo = io::Io<io::local_transport::LocalTransport>;
-  LocalTransportIo &io_;
-
- public:
-  explicit LocalRequestRouterFactory(LocalTransportIo &io) : io_(io) {}
-
-  std::unique_ptr<RequestRouterInterface> CreateRequestRouter(
-      const coordinator::Address &coordinator_address) const override {
-    using TransportType = io::local_transport::LocalTransport;
-
-    auto query_io = io_.ForkLocal(boost::uuids::uuid{boost::uuids::random_generator()()});
-    auto local_transport_io = io_.ForkLocal(boost::uuids::uuid{boost::uuids::random_generator()()});
-
-    return std::make_unique<RequestRouter<TransportType>>(
-        coordinator::CoordinatorClient<TransportType>(query_io, coordinator_address, {coordinator_address}),
-        std::move(local_transport_io));
-  }
-};
-
-class SimulatedRequestRouterFactory : public RequestRouterFactory {
-  io::simulator::Simulator *simulator_;
-
- public:
-  explicit SimulatedRequestRouterFactory(io::simulator::Simulator &simulator) : simulator_(&simulator) {}
-
-  std::unique_ptr<RequestRouterInterface> CreateRequestRouter(
-      const coordinator::Address &coordinator_address) const override {
-    using TransportType = io::simulator::SimulatorTransport;
-    auto actual_transport_handle = simulator_->GetSimulatorHandle();
-
-    boost::uuids::uuid random_uuid;
-    io::Address unique_local_addr_query;
-
-    // The simulated RR should not introduce stochastic behavior.
-    random_uuid = boost::uuids::uuid{3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-    unique_local_addr_query = {.unique_id = boost::uuids::uuid{4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}};
-
-    auto io = simulator_->Register(unique_local_addr_query);
-    auto query_io = io.ForkLocal(random_uuid);
-
-    return std::make_unique<RequestRouter<TransportType>>(
-        coordinator::CoordinatorClient<TransportType>(query_io, coordinator_address, {coordinator_address}),
-        std::move(io));
-  }
-};
-
 }  // namespace memgraph::query::v2

src/query/v2/requests.hpp

@@ -12,6 +12,7 @@
 #pragma once
 
 #include <chrono>
+#include <cstdint>
 #include <iostream>
 #include <map>
 #include <memory>
@@ -25,7 +26,6 @@
 #include "storage/v3/id_types.hpp"
 #include "storage/v3/property_value.hpp"
 #include "storage/v3/result.hpp"
-#include "utils/fnv.hpp"
 
 namespace memgraph::msgs {
 
@@ -571,6 +571,16 @@ struct CommitResponse {
   std::optional<ShardError> error;
 };
 
+struct SplitInfo {
+  PrimaryKey split_key;
+  uint64_t shard_version;
+};
+
+struct PerformSplitDataInfo {
+  PrimaryKey split_key;
+  uint64_t shard_version;
+};
+
 using ReadRequests = std::variant<ExpandOneRequest, GetPropertiesRequest, ScanVerticesRequest>;
 using ReadResponses = std::variant<ExpandOneResponse, GetPropertiesResponse, ScanVerticesResponse>;
 
@@ -580,48 +590,3 @@ using WriteResponses = std::variant<CreateVerticesResponse, DeleteVerticesRespon
                                     CreateExpandResponse, DeleteEdgesResponse, UpdateEdgesResponse, CommitResponse>;
 
 }  // namespace memgraph::msgs
-
-namespace std {
-
-template <>
-struct hash<memgraph::msgs::Value>;
-
-template <>
-struct hash<memgraph::msgs::VertexId> {
-  size_t operator()(const memgraph::msgs::VertexId &id) const {
-    using LabelId = memgraph::storage::v3::LabelId;
-    using Value = memgraph::msgs::Value;
-    return memgraph::utils::HashCombine<LabelId, std::vector<Value>, std::hash<LabelId>,
-                                        memgraph::utils::FnvCollection<std::vector<Value>, Value>>{}(id.first.id,
-                                                                                                     id.second);
-  }
-};
-
-template <>
-struct hash<memgraph::msgs::Value> {
-  size_t operator()(const memgraph::msgs::Value &value) const {
-    using Type = memgraph::msgs::Value::Type;
-    switch (value.type) {
-      case Type::Null:
-        return std::hash<size_t>{}(0U);
-      case Type::Bool:
-        return std::hash<bool>{}(value.bool_v);
-      case Type::Int64:
-        return std::hash<int64_t>{}(value.int_v);
-      case Type::Double:
-        return std::hash<double>{}(value.double_v);
-      case Type::String:
-        return std::hash<std::string>{}(value.string_v);
-      case Type::List:
-        LOG_FATAL("Add hash for lists");
-      case Type::Map:
-        LOG_FATAL("Add hash for maps");
-      case Type::Vertex:
-        LOG_FATAL("Add hash for vertices");
-      case Type::Edge:
-        LOG_FATAL("Add hash for edges");
-    }
-  }
-};
-
-}  // namespace std

src/storage/v3/CMakeLists.txt

@@ -18,6 +18,7 @@ set(storage_v3_src_files
     bindings/typed_value.cpp
     expr.cpp
     vertex.cpp
+    splitter.cpp
     request_helper.cpp)
 
 # ######################

src/storage/v3/bindings/ast/ast.lcp

@@ -146,6 +146,14 @@ cpp<#
     }
     cpp<#)
 
+(defun clone-map (source dest)
+    #>cpp
+    ${dest}.reserve(${source}.size());
+    for (const auto &[config_key, config_val]: ${source}) {
+      ${dest}.emplace(config_key, config_val);
+    }
+    cpp<#)
+
 ;; The following index structs serve as a decoupling point of AST from
 ;; concrete database types. All the names are collected in AstStorage, and can
 ;; be indexed through these instances. This means that we can create a vector
@@ -2705,10 +2713,15 @@ cpp<#
                    #>cpp
                    ${dest} = storage->GetLabelIx(${source}.name);
                    cpp<#))
+
     (schema_type_map "std::vector<std::pair<PropertyIx, common::SchemaType>>"
       :slk-save #'slk-save-property-map
       :slk-load #'slk-load-property-map
       :clone #'clone-schema-property-vector
+      :scope :public)
+
+   (schema_config_map "std::unordered_map<common::SchemaConfigParams, int64_t>"
+      :clone #'clone-map
       :scope :public))
 
   (:public

src/storage/v3/config.hpp

@@ -30,6 +30,10 @@ struct Config {
     io::Duration reclamation_interval{};
   } gc;
 
+  struct Split {
+    uint64_t max_shard_vertex_size{500'000};
+  } split;
+
   struct Items {
     bool properties_on_edges{true};
   } items;

src/storage/v3/delta.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -13,12 +13,14 @@
 
 #include <cstdint>
 #include <memory>
+
 #include "storage/v3/edge_ref.hpp"
 #include "storage/v3/id_types.hpp"
 #include "storage/v3/property_value.hpp"
 #include "storage/v3/vertex.hpp"
 #include "storage/v3/vertex_id.hpp"
 #include "utils/logging.hpp"
+#include "utils/synchronized.hpp"
 
 namespace memgraph::storage::v3 {
 
@@ -27,6 +29,11 @@ struct Edge;
 struct Delta;
 struct CommitInfo;
 
+inline uint64_t GetNextDeltaId() noexcept {
+  static utils::Synchronized<uint64_t, utils::SpinLock> delta_id{0};
+  return delta_id.WithLock([](auto &id) { return id++; });
+}
+
 // This class stores one of three pointers (`Delta`, `Vertex` and `Edge`)
 // without using additional memory for storing the type. The type is stored in
 // the pointer itself in the lower bits. All of those structures contain large
@@ -158,46 +165,54 @@ struct Delta {
   struct RemoveInEdgeTag {};
   struct RemoveOutEdgeTag {};
 
-  Delta(DeleteObjectTag /*unused*/, CommitInfo *commit_info, uint64_t command_id)
-      : action(Action::DELETE_OBJECT), commit_info(commit_info), command_id(command_id) {}
+  Delta(DeleteObjectTag /*unused*/, CommitInfo *commit_info, uint64_t delta_id, uint64_t command_id)
+      : action(Action::DELETE_OBJECT), id(delta_id), commit_info(commit_info), command_id(command_id) {}
 
-  Delta(RecreateObjectTag /*unused*/, CommitInfo *commit_info, uint64_t command_id)
-      : action(Action::RECREATE_OBJECT), commit_info(commit_info), command_id(command_id) {}
+  Delta(RecreateObjectTag /*unused*/, CommitInfo *commit_info, uint64_t delta_id, uint64_t command_id)
+      : action(Action::RECREATE_OBJECT), id(delta_id), commit_info(commit_info), command_id(command_id) {}
 
-  Delta(AddLabelTag /*unused*/, LabelId label, CommitInfo *commit_info, uint64_t command_id)
-      : action(Action::ADD_LABEL), commit_info(commit_info), command_id(command_id), label(label) {}
+  Delta(AddLabelTag /*unused*/, LabelId label, CommitInfo *commit_info, uint64_t delta_id, uint64_t command_id)
+      : action(Action::ADD_LABEL), id(delta_id), commit_info(commit_info), command_id(command_id), label(label) {}
 
-  Delta(RemoveLabelTag /*unused*/, LabelId label, CommitInfo *commit_info, uint64_t command_id)
-      : action(Action::REMOVE_LABEL), commit_info(commit_info), command_id(command_id), label(label) {}
+  Delta(RemoveLabelTag /*unused*/, LabelId label, CommitInfo *commit_info, uint64_t delta_id, uint64_t command_id)
+      : action(Action::REMOVE_LABEL), id(delta_id), commit_info(commit_info), command_id(command_id), label(label) {}
 
   Delta(SetPropertyTag /*unused*/, PropertyId key, const PropertyValue &value, CommitInfo *commit_info,
-        uint64_t command_id)
-      : action(Action::SET_PROPERTY), commit_info(commit_info), command_id(command_id), property({key, value}) {}
+        uint64_t delta_id, uint64_t command_id)
+      : action(Action::SET_PROPERTY),
+        id(delta_id),
+        commit_info(commit_info),
+        command_id(command_id),
+        property({key, value}) {}
 
   Delta(AddInEdgeTag /*unused*/, EdgeTypeId edge_type, VertexId vertex_id, EdgeRef edge, CommitInfo *commit_info,
-        uint64_t command_id)
+        uint64_t delta_id, uint64_t command_id)
       : action(Action::ADD_IN_EDGE),
+        id(delta_id),
         commit_info(commit_info),
         command_id(command_id),
         vertex_edge({edge_type, std::move(vertex_id), edge}) {}
 
   Delta(AddOutEdgeTag /*unused*/, EdgeTypeId edge_type, VertexId vertex_id, EdgeRef edge, CommitInfo *commit_info,
-        uint64_t command_id)
+        uint64_t delta_id, uint64_t command_id)
       : action(Action::ADD_OUT_EDGE),
+        id(delta_id),
         commit_info(commit_info),
         command_id(command_id),
         vertex_edge({edge_type, std::move(vertex_id), edge}) {}
 
   Delta(RemoveInEdgeTag /*unused*/, EdgeTypeId edge_type, VertexId vertex_id, EdgeRef edge, CommitInfo *commit_info,
-        uint64_t command_id)
+        uint64_t delta_id, uint64_t command_id)
       : action(Action::REMOVE_IN_EDGE),
+        id(delta_id),
         commit_info(commit_info),
         command_id(command_id),
         vertex_edge({edge_type, std::move(vertex_id), edge}) {}
 
   Delta(RemoveOutEdgeTag /*unused*/, EdgeTypeId edge_type, VertexId vertex_id, EdgeRef edge, CommitInfo *commit_info,
-        uint64_t command_id)
+        uint64_t delta_id, uint64_t command_id)
       : action(Action::REMOVE_OUT_EDGE),
+        id(delta_id),
         commit_info(commit_info),
         command_id(command_id),
         vertex_edge({edge_type, std::move(vertex_id), edge}) {}
@@ -226,8 +241,10 @@ struct Delta {
     }
   }
 
-  Action action;
+  friend bool operator==(const Delta &lhs, const Delta &rhs) noexcept { return lhs.id == rhs.id; }
 
+  Action action;
+  uint64_t id;
   // TODO: optimize with in-place copy
   CommitInfo *commit_info;
   uint64_t command_id;

src/storage/v3/indices.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -325,7 +325,7 @@ void LabelIndex::RemoveObsoleteEntries(const uint64_t clean_up_before_timestamp)
   }
 }
 
-LabelIndex::Iterable::Iterator::Iterator(Iterable *self, LabelIndexContainer::iterator index_iterator)
+LabelIndex::Iterable::Iterator::Iterator(Iterable *self, IndexContainer::iterator index_iterator)
     : self_(self),
       index_iterator_(index_iterator),
       current_vertex_accessor_(nullptr, nullptr, nullptr, self_->config_, *self_->vertex_validator_),
@@ -353,7 +353,7 @@ void LabelIndex::Iterable::Iterator::AdvanceUntilValid() {
   }
 }
 
-LabelIndex::Iterable::Iterable(LabelIndexContainer &index_container, LabelId label, View view, Transaction *transaction,
+LabelIndex::Iterable::Iterable(IndexContainer &index_container, LabelId label, View view, Transaction *transaction,
                                Indices *indices, Config::Items config, const VertexValidator &vertex_validator)
     : index_container_(&index_container),
       label_(label),
@@ -465,7 +465,7 @@ void LabelPropertyIndex::RemoveObsoleteEntries(const uint64_t clean_up_before_ti
   }
 }
 
-LabelPropertyIndex::Iterable::Iterator::Iterator(Iterable *self, LabelPropertyIndexContainer::iterator index_iterator)
+LabelPropertyIndex::Iterable::Iterator::Iterator(Iterable *self, IndexContainer::iterator index_iterator)
     : self_(self),
       index_iterator_(index_iterator),
       current_vertex_accessor_(nullptr, nullptr, nullptr, self_->config_, *self_->vertex_validator_),
@@ -526,7 +526,7 @@ const PropertyValue kSmallestMap = PropertyValue(std::map<std::string, PropertyV
 const PropertyValue kSmallestTemporalData =
     PropertyValue(TemporalData{static_cast<TemporalType>(0), std::numeric_limits<int64_t>::min()});
 
-LabelPropertyIndex::Iterable::Iterable(LabelPropertyIndexContainer &index_container, LabelId label, PropertyId property,
+LabelPropertyIndex::Iterable::Iterable(IndexContainer &index_container, LabelId label, PropertyId property,
                                        const std::optional<utils::Bound<PropertyValue>> &lower_bound,
                                        const std::optional<utils::Bound<PropertyValue>> &upper_bound, View view,
                                        Transaction *transaction, Indices *indices, Config::Items config,

src/storage/v3/indices.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -18,6 +18,8 @@
 #include <utility>
 
 #include "storage/v3/config.hpp"
+#include "storage/v3/id_types.hpp"
+#include "storage/v3/key_store.hpp"
 #include "storage/v3/property_value.hpp"
 #include "storage/v3/transaction.hpp"
 #include "storage/v3/vertex_accessor.hpp"
@@ -40,12 +42,18 @@ class LabelIndex {
     bool operator==(const Entry &rhs) const { return vertex == rhs.vertex && timestamp == rhs.timestamp; }
   };
 
+  using IndexType = LabelId;
+
  public:
-  using LabelIndexContainer = std::set<Entry>;
+  using IndexContainer = std::set<Entry>;
 
   LabelIndex(Indices *indices, Config::Items config, const VertexValidator &vertex_validator)
       : indices_(indices), config_(config), vertex_validator_{&vertex_validator} {}
 
+  LabelIndex(Indices *indices, Config::Items config, const VertexValidator &vertex_validator,
+             std::map<LabelId, IndexContainer> &data)
+      : index_{std::move(data)}, indices_(indices), config_(config), vertex_validator_{&vertex_validator} {}
+
   /// @throw std::bad_alloc
   void UpdateOnAddLabel(LabelId label, Vertex *vertex, const Transaction &tx);
 
@@ -63,12 +71,12 @@ class LabelIndex {
 
   class Iterable {
    public:
-    Iterable(LabelIndexContainer &index_container, LabelId label, View view, Transaction *transaction, Indices *indices,
+    Iterable(IndexContainer &index_container, LabelId label, View view, Transaction *transaction, Indices *indices,
              Config::Items config, const VertexValidator &vertex_validator);
 
     class Iterator {
      public:
-      Iterator(Iterable *self, LabelIndexContainer::iterator index_iterator);
+      Iterator(Iterable *self, IndexContainer::iterator index_iterator);
 
       VertexAccessor operator*() const { return current_vertex_accessor_; }
 
@@ -81,7 +89,7 @@ class LabelIndex {
       void AdvanceUntilValid();
 
       Iterable *self_;
-      LabelIndexContainer::iterator index_iterator_;
+      IndexContainer::iterator index_iterator_;
       VertexAccessor current_vertex_accessor_;
       Vertex *current_vertex_;
     };
@@ -90,7 +98,7 @@ class LabelIndex {
     Iterator end() { return {this, index_container_->end()}; }
 
    private:
-    LabelIndexContainer *index_container_;
+    IndexContainer *index_container_;
     LabelId label_;
     View view_;
     Transaction *transaction_;
@@ -114,8 +122,35 @@ class LabelIndex {
 
   void Clear() { index_.clear(); }
 
+  std::map<IndexType, IndexContainer> SplitIndexEntries(const PrimaryKey &split_key) {
+    if (index_.empty()) {
+      return {};
+    }
+
+    // Cloned index entries will contain new index entry iterators, but old
+    // vertices address which need to be adjusted after extracting vertices
+    std::map<IndexType, IndexContainer> cloned_indices;
+    for (auto &[index_type_val, index] : index_) {
+      auto entry_it = index.begin();
+      auto &cloned_indices_container = cloned_indices[index_type_val];
+      while (entry_it != index.end()) {
+        // We need to save the next pointer since the current one will be
+        // invalidated after extract
+        auto next_entry_it = std::next(entry_it);
+        if (entry_it->vertex->first > split_key) {
+          [[maybe_unused]] const auto &[inserted_entry_it, inserted, node] =
+              cloned_indices_container.insert(index.extract(entry_it));
+          MG_ASSERT(inserted, "Failed to extract index entry!");
+        }
+        entry_it = next_entry_it;
+      }
+    }
+
+    return cloned_indices;
+  }
+
  private:
-  std::map<LabelId, LabelIndexContainer> index_;
+  std::map<LabelId, IndexContainer> index_;
   Indices *indices_;
   Config::Items config_;
   const VertexValidator *vertex_validator_;
@@ -133,9 +168,10 @@ class LabelPropertyIndex {
     bool operator<(const PropertyValue &rhs) const;
     bool operator==(const PropertyValue &rhs) const;
   };
+  using IndexType = std::pair<LabelId, PropertyId>;
 
  public:
-  using LabelPropertyIndexContainer = std::set<Entry>;
+  using IndexContainer = std::set<Entry>;
 
   LabelPropertyIndex(Indices *indices, Config::Items config, const VertexValidator &vertex_validator)
       : indices_(indices), config_(config), vertex_validator_{&vertex_validator} {}
@@ -159,14 +195,14 @@ class LabelPropertyIndex {
 
   class Iterable {
    public:
-    Iterable(LabelPropertyIndexContainer &index_container, LabelId label, PropertyId property,
+    Iterable(IndexContainer &index_container, LabelId label, PropertyId property,
              const std::optional<utils::Bound<PropertyValue>> &lower_bound,
              const std::optional<utils::Bound<PropertyValue>> &upper_bound, View view, Transaction *transaction,
              Indices *indices, Config::Items config, const VertexValidator &vertex_validator);
 
     class Iterator {
      public:
-      Iterator(Iterable *self, LabelPropertyIndexContainer::iterator index_iterator);
+      Iterator(Iterable *self, IndexContainer::iterator index_iterator);
 
       VertexAccessor operator*() const { return current_vertex_accessor_; }
 
@@ -179,7 +215,7 @@ class LabelPropertyIndex {
       void AdvanceUntilValid();
 
       Iterable *self_;
-      LabelPropertyIndexContainer::iterator index_iterator_;
+      IndexContainer::iterator index_iterator_;
       VertexAccessor current_vertex_accessor_;
       Vertex *current_vertex_;
     };
@@ -188,7 +224,7 @@ class LabelPropertyIndex {
     Iterator end();
 
    private:
-    LabelPropertyIndexContainer *index_container_;
+    IndexContainer *index_container_;
     LabelId label_;
     PropertyId property_;
     std::optional<utils::Bound<PropertyValue>> lower_bound_;
@@ -229,8 +265,35 @@ class LabelPropertyIndex {
 
   void Clear() { index_.clear(); }
 
+  std::map<IndexType, IndexContainer> SplitIndexEntries(const PrimaryKey &split_key) {
+    if (index_.empty()) {
+      return {};
+    }
+
+    // Cloned index entries will contain new index entry iterators, but old
+    // vertices address which need to be adjusted after extracting vertices
+    std::map<IndexType, IndexContainer> cloned_indices;
+    for (auto &[index_type_val, index] : index_) {
+      auto entry_it = index.begin();
+      auto &cloned_index_container = cloned_indices[index_type_val];
+      while (entry_it != index.end()) {
+        // We need to save the next pointer since the current one will be
+        // invalidated after extract
+        auto next_entry_it = std::next(entry_it);
+        if (entry_it->vertex->first > split_key) {
+          [[maybe_unused]] const auto &[inserted_entry_it, inserted, node] =
+              cloned_index_container.insert(index.extract(entry_it));
+          MG_ASSERT(inserted, "Failed to extract index entry!");
+        }
+        entry_it = next_entry_it;
+      }
+    }
+
+    return cloned_indices;
+  }
+
  private:
-  std::map<std::pair<LabelId, PropertyId>, LabelPropertyIndexContainer> index_;
+  std::map<std::pair<LabelId, PropertyId>, IndexContainer> index_;
   Indices *indices_;
   Config::Items config_;
   const VertexValidator *vertex_validator_;

src/storage/v3/mvcc.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -108,7 +108,7 @@ inline bool PrepareForWrite(Transaction *transaction, TObj *object) {
 /// a `DELETE_OBJECT` delta).
 /// @throw std::bad_alloc
 inline Delta *CreateDeleteObjectDelta(Transaction *transaction) {
-  return &transaction->deltas.emplace_back(Delta::DeleteObjectTag(), transaction->commit_info.get(),
+  return &transaction->deltas.emplace_back(Delta::DeleteObjectTag(), transaction->commit_info.get(), GetNextDeltaId(),
                                            transaction->command_id);
 }
 
@@ -119,7 +119,7 @@ template <typename TObj, class... Args>
 requires utils::SameAsAnyOf<TObj, Edge, Vertex>
 inline void CreateAndLinkDelta(Transaction *transaction, TObj *object, Args &&...args) {
   auto delta = &transaction->deltas.emplace_back(std::forward<Args>(args)..., transaction->commit_info.get(),
-                                                 transaction->command_id);
+                                                 GetNextDeltaId(), transaction->command_id);
   auto *delta_holder = GetDeltaHolder(object);
 
   // The operations are written in such order so that both `next` and `prev`

src/storage/v3/request_helper.cpp

@@ -44,7 +44,7 @@ struct VertexIdCmpr {
 };
 
 std::optional<std::map<PropertyId, Value>> PrimaryKeysFromAccessor(const VertexAccessor &acc, View view,
-                                                                   const Schemas::Schema &schema) {
+                                                                   const Schema &schema) {
   std::map<PropertyId, Value> ret;
   auto props = acc.Properties(view);
   auto maybe_pk = acc.PrimaryKey(view);
@@ -53,9 +53,9 @@ std::optional<std::map<PropertyId, Value>> PrimaryKeysFromAccessor(const VertexA
     return std::nullopt;
   }
   auto &pk = maybe_pk.GetValue();
-  MG_ASSERT(schema.second.size() == pk.size(), "PrimaryKey size does not match schema!");
-  for (size_t i{0}; i < schema.second.size(); ++i) {
-    ret.emplace(schema.second[i].property_id, FromPropertyValueToValue(std::move(pk[i])));
+  MG_ASSERT(schema.properties.size() == pk.size(), "PrimaryKey size does not match schema!");
+  for (size_t i{0}; i < schema.properties.size(); ++i) {
+    ret.emplace(schema.properties[i].property_id, FromPropertyValueToValue(std::move(pk[i])));
   }
 
   return ret;
@@ -82,8 +82,7 @@ ShardResult<std::vector<msgs::Label>> FillUpSourceVertexSecondaryLabels(const st
 
 ShardResult<std::map<PropertyId, Value>> FillUpSourceVertexProperties(const std::optional<VertexAccessor> &v_acc,
                                                                       const msgs::ExpandOneRequest &req,
-                                                                      storage::v3::View view,
-                                                                      const Schemas::Schema &schema) {
+                                                                      storage::v3::View view, const Schema &schema) {
   std::map<PropertyId, Value> src_vertex_properties;
 
   if (!req.src_vertex_properties) {
@@ -236,7 +235,7 @@ std::vector<TypedValue> EvaluateEdgeExpressions(DbAccessor &dba, const VertexAcc
 }
 
 ShardResult<std::map<PropertyId, Value>> CollectAllPropertiesFromAccessor(const VertexAccessor &acc, View view,
-                                                                          const Schemas::Schema &schema) {
+                                                                          const Schema &schema) {
   auto ret = impl::CollectAllPropertiesImpl<VertexAccessor>(acc, view);
   if (ret.HasError()) {
     return ret.GetError();
@@ -321,15 +320,12 @@ EdgeFiller InitializeEdgeFillerFunction(const msgs::ExpandOneRequest &req) {
         value_properties.insert(std::make_pair(prop_key, FromPropertyValueToValue(std::move(prop_val))));
       }
       using EdgeWithAllProperties = msgs::ExpandOneResultRow::EdgeWithAllProperties;
-
+      EdgeWithAllProperties edges{ToMsgsVertexId(edge.From()), msgs::EdgeType{edge.EdgeType()}, edge.Gid().AsUint(),
+                                  std::move(value_properties)};
       if (is_in_edge) {
-        result_row.in_edges_with_all_properties.push_back(
-            EdgeWithAllProperties{ToMsgsVertexId(edge.From()), msgs::EdgeType{edge.EdgeType()}, edge.Gid().AsUint(),
-                                  std::move(value_properties)});
+        result_row.in_edges_with_all_properties.push_back(std::move(edges));
       } else {
-        result_row.out_edges_with_all_properties.push_back(
-            EdgeWithAllProperties{ToMsgsVertexId(edge.To()), msgs::EdgeType{edge.EdgeType()}, edge.Gid().AsUint(),
-                                  std::move(value_properties)});
+        result_row.out_edges_with_all_properties.push_back(std::move(edges));
       }
       return {};
     };
@@ -349,15 +345,12 @@ EdgeFiller InitializeEdgeFillerFunction(const msgs::ExpandOneRequest &req) {
         value_properties.emplace_back(FromPropertyValueToValue(std::move(property_result.GetValue())));
       }
       using EdgeWithSpecificProperties = msgs::ExpandOneResultRow::EdgeWithSpecificProperties;
-
+      EdgeWithSpecificProperties edges{ToMsgsVertexId(edge.From()), msgs::EdgeType{edge.EdgeType()},
+                                       edge.Gid().AsUint(), std::move(value_properties)};
       if (is_in_edge) {
-        result_row.in_edges_with_specific_properties.push_back(
-            EdgeWithSpecificProperties{ToMsgsVertexId(edge.From()), msgs::EdgeType{edge.EdgeType()},
-                                       edge.Gid().AsUint(), std::move(value_properties)});
+        result_row.in_edges_with_specific_properties.push_back(std::move(edges));
       } else {
-        result_row.out_edges_with_specific_properties.push_back(
-            EdgeWithSpecificProperties{ToMsgsVertexId(edge.To()), msgs::EdgeType{edge.EdgeType()}, edge.Gid().AsUint(),
-                                       std::move(value_properties)});
+        result_row.out_edges_with_specific_properties.push_back(std::move(edges));
       }
       return {};
     };
@@ -386,7 +379,7 @@ bool FilterOnEdge(DbAccessor &dba, const storage::v3::VertexAccessor &v_acc, con
 ShardResult<msgs::ExpandOneResultRow> GetExpandOneResult(
     Shard::Accessor &acc, msgs::VertexId src_vertex, const msgs::ExpandOneRequest &req,
     const EdgeUniquenessFunction &maybe_filter_based_on_edge_uniqueness, const EdgeFiller &edge_filler,
-    const Schemas::Schema &schema) {
+    const Schema &schema) {
   /// Fill up source vertex
   const auto primary_key = ConvertPropertyVector(src_vertex.second);
   auto v_acc = acc.FindVertex(primary_key, View::NEW);
@@ -429,7 +422,7 @@ ShardResult<msgs::ExpandOneResultRow> GetExpandOneResult(
     VertexAccessor v_acc, msgs::VertexId src_vertex, const msgs::ExpandOneRequest &req,
     std::vector<EdgeAccessor> in_edge_accessors, std::vector<EdgeAccessor> out_edge_accessors,
     const EdgeUniquenessFunction &maybe_filter_based_on_edge_uniqueness, const EdgeFiller &edge_filler,
-    const Schemas::Schema &schema) {
+    const Schema &schema) {
   /// Fill up source vertex
   msgs::Vertex source_vertex = {.id = src_vertex};
   auto maybe_secondary_labels = FillUpSourceVertexSecondaryLabels(v_acc, req);

src/storage/v3/request_helper.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -217,7 +217,7 @@ ShardResult<std::map<PropertyId, Value>> CollectSpecificPropertiesFromAccessor(c
 }
 
 ShardResult<std::map<PropertyId, Value>> CollectAllPropertiesFromAccessor(const VertexAccessor &acc, View view,
-                                                                          const Schemas::Schema &schema);
+                                                                          const Schema &schema);
 namespace impl {
 template <PropertiesAccessor TAccessor>
 ShardResult<std::map<PropertyId, Value>> CollectAllPropertiesImpl(const TAccessor &acc, View view) {
@@ -249,11 +249,11 @@ EdgeFiller InitializeEdgeFillerFunction(const msgs::ExpandOneRequest &req);
 ShardResult<msgs::ExpandOneResultRow> GetExpandOneResult(
     Shard::Accessor &acc, msgs::VertexId src_vertex, const msgs::ExpandOneRequest &req,
     const EdgeUniquenessFunction &maybe_filter_based_on_edge_uniqueness, const EdgeFiller &edge_filler,
-    const Schemas::Schema &schema);
+    const Schema &schema);
 
 ShardResult<msgs::ExpandOneResultRow> GetExpandOneResult(
     VertexAccessor v_acc, msgs::VertexId src_vertex, const msgs::ExpandOneRequest &req,
     std::vector<EdgeAccessor> in_edge_accessors, std::vector<EdgeAccessor> out_edge_accessors,
     const EdgeUniquenessFunction &maybe_filter_based_on_edge_uniqueness, const EdgeFiller &edge_filler,
-    const Schemas::Schema &schema);
+    const Schema &schema);
 }  // namespace memgraph::storage::v3

src/storage/v3/schema_validator.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -44,20 +44,20 @@ ShardResult<void> SchemaValidator::ValidateVertexCreate(LabelId primary_label, c
   }
 
   // Quick size check
-  if (schema->second.size() != primary_properties.size()) {
+  if (schema->properties.size() != primary_properties.size()) {
     return SHARD_ERROR(ErrorCode::SCHEMA_VERTEX_PRIMARY_PROPERTIES_UNDEFINED,
                        "Not all primary properties have been specified for :{} vertex",
                        name_id_mapper_->IdToName(primary_label.AsInt()));
   }
   // Check only properties defined by schema
-  for (size_t i{0}; i < schema->second.size(); ++i) {
+  for (size_t i{0}; i < schema->properties.size(); ++i) {
     // Check schema property type
     if (auto property_schema_type = PropertyTypeToSchemaType(primary_properties[i]);
-        property_schema_type && *property_schema_type != schema->second[i].type) {
+        property_schema_type && *property_schema_type != schema->properties[i].type) {
       return SHARD_ERROR(ErrorCode::SCHEMA_VERTEX_PROPERTY_WRONG_TYPE,
                          "Property {} is of wrong type, expected {}, actual {}",
-                         name_id_mapper_->IdToName(schema->second[i].property_id.AsInt()),
-                         SchemaTypeToString(schema->second[i].type), SchemaTypeToString(*property_schema_type));
+                         name_id_mapper_->IdToName(schema->properties[i].property_id.AsInt()),
+                         SchemaTypeToString(schema->properties[i].type), SchemaTypeToString(*property_schema_type));
     }
   }
 
@@ -72,9 +72,9 @@ ShardResult<void> SchemaValidator::ValidatePropertyUpdate(const LabelId primary_
 
   // Verify that updating property is not part of schema
   if (const auto schema_property = std::ranges::find_if(
-          schema->second,
+          schema->properties,
           [property_id](const auto &schema_property) { return property_id == schema_property.property_id; });
-      schema_property != schema->second.end()) {
+      schema_property != schema->properties.end()) {
     return SHARD_ERROR(ErrorCode::SCHEMA_VERTEX_UPDATE_PRIMARY_KEY,
                        "Cannot update primary property {} of schema on label :{}",
                        name_id_mapper_->IdToName(schema_property->property_id.AsInt()),
@@ -92,7 +92,7 @@ ShardResult<void> SchemaValidator::ValidateLabelUpdate(const LabelId label) cons
   return {};
 }
 
-const Schemas::Schema *SchemaValidator::GetSchema(LabelId label) const { return schemas_->GetSchema(label); }
+const Schema *SchemaValidator::GetSchema(LabelId label) const { return schemas_->GetSchema(label); }
 
 VertexValidator::VertexValidator(const SchemaValidator &schema_validator, const LabelId primary_label)
     : schema_validator{&schema_validator}, primary_label_{primary_label} {}

src/storage/v3/schema_validator.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -33,7 +33,7 @@ class SchemaValidator {
 
   [[nodiscard]] ShardResult<void> ValidateLabelUpdate(LabelId label) const;
 
-  const Schemas::Schema *GetSchema(LabelId label) const;
+  const Schema *GetSchema(LabelId label) const;
 
  private:
   Schemas *schemas_;

src/storage/v3/schemas.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -25,25 +25,26 @@ bool operator==(const SchemaProperty &lhs, const SchemaProperty &rhs) {
 }
 
 Schemas::SchemasList Schemas::ListSchemas() const {
-  Schemas::SchemasList ret;
+  SchemasList ret;
   ret.reserve(schemas_.size());
   std::transform(schemas_.begin(), schemas_.end(), std::back_inserter(ret),
-                 [](const auto &schema_property_type) { return schema_property_type; });
+                 [](const auto &schema_property_type) { return schema_property_type.second; });
   return ret;
 }
 
-const Schemas::Schema *Schemas::GetSchema(const LabelId primary_label) const {
+const Schema *Schemas::GetSchema(const LabelId primary_label) const {
   if (auto schema_map = schemas_.find(primary_label); schema_map != schemas_.end()) {
-    return &*schema_map;
+    return &schema_map->second;
   }
   return nullptr;
 }
 
-bool Schemas::CreateSchema(const LabelId primary_label, const std::vector<SchemaProperty> &schemas_types) {
+bool Schemas::CreateSchema(const LabelId primary_label, const std::vector<SchemaProperty> &schemas_types,
+                           Schema::SchemaConfiguration config) {
   if (schemas_.contains(primary_label)) {
     return false;
   }
-  schemas_.emplace(primary_label, schemas_types);
+  schemas_.insert({primary_label, Schema{.label = primary_label, .properties = schemas_types, .config = config}});
   return true;
 }
 
@@ -51,9 +52,9 @@ bool Schemas::DropSchema(const LabelId primary_label) { return schemas_.erase(pr
 
 bool Schemas::IsPropertyKey(const LabelId primary_label, const PropertyId property_id) const {
   if (const auto schema = schemas_.find(primary_label); schema != schemas_.end()) {
-    return std::ranges::find_if(schema->second, [property_id](const auto &elem) {
+    return std::ranges::find_if(schema->second.properties, [property_id](const auto &elem) {
              return elem.property_id == property_id;
-           }) != schema->second.end();
+           }) != schema->second.properties.end();
   }
   throw utils::BasicException("Schema not found!");
 }

src/storage/v3/schemas.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -11,6 +11,7 @@
 
 #pragma once
 
+#include <cstdint>
 #include <memory>
 #include <optional>
 #include <unordered_map>
@@ -43,12 +44,24 @@ struct SchemaProperty {
   }
 };
 
+struct Schema {
+  LabelId label;
+  std::vector<SchemaProperty> properties;
+  struct SchemaConfiguration {
+    uint64_t replication_factor;
+    uint64_t split_threshold;
+  } config;
+
+  friend bool operator==(const Schema &lhs, const Schema &rhs) noexcept {
+    return lhs.label == rhs.label && lhs.properties == rhs.properties;
+  }
+};
+
 /// Structure that represents a collection of schemas
 /// Schema can be mapped under only one label => primary label
 class Schemas {
  public:
-  using SchemasMap = std::unordered_map<LabelId, std::vector<SchemaProperty>>;
-  using Schema = SchemasMap::value_type;
+  using SchemasMap = std::unordered_map<LabelId, Schema>;
   using SchemasList = std::vector<Schema>;
 
   Schemas() = default;
@@ -64,7 +77,8 @@ class Schemas {
 
   // Returns true if it was successfully created or false if the schema
   // already exists
-  [[nodiscard]] bool CreateSchema(LabelId primary_label, const std::vector<SchemaProperty> &schemas_types);
+  [[nodiscard]] bool CreateSchema(LabelId primary_label, const std::vector<SchemaProperty> &schemas_types,
+                                  Schema::SchemaConfiguration config = {});
 
   // Returns true if it was successfully dropped or false if the schema
   // does not exist

src/storage/v3/shard.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -18,14 +18,14 @@
 #include <memory>
 #include <mutex>
 #include <optional>
-#include <variant>
 
 #include <bits/ranges_algo.h>
-#include <gflags/gflags.h>
 #include <spdlog/spdlog.h>
 
 #include "io/network/endpoint.hpp"
 #include "io/time.hpp"
+#include "storage/v3/delta.hpp"
+#include "storage/v3/edge.hpp"
 #include "storage/v3/edge_accessor.hpp"
 #include "storage/v3/id_types.hpp"
 #include "storage/v3/indices.hpp"
@@ -35,6 +35,7 @@
 #include "storage/v3/property_value.hpp"
 #include "storage/v3/result.hpp"
 #include "storage/v3/schema_validator.hpp"
+#include "storage/v3/schemas.hpp"
 #include "storage/v3/transaction.hpp"
 #include "storage/v3/vertex.hpp"
 #include "storage/v3/vertex_accessor.hpp"
@@ -333,15 +334,69 @@ Shard::Shard(const LabelId primary_label, const PrimaryKey min_primary_key,
       indices_{config.items, vertex_validator_},
       isolation_level_{config.transaction.isolation_level},
       config_{config},
-      uuid_{utils::GenerateUUID()},
-      epoch_id_{utils::GenerateUUID()},
-      global_locker_{file_retainer_.AddLocker()} {
+      shard_splitter_(primary_label, vertices_, edges_, start_logical_id_to_transaction_, indices_, config_, schema,
+                      name_id_mapper_) {
   CreateSchema(primary_label_, schema);
   StoreMapping(std::move(id_to_name));
 }
 
+Shard::Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
+             std::vector<SchemaProperty> schema, VertexContainer &&vertices, EdgeContainer &&edges,
+             std::map<uint64_t, std::unique_ptr<Transaction>> &&start_logical_id_to_transaction, const Config &config,
+             const std::unordered_map<uint64_t, std::string> &id_to_name, const uint64_t shard_version)
+    : primary_label_{primary_label},
+      min_primary_key_{min_primary_key},
+      max_primary_key_{max_primary_key},
+      vertices_(std::move(vertices)),
+      edges_(std::move(edges)),
+      shard_version_(shard_version),
+      schema_validator_{schemas_, name_id_mapper_},
+      vertex_validator_{schema_validator_, primary_label},
+      indices_{config.items, vertex_validator_},
+      isolation_level_{config.transaction.isolation_level},
+      config_{config},
+      start_logical_id_to_transaction_(std::move(start_logical_id_to_transaction)),
+      shard_splitter_(primary_label, vertices_, edges_, start_logical_id_to_transaction_, indices_, config_, schema,
+                      name_id_mapper_) {
+  CreateSchema(primary_label_, schema);
+  StoreMapping(id_to_name);
+}
+
+Shard::Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
+             std::vector<SchemaProperty> schema, VertexContainer &&vertices,
+             std::map<uint64_t, std::unique_ptr<Transaction>> &&start_logical_id_to_transaction, const Config &config,
+             const std::unordered_map<uint64_t, std::string> &id_to_name, const uint64_t shard_version)
+    : primary_label_{primary_label},
+      min_primary_key_{min_primary_key},
+      max_primary_key_{max_primary_key},
+      vertices_(std::move(vertices)),
+      shard_version_(shard_version),
+      schema_validator_{schemas_, name_id_mapper_},
+      vertex_validator_{schema_validator_, primary_label},
+      indices_{config.items, vertex_validator_},
+      isolation_level_{config.transaction.isolation_level},
+      config_{config},
+      start_logical_id_to_transaction_(std::move(start_logical_id_to_transaction)),
+      shard_splitter_(primary_label, vertices_, edges_, start_logical_id_to_transaction_, indices_, config_, schema,
+                      name_id_mapper_) {
+  CreateSchema(primary_label_, schema);
+  StoreMapping(id_to_name);
+}
+
 Shard::~Shard() {}
 
+std::unique_ptr<Shard> Shard::FromSplitData(SplitData &&split_data) {
+  if (split_data.config.items.properties_on_edges) [[likely]] {
+    return std::make_unique<Shard>(split_data.primary_label, split_data.min_primary_key, split_data.max_primary_key,
+                                   split_data.schema, std::move(split_data.vertices), std::move(*split_data.edges),
+                                   std::move(split_data.transactions), split_data.config, split_data.id_to_name,
+                                   split_data.shard_version);
+  }
+  return std::make_unique<Shard>(split_data.primary_label, split_data.min_primary_key, split_data.max_primary_key,
+                                 split_data.schema, std::move(split_data.vertices), std::move(split_data.transactions),
+                                 split_data.config, split_data.id_to_name, split_data.shard_version);
+}
+
 Shard::Accessor::Accessor(Shard &shard, Transaction &transaction)
     : shard_(&shard), transaction_(&transaction), config_(shard_->config_.items) {}
 
@@ -349,8 +404,6 @@ ShardResult<VertexAccessor> Shard::Accessor::CreateVertexAndValidate(
     const std::vector<LabelId> &labels, const PrimaryKey &primary_properties,
     const std::vector<std::pair<PropertyId, PropertyValue>> &properties) {
   OOMExceptionEnabler oom_exception;
-  const auto schema = shard_->GetSchema(shard_->primary_label_)->second;
-
   auto maybe_schema_violation =
       GetSchemaValidator().ValidateVertexCreate(shard_->primary_label_, labels, primary_properties);
   if (maybe_schema_violation.HasError()) {
@@ -436,7 +489,7 @@ ShardResult<std::optional<std::pair<VertexAccessor, std::vector<EdgeAccessor>>>>
   }
 
   std::vector<EdgeAccessor> deleted_edges;
-  const VertexId vertex_id{shard_->primary_label_, *vertex->PrimaryKey(View::OLD)};  // TODO Replace
+  const VertexId vertex_id{shard_->primary_label_, *vertex->PrimaryKey(View::OLD)};
   for (const auto &item : in_edges) {
     auto [edge_type, from_vertex, edge] = item;
     EdgeAccessor e(edge, edge_type, from_vertex, vertex_id, transaction_, &shard_->indices_, config_);
@@ -853,8 +906,8 @@ bool Shard::CreateIndex(LabelId label, const std::optional<uint64_t> /*desired_c
 bool Shard::CreateIndex(LabelId label, PropertyId property,
                         const std::optional<uint64_t> /*desired_commit_timestamp*/) {
   // TODO(jbajic) response should be different when index conflicts with schema
-  if (label == primary_label_ && schemas_.GetSchema(primary_label_)->second.size() == 1 &&
-      schemas_.GetSchema(primary_label_)->second[0].property_id == property) {
+  if (label == primary_label_ && schemas_.GetSchema(primary_label_)->properties.size() == 1 &&
+      schemas_.GetSchema(primary_label_)->properties[0].property_id == property) {
     // Index already exists on primary key
     return false;
   }
@@ -880,10 +933,11 @@ const SchemaValidator &Shard::Accessor::GetSchemaValidator() const { return shar
 
 SchemasInfo Shard::ListAllSchemas() const { return {schemas_.ListSchemas()}; }
 
-const Schemas::Schema *Shard::GetSchema(const LabelId primary_label) const { return schemas_.GetSchema(primary_label); }
+const Schema *Shard::GetSchema(const LabelId primary_label) const { return schemas_.GetSchema(primary_label); }
 
-bool Shard::CreateSchema(const LabelId primary_label, const std::vector<SchemaProperty> &schemas_types) {
-  return schemas_.CreateSchema(primary_label, schemas_types);
+bool Shard::CreateSchema(const LabelId primary_label, const std::vector<SchemaProperty> &schemas_types,
+                         Schema::SchemaConfiguration config) {
+  return schemas_.CreateSchema(primary_label, schemas_types, config);
 }
 
 bool Shard::DropSchema(const LabelId primary_label) { return schemas_.DropSchema(primary_label); }
@@ -1048,6 +1102,24 @@ void Shard::StoreMapping(std::unordered_map<uint64_t, std::string> id_to_name) {
   name_id_mapper_.StoreMapping(std::move(id_to_name));
 }
 
+std::optional<SplitInfo> Shard::ShouldSplit() const noexcept {
+  if (vertices_.size() > config_.split.max_shard_vertex_size) {
+    auto mid_elem = vertices_.begin();
+    // TODO(tyler) the first time we calculate the split point, we should store it so that we don't have to
+    // iterate over half of the entire index each time Cron is run until the split succeeds.
+    std::ranges::advance(mid_elem, static_cast<VertexContainer::difference_type>(vertices_.size() / 2));
+    return SplitInfo{mid_elem->first, shard_version_};
+  }
+  return std::nullopt;
+}
+
+SplitData Shard::PerformSplit(const PrimaryKey &split_key, const uint64_t shard_version) {
+  shard_version_ = shard_version;
+  const auto old_max_key = max_primary_key_;
+  max_primary_key_ = split_key;
+  return shard_splitter_.SplitShard(split_key, old_max_key, shard_version);
+}
+
 bool Shard::IsVertexBelongToShard(const VertexId &vertex_id) const {
   return vertex_id.primary_label == primary_label_ && vertex_id.primary_key >= min_primary_key_ &&
          (!max_primary_key_.has_value() || vertex_id.primary_key < *max_primary_key_);

src/storage/v3/shard.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -14,6 +14,7 @@
 #include <cstdint>
 #include <filesystem>
 #include <map>
+#include <memory>
 #include <numeric>
 #include <optional>
 #include <shared_mutex>
@@ -37,6 +38,7 @@
 #include "storage/v3/result.hpp"
 #include "storage/v3/schema_validator.hpp"
 #include "storage/v3/schemas.hpp"
+#include "storage/v3/splitter.hpp"
 #include "storage/v3/transaction.hpp"
 #include "storage/v3/vertex.hpp"
 #include "storage/v3/vertex_accessor.hpp"
@@ -174,6 +176,11 @@ struct SchemasInfo {
   Schemas::SchemasList schemas;
 };
 
+struct SplitInfo {
+  PrimaryKey split_point;
+  uint64_t shard_version;
+};
+
 /// Structure used to return information about the storage.
 struct StorageInfo {
   uint64_t vertex_count;
@@ -186,9 +193,19 @@ class Shard final {
  public:
   /// @throw std::system_error
   /// @throw std::bad_alloc
-  explicit Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
-                 std::vector<SchemaProperty> schema, Config config = Config(),
-                 std::unordered_map<uint64_t, std::string> id_to_name = {});
+  Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
+        std::vector<SchemaProperty> schema, Config config = Config(),
+        std::unordered_map<uint64_t, std::string> id_to_name = {});
+
+  Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
+        std::vector<SchemaProperty> schema, VertexContainer &&vertices, EdgeContainer &&edges,
+        std::map<uint64_t, std::unique_ptr<Transaction>> &&start_logical_id_to_transaction, const Config &config,
+        const std::unordered_map<uint64_t, std::string> &id_to_name, uint64_t shard_version);
+
+  Shard(LabelId primary_label, PrimaryKey min_primary_key, std::optional<PrimaryKey> max_primary_key,
+        std::vector<SchemaProperty> schema, VertexContainer &&vertices,
+        std::map<uint64_t, std::unique_ptr<Transaction>> &&start_logical_id_to_transaction, const Config &config,
+        const std::unordered_map<uint64_t, std::string> &id_to_name, uint64_t shard_version);
 
   Shard(const Shard &) = delete;
   Shard(Shard &&) noexcept = delete;
@@ -196,6 +213,8 @@ class Shard final {
   Shard operator=(Shard &&) noexcept = delete;
   ~Shard();
 
+  static std::unique_ptr<Shard> FromSplitData(SplitData &&split_data);
+
   class Accessor final {
    private:
     friend class Shard;
@@ -345,9 +364,10 @@ class Shard final {
 
   SchemasInfo ListAllSchemas() const;
 
-  const Schemas::Schema *GetSchema(LabelId primary_label) const;
+  const Schema *GetSchema(LabelId primary_label) const;
 
-  bool CreateSchema(LabelId primary_label, const std::vector<SchemaProperty> &schemas_types);
+  bool CreateSchema(LabelId primary_label, const std::vector<SchemaProperty> &schemas_types,
+                    Schema::SchemaConfiguration = {});
 
   bool DropSchema(LabelId primary_label);
 
@@ -360,6 +380,10 @@ class Shard final {
 
   void StoreMapping(std::unordered_map<uint64_t, std::string> id_to_name);
 
+  std::optional<SplitInfo> ShouldSplit() const noexcept;
+
+  SplitData PerformSplit(const PrimaryKey &split_key, uint64_t shard_version);
+
  private:
   Transaction &GetTransaction(coordinator::Hlc start_timestamp, IsolationLevel isolation_level);
 
@@ -377,6 +401,7 @@ class Shard final {
   // list is used only when properties are enabled for edges. Because of that we
   // keep a separate count of edges that is always updated.
   uint64_t edge_count_{0};
+  uint64_t shard_version_{0};
 
   SchemaValidator schema_validator_;
   VertexValidator vertex_validator_;
@@ -396,41 +421,10 @@ class Shard final {
   // storage.
   std::list<Gid> deleted_edges_;
 
-  // UUID used to distinguish snapshots and to link snapshots to WALs
-  std::string uuid_;
-  // Sequence number used to keep track of the chain of WALs.
-  uint64_t wal_seq_num_{0};
-
-  // UUID to distinguish different main instance runs for replication process
-  // on SAME storage.
-  // Multiple instances can have same storage UUID and be MAIN at the same time.
-  // We cannot compare commit timestamps of those instances if one of them
-  // becomes the replica of the other so we use epoch_id_ as additional
-  // discriminating property.
-  // Example of this:
-  // We have 2 instances of the same storage, S1 and S2.
-  // S1 and S2 are MAIN and accept their own commits and write them to the WAL.
-  // At the moment when S1 commited a transaction with timestamp 20, and S2
-  // a different transaction with timestamp 15, we change S2's role to REPLICA
-  // and register it on S1.
-  // Without using the epoch_id, we don't know that S1 and S2 have completely
-  // different transactions, we think that the S2 is behind only by 5 commits.
-  std::string epoch_id_;
-  // History of the previous epoch ids.
-  // Each value consists of the epoch id along the last commit belonging to that
-  // epoch.
-  std::deque<std::pair<std::string, uint64_t>> epoch_history_;
-
-  uint64_t wal_unsynced_transactions_{0};
-
-  utils::FileRetainer file_retainer_;
-
-  // Global locker that is used for clients file locking
-  utils::FileRetainer::FileLocker global_locker_;
-
   // Holds all of the (in progress, committed and aborted) transactions that are read or write to this shard, but
   // haven't been cleaned up yet
   std::map<uint64_t, std::unique_ptr<Transaction>> start_logical_id_to_transaction_{};
+  Splitter shard_splitter_;
   bool has_any_transaction_aborted_since_last_gc{false};
 };
 

src/storage/v3/shard_rsm.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -12,11 +12,13 @@
 #pragma once
 
 #include <memory>
+#include <optional>
 #include <variant>
 
 #include <openssl/ec.h>
 #include "query/v2/requests.hpp"
 #include "storage/v3/shard.hpp"
+#include "storage/v3/value_conversions.hpp"
 #include "storage/v3/vertex_accessor.hpp"
 
 namespace memgraph::storage::v3 {
@@ -41,6 +43,19 @@ class ShardRsm {
  public:
   explicit ShardRsm(std::unique_ptr<Shard> &&shard) : shard_(std::move(shard)){};
 
+  std::optional<msgs::SplitInfo> ShouldSplit() const noexcept {
+    auto split_info = shard_->ShouldSplit();
+    if (split_info) {
+      return msgs::SplitInfo{conversions::ConvertValueVector(split_info->split_point), split_info->shard_version};
+    }
+    return std::nullopt;
+  }
+
+  std::unique_ptr<Shard> PerformSplit(msgs::PerformSplitDataInfo perform_split) const noexcept {
+    return Shard::FromSplitData(
+        shard_->PerformSplit(conversions::ConvertPropertyVector(perform_split.split_key), perform_split.shard_version));
+  }
+
   // NOLINTNEXTLINE(readability-convert-member-functions-to-static)
   msgs::ReadResponses Read(msgs::ReadRequests requests) {
     return std::visit([&](auto &&request) mutable { return HandleRead(std::forward<decltype(request)>(request)); },

src/storage/v3/splitter.cpp

@@ -0,0 +1,351 @@
+// Copyright 2023 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "storage/v3/splitter.hpp"
+
+#include <algorithm>
+#include <cstdint>
+#include <map>
+#include <memory>
+#include <optional>
+#include <set>
+
+#include "storage/v3/config.hpp"
+#include "storage/v3/delta.hpp"
+#include "storage/v3/id_types.hpp"
+#include "storage/v3/indices.hpp"
+#include "storage/v3/key_store.hpp"
+#include "storage/v3/name_id_mapper.hpp"
+#include "storage/v3/schemas.hpp"
+#include "storage/v3/shard.hpp"
+#include "storage/v3/transaction.hpp"
+#include "storage/v3/vertex.hpp"
+#include "utils/logging.hpp"
+
+namespace memgraph::storage::v3 {
+
+Splitter::Splitter(const LabelId primary_label, VertexContainer &vertices, EdgeContainer &edges,
+                   std::map<uint64_t, std::unique_ptr<Transaction>> &start_logical_id_to_transaction, Indices &indices,
+                   const Config &config, const std::vector<SchemaProperty> &schema, const NameIdMapper &name_id_mapper)
+    : primary_label_(primary_label),
+      vertices_(vertices),
+      edges_(edges),
+      start_logical_id_to_transaction_(start_logical_id_to_transaction),
+      indices_(indices),
+      config_(config),
+      schema_(schema),
+      name_id_mapper_(name_id_mapper) {}
+
+SplitData Splitter::SplitShard(const PrimaryKey &split_key, const std::optional<PrimaryKey> &max_primary_key,
+                               const uint64_t shard_version) {
+  SplitData data{.primary_label = primary_label_,
+                 .min_primary_key = split_key,
+                 .max_primary_key = max_primary_key,
+                 .schema = schema_,
+                 .config = config_,
+                 .id_to_name = name_id_mapper_.GetIdToNameMap(),
+                 .shard_version = shard_version};
+
+  std::set<uint64_t> collected_transactions_;
+  data.vertices = CollectVertices(data, collected_transactions_, split_key);
+  data.edges = CollectEdges(collected_transactions_, data.vertices, split_key);
+  data.transactions = CollectTransactions(collected_transactions_, data.vertices, *data.edges, split_key);
+
+  return data;
+}
+
+void Splitter::ScanDeltas(std::set<uint64_t> &collected_transactions_, Delta *delta) {
+  while (delta != nullptr) {
+    collected_transactions_.insert(delta->commit_info->start_or_commit_timestamp.logical_id);
+    delta = delta->next;
+  }
+}
+
+VertexContainer Splitter::CollectVertices(SplitData &data, std::set<uint64_t> &collected_transactions_,
+                                          const PrimaryKey &split_key) {
+  data.label_indices = indices_.label_index.SplitIndexEntries(split_key);
+  data.label_property_indices = indices_.label_property_index.SplitIndexEntries(split_key);
+
+  VertexContainer splitted_data;
+  auto split_key_it = vertices_.find(split_key);
+  while (split_key_it != vertices_.end()) {
+    // Go through deltas and pick up transactions start_id/commit_id
+    ScanDeltas(collected_transactions_, split_key_it->second.delta);
+
+    auto next_it = std::next(split_key_it);
+
+    const auto &[splitted_vertex_it, inserted, node] = splitted_data.insert(vertices_.extract(split_key_it->first));
+    MG_ASSERT(inserted, "Failed to extract vertex!");
+
+    split_key_it = next_it;
+  }
+  return splitted_data;
+}
+
+std::optional<EdgeContainer> Splitter::CollectEdges(std::set<uint64_t> &collected_transactions_,
+                                                    const VertexContainer &split_vertices,
+                                                    const PrimaryKey &split_key) {
+  if (!config_.items.properties_on_edges) {
+    return std::nullopt;
+  }
+  EdgeContainer splitted_edges;
+  const auto split_vertex_edges = [&](const auto &edges_ref) {
+    // This is safe since if properties_on_edges is true, the this must be a ptr
+    for (const auto &edge_ref : edges_ref) {
+      auto *edge = std::get<2>(edge_ref).ptr;
+      const auto &other_vtx = std::get<1>(edge_ref);
+      ScanDeltas(collected_transactions_, edge->delta);
+      // Check if src and dest edge are both on splitted shard so we know if we
+      // should remove orphan edge, or make a clone
+      if (other_vtx.primary_key >= split_key) {
+        // Remove edge from shard
+        splitted_edges.insert(edges_.extract(edge->gid));
+      } else {
+        splitted_edges.insert({edge->gid, Edge{edge->gid, edge->delta}});
+      }
+    }
+  };
+
+  for (const auto &vertex : split_vertices) {
+    split_vertex_edges(vertex.second.in_edges);
+    split_vertex_edges(vertex.second.out_edges);
+  }
+  return splitted_edges;
+}
+
+std::map<uint64_t, std::unique_ptr<Transaction>> Splitter::CollectTransactions(
+    const std::set<uint64_t> &collected_transactions_, VertexContainer &cloned_vertices, EdgeContainer &cloned_edges,
+    const PrimaryKey &split_key) {
+  std::map<uint64_t, std::unique_ptr<Transaction>> transactions;
+
+  for (const auto &[commit_start, transaction] : start_logical_id_to_transaction_) {
+    // We need all transaction whose deltas need to be resolved for any of the
+    // entities
+    if (collected_transactions_.contains(transaction->commit_info->start_or_commit_timestamp.logical_id)) {
+      transactions.insert({commit_start, start_logical_id_to_transaction_[commit_start]->Clone()});
+    }
+  }
+
+  // It is necessary to clone all the transactions first so we have new addresses
+  // for deltas, before doing alignment of deltas and prev_ptr
+  AdjustClonedTransactions(transactions, cloned_vertices, cloned_edges, split_key);
+  return transactions;
+}
+
+void PruneDeltas(Transaction &cloned_transaction, std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                 const PrimaryKey &split_key) {
+  // Remove delta chains that don't point to objects on splitted shard
+  auto cloned_delta_it = cloned_transaction.deltas.begin();
+
+  while (cloned_delta_it != cloned_transaction.deltas.end()) {
+    const auto prev = cloned_delta_it->prev.Get();
+    switch (prev.type) {
+      case PreviousPtr::Type::DELTA:
+      case PreviousPtr::Type::NULLPTR:
+        ++cloned_delta_it;
+        break;
+      case PreviousPtr::Type::VERTEX: {
+        if (prev.vertex->first < split_key) {
+          // We can remove this delta chain
+          auto *current_next_delta = cloned_delta_it->next;
+          cloned_delta_it = cloned_transaction.deltas.erase(cloned_delta_it);
+
+          while (current_next_delta != nullptr) {
+            auto *next_delta = current_next_delta->next;
+            // Find next delta transaction delta list
+            auto current_transaction_it = std::ranges::find_if(
+                cloned_transactions,
+                [&start_or_commit_timestamp =
+                     current_next_delta->commit_info->start_or_commit_timestamp](const auto &transaction) {
+                  return transaction.second->start_timestamp == start_or_commit_timestamp ||
+                         transaction.second->commit_info->start_or_commit_timestamp == start_or_commit_timestamp;
+                });
+            MG_ASSERT(current_transaction_it != cloned_transactions.end(), "Error when pruning deltas!");
+            // Remove it
+            current_transaction_it->second->deltas.remove_if(
+                [&current_next_delta = *current_next_delta](const auto &delta) { return delta == current_next_delta; });
+
+            current_next_delta = next_delta;
+          }
+        } else {
+          ++cloned_delta_it;
+        }
+        break;
+      }
+      case PreviousPtr::Type::EDGE:
+        ++cloned_delta_it;
+        break;
+    }
+  }
+}
+
+void Splitter::AdjustClonedTransactions(std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                                        VertexContainer &cloned_vertices, EdgeContainer &cloned_edges,
+                                        const PrimaryKey &split_key) {
+  for (auto &[commit_start, cloned_transaction] : cloned_transactions) {
+    AdjustClonedTransaction(*cloned_transaction, *start_logical_id_to_transaction_[commit_start], cloned_transactions,
+                            cloned_vertices, cloned_edges, split_key);
+  }
+  // Prune deltas whose delta chain points to vertex/edge that should not belong on that shard
+  // Prune must be after ajdust, since next, and prev are not set and we cannot follow the chain
+  for (auto &[commit_start, cloned_transaction] : cloned_transactions) {
+    PruneDeltas(*cloned_transaction, cloned_transactions, split_key);
+  }
+}
+
+inline bool IsDeltaHeadOfChain(const PreviousPtr::Type &delta_type) {
+  return delta_type == PreviousPtr::Type::VERTEX || delta_type == PreviousPtr::Type::EDGE;
+}
+
+bool DoesPrevPtrPointsToSplittedData(const PreviousPtr::Pointer &prev_ptr, const PrimaryKey &split_key) {
+  return prev_ptr.type == PreviousPtr::Type::VERTEX && prev_ptr.vertex->first < split_key;
+}
+
+void Splitter::AdjustClonedTransaction(Transaction &cloned_transaction, const Transaction &transaction,
+                                       std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                                       VertexContainer &cloned_vertices, EdgeContainer &cloned_edges,
+                                       const PrimaryKey & /*split_key*/) {
+  auto delta_it = transaction.deltas.begin();
+  auto cloned_delta_it = cloned_transaction.deltas.begin();
+
+  while (delta_it != transaction.deltas.end()) {
+    // We can safely ignore deltas which are not head of delta chain
+    // Dont' adjust delta chain that points to irrelevant data vertices/edges
+    if (const auto delta_prev = delta_it->prev.Get(); !IsDeltaHeadOfChain(delta_prev.type)) {
+      ++delta_it;
+      ++cloned_delta_it;
+      continue;
+    }
+
+    const auto *delta = &*delta_it;
+    auto *cloned_delta = &*cloned_delta_it;
+    Delta *cloned_delta_prev_ptr = cloned_delta;
+    while (delta->next != nullptr) {
+      AdjustEdgeRef(*cloned_delta, cloned_edges);
+
+      // Align next ptr
+      AdjustDeltaNext(*delta, *cloned_delta, cloned_transactions);
+
+      // Align prev ptr
+      if (cloned_delta_prev_ptr != nullptr) {
+        AdjustDeltaPrevPtr(*delta, *cloned_delta_prev_ptr, cloned_transactions, cloned_vertices, cloned_edges);
+      }
+
+      // TODO Next delta might not belong to the cloned transaction and thats
+      // why we skip this delta of the delta chain
+      if (cloned_delta->next != nullptr) {
+        cloned_delta = cloned_delta->next;
+        cloned_delta_prev_ptr = cloned_delta;
+      } else {
+        cloned_delta_prev_ptr = nullptr;
+      }
+      delta = delta->next;
+    }
+    // Align prev ptr
+    if (cloned_delta_prev_ptr != nullptr) {
+      AdjustDeltaPrevPtr(*delta, *cloned_delta_prev_ptr, cloned_transactions, cloned_vertices, cloned_edges);
+    }
+
+    ++delta_it;
+    ++cloned_delta_it;
+  }
+  MG_ASSERT(delta_it == transaction.deltas.end() && cloned_delta_it == cloned_transaction.deltas.end(),
+            "Both iterators must be exhausted!");
+}
+
+void Splitter::AdjustEdgeRef(Delta &cloned_delta, EdgeContainer &cloned_edges) const {
+  switch (cloned_delta.action) {
+    case Delta::Action::ADD_IN_EDGE:
+    case Delta::Action::ADD_OUT_EDGE:
+    case Delta::Action::REMOVE_IN_EDGE:
+    case Delta::Action::REMOVE_OUT_EDGE: {
+      // Find edge
+      if (config_.items.properties_on_edges) {
+        // Only case when not finding is when the edge is not on splitted shard
+        // TODO Do this after prune an move condition into assert
+        if (const auto cloned_edge_it =
+                std::ranges::find_if(cloned_edges, [edge_ptr = cloned_delta.vertex_edge.edge.ptr](
+                                                       const auto &elem) { return elem.second.gid == edge_ptr->gid; });
+            cloned_edge_it != cloned_edges.end()) {
+          cloned_delta.vertex_edge.edge = EdgeRef{&cloned_edge_it->second};
+        }
+      }
+      break;
+    }
+    case Delta::Action::DELETE_OBJECT:
+    case Delta::Action::RECREATE_OBJECT:
+    case Delta::Action::SET_PROPERTY:
+    case Delta::Action::ADD_LABEL:
+    case Delta::Action::REMOVE_LABEL: {
+      // noop
+      break;
+    }
+  }
+}
+
+void Splitter::AdjustDeltaNext(const Delta &original, Delta &cloned,
+                               std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions) {
+  // Get cloned_delta->next transaction, using delta->next original transaction
+  auto cloned_transaction_it = std::ranges::find_if(cloned_transactions, [&original](const auto &elem) {
+    return elem.second->start_timestamp == original.next->commit_info->start_or_commit_timestamp ||
+           elem.second->commit_info->start_or_commit_timestamp == original.next->commit_info->start_or_commit_timestamp;
+  });
+  // TODO(jbajic) What if next in delta chain does not belong to cloned transaction?
+  // MG_ASSERT(cloned_transaction_it != cloned_transactions.end(), "Cloned transaction not found");
+  if (cloned_transaction_it == cloned_transactions.end()) return;
+  // Find cloned delta in delta list of cloned transaction
+  auto found_cloned_delta_it = std::ranges::find_if(
+      cloned_transaction_it->second->deltas, [&original](const auto &elem) { return elem.id == original.next->id; });
+  MG_ASSERT(found_cloned_delta_it != cloned_transaction_it->second->deltas.end(), "Delta with given uuid must exist!");
+  cloned.next = &*found_cloned_delta_it;
+}
+
+void Splitter::AdjustDeltaPrevPtr(const Delta &original, Delta &cloned,
+                                  std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                                  VertexContainer & /*cloned_vertices*/, EdgeContainer &cloned_edges) {
+  auto ptr = original.prev.Get();
+  switch (ptr.type) {
+    case PreviousPtr::Type::NULLPTR: {
+      // noop
+      break;
+    }
+    case PreviousPtr::Type::DELTA: {
+      // Same as for deltas except don't align next but prev
+      auto cloned_transaction_it = std::ranges::find_if(cloned_transactions, [&ptr](const auto &elem) {
+        return elem.second->start_timestamp == ptr.delta->commit_info->start_or_commit_timestamp ||
+               elem.second->commit_info->start_or_commit_timestamp == ptr.delta->commit_info->start_or_commit_timestamp;
+      });
+      MG_ASSERT(cloned_transaction_it != cloned_transactions.end(), "Cloned transaction not found");
+      // Find cloned delta in delta list of cloned transaction
+      auto found_cloned_delta_it =
+          std::ranges::find_if(cloned_transaction_it->second->deltas,
+                               [delta = ptr.delta](const auto &elem) { return elem.id == delta->id; });
+      MG_ASSERT(found_cloned_delta_it != cloned_transaction_it->second->deltas.end(),
+                "Delta with given id must exist!");
+
+      cloned.prev.Set(&*found_cloned_delta_it);
+      break;
+    }
+    case PreviousPtr::Type::VERTEX: {
+      // The vertex was extracted and it is safe to reuse address
+      cloned.prev.Set(ptr.vertex);
+      break;
+    }
+    case PreviousPtr::Type::EDGE: {
+      // We can never be here if we have properties on edge disabled
+      auto *cloned_edge = &*cloned_edges.find(ptr.edge->gid);
+      cloned.prev.Set(&cloned_edge->second);
+      break;
+    }
+  };
+}
+
+}  // namespace memgraph::storage::v3

src/storage/v3/splitter.hpp

@@ -0,0 +1,107 @@
+// Copyright 2023 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+#pragma once
+
+#include <cstdint>
+#include <map>
+#include <memory>
+#include <optional>
+#include <set>
+
+#include "storage/v3/config.hpp"
+#include "storage/v3/delta.hpp"
+#include "storage/v3/edge.hpp"
+#include "storage/v3/id_types.hpp"
+#include "storage/v3/indices.hpp"
+#include "storage/v3/key_store.hpp"
+#include "storage/v3/name_id_mapper.hpp"
+#include "storage/v3/schemas.hpp"
+#include "storage/v3/transaction.hpp"
+#include "storage/v3/vertex.hpp"
+#include "utils/concepts.hpp"
+
+namespace memgraph::storage::v3 {
+
+// If edge properties-on-edges is false then we don't need to send edges but
+// only vertices, since they will contain those edges
+struct SplitData {
+  LabelId primary_label;
+  PrimaryKey min_primary_key;
+  std::optional<PrimaryKey> max_primary_key;
+  std::vector<SchemaProperty> schema;
+  Config config;
+  std::unordered_map<uint64_t, std::string> id_to_name;
+  uint64_t shard_version;
+
+  VertexContainer vertices;
+  std::optional<EdgeContainer> edges;
+  std::map<uint64_t, std::unique_ptr<Transaction>> transactions;
+  std::map<LabelId, LabelIndex::IndexContainer> label_indices;
+  std::map<std::pair<LabelId, PropertyId>, LabelPropertyIndex::IndexContainer> label_property_indices;
+};
+
+class Splitter final {
+ public:
+  Splitter(LabelId primary_label, VertexContainer &vertices, EdgeContainer &edges,
+           std::map<uint64_t, std::unique_ptr<Transaction>> &start_logical_id_to_transaction, Indices &indices,
+           const Config &config, const std::vector<SchemaProperty> &schema, const NameIdMapper &name_id_mapper_);
+
+  Splitter(const Splitter &) = delete;
+  Splitter(Splitter &&) noexcept = delete;
+  Splitter &operator=(const Splitter &) = delete;
+  Splitter operator=(Splitter &&) noexcept = delete;
+  ~Splitter() = default;
+
+  SplitData SplitShard(const PrimaryKey &split_key, const std::optional<PrimaryKey> &max_primary_key,
+                       uint64_t shard_version);
+
+ private:
+  VertexContainer CollectVertices(SplitData &data, std::set<uint64_t> &collected_transactions_start_id,
+                                  const PrimaryKey &split_key);
+
+  std::optional<EdgeContainer> CollectEdges(std::set<uint64_t> &collected_transactions_start_id,
+                                            const VertexContainer &split_vertices, const PrimaryKey &split_key);
+
+  std::map<uint64_t, std::unique_ptr<Transaction>> CollectTransactions(
+      const std::set<uint64_t> &collected_transactions_start_id, VertexContainer &cloned_vertices,
+      EdgeContainer &cloned_edges, const PrimaryKey &split_key);
+
+  static void ScanDeltas(std::set<uint64_t> &collected_transactions_start_id, Delta *delta);
+
+  void AdjustClonedTransaction(Transaction &cloned_transaction, const Transaction &transaction,
+                               std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                               VertexContainer &cloned_vertices, EdgeContainer &cloned_edges,
+                               const PrimaryKey &split_key);
+
+  void AdjustClonedTransactions(std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                                VertexContainer &cloned_vertices, EdgeContainer &cloned_edges,
+                                const PrimaryKey &split_key);
+
+  void AdjustEdgeRef(Delta &cloned_delta, EdgeContainer &cloned_edges) const;
+
+  static void AdjustDeltaNext(const Delta &original, Delta &cloned,
+                              std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions);
+
+  static void AdjustDeltaPrevPtr(const Delta &original, Delta &cloned,
+                                 std::map<uint64_t, std::unique_ptr<Transaction>> &cloned_transactions,
+                                 VertexContainer &cloned_vertices, EdgeContainer &cloned_edges);
+
+  const LabelId primary_label_;
+  VertexContainer &vertices_;
+  EdgeContainer &edges_;
+  std::map<uint64_t, std::unique_ptr<Transaction>> &start_logical_id_to_transaction_;
+  Indices &indices_;
+  const Config &config_;
+  const std::vector<SchemaProperty> schema_;
+  const NameIdMapper &name_id_mapper_;
+};
+
+}  // namespace memgraph::storage::v3

src/storage/v3/transaction.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -31,6 +31,16 @@ struct CommitInfo {
 };
 
 struct Transaction {
+  Transaction(coordinator::Hlc start_timestamp, CommitInfo new_commit_info, std::list<Delta> deltas,
+              uint64_t command_id, bool must_abort, bool is_aborted, IsolationLevel isolation_level)
+      : start_timestamp{start_timestamp},
+        commit_info{std::make_unique<CommitInfo>(new_commit_info)},
+        command_id(command_id),
+        deltas(std::move(deltas)),
+        must_abort(must_abort),
+        is_aborted(is_aborted),
+        isolation_level(isolation_level){};
+
   Transaction(coordinator::Hlc start_timestamp, IsolationLevel isolation_level)
       : start_timestamp(start_timestamp),
         commit_info(std::make_unique<CommitInfo>(CommitInfo{false, {start_timestamp}})),
@@ -54,6 +64,54 @@ struct Transaction {
 
   ~Transaction() {}
 
+  std::list<Delta> CopyDeltas(CommitInfo *commit_info) const {
+    std::list<Delta> copied_deltas;
+    for (const auto &delta : deltas) {
+      switch (delta.action) {
+        case Delta::Action::DELETE_OBJECT:
+          copied_deltas.emplace_back(Delta::DeleteObjectTag{}, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::RECREATE_OBJECT:
+          copied_deltas.emplace_back(Delta::RecreateObjectTag{}, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::ADD_LABEL:
+          copied_deltas.emplace_back(Delta::AddLabelTag{}, delta.label, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::REMOVE_LABEL:
+          copied_deltas.emplace_back(Delta::RemoveLabelTag{}, delta.label, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::ADD_IN_EDGE:
+          copied_deltas.emplace_back(Delta::AddInEdgeTag{}, delta.vertex_edge.edge_type, delta.vertex_edge.vertex_id,
+                                     delta.vertex_edge.edge, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::ADD_OUT_EDGE:
+          copied_deltas.emplace_back(Delta::AddOutEdgeTag{}, delta.vertex_edge.edge_type, delta.vertex_edge.vertex_id,
+                                     delta.vertex_edge.edge, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::REMOVE_IN_EDGE:
+          copied_deltas.emplace_back(Delta::RemoveInEdgeTag{}, delta.vertex_edge.edge_type, delta.vertex_edge.vertex_id,
+                                     delta.vertex_edge.edge, commit_info, delta.id, command_id);
+          break;
+        case Delta::Action::REMOVE_OUT_EDGE:
+          copied_deltas.emplace_back(Delta::RemoveOutEdgeTag{}, delta.vertex_edge.edge_type,
+                                     delta.vertex_edge.vertex_id, delta.vertex_edge.edge, commit_info, delta.id,
+                                     command_id);
+          break;
+        case Delta::Action::SET_PROPERTY:
+          copied_deltas.emplace_back(Delta::SetPropertyTag{}, delta.property.key, delta.property.value, commit_info,
+                                     delta.id, command_id);
+          break;
+      }
+    }
+    return copied_deltas;
+  }
+
+  // This does not solve the whole problem of copying deltas
+  std::unique_ptr<Transaction> Clone() const {
+    return std::make_unique<Transaction>(start_timestamp, *commit_info, CopyDeltas(commit_info.get()), command_id,
+                                         must_abort, is_aborted, isolation_level);
+  }
+
   coordinator::Hlc start_timestamp;
   std::unique_ptr<CommitInfo> commit_info;
   uint64_t command_id;

src/storage/v3/vertex_accessor.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -396,8 +396,8 @@ PropertyValue VertexAccessor::GetPropertyValue(PropertyId property, View view) c
     return value;
   }
   // Find PropertyId index in keystore
-  for (size_t property_index{0}; property_index < schema->second.size(); ++property_index) {
-    if (schema->second[property_index].property_id == property) {
+  for (size_t property_index{0}; property_index < schema->properties.size(); ++property_index) {
+    if (schema->properties[property_index].property_id == property) {
       return vertex_->first[property_index];
     }
   }

tests/benchmark/CMakeLists.txt

@@ -79,3 +79,6 @@ target_link_libraries(${test_prefix}data_structures_contains mg-utils mg-storage
 
 add_benchmark(data_structures_remove.cpp)
 target_link_libraries(${test_prefix}data_structures_remove mg-utils mg-storage-v3)
+
+add_benchmark(storage_v3_split.cpp)
+target_link_libraries(${test_prefix}storage_v3_split mg-storage-v3 mg-query-v2)

tests/benchmark/storage_v3_split.cpp

@@ -0,0 +1,194 @@
+// Copyright 2023 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include <cstdint>
+#include <optional>
+#include <vector>
+
+#include <benchmark/benchmark.h>
+#include <gflags/gflags.h>
+
+#include "storage/v3/id_types.hpp"
+#include "storage/v3/key_store.hpp"
+#include "storage/v3/property_value.hpp"
+#include "storage/v3/shard.hpp"
+#include "storage/v3/vertex.hpp"
+#include "storage/v3/vertex_id.hpp"
+
+namespace memgraph::benchmark {
+
+class ShardSplitBenchmark : public ::benchmark::Fixture {
+ protected:
+  using PrimaryKey = storage::v3::PrimaryKey;
+  using PropertyId = storage::v3::PropertyId;
+  using PropertyValue = storage::v3::PropertyValue;
+  using LabelId = storage::v3::LabelId;
+  using EdgeTypeId = storage::v3::EdgeTypeId;
+  using Shard = storage::v3::Shard;
+  using VertexId = storage::v3::VertexId;
+  using Gid = storage::v3::Gid;
+
+  void SetUp(const ::benchmark::State &state) override {
+    storage.emplace(primary_label, min_pk, std::nullopt, schema_property_vector);
+    storage->StoreMapping(
+        {{1, "label"}, {2, "property"}, {3, "edge_property"}, {4, "secondary_label"}, {5, "secondary_prop"}});
+  }
+
+  void TearDown(const ::benchmark::State &) override { storage = std::nullopt; }
+
+  const PropertyId primary_property{PropertyId::FromUint(2)};
+  const PropertyId secondary_property{PropertyId::FromUint(5)};
+  std::vector<storage::v3::SchemaProperty> schema_property_vector = {
+      storage::v3::SchemaProperty{primary_property, common::SchemaType::INT}};
+  const std::vector<PropertyValue> min_pk{PropertyValue{0}};
+  const LabelId primary_label{LabelId::FromUint(1)};
+  const LabelId secondary_label{LabelId::FromUint(4)};
+  const EdgeTypeId edge_type_id{EdgeTypeId::FromUint(3)};
+  std::optional<Shard> storage;
+
+  coordinator::Hlc last_hlc{0, io::Time{}};
+
+  coordinator::Hlc GetNextHlc() {
+    ++last_hlc.logical_id;
+    last_hlc.coordinator_wall_clock += std::chrono::seconds(1);
+    return last_hlc;
+  }
+};
+
+BENCHMARK_DEFINE_F(ShardSplitBenchmark, BigDataSplit)(::benchmark::State &state) {
+  const auto number_of_vertices{state.range(0)};
+  std::random_device r;
+  std::default_random_engine e1(r());
+  std::uniform_int_distribution<int> uniform_dist(0, number_of_vertices);
+
+  for (int64_t i{0}; i < number_of_vertices; ++i) {
+    auto acc = storage->Access(GetNextHlc());
+    MG_ASSERT(acc.CreateVertexAndValidate({secondary_label}, PrimaryKey{PropertyValue(i)},
+                                          {{secondary_property, PropertyValue(i)}})
+                  .HasValue(),
+              "Failed creating with pk {}", i);
+    if (i > 1) {
+      const auto vtx1 = uniform_dist(e1) % i;
+      const auto vtx2 = uniform_dist(e1) % i;
+
+      MG_ASSERT(acc.CreateEdge(VertexId{primary_label, {PropertyValue(vtx1)}},
+                               VertexId{primary_label, {PropertyValue(vtx2)}}, edge_type_id, Gid::FromUint(i))
+                    .HasValue(),
+                "Failed on {} and {}", vtx1, vtx2);
+    }
+    acc.Commit(GetNextHlc());
+  }
+  for (auto _ : state) {
+    auto data = storage->PerformSplit(PrimaryKey{PropertyValue{number_of_vertices / 2}}, 2);
+  }
+}
+
+BENCHMARK_DEFINE_F(ShardSplitBenchmark, BigDataSplitWithGc)(::benchmark::State &state) {
+  const auto number_of_vertices{state.range(0)};
+  std::random_device r;
+  std::default_random_engine e1(r());
+  std::uniform_int_distribution<int> uniform_dist(0, number_of_vertices);
+
+  for (int64_t i{0}; i < number_of_vertices; ++i) {
+    auto acc = storage->Access(GetNextHlc());
+    MG_ASSERT(acc.CreateVertexAndValidate({secondary_label}, PrimaryKey{PropertyValue(i)},
+                                          {{secondary_property, PropertyValue(i)}})
+                  .HasValue(),
+              "Failed creating with pk {}", i);
+    if (i > 1) {
+      const auto vtx1 = uniform_dist(e1) % i;
+      const auto vtx2 = uniform_dist(e1) % i;
+
+      MG_ASSERT(acc.CreateEdge(VertexId{primary_label, {PropertyValue(vtx1)}},
+                               VertexId{primary_label, {PropertyValue(vtx2)}}, edge_type_id, Gid::FromUint(i))
+                    .HasValue(),
+                "Failed on {} and {}", vtx1, vtx2);
+    }
+    acc.Commit(GetNextHlc());
+  }
+  storage->CollectGarbage(GetNextHlc().coordinator_wall_clock);
+  for (auto _ : state) {
+    auto data = storage->PerformSplit(PrimaryKey{PropertyValue{number_of_vertices / 2}}, 2);
+  }
+}
+
+BENCHMARK_DEFINE_F(ShardSplitBenchmark, BigDataSplitWithFewTransactions)(::benchmark::State &state) {
+  const auto number_of_vertices = state.range(0);
+  const auto number_of_edges = state.range(1);
+  const auto number_of_transactions = state.range(2);
+  std::random_device r;
+  std::default_random_engine e1(r());
+  std::uniform_int_distribution<int> uniform_dist(0, number_of_vertices);
+
+  const auto max_transactions_needed = std::max(number_of_vertices, number_of_edges);
+  for (int64_t vertex_counter{number_of_vertices}, edge_counter{number_of_edges}, i{0};
+       vertex_counter > 0 || edge_counter > 0; --vertex_counter, --edge_counter, ++i) {
+    auto acc = storage->Access(GetNextHlc());
+    if (vertex_counter > 0) {
+      MG_ASSERT(acc.CreateVertexAndValidate({secondary_label}, PrimaryKey{PropertyValue(i)},
+                                            {{secondary_property, PropertyValue(i)}})
+                    .HasValue(),
+                "Failed creating with pk {}", i);
+    }
+    if (edge_counter > 0 && i > 1) {
+      const auto vtx1 = uniform_dist(e1) % std::min(i, number_of_vertices);
+      const auto vtx2 = uniform_dist(e1) % std::min(i, number_of_vertices);
+
+      MG_ASSERT(acc.CreateEdge(VertexId{primary_label, {PropertyValue(vtx1)}},
+                               VertexId{primary_label, {PropertyValue(vtx2)}}, edge_type_id, Gid::FromUint(i))
+                    .HasValue(),
+                "Failed on {} and {}", vtx1, vtx2);
+    }
+
+    acc.Commit(GetNextHlc());
+    if (i == max_transactions_needed - number_of_transactions) {
+      storage->CollectGarbage(GetNextHlc().coordinator_wall_clock);
+    }
+  }
+
+  for (auto _ : state) {
+    // Don't create shard since shard deallocation can take some time as well
+    auto data = storage->PerformSplit(PrimaryKey{PropertyValue{number_of_vertices / 2}}, 2);
+  }
+}
+
+// Range:
+// Number of vertices
+// This run is pessimistic, number of vertices corresponds with number if transactions
+BENCHMARK_REGISTER_F(ShardSplitBenchmark, BigDataSplit)
+    ->RangeMultiplier(10)
+    ->Range(100'000, 100'000)
+    ->Unit(::benchmark::kMillisecond);
+
+// Range:
+// Number of vertices
+// This run is optimistic, in this run there are no transactions
+BENCHMARK_REGISTER_F(ShardSplitBenchmark, BigDataSplitWithGc)
+    ->RangeMultiplier(10)
+    ->Range(100'000, 1'000'000)
+    ->Unit(::benchmark::kMillisecond);
+
+// Args:
+// Number of vertices
+// Number of edges
+// Number of transaction
+BENCHMARK_REGISTER_F(ShardSplitBenchmark, BigDataSplitWithFewTransactions)
+    ->Args({100'000, 100'000, 1'000})
+    ->Args({100'000, 100'000, 10'000})
+    ->Args({1'000'000, 100'000, 1'000})
+    ->Args({1'000'000, 100'000, 10'000})
+    ->Args({100'000, 1'000'000, 1'000})
+    ->Args({1'000'000, 1'00'000, 10'000})
+    ->Unit(::benchmark::kMillisecond);
+
+}  // namespace memgraph::benchmark
+
+BENCHMARK_MAIN();

tests/mgbench/compare_results.py

@@ -14,6 +14,7 @@
 import argparse
 import json
 
+
 FIELDS = [
     {
         "name": "throughput",
@@ -84,32 +85,39 @@ def compare_results(results_from, results_to, fields):
                 if group == "__import__":
                     continue
                 for scenario, summary_to in scenarios.items():
-                    summary_from = recursive_get(results_from, dataset, variant, group, scenario, value={})
-                    if (
-                        len(summary_from) > 0
-                        and summary_to["count"] != summary_from["count"]
-                        or summary_to["num_workers"] != summary_from["num_workers"]
-                    ):
+                    summary_from = recursive_get(
+                        results_from, dataset, variant, group, scenario,
+                        value={})
+                    if len(summary_from) > 0 and \
+                            summary_to["count"] != summary_from["count"] or \
+                            summary_to["num_workers"] != \
+                            summary_from["num_workers"]:
                         raise Exception("Incompatible results!")
-                    testcode = "/".join([dataset, variant, group, scenario, "{:02d}".format(summary_to["num_workers"])])
+                    testcode = "/".join([dataset, variant, group, scenario,
+                                         "{:02d}".format(
+                                             summary_to["num_workers"])])
                     row = {}
                     performance_changed = False
                     for field in fields:
                         key = field["name"]
                         if key in summary_to:
-                            row[key] = compute_diff(summary_from.get(key, None), summary_to[key])
+                            row[key] = compute_diff(
+                                summary_from.get(key, None),
+                                summary_to[key])
                         elif key in summary_to["database"]:
                             row[key] = compute_diff(
-                                recursive_get(summary_from, "database", key, value=None), summary_to["database"][key]
-                            )
+                                recursive_get(summary_from, "database", key,
+                                              value=None),
+                                summary_to["database"][key])
                         else:
                             row[key] = compute_diff(
-                                recursive_get(summary_from, "metadata", key, "average", value=None),
-                                summary_to["metadata"][key]["average"],
-                            )
-                        if "diff" not in row[key] or (
-                            "diff_treshold" in field and abs(row[key]["diff"]) >= field["diff_treshold"]
-                        ):
+                                recursive_get(summary_from, "metadata", key,
+                                              "average", value=None),
+                                summary_to["metadata"][key]["average"])
+                        if "diff" not in row[key] or \
+                                ("diff_treshold" in field and
+                                 abs(row[key]["diff"]) >=
+                                 field["diff_treshold"]):
                             performance_changed = True
                     if performance_changed:
                         ret[testcode] = row
@@ -122,36 +130,29 @@ def generate_remarkup(fields, data):
         ret += "<table>\n"
         ret += "  <tr>\n"
         ret += "    <th>Testcode</th>\n"
-        ret += (
-            "\n".join(
-                map(
-                    lambda x: "    <th>{}</th>".format(x["name"].replace("_", " ").capitalize()),
-                    fields,
-                )
-            )
-            + "\n"
-        )
+        ret += "\n".join(map(lambda x: "    <th>{}</th>".format(
+            x["name"].replace("_", " ").capitalize()), fields)) + "\n"
         ret += "  </tr>\n"
         for testcode in sorted(data.keys()):
             ret += "  <tr>\n"
             ret += "    <td>{}</td>\n".format(testcode)
             for field in fields:
-                result = data[testcode].get(field["name"])
-                if result != None:
-                    value = result["value"] * field["scaling"]
-                    if "diff" in result:
-                        diff = result["diff"]
-                        arrow = "arrow-up" if diff >= 0 else "arrow-down"
-                        if not (field["positive_diff_better"] ^ (diff >= 0)):
-                            color = "green"
-                        else:
-                            color = "red"
-                        sign = "{{icon {} color={}}}".format(arrow, color)
-                        ret += '    <td bgcolor="{}">{:.3f}{} ({:+.2%})</td>\n'.format(
-                            color, value, field["unit"], diff
-                        )
+                result = data[testcode][field["name"]]
+                value = result["value"] * field["scaling"]
+                if "diff" in result:
+                    diff = result["diff"]
+                    arrow = "arrow-up" if diff >= 0 else "arrow-down"
+                    if not (field["positive_diff_better"] ^ (diff >= 0)):
+                        color = "green"
                     else:
-                        ret += '<td bgcolor="blue">{:.3f}{} //(new)// </td>\n'.format(value, field["unit"])
+                        color = "red"
+                    sign = "{{icon {} color={}}}".format(arrow, color)
+                    ret += "    <td>{:.3f}{} //({:+.2%})// {}</td>\n".format(
+                        value, field["unit"], diff, sign)
+                else:
+                    ret += "    <td>{:.3f}{} //(new)// " \
+                           "{{icon plus color=blue}}</td>\n".format(
+                               value, field["unit"])
             ret += "  </tr>\n"
         ret += "</table>\n"
     else:
@@ -160,14 +161,11 @@ def generate_remarkup(fields, data):
 
 
 if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="Compare results of multiple benchmark runs.")
-    parser.add_argument(
-        "--compare",
-        action="append",
-        nargs=2,
-        metavar=("from", "to"),
-        help="compare results between `from` and `to` files",
-    )
+    parser = argparse.ArgumentParser(
+        description="Compare results of multiple benchmark runs.")
+    parser.add_argument("--compare", action="append", nargs=2,
+                        metavar=("from", "to"),
+                        help="compare results between `from` and `to` files")
     parser.add_argument("--output", default="", help="output file name")
     args = parser.parse_args()
 

tests/mgbench/dataset_creator_unwind.py

@@ -1,139 +0,0 @@
-# Copyright 2022 Memgraph Ltd.
-#
-# Use of this software is governed by the Business Source License
-# included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-# License, and you may not use this file except in compliance with the Business Source License.
-#
-# As of the Change Date specified in that file, in accordance with
-# the Business Source License, use of this software will be governed
-# by the Apache License, Version 2.0, included in the file
-# licenses/APL.txt.
-
-import argparse
-import random
-
-import helpers
-
-# Explaination of datasets:
-#   - empty_only_index: contains index; contains no data
-#   - small: contains index; contains data (small dataset)
-#
-# Datamodel is as follow:
-#
-#                               ┌──────────────┐
-#                               │ Permission   │
-#         ┌────────────────┐    │  Schema:uuid │   ┌────────────┐
-#         │:IS_FOR_IDENTITY├────┤  Index:name  ├───┤:IS_FOR_FILE│
-#         └┬───────────────┘    └──────────────┘   └────────────┤
-#          │                                                    │
-#   ┌──────▼──────────────┐                                  ┌──▼────────────────┐
-#   │  Identity           │                                  │ File              │
-#   │   Schema:uuid       │                                  │  Schema:uuid      │
-#   │   Index:email       │                                  │  Index:name       │
-#   └─────────────────────┘                                  │  Index:platformId │
-#                                                            └───────────────────┘
-#
-#   - File: attributes: ["uuid", "name", "platformId"]
-#   - Permission: attributes: ["uuid", "name"]
-#   - Identity: attributes: ["uuid", "email"]
-#
-# Indexes:
-#   - File: [File(uuid), File(platformId), File(name)]
-#   - Permission: [Permission(uuid), Permission(name)]
-#   - Identity: [Identity(uuid), Identity(email)]
-#
-# Edges:
-#   - (:Permission)-[:IS_FOR_FILE]->(:File)
-#   - (:Permission)-[:IS_FOR_IDENTITYR]->(:Identity)
-#
-# AccessControl specific: uuid is the schema
-
-
-def main():
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--number_of_identities", type=int, default=10)
-    parser.add_argument("--number_of_files", type=int, default=10)
-    parser.add_argument("--percentage_of_permissions", type=float, default=1.0)
-    parser.add_argument("--filename", default="dataset.cypher")
-
-    args = parser.parse_args()
-
-    number_of_identities = args.number_of_identities
-    number_of_files = args.number_of_files
-    percentage_of_permissions = args.percentage_of_permissions
-    filename = args.filename
-
-    assert number_of_identities >= 0
-    assert number_of_files >= 0
-    assert percentage_of_permissions > 0.0 and percentage_of_permissions <= 1.0
-    assert filename != ""
-
-    with open(filename, "w") as f:
-        f.write("MATCH (n) DETACH DELETE n;\n")
-
-        # Create the indexes
-        f.write("CREATE INDEX ON :File;\n")
-        f.write("CREATE INDEX ON :Permission;\n")
-        f.write("CREATE INDEX ON :Identity;\n")
-        f.write("CREATE INDEX ON :File(platformId);\n")
-        f.write("CREATE INDEX ON :File(name);\n")
-        f.write("CREATE INDEX ON :Permission(name);\n")
-        f.write("CREATE INDEX ON :Identity(email);\n")
-
-        # Create extra index: in distributed, this will be the schema
-        f.write("CREATE INDEX ON :File(uuid);\n")
-        f.write("CREATE INDEX ON :Permission(uuid);\n")
-        f.write("CREATE INDEX ON :Identity(uuid);\n")
-
-        uuid = 1
-
-        # Create the nodes File
-        f.write("UNWIND [")
-        for index in range(0, number_of_files):
-            if index != 0:
-                f.write(",")
-            f.write(f' {{uuid: {uuid}, platformId: "platform_id", name: "name_file_{uuid}"}}')
-            uuid += 1
-        f.write("] AS props CREATE (:File {uuid: props.uuid, platformId: props.platformId, name: props.name});\n")
-
-        identities = []
-        f.write("UNWIND [")
-        # Create the nodes Identity
-        for index in range(0, number_of_identities):
-            if index != 0:
-                f.write(",")
-            f.write(f' {{uuid: {uuid}, name: "mail_{uuid}@something.com"}}')
-            uuid += 1
-        f.write("] AS props CREATE (:Identity {uuid: props.uuid, name: props.name});\n")
-
-        f.write("UNWIND [")
-        created = 0
-        for outer_index in range(0, number_of_files):
-            for inner_index in range(0, number_of_identities):
-
-                file_uuid = outer_index + 1
-                identity_uuid = number_of_files + inner_index + 1
-
-                if random.random() <= percentage_of_permissions:
-
-                    if created > 0:
-                        f.write(",")
-
-                    f.write(
-                        f' {{permUuid: {uuid}, permName: "name_permission_{uuid}", fileUuid: {file_uuid}, identityUuid: {identity_uuid}}}'
-                    )
-                    created += 1
-                    uuid += 1
-
-                    if created == 5000:
-                        f.write(
-                            "] AS props MATCH (file:File {uuid:props.fileUuid}), (identity:Identity {uuid: props.identityUuid}) CREATE (permission:Permission {uuid: props.permUuid, name: props.permName}) CREATE (permission)-[: IS_FOR_FILE]->(file) CREATE (permission)-[: IS_FOR_IDENTITY]->(identity);\nUNWIND ["
-                        )
-                        created = 0
-        f.write(
-            "] AS props MATCH (file:File {uuid:props.fileUuid}), (identity:Identity {uuid: props.identityUuid}) CREATE (permission:Permission {uuid: props.permUuid, name: props.permName}) CREATE (permission)-[: IS_FOR_FILE]->(file) CREATE (permission)-[: IS_FOR_IDENTITY]->(identity);\n"
-        )
-
-
-if __name__ == "__main__":
-    main()

tests/mgbench/datasets.py

@@ -353,7 +353,7 @@ class AccessControl(Dataset):
 
     def benchmark__create__vertex(self):
         self.next_value_idx += 1
-        query = ("CREATE (:File {uuid: $uuid})", {"uuid": self.next_value_idx})
+        query = (f"CREATE (:File {{uuid: {self.next_value_idx}}});", {})
         return query
 
     def benchmark__create__edges(self):
@@ -379,24 +379,6 @@ class AccessControl(Dataset):
         return query
 
     def benchmark__match__match_all_vertices_with_edges(self):
+        self.next_value_idx += 1
         query = ("MATCH (permission:Permission)-[e:IS_FOR_FILE]->(file:File) RETURN *", {})
         return query
-
-    def benchmark__match__match_users_with_permission_for_files(self):
-        file_uuid_1 = self._get_random_uuid("File")
-        file_uuid_2 = self._get_random_uuid("File")
-        min_file_uuid = min(file_uuid_1, file_uuid_2)
-        max_file_uuid = max(file_uuid_1, file_uuid_2)
-        query = (
-            "MATCH (f:File)<-[ff:IS_FOR_FILE]-(p:Permission)-[fi:IS_FOR_IDENTITY]->(i:Identity) WHERE f.uuid >= $min_file_uuid  AND f.uuid <= $max_file_uuid RETURN *",
-            {"min_file_uuid": min_file_uuid, "max_file_uuid": max_file_uuid},
-        )
-        return query
-
-    def benchmark__match__match_users_with_permission_for_specific_file(self):
-        file_uuid = self._get_random_uuid("File")
-        query = (
-            "MATCH (f:File {uuid: $file_uuid})<-[ff:IS_FOR_FILE]-(p:Permission)-[fi:IS_FOR_IDENTITY]->(i:Identity) RETURN *",
-            {"file_uuid": file_uuid},
-        )
-        return query

tests/mgbench/runners.py

@@ -68,15 +68,6 @@ class Memgraph:
         self._cleanup()
         atexit.unregister(self._cleanup)
 
-    # Returns None if string_value is not true or false, casing doesn't matter
-    def _get_bool_value(self, string_value):
-        lower_string_value = string_value.lower()
-        if lower_string_value == "true":
-            return True
-        if lower_string_value == "false":
-            return False
-        return None
-
     def _get_args(self, **kwargs):
         data_directory = os.path.join(self._directory.name, "memgraph")
         if self._memgraph_version >= (0, 50, 0):
@@ -92,13 +83,7 @@ class Memgraph:
             args_list = self._extra_args.split(" ")
             assert len(args_list) % 2 == 0
             for i in range(0, len(args_list), 2):
-                key = args_list[i]
-                value = args_list[i + 1]
-                maybe_bool_value = self._get_bool_value(value)
-                if maybe_bool_value is not None:
-                    kwargs[key] = maybe_bool_value
-                else:
-                    kwargs[key] = value
+                kwargs[args_list[i]] = args_list[i + 1]
 
         return _convert_args_to_flags(self._memgraph_binary, **kwargs)
 

tests/mgbench/splitfiles/accesscontrol_large.shard_configuration

@@ -1,12 +1,8 @@
-8
+4
 uuid
 email
 name
 platformId
-permUuid
-permName
-fileUuid
-identityUuid
 2
 IS_FOR_IDENTITY
 IS_FOR_FILE

tests/mgbench/splitfiles/accesscontrol_medium.shard_configuration

@@ -1,12 +1,8 @@
-8
+4
 uuid
 email
 name
 platformId
-permUuid
-permName
-fileUuid
-identityUuid
 2
 IS_FOR_IDENTITY
 IS_FOR_FILE

tests/mgbench/splitfiles/accesscontrol_small.shard_configuration

@@ -1,12 +1,8 @@
-8
+4
 uuid
 email
 name
 platformId
-permUuid
-permName
-fileUuid
-identityUuid
 2
 IS_FOR_IDENTITY
 IS_FOR_FILE

tests/simulation/CMakeLists.txt

@@ -17,7 +17,7 @@ function(add_simulation_test test_cpp)
   # requires unique logical target names
   set_target_properties(${target_name} PROPERTIES OUTPUT_NAME ${exec_name})
 
-  target_link_libraries(${target_name} mg-communication mg-utils mg-io mg-io-simulator mg-coordinator mg-query-v2 mg-storage-v3)
+  target_link_libraries(${target_name} mg-storage-v3 mg-communication mg-utils mg-io mg-io-simulator mg-coordinator mg-query-v2)
   target_link_libraries(${target_name} Boost::headers)
   target_link_libraries(${target_name} gtest gtest_main gmock rapidcheck rapidcheck_gtest)
 
@@ -32,5 +32,4 @@ add_simulation_test(trial_query_storage/query_storage_test.cpp)
 add_simulation_test(sharded_map.cpp)
 add_simulation_test(shard_rsm.cpp)
 add_simulation_test(cluster_property_test.cpp)
-add_simulation_test(cluster_property_test_cypher_queries.cpp)
 add_simulation_test(request_router.cpp)

tests/simulation/cluster_property_test_cypher_queries.cpp

@@ -1,64 +0,0 @@
-// Copyright 2023 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-// This test serves as an example of a property-based model test.
-// It generates a cluster configuration and a set of operations to
-// apply against both the real system and a greatly simplified model.
-
-#include <chrono>
-
-#include <gtest/gtest.h>
-#include <rapidcheck.h>
-#include <rapidcheck/gtest.h>
-#include <spdlog/cfg/env.h>
-
-#include "generated_operations.hpp"
-#include "io/simulator/simulator_config.hpp"
-#include "io/time.hpp"
-#include "storage/v3/shard_manager.hpp"
-#include "test_cluster.hpp"
-
-namespace memgraph::tests::simulation {
-
-using io::Duration;
-using io::Time;
-using io::simulator::SimulatorConfig;
-using storage::v3::kMaximumCronInterval;
-
-RC_GTEST_PROP(RandomClusterConfig, HappyPath, (ClusterConfig cluster_config, NonEmptyOpVec ops, uint64_t rng_seed)) {
-  spdlog::cfg::load_env_levels();
-
-  SimulatorConfig sim_config{
-      .drop_percent = 0,
-      .perform_timeouts = false,
-      .scramble_messages = true,
-      .rng_seed = rng_seed,
-      .start_time = Time::min(),
-      .abort_time = Time::max(),
-  };
-
-  std::vector<std::string> queries = {"CREATE (n:test_label{property_1: 0, property_2: 0});", "MATCH (n) RETURN n;"};
-
-  auto [sim_stats_1, latency_stats_1] = RunClusterSimulationWithQueries(sim_config, cluster_config, queries);
-  auto [sim_stats_2, latency_stats_2] = RunClusterSimulationWithQueries(sim_config, cluster_config, queries);
-
-  if (latency_stats_1 != latency_stats_2) {
-    spdlog::error("simulator stats diverged across runs");
-    spdlog::error("run 1 simulator stats: {}", sim_stats_1);
-    spdlog::error("run 2 simulator stats: {}", sim_stats_2);
-    spdlog::error("run 1 latency:\n{}", latency_stats_1.SummaryTable());
-    spdlog::error("run 2 latency:\n{}", latency_stats_2.SummaryTable());
-    RC_ASSERT(latency_stats_1 == latency_stats_2);
-    RC_ASSERT(sim_stats_1 == sim_stats_2);
-  }
-}
-
-}  // namespace memgraph::tests::simulation

tests/simulation/simulation_interpreter.hpp

@@ -1,93 +0,0 @@
-// Copyright 2023 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#include "io/simulator/simulator_handle.hpp"
-#include "machine_manager/machine_config.hpp"
-#include "machine_manager/machine_manager.hpp"
-#include "query/v2/config.hpp"
-#include "query/v2/discard_value_stream.hpp"
-#include "query/v2/frontend/ast/ast.hpp"
-#include "query/v2/interpreter.hpp"
-#include "query/v2/request_router.hpp"
-
-#include <string>
-#include <vector>
-
-// TODO(gvolfing)
-// -How to set up the entire raft cluster with the QE. Also provide abrstraction for that.
-// -Pass an argument to the setup to determine, how many times the retry of a query should happen.
-
-namespace memgraph::io::simulator {
-
-class SimulatedInterpreter {
-  using ResultStream = query::v2::DiscardValueResultStream;
-
- public:
-  explicit SimulatedInterpreter(std::unique_ptr<query::v2::InterpreterContext> interpreter_context)
-      : interpreter_context_(std::move(interpreter_context)) {
-    interpreter_ = std::make_unique<memgraph::query::v2::Interpreter>(interpreter_context_.get());
-  }
-
-  SimulatedInterpreter(const SimulatedInterpreter &) = delete;
-  SimulatedInterpreter &operator=(const SimulatedInterpreter &) = delete;
-  SimulatedInterpreter(SimulatedInterpreter &&) = delete;
-  SimulatedInterpreter &operator=(SimulatedInterpreter &&) = delete;
-  ~SimulatedInterpreter() = default;
-
-  void InstallSimulatorTicker(Simulator &simulator) {
-    interpreter_->InstallSimulatorTicker(simulator.GetSimulatorTickClosure());
-  }
-
-  std::vector<ResultStream> RunQueries(const std::vector<std::string> &queries) {
-    std::vector<ResultStream> results;
-    results.reserve(queries.size());
-
-    for (const auto &query : queries) {
-      results.emplace_back(RunQuery(query));
-    }
-    return results;
-  }
-
- private:
-  ResultStream RunQuery(const std::string &query) {
-    ResultStream stream;
-
-    std::map<std::string, memgraph::storage::v3::PropertyValue> params;
-    const std::string *username = nullptr;
-
-    interpreter_->Prepare(query, params, username);
-    interpreter_->PullAll(&stream);
-
-    return stream;
-  }
-
-  std::unique_ptr<query::v2::InterpreterContext> interpreter_context_;
-  std::unique_ptr<query::v2::Interpreter> interpreter_;
-};
-
-SimulatedInterpreter SetUpInterpreter(Address coordinator_address, Simulator &simulator) {
-  auto rr_factory = std::make_unique<memgraph::query::v2::SimulatedRequestRouterFactory>(simulator);
-
-  auto interpreter_context = std::make_unique<memgraph::query::v2::InterpreterContext>(
-      nullptr,
-      memgraph::query::v2::InterpreterConfig{.query = {.allow_load_csv = true},
-                                             .execution_timeout_sec = 600,
-                                             .replication_replica_check_frequency = std::chrono::seconds(1),
-                                             .default_kafka_bootstrap_servers = "",
-                                             .default_pulsar_service_url = "",
-                                             .stream_transaction_conflict_retries = 30,
-                                             .stream_transaction_retry_interval = std::chrono::milliseconds(500)},
-      std::filesystem::path("mg_data"), std::move(rr_factory), coordinator_address);
-
-  return SimulatedInterpreter(std::move(interpreter_context));
-}
-
-}  // namespace memgraph::io::simulator

tests/simulation/test_cluster.hpp

@@ -36,8 +36,6 @@
 #include "utils/print_helpers.hpp"
 #include "utils/variant_helpers.hpp"
 
-#include "simulation_interpreter.hpp"
-
 namespace memgraph::tests::simulation {
 
 using coordinator::Coordinator;
@@ -281,65 +279,4 @@ std::pair<SimulatorStats, LatencyHistogramSummaries> RunClusterSimulation(const
   return std::make_pair(stats, histo);
 }
 
-std::pair<SimulatorStats, LatencyHistogramSummaries> RunClusterSimulationWithQueries(
-    const SimulatorConfig &sim_config, const ClusterConfig &cluster_config, const std::vector<std::string> &queries) {
-  spdlog::info("========================== NEW SIMULATION ==========================");
-
-  auto simulator = Simulator(sim_config);
-
-  auto machine_1_addr = Address::TestAddress(1);
-  auto cli_addr = Address::TestAddress(2);
-  auto cli_addr_2 = Address::TestAddress(3);
-
-  Io<SimulatorTransport> cli_io = simulator.Register(cli_addr);
-  Io<SimulatorTransport> cli_io_2 = simulator.Register(cli_addr_2);
-
-  auto coordinator_addresses = std::vector{
-      machine_1_addr,
-  };
-
-  ShardMap initialization_sm = TestShardMap(cluster_config.shards - 1, cluster_config.replication_factor);
-
-  auto mm_1 = MkMm(simulator, coordinator_addresses, machine_1_addr, initialization_sm);
-  Address coordinator_address = mm_1.CoordinatorAddress();
-
-  auto mm_thread_1 = std::jthread(RunMachine, std::move(mm_1));
-  simulator.IncrementServerCountAndWaitForQuiescentState(machine_1_addr);
-
-  auto detach_on_error = DetachIfDropped{.handle = mm_thread_1};
-
-  // TODO(tyler) clarify addresses of coordinator etc... as it's a mess
-
-  CoordinatorClient<SimulatorTransport> coordinator_client(cli_io, coordinator_address, {coordinator_address});
-  WaitForShardsToInitialize(coordinator_client);
-
-  auto simulated_interpreter = io::simulator::SetUpInterpreter(coordinator_address, simulator);
-  simulated_interpreter.InstallSimulatorTicker(simulator);
-
-  auto query_results = simulated_interpreter.RunQueries(queries);
-
-  // We have now completed our workload without failing any assertions, so we can
-  // disable detaching the worker thread, which will cause the mm_thread_1 jthread
-  // to be joined when this function returns.
-  detach_on_error.detach = false;
-
-  simulator.ShutDown();
-
-  mm_thread_1.join();
-
-  SimulatorStats stats = simulator.Stats();
-
-  spdlog::info("total messages:     {}", stats.total_messages);
-  spdlog::info("dropped messages:   {}", stats.dropped_messages);
-  spdlog::info("timed out requests: {}", stats.timed_out_requests);
-  spdlog::info("total requests:     {}", stats.total_requests);
-  spdlog::info("total responses:    {}", stats.total_responses);
-  spdlog::info("simulator ticks:    {}", stats.simulator_ticks);
-
-  auto histo = cli_io_2.ResponseLatencies();
-
-  spdlog::info("========================== SUCCESS :) ==========================");
-  return std::make_pair(stats, histo);
-}
-
 }  // namespace memgraph::tests::simulation

tests/unit/CMakeLists.txt

@@ -294,6 +294,9 @@ target_link_libraries(${test_prefix}storage_v3_expr mg-storage-v3 mg-expr)
 add_unit_test(storage_v3_schema.cpp)
 target_link_libraries(${test_prefix}storage_v3_schema mg-storage-v3)
 
+add_unit_test(storage_v3_shard_split.cpp)
+target_link_libraries(${test_prefix}storage_v3_shard_split mg-storage-v3  mg-query-v2)
+
 # Test mg-query-v2
 # These are commented out because of the new TypedValue in the query engine
 # add_unit_test(query_v2_interpreter.cpp ${CMAKE_SOURCE_DIR}/src/glue/v2/communication.cpp)

tests/unit/mock_helpers.hpp

@@ -41,14 +41,8 @@ class MockedRequestRouter : public RequestRouterInterface {
   MOCK_METHOD(std::optional<storage::v3::EdgeTypeId>, MaybeNameToEdgeType, (const std::string &), (const));
   MOCK_METHOD(std::optional<storage::v3::LabelId>, MaybeNameToLabel, (const std::string &), (const));
   MOCK_METHOD(bool, IsPrimaryLabel, (storage::v3::LabelId), (const));
-  MOCK_METHOD(bool, IsPrimaryProperty, (storage::v3::LabelId, storage::v3::PropertyId), (const));
-  MOCK_METHOD((std::optional<std::pair<uint64_t, uint64_t>>), AllocateInitialEdgeIds, (io::Address));
-  MOCK_METHOD(void, InstallSimulatorTicker, (std::function<bool()>));
+  MOCK_METHOD(bool, IsPrimaryKey, (storage::v3::LabelId, storage::v3::PropertyId), (const));
   MOCK_METHOD(const std::vector<coordinator::SchemaProperty> &, GetSchemaForLabel, (storage::v3::LabelId), (const));
-  MOCK_METHOD(int64_t, GetApproximateVertexCount, (), (const));
-  MOCK_METHOD(int64_t, GetApproximateVertexCount, (storage::v3::LabelId label), (const));
-  MOCK_METHOD(int64_t, GetApproximateVertexCount, (storage::v3::LabelId label, storage::v3::PropertyId property),
-              (const));
 };
 
 class MockedLogicalOperator : public plan::LogicalOperator {
@@ -65,7 +59,7 @@ class MockedLogicalOperator : public plan::LogicalOperator {
 class MockedCursor : public plan::Cursor {
  public:
   MOCK_METHOD(bool, Pull, (Frame &, expr::ExecutionContext &));
-  MOCK_METHOD(bool, PullMultiple, (MultiFrame &, expr::ExecutionContext &));
+  MOCK_METHOD(void, PullMultiple, (MultiFrame &, expr::ExecutionContext &));
   MOCK_METHOD(void, Reset, ());
   MOCK_METHOD(void, Shutdown, ());
 };

tests/unit/query_v2_common.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2022 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source

tests/unit/query_v2_create_node_multiframe.cpp

@@ -58,7 +58,7 @@ TEST(CreateNodeTest, CreateNodeCursor) {
   MockedRequestRouter router;
   EXPECT_CALL(router, CreateVertices(_)).Times(1).WillOnce(Return(std::vector<msgs::CreateVerticesResponse>{}));
   EXPECT_CALL(router, IsPrimaryLabel(_)).WillRepeatedly(Return(true));
-  EXPECT_CALL(router, IsPrimaryProperty(_, _)).WillRepeatedly(Return(true));
+  EXPECT_CALL(router, IsPrimaryKey(_, _)).WillRepeatedly(Return(true));
   auto context = MakeContext(ast, symbol_table, &router, &id_alloc);
   auto multi_frame = CreateMultiFrame(context.symbol_table.max_position());
   cursor->PullMultiple(multi_frame, context);

tests/unit/query_v2_expression_evaluator.cpp

@@ -123,25 +123,13 @@ class MockedRequestRouter : public RequestRouterInterface {
 
   bool IsPrimaryLabel(LabelId label) const override { return true; }
 
-  bool IsPrimaryProperty(LabelId primary_label, PropertyId property) const override { return true; }
+  bool IsPrimaryKey(LabelId primary_label, PropertyId property) const override { return true; }
 
-  std::optional<std::pair<uint64_t, uint64_t>> AllocateInitialEdgeIds(io::Address coordinator_address) override {
-    return {};
-  }
-
-  void InstallSimulatorTicker(std::function<bool()> tick_simulator) override {}
   const std::vector<coordinator::SchemaProperty> &GetSchemaForLabel(storage::v3::LabelId /*label*/) const override {
     static std::vector<coordinator::SchemaProperty> schema;
     return schema;
   };
 
-  // TODO(gvolfing) once the real implementation is done make sure these are solved as well.
-  int64_t GetApproximateVertexCount() const override { return 1; }
-  int64_t GetApproximateVertexCount(storage::v3::LabelId label) const override { return 1; }
-  int64_t GetApproximateVertexCount(storage::v3::LabelId label, storage::v3::PropertyId property) const override {
-    return 1;
-  }
-
  private:
   void SetUpNameIdMappers() {
     std::unordered_map<uint64_t, std::string> id_to_name;

tests/unit/query_v2_plan.cpp

@@ -86,7 +86,7 @@ class TestPlanner : public ::testing::Test {};
 
 using PlannerTypes = ::testing::Types<Planner>;
 
-TYPED_TEST_SUITE(TestPlanner, PlannerTypes);
+TYPED_TEST_CASE(TestPlanner, PlannerTypes);
 
 TYPED_TEST(TestPlanner, MatchFilterPropIsNotNull) {
   const char *prim_label_name = "prim_label_one";

tests/unit/storage_v3_schema.cpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2023 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -60,18 +60,22 @@ TEST_F(SchemaTest, TestSchemaCreate) {
     EXPECT_TRUE(schemas.CreateSchema(label2, {schema_prop_string, schema_prop_int}));
     const auto current_schemas = schemas.ListSchemas();
     EXPECT_EQ(current_schemas.size(), 2);
-    EXPECT_THAT(current_schemas,
-                UnorderedElementsAre(Pair(label1, std::vector<SchemaProperty>{schema_prop_string}),
-                                     Pair(label2, std::vector<SchemaProperty>{schema_prop_string, schema_prop_int})));
+    EXPECT_THAT(
+        current_schemas,
+        UnorderedElementsAre(
+            Schema{.label = label1, .properties = std::vector<SchemaProperty>{schema_prop_string}},
+            Schema{.label = label2, .properties = std::vector<SchemaProperty>{schema_prop_string, schema_prop_int}}));
   }
   {
     // Assert after unsuccessful creation, number oif schemas remains the same
     EXPECT_FALSE(schemas.CreateSchema(label2, {schema_prop_int}));
     const auto current_schemas = schemas.ListSchemas();
     EXPECT_EQ(current_schemas.size(), 2);
-    EXPECT_THAT(current_schemas,
-                UnorderedElementsAre(Pair(label1, std::vector<SchemaProperty>{schema_prop_string}),
-                                     Pair(label2, std::vector<SchemaProperty>{schema_prop_string, schema_prop_int})));
+    EXPECT_THAT(
+        current_schemas,
+        UnorderedElementsAre(
+            Schema{.label = label1, .properties = std::vector<SchemaProperty>{schema_prop_string}},
+            Schema{.label = label2, .properties = std::vector<SchemaProperty>{schema_prop_string, schema_prop_int}}));
   }
 }
 
@@ -89,27 +93,29 @@ TEST_F(SchemaTest, TestSchemaList) {
   {
     const auto current_schemas = schemas.ListSchemas();
     EXPECT_EQ(current_schemas.size(), 2);
-    EXPECT_THAT(current_schemas,
-                UnorderedElementsAre(
-                    Pair(label1, std::vector<SchemaProperty>{schema_prop_string}),
-                    Pair(label2, std::vector<SchemaProperty>{{NameToProperty("prop1"), SchemaType::STRING},
+    EXPECT_THAT(
+        current_schemas,
+        UnorderedElementsAre(
+            Schema{.label = label1, .properties = std::vector<SchemaProperty>{schema_prop_string}},
+            Schema{.label = label2,
+                   .properties = std::vector<SchemaProperty>{{NameToProperty("prop}"), SchemaType::STRING},
                                                              {NameToProperty("prop2"), SchemaType::INT},
                                                              {NameToProperty("prop3"), SchemaType::BOOL},
                                                              {NameToProperty("prop4"), SchemaType::DATE},
                                                              {NameToProperty("prop5"), SchemaType::LOCALDATETIME},
                                                              {NameToProperty("prop6"), SchemaType::DURATION},
-                                                             {NameToProperty("prop7"), SchemaType::LOCALTIME}})));
+                                                             {NameToProperty("prop7"), SchemaType::LOCALTIME}}}));
   }
   {
     const auto *const schema1 = schemas.GetSchema(label1);
     ASSERT_NE(schema1, nullptr);
-    EXPECT_EQ(*schema1, (Schemas::Schema{label1, std::vector<SchemaProperty>{schema_prop_string}}));
+    EXPECT_EQ(*schema1, (Schema{.label = label1, .properties = std::vector<SchemaProperty>{schema_prop_string}}));
   }
   {
     const auto *const schema2 = schemas.GetSchema(label2);
     ASSERT_NE(schema2, nullptr);
-    EXPECT_EQ(schema2->first, label2);
-    EXPECT_EQ(schema2->second.size(), 7);
+    EXPECT_EQ(schema2->label, label2);
+    EXPECT_EQ(schema2->properties.size(), 7);
   }
 }
 
@@ -132,7 +138,8 @@ TEST_F(SchemaTest, TestSchemaDrop) {
     const auto current_schemas = schemas.ListSchemas();
     EXPECT_EQ(current_schemas.size(), 1);
     EXPECT_THAT(current_schemas,
-                UnorderedElementsAre(Pair(label2, std::vector<SchemaProperty>{schema_prop_string, schema_prop_int})));
+                UnorderedElementsAre(Schema{
+                    .label = label2, .properties = std::vector<SchemaProperty>{schema_prop_string, schema_prop_int}}));
   }
 
   {
@@ -141,7 +148,8 @@ TEST_F(SchemaTest, TestSchemaDrop) {
     const auto current_schemas = schemas.ListSchemas();
     EXPECT_EQ(current_schemas.size(), 1);
     EXPECT_THAT(current_schemas,
-                UnorderedElementsAre(Pair(label2, std::vector<SchemaProperty>{schema_prop_string, schema_prop_int})));
+                UnorderedElementsAre(Schema{
+                    .label = label2, .properties = std::vector<SchemaProperty>{schema_prop_string, schema_prop_int}}));
   }
 
   EXPECT_TRUE(schemas.DropSchema(label2));

tests/unit/storage_v3_shard_split.cpp

@@ -0,0 +1,501 @@
+// Copyright 2023 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include <cstdint>
+#include <memory>
+
+#include <gmock/gmock-matchers.h>
+#include <gmock/gmock.h>
+#include <gtest/gtest.h>
+
+#include "coordinator/hybrid_logical_clock.hpp"
+#include "query/v2/requests.hpp"
+#include "storage/v3/delta.hpp"
+#include "storage/v3/id_types.hpp"
+#include "storage/v3/key_store.hpp"
+#include "storage/v3/mvcc.hpp"
+#include "storage/v3/property_value.hpp"
+#include "storage/v3/shard.hpp"
+#include "storage/v3/vertex.hpp"
+#include "storage/v3/vertex_id.hpp"
+
+using testing::Pair;
+using testing::UnorderedElementsAre;
+
+namespace memgraph::storage::v3::tests {
+
+class ShardSplitTest : public testing::Test {
+ protected:
+  void SetUp() override {
+    storage.StoreMapping(
+        {{1, "label"}, {2, "property"}, {3, "edge_property"}, {4, "secondary_label"}, {5, "secondary_prop"}});
+  }
+
+  const PropertyId primary_property{PropertyId::FromUint(2)};
+  const PropertyId secondary_property{PropertyId::FromUint(5)};
+  std::vector<storage::v3::SchemaProperty> schema_property_vector = {
+      storage::v3::SchemaProperty{primary_property, common::SchemaType::INT}};
+  const std::vector<PropertyValue> min_pk{PropertyValue{0}};
+  const LabelId primary_label{LabelId::FromUint(1)};
+  const LabelId secondary_label{LabelId::FromUint(4)};
+  const EdgeTypeId edge_type_id{EdgeTypeId::FromUint(3)};
+  Shard storage{primary_label, min_pk, std::nullopt /*max_primary_key*/, schema_property_vector};
+
+  coordinator::Hlc last_hlc{0, io::Time{}};
+
+  coordinator::Hlc GetNextHlc() {
+    ++last_hlc.logical_id;
+    last_hlc.coordinator_wall_clock += std::chrono::seconds(1);
+    return last_hlc;
+  }
+
+  void AssertShardState(auto &shard, const int split_min, const int split_max) {
+    auto acc = shard.Access(GetNextHlc());
+    for (int i{0}; i < split_min; ++i) {
+      EXPECT_FALSE(acc.FindVertex(PrimaryKey{{PropertyValue(i)}}, View::OLD).has_value());
+    }
+    for (int i{split_min}; i < split_max; ++i) {
+      const auto vtx = acc.FindVertex(PrimaryKey{{PropertyValue(i)}}, View::OLD);
+      ASSERT_TRUE(vtx.has_value());
+      EXPECT_TRUE(vtx->InEdges(View::OLD)->size() == 1 || vtx->OutEdges(View::OLD)->size() == 1);
+    }
+  }
+};
+
+void AssertEqVertexContainer(const VertexContainer &actual, const VertexContainer &expected) {
+  ASSERT_EQ(actual.size(), expected.size());
+
+  auto expected_it = expected.begin();
+  auto actual_it = actual.begin();
+  while (expected_it != expected.end()) {
+    EXPECT_EQ(actual_it->first, expected_it->first);
+    EXPECT_EQ(actual_it->second.deleted, expected_it->second.deleted);
+    EXPECT_EQ(actual_it->second.labels, expected_it->second.labels);
+
+    auto *expected_delta = expected_it->second.delta;
+    auto *actual_delta = actual_it->second.delta;
+    // This asserts delta chain
+    while (expected_delta != nullptr) {
+      EXPECT_EQ(actual_delta->action, expected_delta->action);
+      EXPECT_EQ(actual_delta->id, expected_delta->id);
+
+      switch (expected_delta->action) {
+        case Delta::Action::ADD_LABEL:
+        case Delta::Action::REMOVE_LABEL: {
+          EXPECT_EQ(actual_delta->label, expected_delta->label);
+          break;
+        }
+        case Delta::Action::SET_PROPERTY: {
+          EXPECT_EQ(actual_delta->property.key, expected_delta->property.key);
+          EXPECT_EQ(actual_delta->property.value, expected_delta->property.value);
+          break;
+        }
+        case Delta::Action::ADD_IN_EDGE:
+        case Delta::Action::ADD_OUT_EDGE:
+        case Delta::Action::REMOVE_IN_EDGE:
+        case Delta::Action::RECREATE_OBJECT:
+        case Delta::Action::DELETE_OBJECT:
+        case Delta::Action::REMOVE_OUT_EDGE: {
+          break;
+        }
+      }
+
+      const auto expected_prev = expected_delta->prev.Get();
+      const auto actual_prev = actual_delta->prev.Get();
+      switch (expected_prev.type) {
+        case PreviousPtr::Type::NULLPTR: {
+          ASSERT_EQ(actual_prev.type, PreviousPtr::Type::NULLPTR) << "Expected type is nullptr!";
+          break;
+        }
+        case PreviousPtr::Type::DELTA: {
+          ASSERT_EQ(actual_prev.type, PreviousPtr::Type::DELTA) << "Expected type is delta!";
+          EXPECT_EQ(actual_prev.delta->action, expected_prev.delta->action);
+          EXPECT_EQ(actual_prev.delta->id, expected_prev.delta->id);
+          break;
+        }
+        case v3::PreviousPtr::Type::EDGE: {
+          ASSERT_EQ(actual_prev.type, PreviousPtr::Type::EDGE) << "Expected type is edge!";
+          EXPECT_EQ(actual_prev.edge->gid, expected_prev.edge->gid);
+          break;
+        }
+        case v3::PreviousPtr::Type::VERTEX: {
+          ASSERT_EQ(actual_prev.type, PreviousPtr::Type::VERTEX) << "Expected type is vertex!";
+          EXPECT_EQ(actual_prev.vertex->first, expected_prev.vertex->first);
+          break;
+        }
+      }
+
+      expected_delta = expected_delta->next;
+      actual_delta = actual_delta->next;
+    }
+    EXPECT_EQ(expected_delta, nullptr);
+    EXPECT_EQ(actual_delta, nullptr);
+    ++expected_it;
+    ++actual_it;
+  }
+}
+
+void AssertEqDeltaLists(const std::list<Delta> &actual, const std::list<Delta> &expected) {
+  EXPECT_EQ(actual.size(), expected.size());
+  auto actual_it = actual.begin();
+  auto expected_it = expected.begin();
+  while (actual_it != actual.end()) {
+    EXPECT_EQ(actual_it->id, expected_it->id);
+    EXPECT_EQ(actual_it->action, expected_it->action);
+    ++actual_it;
+    ++expected_it;
+  }
+}
+
+void AddDeltaToDeltaChain(Vertex *object, Delta *new_delta) {
+  auto *delta_holder = GetDeltaHolder(object);
+
+  new_delta->next = delta_holder->delta;
+  new_delta->prev.Set(object);
+  if (delta_holder->delta) {
+    delta_holder->delta->prev.Set(new_delta);
+  }
+  delta_holder->delta = new_delta;
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitWithVertices) {
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(1)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(
+      acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(5)}, {{secondary_property, PropertyValue(121)}})
+          .HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+  auto current_hlc = GetNextHlc();
+  acc.Commit(current_hlc);
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 0);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+
+  CommitInfo commit_info{.start_or_commit_timestamp = current_hlc};
+  Delta delta_delete1{Delta::DeleteObjectTag{}, &commit_info, 4, 1};
+  Delta delta_delete2{Delta::DeleteObjectTag{}, &commit_info, 5, 2};
+  Delta delta_remove_label{Delta::RemoveLabelTag{}, secondary_label, &commit_info, 7, 4};
+  Delta delta_set_property{Delta::SetPropertyTag{}, secondary_property, PropertyValue(), &commit_info, 6, 4};
+  Delta delta_delete3{Delta::DeleteObjectTag{}, &commit_info, 8, 3};
+
+  VertexContainer expected_vertices;
+  auto [it_4, inserted1] = expected_vertices.emplace(PrimaryKey{PropertyValue{4}}, VertexData(&delta_delete1));
+  delta_delete1.prev.Set(&*it_4);
+  auto [it_5, inserted2] = expected_vertices.emplace(PrimaryKey{PropertyValue{5}}, VertexData(&delta_delete2));
+  delta_delete2.prev.Set(&*it_5);
+  auto [it_6, inserted3] = expected_vertices.emplace(PrimaryKey{PropertyValue{6}}, VertexData(&delta_delete3));
+  delta_delete3.prev.Set(&*it_6);
+  it_5->second.labels.push_back(secondary_label);
+  AddDeltaToDeltaChain(&*it_5, &delta_set_property);
+  AddDeltaToDeltaChain(&*it_5, &delta_remove_label);
+
+  AssertEqVertexContainer(splitted_data.vertices, expected_vertices);
+
+  // This is to ensure that the transaction that we have don't point to invalid
+  // object on the other shard
+  std::list<Delta> expected_deltas;
+  expected_deltas.emplace_back(Delta::DeleteObjectTag{}, &commit_info, 4, 1);
+  expected_deltas.emplace_back(Delta::DeleteObjectTag{}, &commit_info, 5, 2);
+  expected_deltas.emplace_back(Delta::SetPropertyTag{}, secondary_property, PropertyValue(), &commit_info, 6, 4);
+  expected_deltas.emplace_back(Delta::RemoveLabelTag{}, secondary_label, &commit_info, 7, 4);
+  expected_deltas.emplace_back(Delta::DeleteObjectTag{}, &commit_info, 8, 3);
+  AssertEqDeltaLists(splitted_data.transactions.begin()->second->deltas, expected_deltas);
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitVerticesAndEdges) {
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(1)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(5)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(5)}}, edge_type_id, Gid::FromUint(1))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(4)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(6)}}, edge_type_id, Gid::FromUint(2))
+                   .HasError());
+
+  auto current_hlc = GetNextHlc();
+  acc.Commit(current_hlc);
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 2);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+
+  CommitInfo commit_info{.start_or_commit_timestamp = current_hlc};
+  Delta delta_delete1{Delta::DeleteObjectTag{}, &commit_info, 12, 1};
+  Delta delta_delete2{Delta::DeleteObjectTag{}, &commit_info, 13, 1};
+  Delta delta_delete3{Delta::DeleteObjectTag{}, &commit_info, 14, 1};
+  Delta delta_add_in_edge1{Delta::RemoveInEdgeTag{},
+                           edge_type_id,
+                           VertexId{primary_label, {PropertyValue(1)}},
+                           EdgeRef{Gid::FromUint(1)},
+                           &commit_info,
+                           17,
+                           1};
+  Delta delta_add_out_edge2{Delta::RemoveOutEdgeTag{},
+                            edge_type_id,
+                            VertexId{primary_label, {PropertyValue(6)}},
+                            EdgeRef{Gid::FromUint(2)},
+                            &commit_info,
+                            19,
+                            1};
+  Delta delta_add_in_edge2{Delta::RemoveInEdgeTag{},
+                           edge_type_id,
+                           VertexId{primary_label, {PropertyValue(4)}},
+                           EdgeRef{Gid::FromUint(2)},
+                           &commit_info,
+                           20,
+                           1};
+  VertexContainer expected_vertices;
+  auto [vtx4, inserted4] = expected_vertices.emplace(PrimaryKey{PropertyValue{4}}, VertexData(&delta_delete1));
+  auto [vtx5, inserted5] = expected_vertices.emplace(PrimaryKey{PropertyValue{5}}, VertexData(&delta_delete2));
+  auto [vtx6, inserted6] = expected_vertices.emplace(PrimaryKey{PropertyValue{6}}, VertexData(&delta_delete3));
+  AddDeltaToDeltaChain(&*vtx4, &delta_add_out_edge2);
+  AddDeltaToDeltaChain(&*vtx5, &delta_add_in_edge1);
+  AddDeltaToDeltaChain(&*vtx6, &delta_add_in_edge2);
+
+  AssertEqVertexContainer(splitted_data.vertices, expected_vertices);
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitBeforeCommit) {
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(1)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(5)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(2)}}, edge_type_id, Gid::FromUint(0))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(5)}}, edge_type_id, Gid::FromUint(1))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(4)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(6)}}, edge_type_id, Gid::FromUint(2))
+                   .HasError());
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 2);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitWithCommitedAndOngoingTransactions) {
+  {
+    auto acc = storage.Access(GetNextHlc());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(1)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(5)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+
+    acc.Commit(GetNextHlc());
+  }
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(2)}}, edge_type_id, Gid::FromUint(0))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(3)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(5)}}, edge_type_id, Gid::FromUint(1))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(4)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(6)}}, edge_type_id, Gid::FromUint(2))
+                   .HasError());
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 2);
+  EXPECT_EQ(splitted_data.transactions.size(), 2);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitWithLabelIndex) {
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(1)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(5)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(6)}, {}).HasError());
+  acc.Commit(GetNextHlc());
+  storage.CreateIndex(secondary_label);
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 0);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 1);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+}
+
+TEST_F(ShardSplitTest, TestBasicSplitWithLabelPropertyIndex) {
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(
+      acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(1)}, {{secondary_property, PropertyValue(1)}})
+          .HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(
+      acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(5)}, {{secondary_property, PropertyValue(21)}})
+          .HasError());
+  EXPECT_FALSE(
+      acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(6)}, {{secondary_property, PropertyValue(22)}})
+          .HasError());
+  acc.Commit(GetNextHlc());
+  storage.CreateIndex(secondary_label, secondary_property);
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 0);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 1);
+}
+
+TEST_F(ShardSplitTest, TestSplittingShardsWithGcDestroyOriginalShard) {
+  const auto split_value{4};
+  PrimaryKey splitted_value{{PropertyValue(4)}};
+  std::unique_ptr<Shard> splitted_shard;
+
+  {
+    Shard storage2{primary_label, min_pk, std::nullopt /*max_primary_key*/, schema_property_vector};
+    auto acc = storage2.Access(GetNextHlc());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(1)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(5)}, {}).HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+
+    EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                                VertexId{primary_label, PrimaryKey{PropertyValue(2)}}, edge_type_id, Gid::FromUint(0))
+                     .HasError());
+    EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(3)}},
+                                VertexId{primary_label, PrimaryKey{PropertyValue(5)}}, edge_type_id, Gid::FromUint(1))
+                     .HasError());
+    EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(4)}},
+                                VertexId{primary_label, PrimaryKey{PropertyValue(6)}}, edge_type_id, Gid::FromUint(2))
+                     .HasError());
+    acc.Commit(GetNextHlc());
+
+    auto splitted_data = storage2.PerformSplit({PropertyValue(split_value)}, 2);
+    EXPECT_EQ(splitted_data.vertices.size(), 3);
+    EXPECT_EQ(splitted_data.edges->size(), 2);
+    EXPECT_EQ(splitted_data.transactions.size(), 1);
+    EXPECT_EQ(splitted_data.label_indices.size(), 0);
+    EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+
+    // Create a new shard
+    splitted_shard = Shard::FromSplitData(std::move(splitted_data));
+    // Call gc on old shard
+    storage2.CollectGarbage(GetNextHlc().coordinator_wall_clock);
+    // Destroy original
+  }
+
+  splitted_shard->CollectGarbage(GetNextHlc().coordinator_wall_clock);
+  AssertShardState(*splitted_shard, 4, 6);
+}
+
+TEST_F(ShardSplitTest, TestSplittingShardsWithGcDestroySplittedShard) {
+  PrimaryKey splitted_value{{PropertyValue(4)}};
+
+  auto acc = storage.Access(GetNextHlc());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(1)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(2)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(3)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(4)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(5)}, {}).HasError());
+  EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(6)}, {}).HasError());
+
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(1)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(2)}}, edge_type_id, Gid::FromUint(0))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(3)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(5)}}, edge_type_id, Gid::FromUint(1))
+                   .HasError());
+  EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(4)}},
+                              VertexId{primary_label, PrimaryKey{PropertyValue(6)}}, edge_type_id, Gid::FromUint(2))
+                   .HasError());
+  acc.Commit(GetNextHlc());
+
+  auto splitted_data = storage.PerformSplit({PropertyValue(4)}, 2);
+  EXPECT_EQ(splitted_data.vertices.size(), 3);
+  EXPECT_EQ(splitted_data.edges->size(), 2);
+  EXPECT_EQ(splitted_data.transactions.size(), 1);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 0);
+
+  {
+    // Create a new shard
+    auto splitted_shard = Shard::FromSplitData(std::move(splitted_data));
+    // Call gc on new shard
+    splitted_shard->CollectGarbage(GetNextHlc().coordinator_wall_clock);
+    // Destroy splitted shard
+  }
+
+  storage.CollectGarbage(GetNextHlc().coordinator_wall_clock);
+  AssertShardState(storage, 1, 3);
+}
+
+TEST_F(ShardSplitTest, TestBigSplit) {
+  int pk{0};
+  for (int64_t i{0}; i < 10'000; ++i) {
+    auto acc = storage.Access(GetNextHlc());
+    EXPECT_FALSE(
+        acc.CreateVertexAndValidate({secondary_label}, {PropertyValue(pk++)}, {{secondary_property, PropertyValue(i)}})
+            .HasError());
+    EXPECT_FALSE(acc.CreateVertexAndValidate({}, {PropertyValue(pk++)}, {}).HasError());
+
+    EXPECT_FALSE(acc.CreateEdge(VertexId{primary_label, PrimaryKey{PropertyValue(pk - 2)}},
+                                VertexId{primary_label, PrimaryKey{PropertyValue(pk - 1)}}, edge_type_id,
+                                Gid::FromUint(pk))
+                     .HasError());
+    acc.Commit(GetNextHlc());
+  }
+  storage.CreateIndex(secondary_label, secondary_property);
+
+  const auto split_value = pk / 2;
+  auto splitted_data = storage.PerformSplit({PropertyValue(split_value)}, 2);
+
+  EXPECT_EQ(splitted_data.vertices.size(), 10000);
+  EXPECT_EQ(splitted_data.edges->size(), 5000);
+  EXPECT_EQ(splitted_data.transactions.size(), 5000);
+  EXPECT_EQ(splitted_data.label_indices.size(), 0);
+  EXPECT_EQ(splitted_data.label_property_indices.size(), 1);
+
+  auto shard = Shard::FromSplitData(std::move(splitted_data));
+  AssertShardState(*shard, split_value, split_value * 2);
+}
+
+}  // namespace memgraph::storage::v3::tests