fall 2024 p4 update (#769)

* fall 2024 p4 update

* lint

* disable the public test

* nit: copyright and format

Author: J-HowHuang

src/execution/execution_common.cpp

@@ -1,12 +1,22 @@
+//===----------------------------------------------------------------------===//
+//
+//                         BusTub
+//
+// execution_common.cpp
+//
+// Identification: src/execution/execution_common.cpp
+//
+// Copyright (c) 2024-2024, Carnegie Mellon University Database Group
+//
+//===----------------------------------------------------------------------===//
+
 #include "execution/execution_common.h"
+
 #include "catalog/catalog.h"
-#include "common/config.h"
 #include "common/macros.h"
 #include "concurrency/transaction_manager.h"
 #include "fmt/core.h"
 #include "storage/table/table_heap.h"
-#include "type/value.h"
-#include "type/value_factory.h"
 
 namespace bustub {
 
@@ -23,11 +33,70 @@ auto GenerateSortKey(const Tuple &tuple, const std::vector<OrderBy> &order_bys,
  * You can ignore the remaining part of this file until P4.
  */
 
+/**
+ * @brief Reconstruct a tuple by applying the provided undo logs from the base tuple. All logs in the undo_logs are
+ * applied regardless of the timestamp
+ *
+ * @param schema The schema of the base tuple and the returned tuple.
+ * @param base_tuple The base tuple to start the reconstruction from.
+ * @param base_meta The metadata of the base tuple.
+ * @param undo_logs The list of undo logs to apply during the reconstruction, the front is applied first.
+ * @return An optional tuple that represents the reconstructed tuple. If the tuple is deleted as the result, returns
+ * std::nullopt.
+ */
 auto ReconstructTuple(const Schema *schema, const Tuple &base_tuple, const TupleMeta &base_meta,
                       const std::vector<UndoLog> &undo_logs) -> std::optional<Tuple> {
   UNIMPLEMENTED("not implemented");
 }
 
+/**
+ * @brief Collects the undo logs sufficient to reconstruct the tuple w.r.t. the txn.
+ *
+ * @param rid The RID of the tuple.
+ * @param base_meta The metadata of the base tuple.
+ * @param base_tuple The base tuple.
+ * @param undo_link The undo link to the latest undo log.
+ * @param txn The transaction.
+ * @param txn_mgr The transaction manager.
+ * @return An optional vector of undo logs to pass to ReconstructTuple(). std::nullopt if the tuple did not exist at the
+ * time.
+ */
+auto CollectUndoLogs(RID rid, const TupleMeta &base_meta, const Tuple &base_tuple, std::optional<UndoLink> undo_link,
+                     Transaction *txn, TransactionManager *txn_mgr) -> std::optional<std::vector<UndoLog>> {
+  UNIMPLEMENTED("not implemented");
+}
+
+/**
+ * @brief Generates a new undo log as the transaction tries to modify this tuple at the first time.
+ *
+ * @param schema The schema of the table.
+ * @param base_tuple The base tuple before the update, the one retrieved from the table heap. nullptr if the tuple is
+ * deleted.
+ * @param target_tuple The target tuple after the update. nullptr if this is a deletion.
+ * @param ts The timestamp of the base tuple.
+ * @param prev_version The undo link to the latest undo log of this tuple.
+ * @return The generated undo log.
+ */
+auto GenerateNewUndoLog(const Schema *schema, const Tuple *base_tuple, const Tuple *target_tuple, timestamp_t ts,
+                        UndoLink prev_version) -> UndoLog {
+  UNIMPLEMENTED("not implemented");
+}
+
+/**
+ * @brief Generate the updated undo log to replace the old one, whereas the tuple is already modified by this txn once.
+ *
+ * @param schema The schema of the table.
+ * @param base_tuple The base tuple before the update, the one retrieved from the table heap. nullptr if the tuple is
+ * deleted.
+ * @param target_tuple The target tuple after the update. nullptr if this is a deletion.
+ * @param log The original undo log.
+ * @return The updated undo log.
+ */
+auto GenerateUpdatedUndoLog(const Schema *schema, const Tuple *base_tuple, const Tuple *target_tuple,
+                            const UndoLog &log) -> UndoLog {
+  UNIMPLEMENTED("not implemented");
+}
+
 void TxnMgrDbg(const std::string &info, TransactionManager *txn_mgr, const TableInfo *table_info,
                TableHeap *table_heap) {
   // always use stderr for printing logs...

src/include/execution/execution_common.h

@@ -1,3 +1,14 @@
+//===----------------------------------------------------------------------===//
+//
+//                         BusTub
+//
+// execution_common.h
+//
+// Identification: src/include/execution/execution_common.h
+//
+// Copyright (c) 2014-2024, Carnegie Mellon University Database Group
+//
+//===----------------------------------------------------------------------===//
 #pragma once
 
 #include <string>
@@ -44,6 +55,15 @@ auto GenerateSortKey(const Tuple &tuple, const std::vector<OrderBy> &order_bys,
 auto ReconstructTuple(const Schema *schema, const Tuple &base_tuple, const TupleMeta &base_meta,
                       const std::vector<UndoLog> &undo_logs) -> std::optional<Tuple>;
 
+auto CollectUndoLogs(RID rid, const TupleMeta &base_meta, const Tuple &base_tuple, std::optional<UndoLink> undo_link,
+                     Transaction *txn, TransactionManager *txn_mgr) -> std::optional<std::vector<UndoLog>>;
+
+auto GenerateNewUndoLog(const Schema *schema, const Tuple *base_tuple, const Tuple *target_tuple, timestamp_t ts,
+                        UndoLink prev_version) -> UndoLog;
+
+auto GenerateUpdatedUndoLog(const Schema *schema, const Tuple *base_tuple, const Tuple *target_tuple,
+                            const UndoLog &log) -> UndoLog;
+
 void TxnMgrDbg(const std::string &info, TransactionManager *txn_mgr, const TableInfo *table_info,
                TableHeap *table_heap);
 
@@ -52,11 +72,9 @@ void TxnMgrDbg(const std::string &info, TransactionManager *txn_mgr, const Table
 // To give you a sense of what can be shared across executors / transaction manager, here are the
 // list of helper function names that we defined in the reference solution. You should come up with
 // your own when you go through the process.
-// * CollectUndoLogs
 // * WalkUndoLogs
 // * Modify
 // * IsWriteWriteConflict
-// * GenerateDiffLog
 // * GenerateNullTupleForSchema
 // * GetUndoLogSchema
 //

test/txn/txn_executor_test.cpp

@@ -186,6 +186,112 @@ TEST(TxnExecutorTest, DISABLED_InsertDeleteConflictTest) {  // NOLINT
   WithTxn(txn7, CommitTxn(*bustub, _var, _txn));
 }
 
+TEST(TxnExecutorTest, DISABLED_GenerateUndoLogTest) {
+  {
+    fmt::println(stderr, "--- GenerateUndoLogTest: Simple update ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("b double,c boolean");
+    auto base_tuple = Tuple{{Int(0), Double(0.0), Bool(true)}, schema.get()};
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), &base_tuple, &target_tuple, 0, {});
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, false}, {undo_log});
+    ASSERT_TRUE(tuple.has_value());
+    ASSERT_TRUE(IsTupleContentEqual(*tuple, base_tuple));
+  }
+  {
+    fmt::println(stderr, "--- GenerateUndoLogTest: Simple delete ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto base_tuple = Tuple{{Int(0), Double(0.0), Bool(true)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), &base_tuple, nullptr, 0, {});
+
+    auto tuple = ReconstructTuple(schema.get(), base_tuple, {0, true}, {undo_log});
+    ASSERT_TRUE(tuple.has_value());
+    ASSERT_TRUE(IsTupleContentEqual(*tuple, base_tuple));
+  }
+  {
+    // This case is only relevant after task 4.2, where an insert may happen on a tombstone.
+    // Before task 4.2, insert always generates a new tuple in table heap without any undo logs.
+    fmt::println(stderr, "--- GenerateUndoLogTest: Simple insert ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("");
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), nullptr, &target_tuple, 0, {});
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, false}, {undo_log});
+    ASSERT_FALSE(tuple.has_value());
+  }
+  {
+    fmt::println(stderr, "--- GenerateUndoLogTest: Update twice in a txn ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("b double,c boolean");
+    auto base_tuple = Tuple{{Int(0), Double(0.0), Bool(true)}, schema.get()};
+    auto intermidiate_tuple = Tuple{{Int(0), Double(0.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), &base_tuple, &intermidiate_tuple, 0, {});
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto updated_undo_log = GenerateUpdatedUndoLog(schema.get(), &intermidiate_tuple, &target_tuple, undo_log);
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, false}, {updated_undo_log});
+    ASSERT_TRUE(tuple.has_value());
+    ASSERT_TRUE(IsTupleContentEqual(*tuple, base_tuple));
+  }
+  {
+    fmt::println(stderr, "--- GenerateUndoLogTest: Update then delete in a txn ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("b double,c boolean");
+    auto base_tuple = Tuple{{Int(0), Double(0.0), Bool(true)}, schema.get()};
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), &base_tuple, &target_tuple, 0, {});
+    auto updated_undo_log = GenerateUpdatedUndoLog(schema.get(), &target_tuple, nullptr, undo_log);
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, true}, {updated_undo_log});
+    ASSERT_TRUE(tuple.has_value());
+    ASSERT_TRUE(IsTupleContentEqual(*tuple, base_tuple));
+  }
+  {
+    // This case is only relevant after task 4.2, where an insert may happen on a tombstone.
+    // Before task 4.2, insert always generates a new tuple in table heap without any undo logs.
+    fmt::println(stderr, "--- GenerateUndoLogTest: Insert then update in a txn ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("");
+    auto intermediate_tuple = Tuple{{Int(0), Double(0.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), nullptr, &intermediate_tuple, 0, {});
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto updated_undo_log = GenerateUpdatedUndoLog(schema.get(), &intermediate_tuple, &target_tuple, undo_log);
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, false}, {updated_undo_log});
+    ASSERT_FALSE(tuple.has_value());
+  }
+  {
+    // This case is only relevant after task 4.2, where an insert may happen on a tombstone.
+    // Before task 4.2, insert always generates a new tuple in table heap without any undo logs.
+    fmt::println(stderr, "--- GenerateUndoLogTest: Insert then delete in a txn ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("");
+    auto intermediate_tuple = Tuple{{Int(0), Double(0.0), Bool(false)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), nullptr, &intermediate_tuple, 0, {});
+    auto updated_undo_log = GenerateUpdatedUndoLog(schema.get(), &intermediate_tuple, nullptr, undo_log);
+
+    auto tuple = ReconstructTuple(schema.get(), intermediate_tuple, {0, true}, {updated_undo_log});
+    ASSERT_FALSE(tuple.has_value());
+  }
+  {
+    // This case is only relevant after task 4.2, where an insert may happen on a tombstone.
+    // Before task 4.2, insert always generates a new tuple in table heap without any undo logs.
+    fmt::println(stderr, "--- GenerateUndoLogTest: Delete then insert in a txn ---");
+    auto schema = ParseCreateStatement("a integer,b double,c boolean");
+    auto partial_schema = ParseCreateStatement("b double,c boolean");
+    auto base_tuple = Tuple{{Int(0), Double(0.0), Bool(true)}, schema.get()};
+    auto undo_log = GenerateNewUndoLog(schema.get(), &base_tuple, nullptr, 0, {});
+    auto target_tuple = Tuple{{Int(0), Double(1.0), Bool(false)}, schema.get()};
+    auto updated_undo_log = GenerateUpdatedUndoLog(schema.get(), &base_tuple, &target_tuple, undo_log);
+
+    auto tuple = ReconstructTuple(schema.get(), target_tuple, {0, false}, {updated_undo_log});
+    ASSERT_TRUE(tuple.has_value());
+    ASSERT_TRUE(IsTupleContentEqual(*tuple, base_tuple));
+  }
+}
+
 TEST(TxnExecutorTest, DISABLED_UpdateTest1) {  // NOLINT
   fmt::println(stderr, "--- UpdateTest1: no undo log ---");
   auto bustub = std::make_unique<BusTubInstance>();