Vldbss 2025(merge on demand)

.clang-format

@@ -81,4 +81,3 @@ MacroBlockBegin: "
 END_CATCH_ERROR$"
 ...
 
-

benchmark/pax_storage_concurrency_test.cpp

@@ -117,7 +117,7 @@ class BenchmarkBase : public Fixture
     table_meta_->fields_[1].attr_len_  = 11;
     table_meta_->fields_[1].field_id_  = 1;
     handler_                           = new RecordFileHandler(StorageFormat::PAX_FORMAT);
-    rc                                 = handler_->init(*buffer_pool_, log_handler_, table_meta_);
+    rc                                 = handler_->init(*buffer_pool_, log_handler_, table_meta_, nullptr);
     if (rc != RC::SUCCESS) {
       LOG_WARN("failed to init record file handler. rc=%s", strrc(rc));
       throw runtime_error("failed to init record file handler");

benchmark/record_manager_concurrency_test.cpp

@@ -106,7 +106,7 @@ class BenchmarkBase : public Fixture
     }
 
     handler_ = new RecordFileHandler(StorageFormat::ROW_FORMAT);
-    rc       = handler_->init(*buffer_pool_, log_handler_, nullptr);
+    rc       = handler_->init(*buffer_pool_, log_handler_, nullptr, nullptr);
     if (rc != RC::SUCCESS) {
       LOG_WARN("failed to init record file handler. rc=%s", strrc(rc));
       throw runtime_error("failed to init record file handler");

build.sh

@@ -78,6 +78,7 @@ function do_init
   git -C "deps/3rd/libevent" checkout 112421c8fa4840acd73502f2ab6a674fc025de37 || return
   # git submodule update --remote "deps/3rd/libevent" || return
   git -C "deps/3rd/jsoncpp" checkout 1.9.6 || return
+
   current_dir=$PWD
 
   MAKE_COMMAND="make --silent"

docs/docs/design/miniob-how-to-add-new-datatype.md

@@ -0,0 +1,32 @@
+---
+title: 如何新增一种数据类型
+---
+
+> 本文介绍如何新增一种数据类型。
+MiniOB 的数据类型系统采用分层设计，实现集中在[path](../../../src/observer/common)文件夹下，核心组件包括：
+1. Value 类：统一数据操作接口
+路径：src/observer/common/value.h
+作用：封装实际数据值，提供类型无关的操作方法
+2. Type 工具类：特定类型的操作实现
+路径：src/observer/common/type/
+作用：每种数据类型对应一个工具类，实现具体运算逻辑
+
+以下示例展示 MiniOB 如何处理整数类型数据：
+```cpp
+// 假设解析器识别到整数 "1"
+int val = 1;
+Value value(val);  // 封装为 Value 对象
+// 执行加法运算
+Value result;
+Value::add(value, value, result);  // 调用加法接口
+// Value::add 方法内部会根据类型调用对应工具类
+// 对于 INT 类型，实际调用代码位于：
+// src/observer/common/type/integer_type.cpp
+```
+
+# 若要新增一种数据类型（如 DATE），建议按以下步骤开发：
+1. 在 src/observer/common/type/attr_type.h 中添加新的类型枚举以及对应类型名
+2. 在 src/observer/common/type/data_type.cpp 中添加新的类型实例
+3. 在 src/observer/common/type/ 文件夹下，参照现有工具类，实现 DateType 工具类
+4. 在 Value 类中增加类型处理逻辑，支持date类型的分发，储存date类型值
+5. 必要情况下还需要增加新的词法规则(lex_sql.l)以及语法规则(yacc_sql.y)，支持新类型关键字

docs/docs/design/miniob-sql-execution-process.md

@@ -0,0 +1,169 @@
+如图[image](images/miniob-sql-execution-process.png)
+
+PlantUML时序图使用 https://www.plantuml.com/plantuml 生成
+代码如下：
+```cpp
+@startuml
+title SQL 执行流程时序图\n
+
+skinparam sequence {
+  ArrowColor #003366
+  LifeLineBorderColor #003366
+  LifeLineBackgroundColor #F0F8FF
+  ParticipantBorderColor #003366
+  ParticipantBackgroundColor #E6F7FF
+  ParticipantFontColor #003366
+  NoteBackgroundColor #FFF2E6
+  NoteBorderColor #FFA940
+  BoxPadding 20
+}
+
+actor 客户端 as Client
+participant "Server" as Server
+participant "TaskHandler" as TaskHandler
+participant "SessionStage" as Session
+participant "ParseStage" as Parser
+participant "ResolveStage" as Resolver
+participant "OptimizeStage" as Optimizer
+participant "ExecuteStage" as Executor
+participant "ResultHandler" as Result
+
+autonumber "<b>[000]"
+
+Client -> Server: SQL命令(网络请求)
+note right: 1. 客户端发送SQL到服务器端口\n   - 建立TCP连接\n   - 传输SQL文本
+activate Server
+
+Server -> TaskHandler: handle_event(communicator)
+activate TaskHandler
+note right: 2. 任务处理器初始化\n   - 创建事件对象\n   - 绑定通信器\n   - 准备处理环境
+
+TaskHandler -> Session: handle_request2(event)
+activate Session
+Session --> TaskHandler: 会话就绪
+deactivate Session
+note right Session: 3. 会话管理\n   - 用户身份验证\n   - 设置数据库上下文\n   - 维护会话状态
+
+group "SQL处理核心 (责任链模式)"
+  TaskHandler -> Parser: handle_request(sql_event)
+  activate Parser
+  group "ParseStage.handle_request"
+    Parser -> Parser: parse(sql.c_str())
+    note right: 4.1 启动SQL解析\n   - 准备解析环境
+    Parser -> Parser: yylex_init_extra()
+    note right: 4.2 初始化词法扫描器\n   - 分配扫描器资源\n   - 设置字符串缓冲区
+    Parser -> Parser: yyparse(scanner)
+    note right: 4.3 语法解析核心\n   - Flex词法分析\n   - Bison语法分析\n   - 生成AST节点树
+    Parser -> Parser: yylex_destroy()
+    note right: 4.4 释放扫描器资源\n   - 清理词法分析状态
+    Parser --> TaskHandler: 返回AST
+  end
+  deactivate Parser
+  note right: 5. 解析完成\n   - 输出抽象语法树(AST)\n   - 支持SELECT/INSERT等语句结构
+
+  TaskHandler -> Resolver: handle_request(sql_event)
+  activate Resolver
+  group "ResolveStage.handle_request"
+    Resolver -> Resolver: Stmt::create_stmt()
+    note right: 6.1 语句转换入口\n   - 根据AST节点类型分发
+    Resolver -> Resolver: 递归处理节点
+    note right: 6.2 深度遍历语法树\n   - 处理子表达式\n   - 构建完整语句结构
+    Resolver -> Resolver: 创建具体Stmt对象
+    note right: 6.3 生成可执行语句\n   - SelectStmt\n   - InsertStmt\n   - UpdateStmt等
+    Resolver --> TaskHandler: 返回Stmt
+  end
+  deactivate Resolver
+  note right: 7. 转换完成\n   - AST→可执行Stmt对象\n   - 完成语义分析
+
+  TaskHandler -> Optimizer: handle_request(sql_event)
+  activate Optimizer
+  group "OptimizeStage.handle_request"
+    Optimizer -> Optimizer: create_logical_plan()
+    note right: 8.1 创建初始逻辑计划\n   - 扫描算子\n   - 连接算子\n   - 投影算子
+    Optimizer -> Optimizer: logical_plan_generator.create()
+    note right: 8.2 生成逻辑算子树\n   - 递归处理Stmt对象
+
+    group "重写规则应用"
+      Optimizer -> Optimizer: ComparisonSimplificationRule.rewrite()
+      note right: 9.1 比较表达式简化\n      例: 1=1 → true\n      常量折叠优化
+      Optimizer -> Optimizer: ConjunctionSimplificationRule.rewrite()
+      note right: 9.2 连接表达式简化\n      例: false AND expr → false\n      短路逻辑优化
+      Optimizer -> Optimizer: PredicateRewriteRule.rewrite()
+      note right: 9.3 谓词重写\n      删除恒真表达式\n      简化条件结构
+      Optimizer -> Optimizer: PredicatePushdownRewriter.rewrite()
+      note right: 9.4 谓词下推\n      将过滤条件下推至扫描层\n      减少中间结果集
+    end
+
+    Optimizer -> Optimizer: optimize()
+    note right: 10. 基于代价优化\n   - 选择最优连接顺序\n   - 索引选择\n   - 访问路径优化
+
+    Optimizer -> Optimizer: generate_physical_plan()
+    alt 向量模型
+      Optimizer -> Optimizer: physical_plan_generator.create_vec()
+      note right: 11.1 向量化执行计划\n   - 批量处理(1024行/批)\n   - 列式内存布局\n   - 现代OLAP优化
+    else 火山模型
+      Optimizer -> Optimizer: physical_plan_generator.create()
+      note right: 11.2 迭代式执行计划\n   - 逐行处理\n   - next()接口模型\n   - 传统OLTP优化
+    end
+    Optimizer --> TaskHandler: 返回执行计划
+  end
+  deactivate Optimizer
+  note right: 12. 优化完成\n   - 生成物理执行计划\n   - 准备执行环境
+
+  TaskHandler -> Executor: handle_request(sql_event)
+  activate Executor
+  group "ExecuteStage.handle_request"
+    alt 有物理算子(DML)
+      Executor -> Executor: handle_request_with_physical_operator()
+      note right: 13.1 DML执行路径\n   - SELECT: 执行查询\n   - INSERT: 插入数据\n   - DELETE: 删除数据
+      Executor --> TaskHandler: DML结果
+    else 无物理算子(DDL)
+      Executor -> Executor: command_executor.execute()
+      note right: 13.2 DDL执行路径\n   - CREATE TABLE\n   - CREATE INDEX\n   - ALTER TABLE
+      Executor --> TaskHandler: DDL结果
+    end
+  end
+  deactivate Executor
+  note right: 14. 执行完成\n   - 数据变更生效\n   - 查询结果就绪
+end
+
+TaskHandler -> Result: write_result(event)
+activate Result
+
+group "ResultHandler处理"
+  alt 无结果集(DDL/更新)
+    Result -> Result: write_result_internal()
+    note right: 15.1 简单结果处理\n   - 操作状态(成功/失败)\n   - 影响行数统计
+    Result --> Client: 状态码
+  else 有结果集(查询)
+    Result -> Result: sql_result->open()
+    note right: 15.2 初始化结果集\n   - 准备数据缓冲区\n   - 获取表头信息
+    Result -> Result: 打印表头
+    note right: 15.3 输出列信息\n   - 列名\n   - 数据类型\n   - 列宽
+
+    alt 向量模型
+      Result -> Result: write_chunk_result()
+      note right: 16.1 向量化结果返回\n   - 批量传输数据块\n   - 高效网络利用\n   - 减少序列化开销
+      Result --> Client: 批量数据
+    else 火山模型
+      Result -> Result: write_tuple_result()
+      note right: 16.2 迭代式结果返回\n   - 逐行获取数据\n   - 流式传输模式
+      loop 逐行处理
+        Result --> Client: 单行数据
+      end
+    end
+  end
+end
+
+Result --> TaskHandler: 完成
+deactivate Result
+
+TaskHandler --> Server: 处理完成
+deactivate TaskHandler
+
+Server --> Client: 最终确认
+deactivate Server
+note right: 17. 完整流程结束\n   - 释放所有资源\n   - 维护连接状态\n   - 准备接收下一条SQL
+
+@enduml
+```

docs/docs/how_to_build.md

@@ -160,3 +160,32 @@ git config --global core.autocrlf false
 关于该问题的更多细节，请参考[问题来源](https://ask.oceanbase.com/t/topic/35604437/7)。
 关于该问题的进一步分析，请参考[Linux系统下执行sudo命令环境变量失效现象](https://zhuanlan.zhihu.com/p/669332689)。
 也可以将cmake所在路径添加到sudo的PATH变量中来解决上述问题，请参考[sudo命令下环境变量实效的解决方法](https://www.cnblogs.com/xiao-xiaoyang/p/17444600.html)。
+
+
+### 3. Could not find a package configuration file provided by "Libevent"
+在执行build.sh脚本时，遇到下面的错误
+![cmake error](images/miniob-build-libevent.png)
+
+通常是因为cmake版本原因（版本太高？）导致libevent在init阶段没有编译成功。
+
+***解决方法：***
+
+在[text](../../deps/3rd/libevent/CMakeLists.txt) 中将cmake的最低版本设置
+cmake_minimum_required(VERSION 3.1 FATAL_ERROR)
+改为
+cmake_minimum_required(VERSION 3.1...3.8 FATAL_ERROR)
+之后重新执行
+```bash
+sudo bash build.sh init
+```
+
+如果你成功解决libevent的问题，你大概率会遇到另一个错误：
+![cmake error](images/miniob-build-jsoncpp.png)
+需要在[text](../../deps/3rd/jsoncpp/jsoncppConfig.cmake.in)中将cmake策略
+cmake_policy(VERSION 3.0)
+改为
+cmake_policy(VERSION 3.0...3.8)
+之后重新执行
+```bash
+sudo bash build.sh init
+```

src/common/lang/comparator.cpp

@@ -32,6 +32,19 @@ int compare_int(void *arg1, void *arg2)
   }
 }
 
+int compare_int64(void *arg1, void *arg2)
+{
+  int v1 = *(int64_t *)arg1;
+  int v2 = *(int64_t *)arg2;
+  if (v1 > v2) {
+    return 1;
+  } else if (v1 < v2) {
+    return -1;
+  } else {
+    return 0;
+  }
+}
+
 int compare_float(void *arg1, void *arg2)
 {
   float v1  = *(float *)arg1;

src/common/lang/comparator.h

@@ -17,6 +17,7 @@ See the Mulan PSL v2 for more details. */
 namespace common {
 
 int compare_int(void *arg1, void *arg2);
+int compare_int64(void *arg1, void *arg2);
 int compare_float(void *arg1, void *arg2);
 int compare_string(void *arg1, int arg1_max_length, void *arg2, int arg2_max_length);
 

src/observer/common/type/attr_type.cpp

@@ -31,3 +31,13 @@ AttrType attr_type_from_string(const char *s)
   }
   return AttrType::UNDEFINED;
 }
+
+bool is_numerical_type(AttrType type)
+{
+  return (type == AttrType::INTS || type == AttrType::FLOATS);
+}
+
+bool is_string_type(AttrType type)
+{
+  return (type == AttrType::CHARS);
+}

src/observer/common/type/attr_type.h

@@ -27,3 +27,5 @@ enum class AttrType
 
 const char *attr_type_to_string(AttrType type);
 AttrType    attr_type_from_string(const char *s);
+bool        is_numerical_type(AttrType type);
+bool        is_string_type(AttrType type);

src/observer/common/type/data_type.cpp

@@ -14,6 +14,9 @@ See the Mulan PSL v2 for more details. */
 #include "common/type/data_type.h"
 #include "common/type/vector_type.h"
 
+// Todo: 实现新数据类型
+// your code here
+
 array<unique_ptr<DataType>, static_cast<int>(AttrType::MAXTYPE)> DataType::type_instances_ = {
     make_unique<DataType>(AttrType::UNDEFINED),
     make_unique<CharType>(),

src/observer/common/type/data_type.h

@@ -17,13 +17,15 @@ See the Mulan PSL v2 for more details. */
 #include "common/type/attr_type.h"
 
 class Value;
+class Column;
 
 /**
  * @brief 定义了数据类型相关的操作，比如比较运算、算术运算等
  * @defgroup DataType
  * @details 数据类型定义的算术运算中，比如 add、subtract 等，将按照当前数据类型设置最终结果值的类型。
  * 参与运算的参数类型不一定相同，不同的类型进行运算是否能够支持需要参考各个类型的实现。
  */
+
 class DataType
 {
 public:
@@ -47,6 +49,8 @@ class DataType
    */
   virtual int compare(const Value &left, const Value &right) const { return INT32_MAX; }
 
+  virtual int compare(const Column &left, const Column &right, int left_idx, int right_idx) const { return INT32_MAX; }
+
   /**
    * @brief 计算 left + right，并将结果保存到 result 中
    */

src/observer/common/type/float_type.cpp

@@ -15,16 +15,25 @@ See the Mulan PSL v2 for more details. */
 #include "common/value.h"
 #include "common/lang/limits.h"
 #include "common/value.h"
+#include "storage/common/column.h"
 
 int FloatType::compare(const Value &left, const Value &right) const
 {
-  ASSERT(left.attr_type() == AttrType::FLOATS, "left type is not integer");
+  ASSERT(left.attr_type() == AttrType::FLOATS, "left type is not float");
   ASSERT(right.attr_type() == AttrType::INTS || right.attr_type() == AttrType::FLOATS, "right type is not numeric");
   float left_val  = left.get_float();
   float right_val = right.get_float();
   return common::compare_float((void *)&left_val, (void *)&right_val);
 }
 
+int FloatType::compare(const Column &left, const Column &right, int left_idx, int right_idx) const
+{
+  ASSERT(left.attr_type() == AttrType::FLOATS, "left type is not float");
+  ASSERT(right.attr_type() == AttrType::FLOATS, "right type is not float");
+  return common::compare_float((void *)&((float*)left.data())[left_idx],
+      (void *)&((float*)right.data())[right_idx]);
+}
+
 RC FloatType::add(const Value &left, const Value &right, Value &result) const
 {
   result.set_float(left.get_float() + right.get_float());

src/observer/common/type/float_type.h

@@ -23,6 +23,7 @@ class FloatType : public DataType
   virtual ~FloatType() = default;
 
   int compare(const Value &left, const Value &right) const override;
+  int compare(const Column &left, const Column &right, int left_idx, int right_idx) const override;
 
   RC add(const Value &left, const Value &right, Value &result) const override;
   RC subtract(const Value &left, const Value &right, Value &result) const override;