一、窗口函数：数据分析的利器

1. 窗口函数基础概念

窗口函数(Window Function)是MySQL 8.0引入的强大特性，它可以在不减少行数的情况下对数据进行聚合计算和分析

SELECT 
    employee_name,
    department,
    salary,
    RANK() OVER (PARTITION BY department ORDER BY salary DESC) as dept_rank
FROM employees;

核心组件：
PARTITION BY：定义窗口分区（类似GROUP BY但不聚合）
ORDER BY：确定窗口内行的排序
frame_clause：定义窗口框架（ROWS/RANGE BETWEEN）

2. 常用窗口函数分类

排名函数

ROW_NUMBER()：连续编号（1,2,3…）
RANK()：并列排名会跳过后续序号（1,2,2,4…）
DENSE_RANK()：并列排名不跳号（1,2,2,3…）

聚合函数

SUM()/AVG()/COUNT()/MIN()/MAX() OVER()

分布函数

PERCENT_RANK()：相对排名百分比
CUME_DIST()：累积分布值

前后函数

LAG(column, n)：访问前n行数据
LEAD(column, n)：访问后n行数据
FIRST_VALUE()/LAST_VALUE()：窗口首尾值

3.高级窗口框架控制

SELECT 
    date,
    revenue,
    AVG(revenue) OVER (
        ORDER BY date 
        ROWS BETWEEN 2 PRECEDING AND CURRENT ROW
    ) as moving_avg
FROM sales;

框架类型：
ROWS：物理行偏移
RANGE：逻辑值范围
GROUPS：MySQL 8.0.2+支持，按组偏移

二、Common Table Expressions (CTE)：提升查询可读性

基础CTE语法

WITH department_stats AS (
    SELECT 
        department,
        AVG(salary) as avg_salary,
        COUNT(*) as emp_count
    FROM employees
    GROUP BY department
)
SELECT * FROM department_stats WHERE avg_salary > 5000;

递归CTE实现层次查询

WITH RECURSIVE org_hierarchy AS (
    -- 基础查询（锚成员）
    SELECT id, name, manager_id, 1 as level
    FROM employees
    WHERE manager_id IS NULL
    
    UNION ALL
    
    -- 递归查询（递归成员）
    SELECT e.id, e.name, e.manager_id, h.level + 1
    FROM employees e
    JOIN org_hierarchy h ON e.manager_id = h.id
)
SELECT * FROM org_hierarchy;

应用场景：
组织结构图
产品分类树
社交网络关系

CTE优化技巧

MATERIALIZED：强制物化CTE结果
MERGE：将CTE合并到主查询
限制递归深度：SET @@cte_max_recursion_depth = 100;

三、高级JSON处理：应对半结构化数据

JSON创建与修改

-- 创建JSON
SELECT JSON_OBJECT('name', name, 'salary', salary) as emp_json
FROM employees;

-- 修改JSON
UPDATE products 
SET attributes = JSON_SET(attributes, '$.color', 'blue')
WHERE id = 1001;

JSON路径查询

SELECT 
    product_id,
    JSON_EXTRACT(attributes, '$.dimensions.width') as width,
    attributes->>'$.manufacturer' as manufacturer
FROM products
WHERE JSON_CONTAINS(attributes, '"wireless"', '$.features');

JSON聚合函数

SELECT 
    department,
    JSON_ARRAYAGG(JSON_OBJECT('id', id, 'name', name)) as employees
FROM staff
GROUP BY department;

四、索引优化高级技巧

函数索引（MySQL 8.0+）

-- 创建基于表达式的索引
CREATE INDEX idx_name_lower ON employees ((LOWER(name)));

-- 使用时必须完全匹配索引表达式
SELECT * FROM employees WHERE LOWER(name) = 'john';

不可见索引

-- 创建不可见索引（优化器忽略）
CREATE INDEX idx_temp ON orders (customer_id) INVISIBLE;

-- 测试后决定是否可见
ALTER TABLE orders ALTER INDEX idx_temp VISIBLE;

降序索引优化

-- 创建降序索引
CREATE INDEX idx_created_desc ON log_entries (created_at DESC);

-- 适合ORDER BY ... DESC查询
SELECT * FROM log_entries ORDER BY created_at DESC LIMIT 100;

五、高级事务处理

保存点(Savepoint)控制

START TRANSACTION;
INSERT INTO orders (...) VALUES (...);
SAVEPOINT order_created;
UPDATE inventory SET quantity = quantity - 1;
-- 发生错误时
ROLLBACK TO SAVEPOINT order_created;
COMMIT;

多版本并发控制(MVCC)深度优化

-- 使用特定隔离级别
SET TRANSACTION ISOLATION LEVEL REPEATABLE READ;

-- 优化长时间事务
SET TRANSACTION READ ONLY;

锁优化策略

-- 行锁升级为表锁（慎用）
LOCK TABLES orders WRITE;

-- 使用SKIP LOCKED处理高并发
SELECT * FROM jobs 
WHERE status = 'pending'
ORDER BY priority DESC
LIMIT 1 FOR UPDATE SKIP LOCKED;

六、性能分析高级技术

执行计划深度解读

EXPLAIN FORMAT=JSON 
SELECT * FROM orders WHERE customer_id IN (
    SELECT id FROM customers WHERE region = 'APAC'
);

-- 关键指标分析
/* 
"cost_info": {
    "query_cost": "10.25"  -- 总预估成本
},
"table_scan": {
    "rows_examined_per_scan": 1000,
    "rows_produced_per_join": 100,
    "filtered": "10.00"
}
*/

优化器提示(Optimizer Hints)

SELECT /*+ INDEX(orders idx_customer) */ *
FROM orders FORCE INDEX (idx_customer)
WHERE customer_id = 1001;

性能模式(Performance Schema)监控

-- 分析最耗资源的SQL
SELECT * FROM performance_schema.events_statements_summary_by_digest
ORDER BY SUM_TIMER_WAIT DESC LIMIT 10;

-- 查看锁等待
SELECT * FROM performance_schema.events_waits_current
WHERE EVENT_NAME LIKE '%lock%';

七、实战案例：电商数据分析系统

用户购买路径分析

WITH user_journey AS (
    SELECT 
        user_id,
        event_time,
        event_type,
        LAG(event_type, 1) OVER (PARTITION BY user_id ORDER BY event_time) as prev_event,
        LEAD(event_type, 1) OVER (PARTITION BY user_id ORDER BY event_time) as next_event
    FROM user_events
    WHERE event_date = CURDATE()
)
SELECT 
    prev_event,
    event_type,
    next_event,
    COUNT(*) as transition_count
FROM user_journey
GROUP BY prev_event, event_type, next_event
ORDER BY transition_count DESC;

实时库存预警

WITH inventory_status AS (
    SELECT 
        product_id,
        current_stock,
        AVG(current_stock) OVER (PARTITION BY category_id) as category_avg,
        RANK() OVER (PARTITION BY warehouse_id ORDER BY current_stock) as stock_rank
    FROM inventory
)
SELECT product_id, current_stock
FROM inventory_status
WHERE 
    current_stock < (0.2 * category_avg) OR
    stock_rank <= 5; -- 每个仓库库存最低的5个商品

建议

渐进式优化：先确保SQL正确性，再逐步应用高级优化
测试验证：所有优化必须通过真实数据验证
监控迭代：持续监控执行计划变化
适度使用：避免过度复杂化SQL逻辑
版本特性：充分利用MySQL 8.0+的新特性