313 lines
9.6 KiB
Markdown
313 lines
9.6 KiB
Markdown
---
|
||
tags: [MySQL, SELECT, 查询执行顺序, GROUP BY, 分页, 窗口函数, CTE]
|
||
create time: 2026-05-16 00:00
|
||
---
|
||
|
||
# DQL — SELECT 全解析
|
||
|
||
## 概述
|
||
|
||
SELECT 是 MySQL 中使用频率最高的语句，也是最容易被误解的语句。理解其内部执行顺序是写出高性能查询的第一步。
|
||
|
||
## SQL 书写顺序 vs 执行顺序
|
||
|
||
```mermaid
|
||
flowchart LR
|
||
W["⑥ SELECT"] --> A["① FROM"]
|
||
A --> B["② JOIN"]
|
||
B --> C["③ ON"]
|
||
C --> D["④ WHERE"]
|
||
D --> E["⑤ GROUP BY"]
|
||
E --> F["⑦ HAVING"]
|
||
F --> G["⑧ DISTINCT"]
|
||
G --> H["⑨ ORDER BY"]
|
||
H --> I["⑩ LIMIT / OFFSET"]
|
||
|
||
style W fill:#C44569,color:#fff
|
||
style A fill:#00B6BC,color:#fff
|
||
style I fill:#4FC08D,color:#fff
|
||
```
|
||
|
||
> [!QUESTION] 为什么理解这个很重要？
|
||
> 因为 SQL 不是按照你写的顺序执行的——而是按数字顺序！这意味着：
|
||
> - `WHERE` 在 `SELECT` 之前执行 → 不能在 WHERE 中使用 SELECT 定义的别名
|
||
> - `GROUP BY` 在 `HAVING` 之前 → WHERE 过滤行，HAVING 过滤分组
|
||
> - `ORDER BY` 在 `LIMIT` 之前 → 先排序再截断
|
||
|
||
### 一个完整的例子
|
||
|
||
```sql
|
||
SELECT
|
||
DATE(created_at) AS order_date, -- ⑥ 计算列
|
||
COUNT(*) AS cnt, -- 聚合函数
|
||
SUM(amount) AS total -- 聚合函数
|
||
FROM orders -- ① 确定数据源
|
||
WHERE status = 'paid' -- ④ 先过滤行
|
||
AND created_at >= '2026-01-01' -- 过滤条件
|
||
GROUP BY DATE(created_at) -- ⑤ 按日期分组
|
||
HAVING COUNT(*) > 5 -- ⑦ 过滤分组
|
||
ORDER BY total DESC -- ⑨ 排序
|
||
LIMIT 10; -- ⑩ 取前 10 页
|
||
```
|
||
|
||
## SELECT 关键字详解
|
||
|
||
### DISTINCT
|
||
|
||
```sql
|
||
-- 去重查询
|
||
SELECT DISTINCT department FROM employees;
|
||
|
||
-- DISTINCT 作用于所有选中的列组合
|
||
SELECT DISTINCT country, city FROM customers;
|
||
-- 返回的是 (country, city) 的唯一组合
|
||
```
|
||
|
||
> [!QUESTION] DISTINCT 一定快吗？
|
||
> 不一定。`DISTINCT` 本质上是 `GROUP BY` 的简化版——MySQL 内部可能通过 temporary table + filesort 去重。当数据量大时，它的代价不亚于一次普通的分组聚合。
|
||
>
|
||
> **替代方案**：如果去重目的是为下拉框提供选项，可以用 `SELECT DISTINCT department FROM employees LIMIT 50;` 限制返回量；更优的做法是在应用层缓存选项列表。
|
||
|
||
### GROUP BY 优化
|
||
|
||
```sql
|
||
-- ✅ 好：GROUP BY 走索引
|
||
-- idx_status_dept = (status, department)，可以直接按 department 分组
|
||
SELECT department, COUNT(*)
|
||
FROM employees
|
||
WHERE status = 'active'
|
||
GROUP BY department;
|
||
|
||
-- ❌ 差：GROUP BY 无法利用索引（LIKE 左模糊导致索引失效）
|
||
SELECT department, COUNT(*)
|
||
FROM employees
|
||
WHERE name LIKE '%chen%' -- 左模糊导致索引失效
|
||
GROUP BY department;
|
||
-- EXPLAIN: Using where; Using temporary
|
||
```
|
||
|
||
### HAVING 与 WHERE 的选择
|
||
|
||
```sql
|
||
-- ✅ WHERE 过滤行（早过滤，减少数据量）
|
||
SELECT department, AVG(salary)
|
||
FROM employees
|
||
WHERE hire_date >= '2024-01-01' -- 先筛选最近入职的人
|
||
GROUP BY department
|
||
HAVING AVG(salary) > 15000; -- 再过滤平均工资
|
||
|
||
-- ❌ 把能放 WHERE 的条件放到 HAVING 里
|
||
-- 虽然结果一样，但效率更低
|
||
SELECT department, AVG(salary)
|
||
FROM employees
|
||
GROUP BY department
|
||
HAVING hire_date >= '2024-01-01' -- 错！HAVING 不能用非聚合列
|
||
AND AVG(salary) > 15000;
|
||
```
|
||
|
||
## 窗口函数 (WINDOW FUNCTIONS)
|
||
|
||
窗口函数是 MySQL 8.0+ 引入的分析利器——它能在**不减少行数**的前提下进行聚合计算。
|
||
|
||
### 排名函数
|
||
|
||
```sql
|
||
-- ROW_NUMBER() / RANK() / DENSE_RANK()
|
||
-- 按部门内工资排名（处理并列名次的三种方式）
|
||
SELECT name, department, salary,
|
||
ROW_NUMBER() OVER(PARTITION BY department ORDER BY salary DESC) AS rn,
|
||
RANK() OVER(PARTITION BY department ORDER BY salary DESC) AS rnk,
|
||
DENSE_RANK() OVER(PARTITION BY department ORDER BY salary DESC) AS drnk
|
||
FROM employees;
|
||
|
||
-- 结果示例：
|
||
-- | name | dept | salary | rn | rnk | drnk |
|
||
-- | Alice | sales | 20000 | 1 | 1 | 1 |
|
||
-- | Bob | sales | 20000 | 2 | 1 | 1 |
|
||
-- | Carol | sales | 18000 | 3 | 3 | 2 |
|
||
```
|
||
|
||
> [!TIP] RANK vs DENSE_RANK 的区别
|
||
> - `RANK(20000, 20000, 18000)` → **1, 1, 3**（跳过第二名）
|
||
> - `DENSE_RANK(20000, 20000, 18000)` → **1, 1, 2**（不跳号）
|
||
> - `ROW_NUMBER` → **1, 2, 3**（永远无并列）
|
||
|
||
### 前后行访问 — LAG / LEAD
|
||
|
||
```sql
|
||
SELECT order_date, amount,
|
||
LAG(amount, 1) OVER(ORDER BY order_date) AS prev_amount, -- 前一天的金额
|
||
LEAD(amount, 1) OVER(ORDER BY order_date) AS next_amount -- 后一天的金额
|
||
FROM daily_sales;
|
||
|
||
-- 计算日环比增长率
|
||
SELECT order_date, amount,
|
||
ROUND((amount - LAG(amount) OVER(ORDER BY order_date)) / LAG(amount) OVER(ORDER BY order_date) * 100, 2) AS growth_pct
|
||
FROM daily_sales;
|
||
```
|
||
|
||
### 累计计算
|
||
|
||
```sql
|
||
-- 累计求和 (Running Total)
|
||
SELECT order_date, amount,
|
||
SUM(amount) OVER(ORDER BY order_date) AS running_total
|
||
FROM daily_sales;
|
||
|
||
-- 当前分区内占比
|
||
SELECT department, name, salary,
|
||
ROUND(salary * 100.0 / SUM(salary) OVER(PARTITION BY department), 2) AS dept_pct
|
||
FROM employees;
|
||
```
|
||
|
||
> [!NOTE] 窗口函数执行时机
|
||
> - 执行顺序在 WHERE、GROUP BY、HAVING **之后**，ORDER BY **之前**
|
||
> - 因此不能用 WHERE 直接过滤窗口函数的结果——需要套一层子查询：
|
||
> ```sql
|
||
> SELECT * FROM (
|
||
> SELECT *, ROW_NUMBER() OVER(PARTITION BY user_id ORDER BY created_at DESC) AS rn
|
||
> FROM orders
|
||
> ) ranked WHERE rn = 1;
|
||
> -- 作用：每个用户的最新一条订单记录
|
||
> ```
|
||
|
||
## 子查询与 CTE
|
||
|
||
### 标量子查询 (Scalar Subquery)
|
||
|
||
返回单一值，可以像普通列一样使用：
|
||
|
||
```sql
|
||
-- WHERE 中的标量子查询
|
||
SELECT name, salary
|
||
FROM employees
|
||
WHERE salary > (SELECT AVG(salary) FROM employees); -- 工资高于公司平均的人
|
||
```
|
||
|
||
### 行子查询 (Row Subquery / IN)
|
||
|
||
```sql
|
||
-- IN 子查询
|
||
SELECT name, department
|
||
FROM employees
|
||
WHERE department IN (SELECT id FROM departments WHERE region = 'APAC');
|
||
|
||
-- EXISTS：关注"是否存在"而非具体数据，比 IN 更高效
|
||
SELECT d.name
|
||
FROM departments d
|
||
WHERE EXISTS (SELECT 1 FROM employees e WHERE e.dept_id = d.id);
|
||
```
|
||
|
||
### CTE (Common Table Expression) — WITH 语法
|
||
|
||
MySQL 8.0+ 推荐用法，比嵌套子查询可读性强很多：
|
||
|
||
```sql
|
||
-- 简单 CTE
|
||
WITH dept_stats AS (
|
||
SELECT department, COUNT(*) AS emp_count, AVG(salary) AS avg_salary
|
||
FROM employees
|
||
GROUP BY department
|
||
)
|
||
SELECT * FROM dept_stats WHERE avg_salary > 15000;
|
||
|
||
-- 递归 CTE：处理层级数据（组织架构、分类树等）
|
||
WITH RECURSIVE org_chart AS (
|
||
-- 锚点成员：根节点
|
||
SELECT id, name, manager_id, 1 AS level
|
||
FROM employees
|
||
WHERE manager_id IS NULL
|
||
|
||
UNION ALL
|
||
|
||
-- 递归成员：逐层展开
|
||
SELECT e.id, e.name, e.manager_id, oc.level + 1
|
||
FROM employees e
|
||
INNER JOIN org_chart oc ON e.manager_id = oc.id
|
||
)
|
||
SELECT * FROM org_chart ORDER BY level, name;
|
||
```
|
||
|
||
> [!QUESTION] CTE vs 派生表？
|
||
> - **可读性**：CTE 命名清晰，逻辑分层；派生表层层嵌套，括号匹配困难
|
||
> - **性能**：MySQL 会将非递归 CTE 优化为临时表或内联展开——多数情况下两者性能一致
|
||
> - **复用**：同一个 CTE 可在一个语句中多次引用（派生表不行）
|
||
|
||
---
|
||
|
||
## 深分页问题
|
||
|
||
这是 MySQL 最著名的性能陷阱之一。
|
||
|
||
```sql
|
||
-- ❌ 灾难级写法：扫描 100 万行后丢弃前 999,990 行
|
||
SELECT * FROM orders LIMIT 999990, 10;
|
||
-- MySQL 需要先定位到第 999990 行，才返回接下来的 10 行
|
||
|
||
-- ✅ 方案一：延迟关联（Deferred Join）
|
||
SELECT o.* FROM orders o
|
||
INNER JOIN (
|
||
SELECT id FROM orders ORDER BY id LIMIT 999990, 10
|
||
) AS tmp ON o.id = tmp.id;
|
||
-- 子查询只扫主键索引（极紧凑），外层再 JOIN 拿完整数据
|
||
```
|
||
|
||
```mermaid
|
||
flowchart LR
|
||
subgraph "传统 LIMIT 999990,10"
|
||
A1["扫描聚簇索引<br/>跳过 999990 行"] --> A2["取出 10 行数据"]
|
||
end
|
||
|
||
subgraph "延迟关联方案"
|
||
B1["扫描聚簇索引<br/>跳过 999990 行"] --> B2["仅提取 10 个主键"]
|
||
B2 --> B3["JOIN 回聚簇索引<br/>精确查找 10 个主键"]
|
||
B3 --> B4["返回结果"]
|
||
end
|
||
|
||
A1 --> A2
|
||
style B1 fill:#00B6BC,color:#fff
|
||
style B2 fill:#00D866,color:#fff
|
||
```
|
||
|
||
### 游标分页（推荐）
|
||
|
||
```sql
|
||
-- 上一页最后一条记录的 id = 999985
|
||
SELECT * FROM orders
|
||
WHERE id > 999985
|
||
ORDER BY id ASC
|
||
LIMIT 10;
|
||
```
|
||
|
||
> [!TIP] 为什么游标分页更优？
|
||
> - `WHERE id > ?` 走索引范围扫描，复杂度 O(log N) 而非 O(N)
|
||
> - 无论在第几页，查询时间恒定
|
||
> - 需要前端传「上一页最后一个 ID」作为下一页的游标
|
||
>
|
||
> **局限性**：不支持跳页（不能直接跳到第 100 页），但这对瀑布流场景足够。
|
||
|
||
## 性能小贴士
|
||
|
||
```sql
|
||
-- ❌ 避免 SELECT *
|
||
SELECT * FROM users WHERE status = 1;
|
||
|
||
-- ✅ 只查需要的列
|
||
SELECT id, username, email FROM users WHERE status = 1;
|
||
-- 好处：减少网络传输、提高 Buffer Pool 命中率、可能触发 Covering Index
|
||
|
||
-- ❌ 函数包裹索引列
|
||
SELECT * FROM users WHERE YEAR(created_at) = 2026;
|
||
|
||
-- ✅ 用范围替代函数
|
||
SELECT * FROM users
|
||
WHERE created_at >= '2026-01-01'
|
||
AND created_at < '2027-01-01';
|
||
```
|
||
|
||
## 关联笔记
|
||
|
||
- [[hhs/MySQL/12-JOIN 原理与优化]] — 深入理解 JOIN 的内部执行机制
|
||
- [[hhs/MySQL/13-子查询与派生表]] — 与 SELECT 密切相关的子查询技术
|
||
- [[hhs/GORM/06-排序与分页]] — GORM 分页 API 与原生 SQL 的差异
|