今天继续给大家分享一个 SQL 优化案例。

问题描述

已知表结构如下：

CREATE TABLE `customer` (
`C_CUSTKEY` int NOT NULL,
`C_NAME` varchar(25) NOT NULL,
`C_ADDRESS` varchar(40) NOT NULL,
`C_NATIONKEY` int NOT NULL,
`C_PHONE` char(15) NOT NULL,
`C_ACCTBAL` decimal(15,2) NOT NULL,
`C_MKTSEGMENT` char(10) NOT NULL,
`C_COMMENT` varchar(117) NOT NULL,
PRIMARY KEY (`C_CUSTKEY`)
) ENGINE=InnoDB;

CREATE TABLE `orders` (
`O_ORDERKEY` int NOT NULL,
`O_CUSTKEY` int NOT NULL,
`O_ORDERSTATUS` char(1) NOT NULL,
`O_TOTALPRICE` decimal(15,2) NOT NULL,
`O_ORDERDATE` date NOT NULL,
`O_ORDERPRIORITY` char(15) NOT NULL,
`O_CLERK` char(15) NOT NULL,
`O_SHIPPRIORITY` int NOT NULL,
`O_COMMENT` varchar(79) NOT NULL
PRIMARY KEY (`C_ORDERKEY`)
) ENGINE=InnoDB;

请问下面的查询语句应该如何优化？

SELECT o_custkey, c_name, sum(o.o_totalprice)
FROM customer c, orders o 
WHERE o_custkey = c_custkey 
GROUP BY o_custkey, c_name
ORDER BY c_name;

思路分析

首先，这个查询的作用是统计每个客户的订单金额汇总，数据来自 orders 表；最后按照客户名称排序，名称来自 customer 表，两个表通过 c_custkey 字段关联。看一下执行计划：

EXPLAIN
select c_custkey, c_name, sum(o.o_totalprice)
from customer c, orders o 
where o_custkey = c_custkey 
group by c_name, c_custkey
order by c_name;

id|select_type|table|partitions|type  |possible_keys|key    |key_len|ref           |rows|filtered|Extra                          |
--+-----------+-----+----------+------+-------------+-------+-------+--------------+----+--------+-------------------------------+
1|SIMPLE     |o    |          |ALL   |             |       |       |              |   1|   100.0|Using temporary; Using filesort|
1|SIMPLE     |c    |          |eq_ref|PRIMARY      |PRIMARY|4      |hr.o.O_CUSTKEY|   1|   100.0|                               |

orders 表使用了 ALL 全表扫描，由于 group by，所以 Extra 包含了临时表（Using temporary）；由于 order by，所以 Extra 包含了排序操作（Using filesort）。这些都会导致性能问题。

第一步，考虑到每个客户的订单金额汇总只需要 orders 表中的数据，可以单独按照 o_custkey 分组，并且基于 o_custkey 和 o_totalprice 创建复合索引，利用索引覆盖扫描优化；然后再和 customer 表关联。另外，c_custkey 是主键，决定了 c_name，延迟关联可以避免 group by 中的 c_name。

-- 创建索引
CREATE INDEX idx_orders_custkey_totalprice ON orders(o_custkey, o_totalprice);

EXPLAIN
select o.o_custkey, c.c_name, o.sumprice
from
(select o_custkey, sum(o_totalprice) sumprice
from orders 
group by o_custkey) o, customer c
where o_custkey = c_custkey 
order by c_name;

id|select_type|table     |partitions|type |possible_keys                |key                          |key_len|ref           |rows|filtered|Extra         |
--+-----------+----------+----------+-----+-----------------------------+-----------------------------+-------+--------------+----+--------+--------------+
1|PRIMARY    |c         |          |ALL  |PRIMARY                      |                             |       |              |   1|   100.0|Using filesort|
1|PRIMARY    |<derived2>|          |ref  |<auto_key0>                  |<auto_key0>                  |4      |hr.c.C_CUSTKEY|   2|   100.0|              |
2|DERIVED    |orders    |          |index|idx_orders_custkey_totalprice|idx_orders_custkey_totalprice|11     |              |   1|   100.0|Using index   |

子查询 o 已经优化完成，Extra 中的 Using index 显示了索引覆盖扫描。剩下的问题就是 customer 表上的排序操作，对应 Extra 中的 Using filesort。

第二步，基于 c_name 字段创建索引，避免排序操作。

-- 创建索引
CREATE INDEX idx_customer_name ON customer(c_name);

EXPLAIN
select c.c_custkey, c.c_name, o.sumprice
from
(select o_custkey, sum(o_totalprice) sumprice
from orders 
group by o_custkey) o, customer c
where o_custkey = c_custkey 
order by c_name;

id|select_type|table     |partitions|type |possible_keys                |key                          |key_len|ref           |rows|filtered|Extra      |
--+-----------+----------+----------+-----+-----------------------------+-----------------------------+-------+--------------+----+--------+-----------+
1|PRIMARY    |c         |          |index|PRIMARY                      |idx_customer_name            |102    |              |   1|   100.0|Using index|
1|PRIMARY    |<derived2>|          |ref  |<auto_key0>                  |<auto_key0>                  |4      |hr.c.C_CUSTKEY|   2|   100.0|           |
2|DERIVED    |orders    |          |index|idx_orders_custkey_totalprice|idx_orders_custkey_totalprice|11     |              |   1|   100.0|Using index|

可以看到，对于 customer 表的访问，也使用了索引覆盖扫描，因为二级索引包含了主键字段（c_custkey）。

额外补充，还有一种写法就是调整 group by 字段的顺序，把 c_name 放在第一位，并且把 c_custkey 改成 o_custkey：

EXPLAIN 
select o_custkey, c_name, sum(o.o_totalprice)
from customer c, orders o 
where o_custkey = c_custkey 
group by c_name, o_custkey
order by c_name;

id|select_type|table|partitions|type |possible_keys                |key                          |key_len|ref           |rows|filtered|Extra      |
--+-----------+-----+----------+-----+-----------------------------+-----------------------------+-------+--------------+----+--------+-----------+
1|SIMPLE     |c    |          |index|PRIMARY,idx_customer_name    |idx_customer_name            |102    |              |   1|   100.0|Using index|
1|SIMPLE     |o    |          |ref  |idx_orders_custkey_totalprice|idx_orders_custkey_totalprice|4      |hr.c.C_CUSTKEY|   1|   100.0|Using index|

留个作业，为什么呢？