Mysql中in和not in子查询的优化

可以把in改为关联查询，也可以把它改成exists

如需求：查询所有部门的所有员工：

#in子查询
SELECT * FROM tb_user WHERE dept_id IN (SELECT id FROM tb_dept);

#这样写等价于：

#先查询部门表
SELECT id FROM tb_dept

#再由部门dept_id，查询tb_user的员工
SELECT * FROM tb_user u,tb_dept d WHERE u.dept_id = d.id

not in使用 not exists 进行替代

# 比如语句
SELECT * FROM tb_case WHERE SYMBOL!= 9 AND 
CASENO NOT IN (SELECT CASENO FROM tb_carout WHERE STATUS=25 AND SYMBOL!=9)


# 该语句之所以效率低下，很大一部分原因是Not IN不走索引。
# 1：可以优化成这样
SELECT COUNT(*) FROM tb_case a WHERE SYMBOL!= 9 and 
NOT exists 
(SELECT CASENO FROM tb_carout b WHERE a.caseno=b.caseno and STATUS=25 AND SYMBOL!=9)


# 也可以使用left join进行优化
SELECT * FROM tb_case a 
left join 
    tb_carout b 
    on a.caseno=b.caseno and b.STATUS=25 and b.SYMBOL!=9 where b.caseno is null and a.symbol!=9 

循环嵌套查询执行原理
循环由外向内，外层循环执行一次，内层循环则需要完整的执行一次，内层执行执行完后返回执行结果，外层循环继续执行，直到外层循环完全执行完成

循环优化策略
有了上面的执行原理的说明，我们明白了一个道理：内层循环次数的多少不会影响到外层的次数，但是外层循环的次数直接会影响到内层循环的次数，外层循环每多一次，内层循环就需要多完整的一次循环，所以我们优化的目标其实就是使外层的循环次数尽量少，总结来说：小表驱动大表。小表就是外层循环，大表就是内层循环，也就是尽量减少外层循环的次数

IN和Exists的区别

exists ： 外表先进行循环查询，将查询结果放入exists的子查询中进行条件验证，确定外层查询数据是否保留

in ： 先查询内表，将内表的查询结果当做条件提供给外表查询语句进行比较

结论：
 

外层小表，内层大表（或者将sql从左到由来看：左面小表，右边大表）： exists 比 in 的效率高
外层大表，内层小表（或者将sql从左到由来看：左面大表，右边小表）： in 比 exists 的效率高
 

比如

SELECT
	TB_CASE.CASENO AS CASENO,
	TB_CASE.LAWGROUP AS LAWGROUP,
	S1.ROLE_NAME AS LGNAME,
	TB_CASE.CARPARK AS CARPARK,
	S2.ROLE_NAME AS CARPARKNAME,
	TB_CASE.LAWTIME AS LAWTIME,
	TB_CASE.CARTYPE AS CARTYPE,
	TB_CASE.CARNO AS CARNO,
	TB_CASE.STATUS AS STATUS,
	TB_CASE.IS_CLONE AS IS_CLONE,
	TB_CASE.BCFS AS BCFS 
FROM
	TB_CASE,
	SYS_ROLE S1,
	SYS_ROLE S2 
WHERE
	((
			NOT (
				TB_CASE.CASENO IN (
				SELECT
					TB_CAROUT.CASENO 
				FROM
					TB_CAROUT 
				WHERE
					((
							TB_CAROUT.STATUS = '25' 
							) 
					AND ( TB_CAROUT.SYMBOL <> 9 ))))) 
	AND ( TB_CASE.LAWGROUP = S1.ROLE_ID ) 
	AND ( TB_CASE.CARPARK = S2.ROLE_ID ));

他的left join写法

SELECT
	T1.CASENO AS CASENO,
	T1.LAWGROUP AS LAWGROUP,
	S1.ROLE_NAME AS LGNAME,
	T1.CARPARK AS CARPARK,
	S2.ROLE_NAME AS CARPARKNAME,
	T1.LAWTIME AS LAWTIME,
	T1.CARTYPE AS CARTYPE,
	T1.CARNO AS CARNO,
	T1.STATUS AS STATUS,
	T1.IS_CLONE AS IS_CLONE,
	T1.BCFS AS BCFS 
FROM
	TB_CASE T1
	LEFT JOIN TB_CAROUT T2 ON T1.CASENO = T2.CASENO 
	    AND T2.STATUS = '25' 
	    AND T2.SYMBOL != 9
	LEFT JOIN SYS_ROLE S1 ON T1.LAWGROUP = S1.ROLE_ID
	LEFT JOIN SYS_ROLE S2 ON T1.CARPARK = S2.ROLE_ID 
WHERE
	T2.CASENO IS NULL;