多个左联接优化

Question

表格：

• 员工
• employee_orgn
  • 联合主键（employee_id，orgn_id）
  • 两个索引：key1：employee_id，index2：orgn_id
• orgn

有些员工没有组织。

sql：

explain SELECT DISTINCT
    e.*
FROM
    employee e
        LEFT JOIN
    employee_orgn eo ON eo.employee_id = e.id
        LEFT JOIN
    orgn o ON o.id = eo.orgn_id
WHERE
    e.state != 'deleted'
        AND e.state != 'hidden'
        AND (o.state != 'hidden' OR o.state IS NULL)
ORDER BY e.id DESC

解释：

| id   | select_type    |  table | type | possible_keys | key  | key_len | ref  | rows | Extra |

| 1        | SIMPLE     |   e    |all        | NULL     |   NULL    |NULL        | NULL      |   12792    |Using where;USing tempory;Using filesort       | 

| 1        | SIMPLE     |   eo    |index        | PRIMARY     |   idx_orgn_id    |8        | NULL     |   13226  |Using index:Distinct  | 

| 1        | SIMPLE     |   o    |eq_ref        | PRIMARY      |   PRIMARY    |8        | eo.orgn_id      |   1    |Using where:Distinct           | 

问：

1. 在这里左联接，mysql嵌套循环查询10个数量级8？
2. 为什么有临时表，为什么排序是文件排序？
3. 为什么第二行是覆盖索引
4. 我希望有人能解释这个解释结果并优化分析。

谢谢。

Answer 1

我建议重写查询-摆脱select distinct。我认为这是您想要的逻辑：

SELECT e.*
FROM employee e
WHERE e.state not in ('deleted', 'hidden')
      NOT EXISTS (SELECT 1
                  FROM employee_orgn eo JOIN
                       orgn o
                       ON o.id = eo.orgn_id AND o.state = 'hidden'
                  WHERE eo.employee_id = e.id
                 )
ORDER BY e.id DESC;

对于此查询，您需要在employee_orgn(employee_id, orgn_id)和orgn(id, state)上建立索引。

Answer 2

您不应在WHERE子句中保留联接的表上放置条件。而是将它们放在ON子句中。然后，您无需使用OR o.state IS NULL，这会导致优化程序出现问题。

SELECT DISTINCT
    e.*
FROM
    employee e
        LEFT JOIN
    employee_orgn eo ON eo.employee_id = e.id
        LEFT JOIN
    orgn o ON o.id = eo.orgn_id AND o.state != hidden
WHERE
    e.state NOT IN ('deleted', 'hidden')
ORDER BY e.id DESC