SQL 使用 OR 子句的查询性能

Question

这个 SQL 查询需要 10 秒：

SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON((e.perimeterId = a.idA AND e.level = 3) 
          OR (e.perimeterId = b.idB AND e.level = 2) 
          OR (e.perimeterId = c.idC AND e.level = 1)
          OR (e.perimeterId = d.idD AND e.level = 0))

将 OR 子句更改为 CASE，查询需要 3 秒：

SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON (
               CASE e.level
                    when 3 then a.idA
                    when 2 then b.idB
                    when 1 then c.idC
                    when 0 then d.idD
               END
             ) = e.perimeterId

如果我用单独的 OR 子句执行查询，查询会立即执行使用我们的 OR 子句：

SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON (e.perimeterId = a.idX AND e.level = X)

如何重写我的查询以立即执行或以高性能执行？

Answer 1

您只需要来自 e 的列，所以我会使用 exists:

来编写

SELECT e.*
FROM E e
WHERE EXISTS (SELECT 1
              FROM a
              WHERE e.perimeterId = a.idA AND e.level = 3
             ) OR
      EXISTS (SELECT 1
              FROM a JOIN
                   b
                   ON a.idB = b.idB
              WHERE e.perimeterId = b.idA AND e.level = 2
             ) OR
      EXISTS (SELECT 1
              FROM a JOIN
                   b
                   ON a.idB = b.idB JOIN
                   c
                   ON c.idC = b.idC
              WHERE e.perimeterId = c.idA AND e.level = 1
             ) OR
      EXISTS (SELECT 1
              FROM a JOIN
                   b
                   ON a.idB = b.idB JOIN
                   c
                   ON c.idC = b.idC JOIN
                   d
                   ON d.idD = c.idD
              WHERE e.perimeterId = d.idA AND e.level = 0
             );

我不是 100% 确定子查询中需要 JOIN，但无论如何我都保留了它们。

Answer 2

1)在a.idB、b.idB、b.idC、c.idC、c.idD、d.idD、[=21=上创建索引], e.level

2) 试试这个

SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON(e.perimeterId = a.idA AND e.level = 3) 
UNION ALL
SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON(e.perimeterId = a.idB AND e.level = 2) 
UNION ALL
SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON(e.perimeterId = a.idC AND e.level = 1) 
UNION ALL
SELECT * FROM A a
JOIN B b on a.idB = b.idB
JOIN C c on b.idC = c.idC
JOIN D d on c.idD = d.idD
JOIN E e  ON(e.perimeterId = a.idD AND e.level = 0)

Answer 3

大多数性能问题都可以通过构建良好的索引来解决，如果您不必创建相应的过滤器列，则在您的子句中始终尝试使用主索引。

SQL 使用 OR 子句的查询性能

SQL query performance with OR clause

sql

sql-server-2017