select 包含 Union 的优化

Question

我很容易 select:

define account_id = 7
select * from A where ACCOUNT_ID = &account_id
UNION
select * from B where ACCOUNT_ID = &account_id;

我想将 account_id 作为另一个 select 的输入，我是这样做的：

select * from A where ACCOUNT_ID in(select accound_id from ACCOUNTS where EMAIL like 'aa@aa.com') -- id 7 returned
UNION
select * from B where ACCOUNT_ID in(select accound_id from ACCOUNTS where EMAIL like 'aa@aa.com')

如何将其优化为仅调用 select accound_id from ACCOUNTS where EMAIL like 'aa@aa.com' 一次？

Answer 1

这就是 WITH 派上用场的地方

WITH ids AS (select account_id from ACCOUNTS where EMAIL like 'aa@aa.com')
select * from A where ACCOUNT_ID in ids
UNION ALL
select * from B where ACCOUNT_ID in ids;

我也改成了UNION ALL，因为这样快多了。

Answer 2

我的第一个问题是 union 是否可以用 union all 代替。所以，我的第一次尝试是使用 exists 和 union all:

select a.*
from a
where exists (select 1
              from accounts aa
              where aa.account_id = a.account_id and
                    aa.email = 'aa@aa.com'
             )
union all
select b.*
from b
where exists (select 1
              from accounts aa
              where aa.account_id = b.account_id and
                    aa.email = 'aa@aa.com'
             );

对于此结构，您需要 accounts(account_id, email) 上的索引。 exists 只是在索引中查找值。这确实需要扫描 a 和 b.

如果查询返回少量行并且您想删除重复项，则 union 并替换 union all。如果它返回大量行——并且每个 table 中没有重复项，并且有一种简单的方法来识别重复项——那么你可以改为：

with cte_a as (
      select a.*
      from a
      where exists (select 1
                    from accounts aa
                    where aa.account_id = a.account_id and
                          aa.email = 'aa@aa.com'
                   )
       )
select cte_a.*
from ctea_a
union all
select b.*
from b
where exists (select 1
              from accounts aa
              where aa.account_id = b.account_id and
                    aa.email = 'aa@aa.com'
             ) and
      not exists (select 1
                  from cte_a
                  where cte_a.? = b.?  -- whatever is needed to identify duplicates
                 );

select 包含 Union 的优化

Optimalization of select containing Union

sql

oracle

union

query-performance