优化 where 子句中的 SQL 子查询

Question

我有 sql 使用 Oracle SQL Developer 可以正常工作，但我担心更大的实时数据库的性能。

请原谅所有加入 tables，这显示了与保存审计历史记录的 _aud table 和 revision_table 的连接。

select cust.forename, cust.surname
from customer cust
join section_details sd on cust.section = sd.section
where
-- this value is substituted programatically
sd.type = 5 and
(
  (select max(rt.timestamp)
  from
  customer cust_tmp1
  join section_details sd on cust_tmp1.section = sd.section
  join measure msr on sd.measure = msr.id
  join measure_aud msr_a on msr.id = msr_a.id
  join revision_table rt on msr_a.rev = rt.id
  where
  cust_tmp1.id = cust.id and msr.measure = 'Last Contact Date')
> 
  (select max(rt.timestamp)
  from
  customer cust_tmp2
  join section_details sd on cust_tmp2.section = sd.section
  join measure msr on sd.measure = msr.id
  join measure_aud msr_a on msr.id = msr_a.id
  join revision_table rt on msr_a.rev = rt.id
  where
  cust_tmp2.id = cust.id and  msr.measure = 'Last Order Date')
);

简而言之，如果 'Last Contact Date' 比 'Last Order Date' 更新，我只想检索客户详细信息。我最终检查了一个 select 的 max(timestamp) 是否大于另一个 select.

的 max(timestamp)

一切正常，我得到了我期待的结果。

除了 msr.measure = 'Last Contact Date' 或 'Last Order Date'.

之外，时间戳比较的每一侧都是重复的

我尝试了一些从未真正起作用的替代方案，因为它们导致了多级嵌套子查询，而且我无法传入外部客户记录。

任何进一步的想法将不胜感激。

Answer 1

您可以将其切换为 exists:

exists (select max(rt.timestamp)
        from customer cust_tmp1 join
             section_details sd
             on cust_tmp1.section = sd.section join
             measure msr
             on sd.measure = msr.id join
             measure_aud msr_a
             on msr.id = msr_a.id join
             revision_table rt
             on msr_a.rev = rt.id
        where cust_tmp1.id = cust.id and
              msr.measure in ( 'Last Contact Date', 'Last Order Date')
        having max(case when msr.measure = 'Last Contact Date' then rt.timestamp end) >
               max(case when msr.mesure = 'Last Order Date' then rt.timestamp end)
       );

我不确定是否会有很大的性能提升。如果您在数据库中设置了正确的索引，那么您的原始版本应该非常快且可扩展。

Answer 2

您确定需要所有这些联接吗？

具有不同 WHERE 条件的相同查询通常转换为 CASE：

select cust.forename, cust.surname
from customer cust
where
  -- this value is substituted programatically
   sd.type = 5 
and exists
 ( select *
   from
   customer cust_tmp1
   join section_details sd on cust_tmp1.section = sd.section
   join measure msr on sd.measure = msr.id
   join measure_aud msr_a on msr.id = msr_a.id
   join revision_table rt on msr_a.rev = rt.id
   where cust_tmp1.id = cust.id

   having max(case when msr.measure = 'Last Contact Date' then rt.timestamp end)
        > max(case when msr.measure = 'Last Order Date'   then rt.timestamp end)
 )

或简化删除子查询：

select cust.forename, cust.surname
from customer cust
join section_details sd on cust_tmp1.section = sd.section
join measure msr on sd.measure = msr.id
join measure_aud msr_a on msr.id = msr_a.id
join revision_table rt on msr_a.rev = rt.id
where
  -- this value is substituted programatically
   sd.type = 5 
group by cust.forename, cust.surname
having max(case when msr.measure = 'Last Contact Date' then rt.timestamp end)
     > max(case when msr.measure = 'Last Order Date'   then rt.timestamp end)

Answer 3

我尝试了此处发布的解决方案，它们似乎都有效，非常感谢您的回复 - 我之前没有研究过 having 子句。

我添加了所有必需的索引并运行解释了所有选项的计划，我的原始查询成本最低。所以我想我会继续使用这个选项。

优化 where 子句中的 SQL 子查询

Optimising SQL sub query in where clause

sql

oracle

query-optimization

oracle11g