Question

我有以下更新声明

UPDATE CONTRACTSF CF
SET CF.PAYDATE = (SELECT MAX(FLHTEMP.PAYDATE)
                      FROM FLHTEMP
                      WHERE
                      FLHTEMP.FLHDATE = CF.FLHDATE
                      AND FLHTEMP.FLHCOD = CF.FLHCOD
                      AND FLHTEMP.FLHUN = CF.FLHUN
                      AND FLHTEMP.FLHCONTRACT = CF.FLHCONTRACT)
WHERE CF.FLHUN || CF.FLHCONTRACT || CF.FLHDATE 
        IN (SELECT DISTINCT FT.FLHUN || FT.FLHCONTRACT || FT.FLHDATE 
            FROM FLHTEMP FT)

当FLHTEMP表有大量记录（6M记录）时，需要很长时间才能完成。

我尝试将IN用于多列，而不是将其与||连接但没有区别。

我创建了一个使用的所有列的索引（FLHUN，FLHCONTRACT，FLHCOD，FLHDATE），但是Plan解释显示我没有使用它。

计划解释中显示的代价高昂的操作是散列连接，排序唯一和表格完全访问

Answer 1

查询的性能取决于多种因素。如

表格的数据量
已加入列的访问类型
表格访问类型（全扫描/索引扫描/..)

在不知道表结构和解释计划的情况下很难优化查询。

此处的WHERE子句是不必要的，如果您不想更新三列不匹配的记录，那么

尝试更新CONTRACTSF的所有行：NVL将保留以后加入条件不匹配的值。

UPDATE CONTRACTSF CF SET CF.PAYDATE = (SELECT NVL(MAX(FLHTEMP.PAYDATE),CF.PAYDATE) FROM FLHTEMP WHERE FLHTEMP.FLHDATE = CF.FLHDATE AND FLHTEMP.FLHCOD = CF.FLHCOD AND FLHTEMP.FLHUN = CF.FLHUN AND FLHTEMP.FLHCONTRACT = CF.FLHCONTRACT)
或使用EXISTS代替IN

UPDATE CONTRACTSF CF SET CF.PAYDATE = (SELECT MAX (FLHTEMP.PAYDATE) FROM FLHTEMP WHERE FLHTEMP.FLHDATE = CF.FLHDATE AND FLHTEMP.FLHCOD = CF.FLHCOD AND FLHTEMP.FLHUN = CF.FLHUN AND FLHTEMP.FLHCONTRACT = CF.FLHCONTRACT) WHERE EXISTS (SELECT 1 FROM FLHTEMP FT WHERE FLHTEMP.FLHDATE = CF.FLHDATE AND FLHTEMP.FLHUN = CF.FLHUN AND FLHTEMP.FLHCONTRACT = CF.FLHCONTRACT)

Answer 2

试试这个。它需要一次flhemp扫描，一次聚合，然后一次散列连接。

merge
 into contractsf cf
using (select flhdate, flhcod, flhun, flhcontract
             ,max(flhtemp.paydate) as max_paydate
         from flhtemp
        group by flhdate, flhcod, flhun, flhcontract
      ) fl
    on(    cf.flhdate     = fl.flhdate
       and cf.flhcod      = fl.flhcod
       and cf.flhun       = fl.flhun
       and cf.flhcontract = fl.flhcontract
      )    
when matched then
   update
      set cf.paydate = fl.max_paydate;

如何优化此Update语句？

2 个答案: