如何使用多个连接和冗余标准优化查询？

时间：2013-03-27 06:59:10

标签： sql sql-server sql-server-2008 tsql

我真的希望你能帮我这个。这是我目前的代码，到目前为止，它是我能做的最好的代码。

 SELECT
      max(t2.pmskey) as pmskey, max(cast(t1.recdate as datetime)) as recdate,
      max(t2.mrtype) as mrtype, max(cast(t1.mrdate as datetime)) as mrdate,
      max(t2.CaseNo) as CaseNo, max(t2.pmclin) as pmclin,
      max(cast(t1.nexteval as datetime)) as nexteval,
      max(cast(t1.repdate as datetime)) as repdate,
      max(t3.mrprocedure) as mrprocedure, max(t2.med_stat) as med_stat,
      max(cast(t1.med_stateff as datetime)) as med_stateff,
      max(t2.clincontact) as clincontact,
      max(cast(t1.datemodf as datetime)) as datemodf, max(t2.modfby) as modfby,
      max(cast(t1.inceptiondate as datetime)) as inceptiondate,
      max(t2.createdby) as createdby, max(cast(t1.date_ent as datetime)) as date_ent,
      max(t2.ppihandler) as ppihandler

 FROM
      tblpms as t1

 JOIN
      (
      select * from tblpms where lower(CaseNo) like '%tr13-011%'

 AND
      cast(mrdate as datetime) IN (select max(cast(mrdate as datetime))
      from tblpms where lower(CaseNo) like '%tr13-011%')
      ) as t2

      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

 JOIN
      (
      select * from tblpms where lower(CaseNo) like '%tr13-011%'

 AND
      lower(mrprocedure) is not null and cast(nexteval as datetime)
      in (select max(cast(nexteval as datetime)) from
      tblpms where lower(CaseNo) like '%tr13-011%')
      ) as t3

      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

      and lower(t2.CaseNo) like '%tr13-011%'

lower(CaseNo)的条件在所有联接中被复制，我不知道如何减少这些条件。我相信有一些事情可以做。如果还有其他可以优化的内容，请将其包含在答案中。

这是我目前基于以下答案的查询。

 SELECT
      max(t2.pmskey) as pmskey, max(cast(t1.recdate as datetime)) as recdate,
      max(t2.mrtype) as mrtype, max(cast(t1.mrdate as datetime)) as mrdate,
      max(t2.CaseNo) as CaseNo, max(t2.pmclin) as pmclin,
      max(cast(t1.nexteval as datetime)) as nexteval,
      max(cast(t1.repdate as datetime)) as repdate,
      max(t3.mrprocedure) as mrprocedure, max(t2.med_stat) as med_stat,
      max(cast(t1.med_stateff as datetime)) as med_stateff,
      max(t2.clincontact) as clincontact,
      max(cast(t1.datemodf as datetime)) as datemodf, max(t2.modfby) as modfby,
      max(cast(t1.inceptiondate as datetime)) as inceptiondate,
      max(t2.createdby) as createdby, max(cast(t1.date_ent as datetime)) as date_ent,
      max(t2.ppihandler) as ppihandler

 FROM
      tblpms as t1

 JOIN
      (
      select * from tblpms where CaseNo = 'TR13-011-CRW'

 AND
      cast(mrdate as datetime) IN (select max(cast(mrdate as datetime))
      from tblpms where CaseNo = 'TR13-011-CRW')
      ) as t2

      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

 JOIN
      (
      select * from tblpms where CaseNo = 'TR13-011-CRW'

 AND
      lower(mrprocedure) is not null and cast(nexteval as datetime)
      in (select max(cast(nexteval as datetime)) from
      tblpms where CaseNo = 'TR13-011-CRW')
      ) as t3

      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

      and CaseNo = 'TR13-011-CRW'

1 个答案:

答案 0 :(得分：3)

与[{1}}匹配时，lower()通常不需要LIKE，因为它case-insensitive by default。

此外，当在字符串中间执行搜索（例如LIKE '%whatever%）时，不使用索引。 顺便问一下，你有任何索引吗？您是否尝试过构建查询计划？

正如我所看到的，这个问题的解决方案很少：

  1. Use full text search

  2. 添加另一个索引字段并在查询之前填充一次：

    update tblpms set tr13_flag=1 where CaseNo like '%tr13-011%'

    然后用where lower(CaseNo) like '%tr13-011%'

    3. 替换每个where tr13_flag=1语句

  3. 上述方差：使用索引PERSISTED computed column，因此您可以避免手动计算标记。

  4. 将所有有价值的行抓到临时表中一次：

    选择* 进入＃tr13011 来自tblpms，其中较低（CaseNo）喜欢'％tr13-011％'

    5. 并加入它，可能在有趣的字段上添加索引。

    此外，您cast持续：cast(nexteval as datetime)。使用答案中的第4个版本并同时转换为#temp表。

    这种情况对我来说毫无用处：and lower(mrprocedure) is not null and cast(nexteval as datetime)。为什么不简单and mrprocedure is not null？

    版本4的示例 

    select
    CaseNo,
    mrprocedure,
    cast(mrdate as datetime) as mrdate,
    cast(nexteval as datetime) as nexteval
    -- other required fields from tblpms table
into #tr13
from tblpms where CaseNo = 'TR13-011-CRW'

declare @max_mrdate datetime
declare @max_nexteval datetime

select @max_mrdate=max(mrdate), @max_nexteval=max(nexteval) from #tr13

SELECT
      max(t2.pmskey) as pmskey, max(cast(t1.recdate as datetime)) as recdate,
      max(t2.mrtype) as mrtype, max(cast(t1.mrdate as datetime)) as mrdate,
      max(t2.CaseNo) as CaseNo, max(t2.pmclin) as pmclin,
      max(cast(t1.nexteval as datetime)) as nexteval,
      max(cast(t1.repdate as datetime)) as repdate,
      max(t3.mrprocedure) as mrprocedure, max(t2.med_stat) as med_stat,
      max(cast(t1.med_stateff as datetime)) as med_stateff,
      max(t2.clincontact) as clincontact,
      max(cast(t1.datemodf as datetime)) as datemodf, max(t2.modfby) as modfby,
      max(cast(t1.inceptiondate as datetime)) as inceptiondate,
      max(t2.createdby) as createdby, max(cast(t1.date_ent as datetime)) as date_ent,
      max(t2.ppihandler) as ppihandler

 FROM
      tblpms as t1

 JOIN
      (
      select * from #tr13 where mrdate = @max_mrdate
      ) as t2
      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

 JOIN
      (
      select * from #tr13
        where mrprocedure is not null and nexteval = @max_nexteval
      ) as t3
      on t1.CaseNo COLLATE DATABASE_DEFAULT = t2.CaseNo COLLATE DATABASE_DEFAULT

drop table #tr13

    我不确定为什么你需要COLLATE DATABASE_DEFAULT，因为它总是在同一张桌子上。 逻辑对我来说不是很清楚，但似乎产生相同的结果，同时更具可读性（并且可能表现更好）：

    select
    CaseNo,
    mrprocedure,
    cast(mrdate as datetime) as mrdate,
    cast(nexteval as datetime) as nexteval,
    pmskey,
    mrtype,
    med_stat,
    clincontact,
    modfby,
    createdby,
    ppihandler
    -- other required fields from tblpms table
into #tr13
from tblpms where CaseNo = 'TR13-011-CRW'

declare @max_mrdate datetime
declare @max_nexteval datetime

select @max_mrdate=max(mrdate), @max_nexteval=max(nexteval) from #tr13

SELECT
      max(t2.pmskey) as pmskey, max(cast(t1.recdate as datetime)) as recdate,
      max(t2.mrtype) as mrtype, max(cast(t1.mrdate as datetime)) as mrdate,
      max(t2.CaseNo) as CaseNo, max(t2.pmclin) as pmclin,
      max(cast(t1.nexteval as datetime)) as nexteval,
      max(cast(t1.repdate as datetime)) as repdate,
      max(t2.mrprocedure) as mrprocedure, max(t2.med_stat) as med_stat,
      max(cast(t1.med_stateff as datetime)) as med_stateff,
      max(t2.clincontact) as clincontact,
      max(cast(t1.datemodf as datetime)) as datemodf, max(t2.modfby) as modfby,
      max(cast(t1.inceptiondate as datetime)) as inceptiondate,
      max(t2.createdby) as createdby, max(cast(t1.date_ent as datetime)) as date_ent,
      max(t2.ppihandler) as ppihandler

 FROM tblpms as t1
 JOIN #tr13 t2 on t1.CaseNo = t2.CaseNo
where
    t2.mrdate = @max_mrdate
    and t2.mrprocedure is not null
    and t2.nexteval = @max_nexteval

drop table #tr13

    <强>更新 我们设法将其减少为以下代码：

    declare @max_mrdate datetime
declare @max_nexteval datetime

select
    @max_mrdate=max(cast(mrdate as datetime)),
    @max_nexteval=max(cast(nexteval as datetime))
from tblpms
where CaseNo = 'TR13-011-CRW'

select *
from tblpms where CaseNo = 'TR13-011-CRW'
and (
    cast(mrdate as datetime) = @max_mrdate
    or
    (mrprocedure is not null and cast(nexteval as datetime) = @max_nexteval)
)
相关问题
最新问题