Question

PostgreSQL的新手问题。我有以下查询，其中我试图返回符合以下条件的所有记录的联合：

所有dx在tencounter之前都标记为慢性，
所有经常性的dx（即，标记处理的dx记录比之前的任何分辨率更新，
任何标记为已处理但尚未解决的dx记录

有没有更好的方法（可能使用PostgreSQL的WITH子句）？我没有给我们读过＆＃34; NOT IN＆＃34;在PostgreSQL中，这样做有多好？你如何＆＃34;优化＆＃34;这个东西？

   CREATE OR REPLACE FUNCTION f_getactivedx(groupid character varying, tencounter timestamp without time zone)
   RETURNS SETOF view_dx AS
$BODY$

    select max(dx.recid) as recid, dx.cicd9, dx.cdesc, max( dx.tposted) as tposted, 
    bool_or(dx.resolved) as resolved, bool_or(dx.treated) as treated, bool_or(dx.chronic),
    dx.groupid 
        from dx 
    where dx.chronic = true 
    and dx.groupid = $1 
    and date_trunc('day',dx.tposted) <= date_trunc('day',$2)
    group by dx.cicd9, dx.cdesc, dx.groupid

    union


    select max(dx.recid) as recid, dx.cicd9, dx.cdesc, max( dx.tposted) as tposted, 
    bool_and(dx.resolved), bool_and(dx.treated), bool_and(dx.chronic), dx.groupid 
    from dx 
            join    (select cdesc, max(tposted) as tposted from dx  
                       where groupid =$1  and resolved = true and 
                       date_trunc('day',tposted) <=  date_trunc('day', $2)
                       group by cdesc) j 
              on (dx.cdesc = j.cdesc and dx.tposted > j.tposted) 
              where groupid = $1 and treated = true
              and date_trunc('day',dx.tposted)  <= date_trunc('day', $2)
              group by dx.cicd9, dx.cdesc, dx.groupid 

    union 

    select max(dx.recid) as recid, dx.cicd9, dx.cdesc, max( dx.tposted), 
       bool_and(dx.resolved), bool_and(dx.treated), bool_and(dx.chronic), dx.groupid 
              from dx 
              where dx.cdesc NOT IN
                (select cdesc from dx  
                       where groupid =$1  and resolved = true and 
                       date_trunc('day',tposted) <=  date_trunc('day', $2)
                       group by cdesc) 
             and groupid =$1 and treated = true and
              date_trunc('day',tposted)   <= date_trunc('day', $2)        
              group by dx.cicd9, dx.cdesc, dx.groupid

    order by tposted desc, treated desc, resolved desc, cdesc asc

Answer 1

NOT IN可以，你只需仔细考虑一下NULL，反连接通常是更好的选择。

对于任何查询：

SELECT ...
FROM t
WHERE col NOT IN (SELECT col2 FROM t2 WHERE col2 IS NOT NULL AND ...predicate...)

你可以等同地写：

SELECT ...
FROM t LEFT OUTER JOIN t2 ON (t.col = t2.col2 AND ...predicate...)
WHERE t2.col2 IS NULL;

被称为“左反连接”。

PostgreSQL可能会为两者生成相同的查询计划。

除非您知道由于某些原因导致NOT IN使用导致性能问题，否则经过explain analyze的正确检查后，我强烈建议您不要单独使用。

Answer 2

使用NOT EXISTS通常比使用NOT IN更有效。

SELECT A.*
FROM A
WHERE NOT EXISTS (
  SELECT 1
  FROM B
  WHERE A.id = B.id
)

如何优化PostrgreSQL查询并避免使用＆＃34; NOT IN＆＃34;？

2 个答案: