Question

以下查询需要将近1分钟才能执行。该表有200万条记录。它可以被重写以更快地执行吗？

SELECT  CLIENT_FNM  FROM (SELECT DISTINCT  TRIM(A.CLIENT_FNM)  AS client_fnm 
FROM gen_clientvendor_m A JOIN gen_clientvendor_m B ON (TRIM(A.CLIENT_FNM) = TRIM(B.CLIENT_FNM))
WHERE A.CLIENT_VENDOR_ID <> B.CLIENT_VENDOR_ID
AND A.PARTY_TYPE  = 'C'
AND A.status   = 'A'
ORDER BY TRIM(CLIENT_FNM))
WHERE ROWNUM <= 5

Answer 1

试试这个：

SELECT  a.CLIENT_FNM   
FROM (select distinct TRIM(CLIENT_FNM)CLIENT_FNM, CLIENT_VENDOR_ID, PARTY_TYPE, status from gen_clientvendor_m) A 
JOIN gen_clientvendor_m B ON  A.CLIENT_FNM  = TRIM(B.CLIENT_FNM)
WHERE A.CLIENT_VENDOR_ID <> B.CLIENT_VENDOR_ID
AND A.PARTY_TYPE  = 'C'
AND A.status   = 'A'
ORDER BY TRIM(CLIENT_FNM))
WHERE ROWNUM <= 5

Answer 2

我会聚合TRIM(client_fnm)，以便只读一次表。

select client_fnm
from
(
  select
    trim(client_fnm) as client_fnm,
    row_number() over (order by trim(client_fnm)) as rn
  from gen_clientvendor_m
  group by trim(client_fnm)
  having count(distinct client_vendor_id) > 1
  and count(case when party_type = 'C' and status = 'A' then 1 end) > 0
)
where rn <= 5;

我用标准SQL的ROWNUM替换了Oracle的ROW_NUMBER。（最好是FETCH FIRST n ROWS ONLY，但这只能从Oracle 12c开始提供。）

要快速获得此功能，您应该拥有以下功能索引：

create index idx1 on gen_clientvendor_m(trim(client_fnm));

甚至是覆盖指数：

create index idx2 on gen_clientvendor_m(trim(client_fnm), party_type, status, client_vendor_id);

Answer 3

对于您的查询，我提供以下两个索引：

create index index1 on gen_clientvendor_m(party_type, status, trim(client_fnm), client_vendor_id);

快速找到PARTY_TYPE = 'C' AND status = 'A'的记录，并为加入提供trim(client_fnm)和client_vendor_id。

和

create index index2 on gen_clientvendor_m(trim(client_fnm), client_vendor_id);

快速自我加入这两个领域。

如何在同一个表上使用join从表中仅获取5条记录

3 个答案: