查询表中最新纪录-减少表扫描量
今天开发给了一条sql，查询表中最近更新的一条记录：
select id,
type,
object_id,
nick,
isv_id,
next_auditor_role,
status,
apply_parameters,
comments,
gmt_create,
gmt_modified
from (select id,
type,
object_id,
nick,
isv_id,
next_auditor_role,
status,
apply_parameters,
comments,
gmt_create,
gmt_modified
from tops_apply
where object_id = 552
and status = 1
order by gmt_create desc)
where rownum = 1
一看这条sql就有优化的余地，内层查询查询了大量无用的数据。
2.创建索引：
17:02:07 SQL> create index ind_tops_apply_oid_sts_create1 on tops_apply(object_id,status,gmt_create)
17:02:30 2 tablespace tbs_top_ind
17:02:30 3 compute statistics;
3.执行sql
17:02:31 SQL> select id,
17:02:42 2 type,
17:02:42 3 object_id,
17:02:42 4 nick,
17:02:42 5 isv_id,
17:02:42 6 next_auditor_role,
17:02:42 7 status,
17:02:42 8 apply_parameters,
17:02:42 9 comments,
17:02:42 10 gmt_create,
17:02:42 11 gmt_modified
17:02:42 12 from (select id,
17:02:42 13 type,
17:02:42 14 object_id,
17:02:42 15 nick,
17:02:42 16 isv_id,
17:02:42 17 next_auditor_role,
17:02:42 18 status,
17:02:42 19 apply_parameters,
17:02:42 20 comments,
17:02:42 21 gmt_create,
17:02:42 22 gmt_modified
17:02:42 23 from tops_apply
17:02:42 24 where object_id = 552
17:02:42 25 and status = 1
17:02:42 26 order by gmt_create desc)
17:02:42 27 where rownum = 1;
13 consistent gets
4.查询有13个逻辑读,改写sql，换成rowid形式来取结果：
17:02:44 SQL> select id,
17:02:59 2 type,
17:02:59 3 object_id,
17:02:59 4 nick,
17:02:59 5 isv_id,
17:02:59 6 next_auditor_role,
17:02:59 7 status,
17:02:59 8 apply_parameters,
17:02:59 9 comments,
17:02:59 10 gmt_create,
17:02:59 11 gmt_modified
17:02:59 12 from tops_apply t, (select rid from (select rowid rid
17:02:59 13 from tops_apply
17:02:59 14 where object_id = 552
17:02:59 15 and status = 1
17:02:59 16 order by gmt_create desc) where rownum<2)tt where tt.rid=t.rowid;
3 consistent gets
5.改写查询后，只有3个逻辑读，但是一个很简单的查询，但写出来的sql有些复杂，能不能有办法改进一下；在执行计划中看到了：
INDEX RANGE SCAN DESCENDING| IND_TOPS_APPLY_OID_STS_CREATE1
oracle很聪明的在order by gmt_create desc的时候对索引进行了descending扫描，在仔细想一想这条sql的目的是找出表中的最新记录， 而索引在创建的时候是有序的，这样就可以利用索引的有序性来查找出结果，oracle也只需要扫描索引的第一行（descending建索引）或者索引最后一行（默认 ASCEDING建索引 ），改写sql：
A.select id,
type,
object_id,
nick,
isv_id,
next_auditor_role,
status,
apply_parameters,
comments,
gmt_create,
gmt_modified
from tops_apply
where object_id = 552
and status = 1 and rownum<2 order by gmt_create desc ;
(index(object_id,status,gmt_create))
执行计划:INDEX RANGE SCAN DESCENDING)
B.select id,
type,
object_id,
nick,
isv_id,
next_auditor_role,
status,
apply_parameters,
comments,
gmt_create,
gmt_modified
from tops_apply
where object_id = 552 and status = 1 and rownum<2;
(index(object_id,status,gmt_create desc))
执行计划:INDEX RANGE SCAN)