网友QQ发来如下信息，问我下面这个SQL是否能有性能提升的地方，他们监控到这个SQL磁盘读很高
SQL> set autotrace on
SQL> select * from( select GRDL_ID qyid,KHMC nsrmc,KHBM nsrsbh,KHBM six_nsrsbh,'' six_dssh,gjc,fzgs_dm,'2' khlx from khgl_grdlxx
2 union all
3 select dwkh_id qyid, khmc nsrmc,nvl(nsrsbh,dssh) nsrsbh,nvl(six_nsrsbh,six_dssh) six_nsrsbh,six_dssh,gjc,fzgs_dm,'0' khlx from KHGL_DWKH_COREINFO
4 union all
5 select DLS_BM qyid,DLS_MC nsrmc,DLS_BM nsrsbh,DLS_BM six_nsrsbh,'' six_dssh,gjc,fzgs_dm,'1' khlx from KHGL_DLSJBXX
6 )
7 where (six_nsrsbh = '706773' or six_dssh = '706773') and rownum<11;
QYID NSRMC
-------------------------------- -------------------------------------------------------------------------------------
a4af925f2a224bc4a8ac42dc87bb5192 农一师塔里木广告信息公司
b59f82aa67ae4b88b331a4042e1ced43 保定市丽景园林绿化有限公司
6db2f2cc8d8446a2b80f190fffd1ff72 保定市绿景园林绿化有限公司
执行计划
----------------------------------------------------------
Plan hash value: 112825667
-------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 10 | 1320 | 18 (0)| 00:00:01 |
|* 1 | COUNT STOPKEY | | | | | |
| 2 | VIEW | | 32 | 4224 | 18 (0)| 00:00:01 |
| 3 | UNION-ALL | | | | | |
|* 4 | TABLE ACCESS FULL| KHGL_GRDLXX | 11 | 803 | 5 (0)| 00:00:01 |
|* 5 | TABLE ACCESS FULL| KHGL_DWKH_COREINFO | 11 | 1144 | 7 (0)| 00:00:01 |
|* 6 | TABLE ACCESS FULL| KHGL_DLSJBXX | 10 | 380 | 6 (0)| 00:00:01 |
-------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<11)
4 - filter(''='706773' OR "KHBM"='706773')
5 - filter(NVL("SIX_NSRSBH","SIX_DSSH")='706773' OR "SIX_DSSH"='706773')
6 - filter(''='706773' OR "DLS_BM"='706773')
统计信息
----------------------------------------------------------
0 recursive calls
0 db block gets
27226 consistent gets
26937 physical reads
0 redo size
1288 bytes sent via SQL*Net to client
469 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
3 rows processed
根据SQL的执行计划和统计信息可以看出，这个SQL效率确非常低下，3个表都采用了全表扫描，这个必然导致物理读高。
仔细分析执行计划以及SQL语句，我真的无语了，这个SQL语句写得之烂让我情何以堪，于是我做如下改写
select *
from (select GRDL_ID qyid,
KHMC nsrmc,
KHBM nsrsbh,
KHBM six_nsrsbh,
'' six_dssh,
gjc,
fzgs_dm,
'2' khlx
from khgl_grdlxx
where KHBM = '706773'
union all
select dwkh_id qyid,
khmc nsrmc,
nvl(nsrsbh, dssh) nsrsbh,
nvl(six_nsrsbh, six_dssh) six_nsrsbh,
six_dssh,
gjc,
fzgs_dm,
'0' khlx
from KHGL_DWKH_COREINFO where (six_nsrsbh = '706773' or
six_dssh = '706773')
union all
select DLS_BM qyid,
DLS_MC nsrmc,
DLS_BM nsrsbh,
DLS_BM six_nsrsbh,
'' six_dssh,
gjc,
fzgs_dm,
'1' khlx
from KHGL_DLSJBXX
where DLS_BM = '706773')
where rownum < 11;
改写之后，SQL执行计划如下：
执行计划
----------------------------------------------------------
Plan hash value: 1863808260
----------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time
----------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 8 | 1056 | 12 (0)| 00:00:
|* 1 | COUNT STOPKEY | | | | |
| 2 | VIEW | | 8 | 1056 | 12 (0)| 00:00:
| 3 | UNION-ALL | | | | |
| 4 | TABLE ACCESS BY INDEX ROWID | KHGL_GRDLXX | 1 | 73 | 2 (0)| 00:00:
|* 5 | INDEX UNIQUE SCAN | IDX_GRDLXX_KHBM | 1 | | 1 (0)| 00:00:
| 6 | TABLE ACCESS BY INDEX ROWID | KHGL_DWKH_COREINFO | 6 | 624 | 8 (0)| 00:00:
| 7 | BITMAP CONVERSION TO ROWIDS | | | | |
| 8 | BITMAP OR | | | | |
| 9 | BITMAP CONVERSION FROM ROWIDS| | | | |
|* 10 | INDEX RANGE SCAN | IDX_DWKHCORE_SIX_DSSH | | | 3 (0)| 00:00:
| 11 | BITMAP CONVERSION FROM ROWIDS| | | | |
|* 12 | INDEX RANGE SCAN | IDX_DWKHCORE_SIXNSRSBH | | | 3 (0)| 00:00:
| 13 | TABLE ACCESS BY INDEX ROWID | KHGL_DLSJBXX | 1 | 38 | 2 (0)| 00:00:
|* 14 | INDEX UNIQUE SCAN | PK_KHGL_DLSJBXX | 1 | | 1 (0)| 00:00:
----------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(ROWNUM<11)
5 - access("KHBM"='706773')
10 - access("SIX_DSSH"='706773')
12 - access("SIX_NSRSBH"='706773')
14 - access("DLS_BM"='706773')
统计信息
----------------------------------------------------------
15 recursive calls
0 db block gets
17 consistent gets
0 physical reads
0 redo size
1288 bytes sent via SQL*Net to client
469 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
3 rows processed
为什么要这么改写呢？ 因为不改写的执行计划里面有''='706773' 这个会限制使用索引
其实这里ORACLE用到的CBO转换技术有1个，叫做Pushing Predicate(谓词推入)，但是在第一个SQL语句中，由于有这样的SELECT 条件'' six_dssh
导致谓词推入的时候过滤也成了''='706773' ，从这里也知道CBO还是不够智能，如果能智能一点，帮我们把''='706773' 这样的过滤省略了岂不是更好
