这是查询
UPDATE hrs.rns_recon_ho c SET c.refr_numb =
(
SELECT seqn_numb FROM hrs.rns_recon_ho p WHERE p.narr_1 = c.narr_1
AND p.seqn_numb = p.refr_numb AND p.prod_code = 0
)
Where c.prod_code = 0 And c.refr_numb = 0
AND c.narr_1 = '3/13/201211013198693442091'
这是执行计划
查询的哪一部分导致TABLE ACCESS(FULL)
?
修改
我的错误,我应该更清楚。现在让我清楚一点,我需要知道如何将执行计划的各个部分与查询的各个部分联系起来。例如
TABLE ACCESS(FULL)
TABLE ACCESS(BY GLOBAL INDEX ROWID)
但是,显然INDEX(...)...PROD_CODE
是由PROD_CODE字段引起的。
是否有任何经验法则,指南或技巧可以识别?或者,我们无法知道,因为优化器不会告诉我们为什么选择某条路径。
答案 0 :(得分:0)
也许可以通过提示在一张桌子上强制进行全表扫描,例如
UPDATE hrs.rns_recon_ho c SET c.refr_numb =
(
SELECT /*+ FULL(p) */ seqn_numb FROM hrs.rns_recon_ho p WHERE p.narr_1 = c.narr_1
AND p.seqn_numb = p.refr_numb AND p.prod_code = 0
)
Where c.prod_code = 0 And c.refr_numb = 0
AND c.narr_1 = '3/13/201211013198693442091'
或
UPDATE /*+ FULL(c) */ hrs.rns_recon_ho c SET c.refr_numb =
(
SELECT seqn_numb FROM hrs.rns_recon_ho p WHERE p.narr_1 = c.narr_1
AND p.seqn_numb = p.refr_numb AND p.prod_code = 0
)
Where c.prod_code = 0 And c.refr_numb = 0
AND c.narr_1 = '3/13/201211013198693442091'
如果您的执行计划显示两个“全扫描”,您知道它最初是由另一个引起的。
答案 1 :(得分:0)
让我重新创作:
CREATE TABLE RNS_RECON_HO ( PROD_CODE NUMBER,
REFR_NUMB NUMBER,
NARR_1 VARCHAR2 ( 25 ),
SEQN_NUMB NUMBER primary key );
CREATE INDEX TESTTT ON RNS_RECON_HO (PROD_CODE);
CREATE INDEX TESTTT1 ON RNS_RECON_HO (REFR_NUMB);
CREATE INDEX TESTTT2 ON RNS_RECON_HO (NARR_1);
现在复制你的计划:
SET AUTOTRACE ON
UPDATE
RNS_RECON_HO C
SET
C.REFR_NUMB =
( SELECT
SEQN_NUMB
FROM
RNS_RECON_HO P
WHERE
P.NARR_1 = C.NARR_1
AND P.SEQN_NUMB = P.REFR_NUMB
AND P.PROD_CODE = PROD_CODE )
WHERE
C.PROD_CODE = 0
AND C.REFR_NUMB = 0
AND C.NARR_1 = '3/13/201211013198693442091';
0 rows updated.
Execution Plan
----------------------------------------------------------
0 UPDATE STATEMENT Optimizer Mode=ALL_ROWS (Cost=2 Card=1 Bytes=78)
1 0 UPDATE RNS_RECON_HO
2 1 TABLE ACCESS FULL RNS_RECON_HO (Cost=2 Card=1 Bytes=78)
3 1 TABLE ACCESS BY INDEX ROWID RNS_RECON_HO (Cost=5 Card=1 Bytes=91)
4 3 INDEX RANGE SCAN TESTTT (Cost=1 Card=1)
Statistics
----------------------------------------------------------
190 recursive calls
0 spare statistic 3
0 gcs messages sent
25 db block gets from cache
0 physical reads direct (lob)
0 queue position update
0 queue single row
0 queue ocp pages
0 HSC OLTP Compressed Blocks
0 HSC IDL Compressed Blocks
0 rows processed
<强>原因强>:
如果索引比表扫描更便宜(读取次数更少),优化器将只选择使用索引。这通常意味着WHERE子句标准需要映射到索引的前导(即最左边)列。因此,在where子句列上添加索引(PROD_CODE,REFR_NUMB,NARR_1)将避免这种情况。
确认检查此
CREATE INDEX TESTTT3 ON RNS_RECON_HO (PROD_CODE, REFR_NUMB, NARR_1);
SET AUTOTRACE ON
UPDATE
RNS_RECON_HO C
SET
C.REFR_NUMB =
( SELECT
SEQN_NUMB
FROM
RNS_RECON_HO P
WHERE
P.NARR_1 = C.NARR_1
AND P.SEQN_NUMB = P.REFR_NUMB
AND P.PROD_CODE = PROD_CODE )
WHERE
C.PROD_CODE = 0
AND C.REFR_NUMB = 0
AND C.NARR_1 = '3/13/201211013198693442091';
0 rows updated.
Execution Plan
----------------------------------------------------------
0 UPDATE STATEMENT Optimizer Mode=ALL_ROWS (Cost=1 Card=1 Bytes=78)
1 0 UPDATE RNS_RECON_HO
2 1 INDEX RANGE SCAN TESTTT3 (Cost=1 Card=1 Bytes=78)
3 1 TABLE ACCESS BY INDEX ROWID RNS_RECON_HO (Cost=27 Card=1 Bytes=91)
4 3 INDEX FULL SCAN TESTTT3 (Cost=26 Card=1)
Statistics
----------------------------------------------------------
1 recursive calls
0 spare statistic 3
0 gcs messages sent
0 db block gets from cache
0 physical reads direct (lob)
0 queue position update
0 queue single row
0 queue ocp pages
0 HSC OLTP Compressed Blocks
0 HSC IDL Compressed Blocks
0 rows processed
但这只是为了改变计划,但从未承诺任何性能改进,因为您提供的输入有限。
关注:
外部查询正在进行FTS,因为它忽略了索引。因此,当我们放置一个复合键时,索引对于优化器来说是直截了当的,并且使用相同的索引。子查询使用索引,因为它与索引列上的外部查询连接。
如果您可以通过遵循核心规则接近可能的执行计划,那么如何确定准确的计划无法100%实现。