oracle扩展统计信息
SQL> create table t as select mod(rownum,50) as n1,mod(rownum,50) as n2 from dual connect by level<=1000;
SQL> exec dbms_stats.gather_table_stats(user,'t');
SQL> select column_name,histogram from user_tab_col_statistics where table_name='T';
COLUMN_NAME HISTOGRAM
N1 NONE
N2 NONE
看到是没有直方图信息的
SQL> select count(*) from t where n1=42 and n2=42;
COUNT(*)
----------
20
SQL> explain plan for select * from t where n1=42 and n2=42;
已解释。
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 1601196873
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 6 | 3 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| T | 1 | 6 | 3 (0)| 00:00:01 |
--------------------------------------------------------------------------
在评估结果集的时候，oracle认为返回了1行记录，但是实际上是返回的20条记录
SQL> SELECT dbms_stats.create_extended_stats(ownname=>user, tabname=>'t', extension=>'(n1,n2)') from
DBMS_STATS.CREATE_EXTENDED_STATS(OWNNAME=>USER,TABNAME=>'T',EXTENSION=>'(N1,N2)'
--------------------------------------------------------------------------------
SYS_STUBZH0IHA7K$KEBJVXO5LOHAS
SQL> execute dbms_stats.gather_table_stats(user,'t');
PL/SQL 过程已成功完成。
SQL> explain plan for select * from t where n1=42 and n2=42;
已解释。
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 1601196873
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 6 | 3 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| T | 1 | 6 | 3 (0)| 00:00:01 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
1 - filter("N1"=42 AND "N2"=42)
在创建了扩展统计信息后，重新收集，依然还是评估返回了1行
SQL> select table_name,column_name,density,histogram from user_tab_col_statistics where table_name='T';
TABLE_NAME COLUMN_NAME DENSITY HISTOGRAM
-------------------- -------------------- ---------- --------------------
T N1 .0005 FREQUENCY
T N2 .0005 FREQUENCY
T SYS_STUBZH0IHA7K$KEB .02 NONE
JVXO5LOHAS
看到在扩展列上没有直方图
SQL> explain plan for select * from t where n1=42 and n2=42;
已解释。
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 1601196873
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 20 | 120 | 3 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| T | 20 | 120 | 3 (0)| 00:00:01 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
1 - filter("N1"=42 AND "N2"=42)
已选择13行。
再次收集了统计信息，这次的评估是返回了20行，这个是正确的
SQL> select column_name,histogram from user_tab_col_statistics where table_name='T';
COLUMN_NAME HISTOGRAM
-------------------- --------------------
N1 FREQUENCY
N2 FREQUENCY
SYS_STUBZH0IHA7K$KEB FREQUENCY
JVXO5LOHAS
现在扩展列上也有了直方图信息
注意的就是扩展列上要有直方图信息，如果没有的化，优化器可能对返回的记录数的评估还是错误的
扩展列是11g的特性，10g是没有的，但是如果10g上出现了这种2个字段的and出现了平谷错误，要怎么处理呢？
可以加大动态采样的级别，采集更多的样本，达到正确评估数据的目的。参考链接如下：
http://scn.sap.com/community/oracle/blog/2013/06/04/oracle-db-optimizer-part-v--introduction-of-dynamic-sampling-and-why-is-it-used-in-sap-bi-environments-by-sap-default