本实验基于Oracle10g
先建立一个组合分区表
create table t_sub(id number, data varchar2(20))
partition by range(id)subpartition by hash(data)
(partition p1 values less than (100)
(subpartition p1_s_1,
subpartition p1_s_2
),
partition p2 values less than (maxvalue)
(subpartition p2_s_1,
subpartition p2_s_2
)
);
插入数据
BEGIN
FOR i in 1..200 loop
insert into t_sub values(i,i||'A');
end loop;
END;
SQL> create index i_id on t_sub(id) local nologging;
Index created
SQL> select a.index_name,
2 decode(partitioned,'YES',b.partition_name,'NO')
3 partition,b.subpartition_name subpartition_name,b.num_rows,
4 b.distinct_keys,b.num_rows/b.distinct_keys
5 avg_row_per_key,b.distinct_keys/b.num_rows SELECTIVITY,
6 b.clustering_factor from dba_indexes
7 a,dba_ind_statistics b where a.owner=b.owner
8 and a.index_name=b.index_name and a.owner='&owner' and
9 a.index_name='&index_name';
INDEX_NAME PARTITION SUBPARTITI NUM_ROWS DISTINCT_KEYS AVG_ROW_PER_KEY SELECTIVITY CLUSTERING_FACTOR
---------- ---------- ---------- ---------- ------------- --------------- ----------- -----------------
I_ID P2 P2_S_2 58 58 1 1 1
I_ID P2 P2_S_1 43 43 1 1 1
I_ID P1 P1_S_2 44 44 1 1 1
I_ID P1 P1_S_1 55 55 1 1 1
I_ID P2 101 58 1.7413793103448 0.574257425 2
I_ID P1 99 55 1.8 0.555555555 2
I_ID 200 58 3.4482758620689 0.29 4
7 rows selected
我们都知道，创建索引的时候会自动收集统计信息，以前创建索引如果要收集统计信息还需要加上compute statistics，不过现在不需要了，已经默认会收集。从查询看出，子分区的统计信息 是准确的， 不过分区，以及全局的统计信息 是不准确 的。
注意看 distinct_keys。我插入的数据都没有重复的，不过Oracle却统计出有重复数据。
好了现在我收集统计信息
SQL> BEGIN
2 DBMS_STATS.GATHER_TABLE_STATS(ownname => 'ROBINSON',
3 tabname => 'T_SUB',
4 estimate_percent => DBMS_STATS.AUTO_SAMPLE_SIZE,
5 method_opt => 'for all columns size repeat',
6 degree => DBMS_STATS.AUTO_DEGREE,
7 granularity => 'ALL',
8 cascade=>TRUE
9 );
10 END;
11 /
PL/SQL procedure successfully completed
再次查询 统计信息
SQL> select a.index_name,
2 decode(partitioned,'YES',b.partition_name,'NO')
3 partition,b.subpartition_name subpartition_name,b.num_rows,
4 b.distinct_keys,b.num_rows/b.distinct_keys
5 avg_row_per_key,b.distinct_keys/b.num_rows SELECTIVITY,
6 b.clustering_factor from dba_indexes
7 a,dba_ind_statistics b where a.owner=b.owner
8 and a.index_name=b.index_name and a.owner='&owner' and
9 a.index_name='&index_name';
INDEX_NAME PARTITION SUBPARTITI NUM_ROWS DISTINCT_KEYS AVG_ROW_PER_KEY SELECTIVITY CLUSTERING_FACTOR
---------- ---------- ---------- ---------- ------------- --------------- ----------- -----------------
I_ID P2 P2_S_2 58 58 1 1 1
I_ID P2 P2_S_1 43 43 1 1 1
I_ID P1 P1_S_2 44 44 1 1 1
I_ID P1 P1_S_1 55 55 1 1 1
I_ID P2 101 101 1 1 2
I_ID P1 99 99 1 1 2
I_ID 200 200 1 1 4
7 rows selected
现在准确了。
从这个实验，可以看出如果表是组合分区表， 那么创建索引之后，最好收集索引的统计信息。