Oracle 9i和10g在create index和rebuild index的统计信息的区别
在9.2.0.6和10.2.0.4做了个小小的试验，演示9i和10g对于create index和rebuild index时统计信息的区别。这里列出试验的过程，由于结果比较明显，就懒得写太多文字做说明了。其中tbsql是一个常用脚本的集成环境，tbsql tabstat用户输出一个表以及列和索引的信息，其实就是关联dba_tables/dba_indexes/dba_tab_columns的一个查询。
先来看9.2.0.6的情况：
SQL> create table t as select * from all_objects; Table created. SQL> create index t_id on t(object_id); Index created.
$tbsql tabstat t Table Number Empty Average Chain Average OWNER Name of Rows Blocks Blocks Space Count Row Len ---------- ------ -------- ------- ------- -------- -------- -------- SYS T Column Column Distinct Name Details Values Density ------------------ ------------------------ -------- ------- OWNER VARCHAR2(30) NOT NULL OBJECT_NAME VARCHAR2(30) NOT NULL SUBOBJECT_NAME VARCHAR2(30) OBJECT_ID NUMBER(22) NOT NULL DATA_OBJECT_ID NUMBER(22) OBJECT_TYPE VARCHAR2(18) CREATED DATE NOT NULL LAST_DDL_TIME DATE NOT NULL TIMESTAMP VARCHAR2(19) STATUS VARCHAR2(7) TEMPORARY VARCHAR2(1) GENERATED VARCHAR2(1) SECONDARY VARCHAR2(1) 13 rows selected. B Average Average Index Tree Leaf Distinct Leaf Blocks Data Blocks Cluster Name Unique Level Blks Keys Per Key Per Key Factor --------- --------- ----- ----- -------------- ----------- ----------- ------- T_ID NONUNIQUE Index Column Col Column Name Name Pos Details --------- --------------- ---- ------------------------ T_ID OBJECT_ID 1 NUMBER(22) NOT NULL
可以看到到表，列和索引都没有统计信息。
SQL> alter index t_id rebuild online compute statistics; Index altered.
$tbsql tabstat t Table Number Empty Average Chain Average OWNER Name of Rows Blocks Blocks Space Count Row Len ------- ------ -------- ------- ------- -------- -------- -------- SYS T 25,420 348 0 0 0 100 Column Column Distinct Name Details Values Density ----------------- ------------------------ --------- ------- OWNER VARCHAR2(30) NOT NULL OBJECT_NAME VARCHAR2(30) NOT NULL SUBOBJECT_NAME VARCHAR2(30) OBJECT_ID NUMBER(22) NOT NULL 25,420 0 DATA_OBJECT_ID NUMBER(22) OBJECT_TYPE VARCHAR2(18) CREATED DATE NOT NULL LAST_DDL_TIME DATE NOT NULL TIMESTAMP VARCHAR2(19) STATUS VARCHAR2(7) TEMPORARY VARCHAR2(1) GENERATED VARCHAR2(1) SECONDARY VARCHAR2(1) 13 rows selected. B Average Average Index Tree Leaf Distinct Leaf Blocks Data Blocks Cluster Name Unique Level Blks Keys Per Key Per Key Factor -------- --------- ----- ----- --------- ----------- ----------- ------- T_ID NONUNIQUE 1 56 25,420 1 1 22,731 Index Column Col Column Name Name Pos Details ------- ---------- ---- ------------------------ T_ID OBJECT_ID 1 NUMBER(22) NOT NULL
注意到表，object_id列和索引都有统计信息了
再来看10.2.0.4的情况
SQL> create table t as select * from all_objects; Table created. SQL> create index t_id on t(object_id); Index created.
$tbsql tabstat t Table Number Empty Average Chain Average OWNER Name of Rows Blocks Blocks Space Count Row Len ---------- ------- -------- ------- ------- -------- ------ -------- SYS T Column Column Distinct Name Details Values Density ---------------- ----------------------- --------- ------- OWNER VARCHAR2(30) NOT NULL OBJECT_NAME VARCHAR2(30) NOT NULL SUBOBJECT_NAME VARCHAR2(30) OBJECT_ID NUMBER(22) NOT NULL DATA_OBJECT_ID NUMBER(22) OBJECT_TYPE VARCHAR2(19) CREATED DATE NOT NULL LAST_DDL_TIME DATE NOT NULL TIMESTAMP VARCHAR2(19) STATUS VARCHAR2(7) TEMPORARY VARCHAR2(1) GENERATED VARCHAR2(1) SECONDARY VARCHAR2(1) 13 rows selected. B Average Average Index Tree Leaf Distinct Leaf Blocks Data Blocks Cluster Name Unique Level Blks Keys Per Key Per Key Factor ------- --------- ----- ------ --------- ----------- ----------- -------- T_ID NONUNIQUE 1 21 9,610 1 1 134 Index Column Col Column Name Name Pos Details --------- ------------ ---- ------------------------ T_ID OBJECT_ID 1 NUMBER(22) NOT NULL
注意到表和列都没有统计信息，而索引有统计信息
SQL> alter index t_id rebuild online compute statistics; Index altered.
$tbsql tabstat t Table Number Empty Average Chain Average OWNER Name of Rows Blocks Blocks Space Count Row Len ------- ------- -------- ------- ------- -------- ------- -------- SYS T Column Column Distinct Name Details Values Density ----------------- ------------------------ ------------ ------- OWNER VARCHAR2(30) NOT NULL OBJECT_NAME VARCHAR2(30) NOT NULL SUBOBJECT_NAME VARCHAR2(30) OBJECT_ID NUMBER(22) NOT NULL DATA_OBJECT_ID NUMBER(22) OBJECT_TYPE VARCHAR2(19) CREATED DATE NOT NULL LAST_DDL_TIME DATE NOT NULL TIMESTAMP VARCHAR2(19) STATUS VARCHAR2(7) TEMPORARY VARCHAR2(1) GENERATED VARCHAR2(1) SECONDARY VARCHAR2(1) 13 rows selected. B Average Average Index Tree Leaf Distinct Leaf Blocks Data Blocks Cluster Name Unique Level Blks Keys Per Key Per Key Factor -------- --------- ----- ------ ---------- ----------- ----------- -------- T_ID NONUNIQUE 1 21 9,610 1 1 134 Index Column Col Column Name Name Pos Details ---------- ------------- ---- ------------------------ T_ID OBJECT_ID 1 NUMBER(22) NOT NULL
注意到表和列还是没有统计信息，索引有统计信息。
应该说，10g的处理方式要更加合理一些。我只是alert index rebuild online compute statistics，9i却同时收集了表，列和索引的统计信息，这有点属于自做主张，而10g则正确的按照语法，只计算了索引的统计信息。有个时候，收集列的统计信息，可能导致执行计划选择错误，尤其是这个列是递增列的时候，一定要注意到9i和10g的这个差异。