[20180926]查询相似索引.txt
[20180926]查询相似索引.txt
--//有时候在表上建立索引比如A,B字段,可能又建立B字段索引,甚至A字段索引以及B,A字段索引,或者还建立C,A字段索引,
--//需要有1个脚本查询这些索引,可能还有必要删除一些索引,统一协调建立合适的索引.
--//优化需要,做一个记录.
1.环境:
SCOTT@test01p> @ ver1
PORT_STRING VERSION BANNER CON_ID
------------------------------ -------------- -------------------------------------------------------------------------------- ----------
IBMPC/WIN_NT64-9.1.0 12.1.0.1.0 Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production 0
2.建立测试例子:
SCOTT@test01p> create table t (a number,b number,c number);
Table created.
SCOTT@test01p> create index i_t_a_b on t(a,b);
Index created.
SCOTT@test01p> create index i_t_c_b on t(c,b);
Index created.
--//网上找到的例子:
SELECT TABLE_OWNER, TABLE_NAME, INDEX_NAME, COLUMN_NAME
FROM ALL_IND_COLUMNS
WHERE COLUMN_POSITION = 1
AND TABLE_OWNER = UPPER ('&&1')
AND (TABLE_OWNER, TABLE_NAME, COLUMN_NAME) IN (
SELECT TABLE_OWNER, TABLE_NAME, COLUMN_NAME
FROM (
SELECT TABLE_OWNER, TABLE_NAME, COLUMN_NAME, COUNT (*) TCOUNT
FROM ALL_IND_COLUMNS
WHERE TABLE_OWNER NOT IN ('SYS','SYSTEM','OUTLN','DBSNMP')
HAVING COUNT (*) > 1
GROUP BY TABLE_OWNER, TABLE_NAME, COLUMN_NAME))
ORDER BY TABLE_OWNER, TABLE_NAME, COLUMN_NAME, INDEX_NAME;
--//实际上这个例子在我们生产系统根本无法执行,表N多,1个小时都没有查询出来.而且像上面建立的索引是无法找到的.
--//因为它仅仅针对COLUMN_POSITION = 1的情况.
--//使用with改写如下:
/* Formatted on 2018/9/25 22:19:20 (QP5 v5.227.12220.39754) */
WITH t1
AS (SELECT TABLE_OWNER
,TABLE_NAME
,INDEX_NAME
,COLUMN_NAME
,COLUMN_POSITION
FROM ALL_IND_COLUMNS
WHERE TABLE_OWNER = UPPER ('&&1'))
,t2
AS ( SELECT DISTINCT TABLE_OWNER
,TABLE_NAME
,INDEX_NAME
,COLUMN_NAME
FROM t1
WHERE (TABLE_OWNER, TABLE_NAME, COLUMN_NAME) IN
(SELECT TABLE_OWNER, TABLE_NAME, COLUMN_NAME
FROM ( SELECT TABLE_OWNER
,TABLE_NAME
,COLUMN_NAME
,COUNT (*) TCOUNT
FROM T1
HAVING COUNT (*) > 1
GROUP BY TABLE_OWNER, TABLE_NAME, COLUMN_NAME))
ORDER BY TABLE_OWNER
,TABLE_NAME
,COLUMN_NAME
,INDEX_NAME)
,t3
AS ( SELECT TABLE_OWNER
,TABLE_NAME
,INDEX_NAME
,LISTAGG (column_name, ', ')
WITHIN GROUP (ORDER BY column_position)
AS column_group
FROM t1
GROUP BY TABLE_OWNER, TABLE_NAME, INDEX_NAME)
SELECT TABLE_OWNER
,TABLE_NAME
,INDEX_NAME
,column_group
FROM t3
WHERE (TABLE_OWNER, TABLE_NAME, INDEX_NAME) IN
(SELECT TABLE_OWNER, TABLE_NAME, INDEX_NAME FROM t2);
TABLE_OWNER TABLE_NAME INDEX_NAME COLUMN_GROUP
-------------------- -------------------- -------------------- ----------------------
SCOTT T I_T_A_B A, B
SCOTT T I_T_C_B C, B
--//补充:在生产系统使用不到1秒就执行完成.
--//换一个参数OE.
Enter value for 1: OE
old 8: WHERE TABLE_OWNER = UPPER ('&&1'))
new 8: WHERE TABLE_OWNER = UPPER ('OE'))
TABLE_OWNER TABLE_NAME INDEX_NAME COLUMN_GROUP
-------------------- -------------------- -------------------- -------------------------------------------------
OE INVENTORIES INVENTORY_IX WAREHOUSE_ID, PRODUCT_ID
OE INVENTORIES INV_PRODUCT_IX PRODUCT_ID
OE ORDER_ITEMS ITEM_ORDER_IX ORDER_ID
OE ORDER_ITEMS ORDER_ITEMS_PK ORDER_ID, LINE_ITEM_ID
OE ORDER_ITEMS ORDER_ITEMS_UK ORDER_ID, PRODUCT_ID
OE ORDER_ITEMS ITEM_PRODUCT_IX PRODUCT_ID
6 rows selected.
--//如果使用网上的脚本结果如下:
SCOTT@test01p> SELECT TABLE_OWNER, TABLE_NAME, INDEX_NAME, COLUMN_NAME
2 FROM ALL_IND_COLUMNS
3 WHERE COLUMN_POSITION = 1
4 AND TABLE_OWNER = UPPER ('&&1')
5 AND (TABLE_OWNER, TABLE_NAME, COLUMN_NAME) IN (
6 SELECT TABLE_OWNER, TABLE_NAME, COLUMN_NAME
7 FROM (
8 SELECT TABLE_OWNER, TABLE_NAME, COLUMN_NAME, COUNT (*) TCOUNT
9 FROM ALL_IND_COLUMNS
10 WHERE TABLE_OWNER NOT IN ('SYS','SYSTEM','OUTLN','DBSNMP')
11 HAVING COUNT (*) > 1
12 GROUP BY TABLE_OWNER, TABLE_NAME, COLUMN_NAME))
13 ORDER BY TABLE_OWNER, TABLE_NAME, COLUMN_NAME, INDEX_NAME;
old 4: AND TABLE_OWNER = UPPER ('&&1')
new 4: AND TABLE_OWNER = UPPER ('OE')
TABLE_OWNER TABLE_NAME INDEX_NAME COLUMN_NAME
-------------------- -------------------- -------------------- --------------------
OE INVENTORIES INV_PRODUCT_IX PRODUCT_ID
OE ORDER_ITEMS ITEM_ORDER_IX ORDER_ID
OE ORDER_ITEMS ORDER_ITEMS_PK ORDER_ID
OE ORDER_ITEMS ORDER_ITEMS_UK ORDER_ID
OE ORDER_ITEMS ITEM_PRODUCT_IX PRODUCT_ID
--//1.结果不同,存在遗漏.
--//2.明显感觉执行很慢.
--//3.显示不直观.
