Question

我一直在尝试通过添加多个索引来优化查询。我在大多数人中取得了成功，但是有一个查询，我创建的列从未使用过。查询看起来像这样：

SELECT DISTINCT
  c.containername
  , hml.qty
  , CAST(hml.txndate AS TIMESTAMP) txndate
  , hml.comments
  , e.fullname
  , 'ICO' txn
FROM con c
JOIN hml hml ON c.containerid = hml.historyid
JOIN emp e ON hml.employeeid = e.employeeid
WHERE hml.comments LIKE '%ICO%'
      AND hml.txndate >= to_date(:ddate, 'MM/DD/YYYY HH:MI:SS PM')
      AND e.notes IN ('DBP', 'FTD', 'MH', 'SUPERVISOR')
ORDER BY CAST(hml.txndate AS TIMESTAMP);

执行计划

    Plan hash value: 267728100

---------------------------------------------------------------------------------------------------------
| Id  | Operation                        | Name                 | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                 |                      |   116 | 14616 |   248K  (1)| 00:49:40 |
|   1 |  SORT ORDER BY                   |                      |   116 | 14616 |   248K  (1)| 00:49:40 |
|   2 |   HASH UNIQUE                    |                      |   116 | 14616 |   248K  (1)| 00:49:40 |
|   3 |    NESTED LOOPS                  |                      |       |       |            |          |
|   4 |     NESTED LOOPS                 |                      |   116 | 14616 |   248K  (1)| 00:49:40 |
|*  5 |      HASH JOIN                   |                      |   116 | 11252 |   248K  (1)| 00:49:37 |
|*  6 |       TABLE ACCESS FULL          | EMPLOYEE             |     8 |   304 |    25   (0)| 00:00:01 |
|*  7 |       TABLE ACCESS BY INDEX ROWID| HISTORYMAINLINE      | 38373 |  2210K|   248K  (1)| 00:49:37 |
|*  8 |        INDEX RANGE SCAN          | HISTORYMAINLINEMX005 |   347K|       |  1248   (1)| 00:00:15 |
|*  9 |      INDEX UNIQUE SCAN           | CONTAINER450         |     1 |       |     1   (0)| 00:00:01 |
|  10 |     TABLE ACCESS BY INDEX ROWID  | CONTAINER            |     1 |    29 |     2   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   5 - access("HML"."EMPLOYEEID"="E"."EMPLOYEEID")
   6 - filter("E"."NOTES"='DBP' OR "E"."NOTES"='FTD' OR "E"."NOTES"='MH' OR 
              "E"."NOTES"='SUPERVISOR')
   7 - filter("HML"."COMMENTS" IS NOT NULL AND "HML"."COMMENTS" LIKE '%ICO%')
   8 - access("HML"."TXNDATE">=TO_DATE(:DDATE,'MM/DD/YYYY HH:MI:SS AM'))
   9 - access("C"."CONTAINERID"="HML"."HISTORYID")

费用为248,283。与我投入指数之前的成本相同。此外，当我将hml.comments LIKE '%ICO%'更改为hml.comments = '%ICO%'时，费用会很好地提高(only 14!)，因为我使用了创建的索引。不幸的是，我仍然需要使用LIKE，因为我不知道在评论栏中使用标签ICO的位置。

有没有办法可以改进查询？我正在使用Oracle 11g，而hml table总共有40,193,106行。提前感谢所有需要时间回答的人。

在使用LIKE运算符的列上不使用Oracle索引

0 个答案: