標籤：style io color os 使用 sp for strong 資料
今天看到一條SQL大量使用了with，開發人員為了邏輯清晰，把一些結果集先用with緩衝起來，後面還有很多地方用到這個結果集，原始的SQL需要執行2個多小時。最佳化方法是將把最先緩衝的SQL放到用的地方,最佳化後12s。
下面來類比這個情境，不用糾結SQL的意義，把當時的SQL抽象就是這樣的。可以看到SQL1中先把語句a中的結果緩衝起來，當語句b要用的時候，object_id上的索引是用不到的，它只有在結果集中過濾。簡單點說，SQL1相對於SQL2來說，謂詞沒有推進。所以在SQL的寫法上，追求書面上的清晰和效能要做一個平衡。
--製造資料
SQL> drop table test purge;
SQL> create table test as select * from dba_objects;
SQL> create index ind_t_object_id on test(object_id) nologging;
SQL> exec dbms_stats.gather_table_stats(user,‘test‘,cascade => true);
SQL> set autotrace traceonly
--SQL1，最佳化前沒有謂詞推進
SQL> with a as(select * from test where object_type=‘TABLE‘),
b as (select count(1) from a where object_id<10),
c as (select count(1) from a where object_id>=10 and object_id<20)
select (select * from b) bc,
(select * from c) cc
from dual;
執行計畫
----------------------------------------------------------
Plan hash value: 2659770981
----------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | 198 (1)| 00:00:03 |
| 1 | VIEW | | 1 | 13 | 6 (0)| 00:00:01 |
| 2 | SORT AGGREGATE | | 1 | 13 | | |
|* 3 | VIEW | | 1600 | 20800 | 6 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL | SYS_TEMP_0FD9D66| 1600 | 154K| 6 (0)| 00:00:01 |
| 5 | VIEW | | 1 | 13 | 6 (0)| 00:00:01 |
| 6 | SORT AGGREGATE | | 1 | 13 | | |
|* 7 | VIEW | | 1600 | 20800 | 6 (0)| 00:00:01 |
| 8 | TABLE ACCESS FULL | SYS_TEMP_0FD9D66| 1600 | 154K| 6 (0)| 00:00:01 |
| 9 | TEMP TABLE TRANSFORMATION | | | | | |
| 10 | LOAD AS SELECT | SYS_TEMP_0FD9D66| | | | |
|* 11 | TABLE ACCESS FULL | TEST | 1600 | 154K| 196 (1)| 00:00:03 |
| 12 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
----------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - filter("OBJECT_ID"<10)
7 - filter("OBJECT_ID">=10 AND "OBJECT_ID"<20)
11 - filter("OBJECT_TYPE"=‘TABLE‘)
統計資訊
----------------------------------------------------------
394 recursive calls
22 db block gets
589 consistent gets
15 physical reads
600 redo size
381 bytes sent via SQL*Net to client
337 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
--SQL2，最佳化後
SQL> with b as (select count(1) from test where object_id<10),
c as (select count(1) from test where object_id>=10 and object_id<20)
select (select * from b) bc,
(select * from c) cc
from dual;
執行計畫
----------------------------------------------------------
Plan hash value: 1155001961
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | 2 (0)| 00:00:01 |
| 1 | VIEW | | 1 | 13 | 2 (0)| 00:00:01 |
| 2 | SORT AGGREGATE | | 1 | 5 | | |
|* 3 | INDEX RANGE SCAN| IND_T_OBJECT_ID | 6 | 30 | 2 (0)| 00:00:01 |
| 4 | VIEW | | 1 | 13 | 2 (0)| 00:00:01 |
| 5 | SORT AGGREGATE | | 1 | 5 | | |
|* 6 | INDEX RANGE SCAN| IND_T_OBJECT_ID | 9 | 45 | 2 (0)| 00:00:01 |
| 7 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - access("OBJECT_ID"<10)
6 - access("OBJECT_ID">=10 AND "OBJECT_ID"<20)
統計資訊
----------------------------------------------------------
1 recursive calls
0 db block gets
4 consistent gets
0 physical reads
0 redo size
381 bytes sent via SQL*Net to client
337 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
Oracle 使用with要小心了--謂詞不能推進
