Oracle9i新增了WITH語法功能，可以將查詢中的子查詢命名，放到SELECT語句的最前面。 下面看一個簡單的例子: SQL> WITH2 SEG AS (SELECT SEGMENT_NAME, SUM(BYTES)/1024 K FROM USER_SEGMENTS GROUP BY SEGMENT_NAME),3 OBJ AS (SELECT OBJECT_NAME, OBJECT_TYPE FROM USER_OBJECTS)4 SELECT O.OBJECT_NAME, OBJECT_TYPE, NVL(S.K, 0) SIZE_K5 FROM OBJ O, SEG S6 WHERE O.OBJECT_NAME = S.SEGMENT_NAME (+)7 ;OBJECT_NAME OBJECT_TYPE SIZE_KDAIJC_TEST TABLE 128P_TEST PROCEDURE 0IND_DAIJC_TEST_C1 INDEX 128通過WITH語句定義了兩個子查詢SEG和OBJ，在隨后的SELECT語句中可以直接對預定義的子查詢進行查詢。從上面的例子也可以看出，使用WITH語句，將一個包含聚集、外連接等操作SQL清楚的展現出來。 WITH定義的子查詢不僅可以使查詢語句更加簡單、清楚，而且WITH定義的子查詢還具有在SELECT語句的任意層均可見的特點。 即使是在WITH的定義層中，后定義的子查詢都可以使用前面已經定義好的子查詢: SQL> WITH2 Q1 AS (SELECT 3 + 5 S FROM DUAL),3 Q2 AS (SELECT 3 * 5 M FROM DUAL),4 Q3 AS (SELECT S, M, S + M, S * M FROM Q1, Q2)5 SELECT * FROM Q3;S M S+M S*M8 15 23 120利用WITH定義查詢中出現多次的子查詢還能帶來性能提示。Oracle會對WITH進行性能優化，當需要多次訪問WITH定義的子查詢時，Oracle會將子查詢的結果放到一個臨時表中，避免同樣的子查詢多次執行，從而有效的減少了查詢的IO數量。 看一個簡單的例子，首先構造一張大表，現在要取出大表中ID最小、ID最大以及ID等于平均值的記錄，看看普通寫法和WITH語句的區別: SQL> CREATE TABLE T_WITH AS SELECT ROWNUM ID, A.* FROM DBA_SOURCE A WHERE ROWNUM < 100001; 表已創建。 SQL> SET TIMING ON SQL> SET AUTOT ON SQL> SELECT ID, NAME FROM T_WITH2 WHERE ID IN 3 (4 SELECT MAX(ID) FROM T_WITH 5 UNION ALL6 SELECT MIN(ID) FROM T_WITH7 UNION ALL8 SELECT TRUNC(AVG(ID)) FROM T_WITH9 );ID NAME1 STANDARD50000 DBMS_BACKUP_RESTORE100000 INITJVMAUX已用時間: 00: 00: 00.09執行計劃Plan hash value: 647530712----------------------------------------------------------- Id Operation Name Rows Bytes ----------------------------------------------------------- 0 SELECT STATEMENT 3 129 * 1 HASH JOIN 3 129 2 VIEW VW_NSO_1 3 39 3 HASH UNIQUE 3 39 4 UNION-ALL 5 SORT AGGREGATE 1 13 6 TABLE Access FULL T_WITH 112K 1429K 7 SORT AGGREGATE 1 13 8 TABLE ACCESS FULL T_WITH 112K 1429K 9 SORT AGGREGATE 1 13 10 TABLE ACCESS FULL T_WITH 112K 1429K 11 TABLE ACCESS FULL T_WITH 112K 3299K-----------------------------------------------------------Predicate Information (identified by operation id):---------------------------------------------------1 - access('ID'='$nso_col_1')Note------ dynamic sampling used for this statement統計信息----------------------------------------------------------0 recursive calls0 db block gets5529 consistent gets0 physical reads0 redo size543 bytes sent via SQL*Net to client385 bytes received via SQL*Net from client2 SQL*Net roundtrips to/from client0 sorts (memory)0 sorts (disk)3 rows processed為了避免第一次執行時物理讀的影響，查詢結果選取了SQL的第三次運行，物理讀為0時的統計信息。 觀察執行計劃可以看到，先后對T_WITH表進行了4次全表掃描，并產生了5529個邏輯讀。下面看看WITH語句的表現: SQL> WITH2 AGG AS (SELECT MAX(ID) MAX, MIN(ID) MIN, TRUNC(AVG(ID)) AVG FROM T_WITH)3 SELECT ID, NAME FROM T_WITH 4 WHERE ID IN 5 (6 SELECT MAX FROM AGG 7 UNION ALL 8 SELECT MIN FROM AGG 9 UNION ALL 10 SELECT AVG FROM AGG11 );ID NAME---------- ------------------------------1 STANDARD50000 DBMS_BACKUP_RESTORE100000 INITJVMAUX已用時間: 00: 00: 00.07執行計劃----------------------------------------------------------Plan hash value: 1033356310---------------------------------------------------------------------------------- Id Operation Name Rows Bytes ---------------------------------------------------------------------------------- 0 SELECT STATEMENT 3 129 1 TEMP TABLE TRANSFORMATION 2 LOAD AS SELECT T_WITH 3 SORT AGGREGATE 1 13 4 TABLE ACCESS FULL T_WITH 112K 1429K* 5 HASH JOIN 3 129 6 VIEW VW_NSO_1 3 39 7 HASH UNIQUE 3 39 8 UNION-ALL 9 VIEW 1 13 10 TABLE ACCESS FULL SYS_TEMP_0FD9D662E_BF2EDF12 1 13 11 VIEW 1 13 12 TABLE ACCESS FULL SYS_TEMP_0FD9D662E_BF2EDF12 1 13 13 VIEW 1 13 14 TABLE ACCESS FULL SYS_TEMP_0FD9D662E_BF2EDF12 1 13 15 TABLE ACCESS FULL T_WITH 112K 3299K----------------------------------------------------------------------------------Predicate Information (identified by operation id):---------------------------------------------------5 - access('ID'='$nso_col_1')Note------ dynamic sampling used for this statement統計信息----------------------------------------------------------2 recursive calls8 db block gets2776 consistent gets1 physical reads648 redo size543 bytes sent via SQL*Net to client385 bytes received via SQL*Net from client2 SQL*Net roundtrips to/from client0 sorts (memory)0 sorts (disk)3 rows processed觀察這次的執行計劃，發現只對T_WITH表進行了兩次全表掃描，而從邏輯讀上也可以觀察到，這次只產生了2776的邏輯讀，正好是上面不使用WITH語句的一半。 通過分析執行計劃，Oracle執行了WITH子查詢一次，并將結果放到了臨時表中，在隨后對子查詢的多次訪問中，都從臨時表中直接讀取了數據，這應該也是那1個物理讀的由來。 通過上面的例子可以看到，將子查詢放到WITH語句中不僅可以簡化查詢語句的結構，對于子查詢需要多次執行的情況，還有可能提示查詢的性能。 可惜的是，WITH語句只能用在SELECT語句中，UPDATE和DELETE語句不支持WITH語法: SQL> SET AUTOT OFFSQL> SET TIMING OFFSQL> WITH SUBQ AS (SELECT 1 FROM DUAL)2 SELECT ID, NAME FROM T_WITH WHERE ID IN (SELECT * FROM SUBQ);ID NAME---------- ------------------------------1 STANDARDSQL> WITH SUBQ AS (SELECT 1 FROM DUAL)2 UPDATE T_WITH SET ID = 1 WHERE ID IN (SELECT * FROM SUBQ);UPDATE T_WITH SET ID = 1 WHERE ID IN (SELECT * FROM SUBQ)*第 2 行出現錯誤:ORA-00928: 缺失 SELECT 要害字SQL> WITH SUBQ AS (SELECT 1 FROM DUAL)2 DELETE T_WITH WHERE ID IN (SELECT * FROM SUBQ);DELETE T_WITH WHERE ID IN (SELECT * FROM SUBQ)*第 2 行出現錯誤:ORA-00928: 缺失 SELECT 要害字