PL/SQL引入过程化语言的相应元素,比如条件分支或者循环,不过,SQL本身仍然作为主体嵌套于其中,由于需要SQL引擎才能执行SQL命令,对于PL/SQL程序,往往存在许多PL/SQL引擎 - SQL引擎之间的交互,过多这样的交互会对性能产生负面影响。
Oracle在PL/SQL中引入了BULK SQL,用于尽量减少PL/SQL – SQL引擎之间的交互,以期提高性能。具体而言,Oracle BULK SQL包括FORALL语句、BULK COLLECT子句。前者将多条语句(通常是DML)一次性发送给SQL引擎;后者将SQL引擎所获得的结果一次性返回给PL/SQL引擎。
(1)FORALL
下面的两个例子对比了FORALL与FOR循环之间的区别:
SQL> desc t_bulk;
Name Type Nullable Default Comments
-------------- ------------ -------- ------- --------
EMPLOYEE_ID NUMBER(6) Y
FIRST_NAME VARCHAR2(20) Y
LAST_NAME VARCHAR2(25)
EMAIL VARCHAR2(25)
PHONE_NUMBER VARCHAR2(20) Y
HIRE_DATE DATE
JOB_ID VARCHAR2(10)
SALARY NUMBER(8,2) Y
COMMISSION_PCT NUMBER(2,2) Y
MANAGER_ID NUMBER(6) Y
DEPARTMENT_ID NUMBER(4) Y
DECLARE
TYPE NumList IS VARRAY(20) OF NUMBER;
depts NumList := NumList(10, 30, 70);
BEGIN
FOR i IN depts.FIRST..depts.LAST LOOP
DELETE FROM t_bulk
WHERE department_id = depts(i);
END LOOP;
END;
DECLARE
TYPE NumList IS VARRAY(20) OF NUMBER;
depts NumList := NumList(10, 30, 70); -- department numbers
BEGIN
FORALL i IN depts.FIRST..depts.LAST
DELETE FROM t_bulk
WHERE department_id = depts(i);
END;
虽然从内部执行机制上来说,两个循环有很大的区别,但从语法上来说,还是非常类似的。不过有一个小细节需要注意,就是FORALL语句并没有对应的END语句。
我们再来看看使用FORALL的情况下对异常的处理:
DECLARE
TYPE type1 IS TABLE OF NUMBER;
v type1:=type1(1, 2, 3000, 4, 5, 6, 77777, 8, 9, 10001);
BEGIN
EXECUTE IMMEDIATE 'TRUNCATE TABLE t_bulk2';
--
FORALL idx IN v.FIRST..v.LAST
INSERT INTO t_bulk2 VALUES(v(idx));
--
EXCEPTION
WHEN OTHERS THEN
DBMS_OUTPUT.PUT_LINE(SQLERRM);
END;
ORA-01438: 值大于为此列指定的允许精度
PL/SQL procedure successfully completed
SQL> SELECT * FROM t_bulk2;
F1
----
1
2
Oracle 9i中引入了SAVE EXCEPTIONS语法及与之对应的“ORA-24381: error(s) in array DML”异常,使用它们,我们可以跳过FORALL中出现异常的语句,并将异常保存在SQL%BULK_EXCEPTIONSP这个集合中:
TYPE type1 IS TABLE OF NUMBER;
v type1:=type1(1, 2, 3000, 4, 5, 6, 77777, 8, 9, 10001);
--
BULK_ERROR EXCEPTION;
PRAGMA EXCEPTION_INIT(BULK_ERROR, -24381);
BEGIN
EXECUTE IMMEDIATE 'TRUNCATE TABLE t_bulk2';
--
FORALL idx IN v.FIRST..v.LAST SAVE EXCEPTIONS
INSERT INTO t_bulk2 VALUES(v(idx));
--
EXCEPTION
WHEN BULK_ERROR THEN
FOR i IN 1..SQL%BULK_EXCEPTIONS.COUNT LOOP
DBMS_OUTPUT.PUT_LINE(SQLERRM(-SQL%BULK_EXCEPTIONS(i).ERROR_CODE)||', Statement: #'||SQL%BULK_EXCEPTIONS(i).ERROR_INDEX);
END LOOP;
WHEN OTHERS THEN
NULL;
END;
ORA-01438: 值大于为此列指定的允许精度, Statement: #3
ORA-01438: 值大于为此列指定的允许精度, Statement: #7
ORA-01438: 值大于为此列指定的允许精度, Statement: #10
PL/SQL procedure successfully completed
SQL> SELECT * FROM t_bulk2;
F1
----
1
2
4
5
6
8
9
(注意使用ERROR_CODE时要加上负号。)
下面介绍如何获取第一条语句所影响的行数,这需要使用SQL%BULK_ROWCOUNT:
insert into t_bulk3 values(1);
insert into t_bulk3 values(2);
insert into t_bulk3 values(2);
insert into t_bulk3 values(3);
insert into t_bulk3 values(3);
insert into t_bulk3 values(3);
insert into t_bulk3 values(3);
DECLARE
TYPE type1 IS TABLE OF NUMBER;
v type1:=type1(1, 2, 3, 4);
--
BULK_ERROR EXCEPTION;
PRAGMA EXCEPTION_INIT(BULK_ERROR, -24381);
BEGIN
FORALL idx IN v.FIRST..v.LAST SAVE EXCEPTIONS
DELETE FROM t_bulk3 WHERE fid=v(idx);
--
FOR idx IN v.FIRST..v.LAST LOOP
DBMS_OUTPUT.PUT_LINE('Statement: #'||idx||', '||SQL%BULK_ROWCOUNT(idx)||' rows were impacted.');
END LOOP;
--
EXCEPTION
WHEN BULK_ERROR THEN
FOR i IN 1..SQL%BULK_EXCEPTIONS.COUNT LOOP
DBMS_OUTPUT.PUT_LINE(SQLERRM(-SQL%BULK_EXCEPTIONS(i).ERROR_CODE)||', Statement: #'||SQL%BULK_EXCEPTIONS(i).ERROR_INDEX);
END LOOP;
WHEN OTHERS THEN
NULL;
END;
Statement: #1, 1 rows were impacted.
Statement: #2, 2 rows were impacted.
Statement: #3, 4 rows were impacted.
Statement: #4, 0 rows were impacted.
PL/SQL procedure successfully completed
(2)BULK COLLECT
假设有一条SQL查询,返回的记录中包含5行,那么如果在PL/SQL中执行此查询,会有5次的PL/SQL – SQL交互,如果使用BULK COLLECT,可以降低到1次。
BULK COLLECT子句可以出现在以下语句中:
SELECT INTO
FETCH
RETURNING INTO
insert into t_bulk4 values(1,'abc');
insert into t_bulk4 values(2,'def');
insert into t_bulk4 values(3,'xyz');
insert into t_bulk4 values(4,'xxx');
insert into t_bulk4 values(5,'123');
commit;
DECLARE
TYPE type1 IS TABLE OF t_bulk4%ROWTYPE;
v type1;
BEGIN
SELECT * BULK COLLECT INTO v FROM t_bulk4;
--
FOR i IN 1..v.COUNT LOOP
DBMS_OUTPUT.PUT_LINE(v(i).fid||' '||v(i).fval);
END LOOP;
END;
1 abc
2 def
3 xyz
4 xxx
5 123
PL/SQL procedure successfully completed
另一个INDEX BY集合的示例(实际上使用跟上例一样的FOR循环也可以):
TYPE type1 IS TABLE OF t_bulk4%ROWTYPE INDEX BY PLS_INTEGER;
v type1;
idx PLS_INTEGER;
BEGIN
SELECT * BULK COLLECT INTO v FROM t_bulk4;
--
idx:=v.FIRST;
WHILE(idx IS NOT NULL) LOOP
DBMS_OUTPUT.PUT_LINE(v(idx).fid||' '||v(idx).fval);
idx := v.NEXT(idx);
END LOOP;
END;
FETCH cursor BULK COLLECT INTO的使用与上述例子都类似,不多写了。
关于RETURNING INTO + BULK COLLECT,我们来一个综合的例子:
DECLARE
TYPE type1 IS TABLE OF NUMBER;
v type1:=type1(1, 2, 3, 5);
--
TYPE type2 IS TABLE OF t_bulk5.fid%TYPE;
v2 type2;
BEGIN
FORALL idx IN v.FIRST..v.LAST
INSERT INTO t_bulk5 VALUES(v(idx)) RETURNING fid BULK COLLECT INTO v2;
--
DBMS_OUTPUT.PUT_LINE(v2.COUNT);
END;
4
PL/SQL procedure successfully completed
顺便比较一下使用FOR循环时是什么结果:
TYPE type1 IS TABLE OF NUMBER;
v type1:=type1(1, 2, 3, 5);
--
TYPE type2 IS TABLE OF t_bulk5.fid%TYPE;
v2 type2;
BEGIN
FOR idx IN v.FIRST..v.LAST LOOP
INSERT INTO t_bulk5 VALUES(v(idx)) RETURNING fid BULK COLLECT INTO v2;
END LOOP;
--
DBMS_OUTPUT.PUT_LINE(v2.COUNT);
END;
1
PL/SQL procedure successfully completed
这实际上也好理解,因为上面的INSERT语句每次影响的只有一行,所以第二例中,保留的是循环中最后一次执行的INSERT所影响的行数,当然是1;而由于FORALL语句会将所有语句一次性提交到数据库,这也使得我们可以使用RETURNING INTO + BULK COLLECT获取所有插入的数据。