oracle update 分次提交

这里我插入1999条记录,然后把记录中xxx为非33的全部更新为33,分20次提交,1次提交100条,比较下更新前后xxx为33的记录个数就知道对不对了

SQL> CREATE TABLE test (ID VARCHAR2(20),xxx NUMBER);

Table created

SQL> INSERT INTO test SELECT lpad(ROWNUM,4,'0'),mod(ROWNUM,34) FROM dual CONNECT BY LEVEL < 2000;

1999 rows inserted

Oracle lpad函数将左边的字符串填充一些特定的字符,其语法格式如下:
lpad( string1, padded_length, [ pad_string ] )
其中string1是需要粘贴字符的字符串
padded_length是返回的字符串的数量,如果这个数量比原字符串的长度要短,lpad函数将会把字符串截取成padded_length;
pad_string是个可选参数,这个字符串是要粘贴到string1的左边,如果这个参数未写,lpad函数将会在string1的左边粘贴空格。

例如:

lpad('tech', 7); 将返回' tech'

lpad('tech', 2); 将返回'te'

lpad('tech', 8, '0'); 将返回'0000tech'

lpad('tech on the net', 15, 'z'); 将返回 'tech on the net'

lpad('tech on the net', 16, 'z'); 将返回 'ztech on the net'

SQL> commit;

Commit complete

SQL> set serverout on;
SQL> select count(*) from test where id = 33;

COUNT(*)
----------
1

SQL> select count(*) from test where xxx = 33;

COUNT(*)
----------
58

SQL> select count(*) from test where xxx <> 33;

COUNT(*)
----------
1941

SQL>
SQL> declare
2 -- 数组变量,保存查询条件
3 TYPE t_id IS TABLE OF test.id%TYPE;
4 v_t_id t_id;
5
6 CURSOR c IS
7 SELECT id FROM test where xxx <> 33;
8 -- 循环次数
9 cnt NUMBER := 0;
10 BEGIN
11 OPEN c;
12 LOOP
13 cnt := cnt + 1;
14 -- 批量更新,一次更新100条数据 ,取出数据放到变量v_t_id中
15 fetch c bulk collect into v_t_id LIMIT 100;
16 -- 这里用forall效率更高
17 FORALL i IN 1 .. v_t_id.COUNT
18 UPDATE test SET xxx = 33 WHERE id = v_t_id(i);
19
20 -- 提交
21 COMMIT;
22 -- 循环退出
23 exit when c%NOTFOUND;
24
25 END LOOP;
26
27 dbms_output.put_line('循环次数:' || cnt);
28
29 CLOSE c;
30 COMMIT;
31 end;
32 /
循环次数:20

PL/SQL procedure successfully completed

SQL> select count(*) from test where xxx = 33;

COUNT(*)
----------
1999

SQL> select count(*) from test where xxx <> 33;

COUNT(*)
----------
0

SQL>
原文地址:https://www.cnblogs.com/danghuijian/p/4400524.html