这里主要对比以下三种格式的游标循环:
1.单条处理
open 游标;
LOOP
FETCH 游标 INTO 变量;
EXIT WHEN 条件;
END LOOP;
CLOSE 游标;
2.批量处理
open 游标;
FETCH 游标 BULK COLLECT INTO 集合变量;
CLOSE 游标;
3.隐式游标
for x in (sql语句) loop
...--逻辑处理
end loop;
以上为工作中常见的几种游标处理方式,一般来说批量处理的速度要最好,隐式游标的次之,单条处理的最差,但是在我的实际工作中发现大部分使用的还是第一种游标处理。
归其原因竟是对集合变量及批量处理的效率等问题不了解所致。
推荐阅读：
PL/SQL下连接远程Oracle数据库
PL/SQL“ ORA-14551: 无法在查询中执行 DML 操作”解决
这里简单的测试一下以上三种游标的效率,并分析trace文件来查看这3种处理方式的本质。
--创建测试大表
00:09:54 SCOTT@orcl> create table big_data as select 'Cc'||mod(level,8) a,'Dd'||
mod(level,13) b from dual connect by level select count(*) from big_data;
COUNT(*)
----------
999999
1 row selected.
Elapsed: 00:00:00.07
--分别执行以上三种方式的游标处理的plsql块
00:11:21 SCOTT@orcl> declare
00:17:54 2 cursor c_a is
00:17:54 3 select a from big_data;
00:17:54 4
00:17:54 5 v_a big_data.a%type;
00:17:54 6 begin
00:17:54 7 open c_a;
00:17:54 8 loop
00:17:54 9 fetch c_a into v_a;
00:17:54 10 exit when c_a%notfound;
00:17:54 11 end loop;
00:17:54 12 close c_a;
00:17:54 13 end;
00:17:56 14 /
PL/SQL procedure successfully completed.
Elapsed: 00:00:07.42
00:18:05 SCOTT@orcl> declare
00:19:56 2 cursor c_a is
00:19:56 3 select a from big_data;
00:19:56 4
00:19:56 5 type t_a is table of c_a%rowtype;
00:19:56 6 v_a t_a;
00:19:56 7 begin
00:19:56 8 open c_a;
00:19:56 9 --批量处理
00:19:56 10 fetch c_a bulk collect into v_a;
00:19:56 11 close c_a;
00:19:56 12 end;
00:19:57 13 /
PL/SQL procedure successfully completed.
Elapsed: 00:00:00.64
00:22:55 SCOTT@orcl> declare
00:23:18 2 v_a big_data.a%type;
00:23:18 3 begin
00:23:18 4 --批量处理
00:23:18 5 for x in (select a from big_data) loop
00:23:18 6 v_a:=x.a;
00:23:18 7 end loop;
00:23:18 8 end;
00:23:18 9 /
PL/SQL procedure successfully completed.
Elapsed: 00:00:00.79
注:请保证plsql_optimize_level的参数为2,如果不为2,可能测试结果会不一样.
注意对比消耗时间,1为7.42s, 2为0.64s, 3为0.79s
请继续阅读：