2014-09-25 BaoXinjian
一、摘要
客户端Arraysize的设置实际上可以理解为设置数组的大小来保存sql调用的返回值。设置适当的arraysize有几个好处：
1.减少读data block的次数，也就是consistent gets的次数。
假设一个block有n行，如果每次取一行，则对该block的逻辑读次数就是n，尽管逻辑读是内存读，速度很快，但并发量很大时也会使数据库的压力很大。
如果每次取>n行，则该block只需读一次。因此合理设置arraysize能大大减轻数据库压力。
2.在PLSQL中减少plsql引擎和sql引擎的切换次数，提高效率
3.通过数组存放结果，减少客户端和数据库服务器的调用次数，减少往返时间
二、SQLPLUS设置数组访问的方法
1. Sqlplus的arraysize Sqlplus的arraysize参数表示数据库一次返回给客户端的行数，缺省是15。如果返回数据量大，可以调大此参数:
Set arraysize nnnn
该参数最大为5000
测试一：缺省arraysize
select * from test_table where rownum <100000;
99999 rows selected.
lapsed: 00:00:06.12
Statistics
----------------------------------------------------------
4573
recursive calls
测试二：SQL> set arraysize 1000
select * from test_table where rownum <100000;
99999 rows selected.
三、PLSQL Bulk Collect设置数组访问的方法
测试一：不使用bulk collect,一次fetch一条记录
select a.name name, b.value
from v$statname a, v$mystat b
where a.statistic# = b.statistic# and (a.name) in( ‘consistent gets‘)
order by name;
100759
declare
cursor c1 is select user_id from test_table where rownum<100000;
v_id varchar2(30);
begin
open c1;
loop
FETCH c1 INTO v_id;
exit when c1%notfound;
null;
end loop;
close c1;
end;
/
select a.name name, b.value
from v$statname a, v$mystat b
where a.statistic# = b.statistic# and (a.name) in( ‘consistent gets‘)
order by name;
测试二：设置bulk collect，一次取100条 select a.name name, b.value
from v$statname a, v$mystat b
where a.statistic# = b.statistic# and (a.name) in( ‘consistent gets‘) order by name; 110129 declare
TYPE VarcharTab IS TABLE OF varchar2(30) INDEX BY BINARY_INTEGER; cursor c1 is select user_id from test_table where rownum<100000; v_id VarcharTab; begin
open c1; loop
FETCH c1 BULK COLLECT INTO v_id LIMIT 100; exit when c1%notfound;
null; end loop; close c1; end; /
Elapsed: 00:00:00.09
select a.name name, b.value
from v$statname a, v$mystat b
where a.statistic# = b.statistic# and (a.name) in( ‘consistent gets‘) order by name; 113001
测试三：设置bulk collect，一次取5000条 declare
TYPE VarcharTab IS TABLE OF varchar2(30) INDEX BY BINARY_INTEGER; cursor c1 is select user_id from test_table where rownum<100000; v_id VarcharTab; begin
open c1; loop
FETCH c1 BULK COLLECT INTO v_id LIMIT 5000; exit when c1%notfound;
null; end loop; close c1; end; /
Elapsed: 00:00:00.08
select a.name name, b.value
from v$statname a, v$mystat b
where a.statistic# = b.statistic# and (a.name) in( ‘consistent gets‘) order by name;
114905
********************作者：鲍新建********************