材料化视图“ MV_AMP”:
CREATE MATERIALIZED VIEW MV_AMP
NOLOGGING
BUILD IMMEDIATE
REFRESH FORCE
ON DEMAND
AS
Select a, b, c from amp;
依赖于“ MV_AMP”的材料化视图“ MV_BOT”:
CREATE MATERIALIZED VIEW MV_BOT
NOLOGGING
BUILD IMMEDIATE
REFRESH FORCE
ON DEMAND
AS
SELECT bot.x, bot.y, mv_amp
FROM bot, mv_amp
WHERE bot.a = mv_amp.a;
并在mv_bot中创建唯一索引:
CREATE UNIQUE INDEX mv_bot_idx001 ON mv_bot(x, a);
视图和索引创建成功后,假设我添加一个重复值,由于唯一索引,在刷新mv_bot时会导致类似(dup_val_on_index)的错误。
因此,我使用nested = TRUE在MV_AMP(主视图)中进行刷新,并且Oracle没有引发错误:
BEGIN
dbms_mview.refresh_dependent(number_of_failures => n_failures,
list => 'MV_AMP',
atomic_refresh => TRUE,
nested => TRUE);
EXCEPTION
WHEN
OTHERS THEN
-- it never reach this code
dbms_output.put_line('Errors: '||SQLERRM);
END;
n_failures返回:0,并且永远不会到达异常内部的dbms_output。
当oracle尝试更新嵌套的MV并登录表时,我需要捕获错误。
使用Oracle 11g。
答案 0 :(得分:1)
只有在实例化视图EXCEPTION
中存在重复行的情况下,MV_BOT
块才达到-情况并非如此。
您可能会问为什么;最有说服力的答案是,您还需要刷新实例化视图MV_AMP
,以便在MV_BOT
的联接中获得dup。
在阅读dbms_mview.refresh_dependent
的文档时,您必须从表AMP
开始才能刷新两个MV(从MV_AMP
开始仅刷新依赖的MV < / em> MV_BOT
)
测试用例
create table amp (
a number,
b number,
c number);
create table bot (
a number,
x number,
y number);
CREATE MATERIALIZED VIEW MV_AMP
NOLOGGING
BUILD IMMEDIATE
REFRESH FORCE
ON DEMAND
AS
Select a, b, c from amp;
CREATE MATERIALIZED VIEW MV_BOT
NOLOGGING
BUILD IMMEDIATE
REFRESH FORCE
ON DEMAND
AS
SELECT bot.x, bot.y, mv_amp.a
FROM bot, mv_amp
WHERE bot.a = mv_amp.a;
CREATE UNIQUE INDEX mv_bot_idx001 ON mv_bot(x, a);
insert into amp(a,b,c) values(1,1,1);
insert into bot(a,x,y) values(1,1,1);
insert into bot(a,x,y) values(1,1,3);
commit;
DECLARE
n_failures NUMBER;
BEGIN
dbms_mview.refresh_dependent(number_of_failures => n_failures,
list => 'AMP',
atomic_refresh => TRUE,
nested => TRUE);
dbms_output.put_line('Failures: '||n_failures);
EXCEPTION
WHEN
OTHERS THEN
dbms_output.put_line('Errors: '||SQLERRM);
END;
/
--> Errors: ORA-12008: error in materialized view refresh path
--> ORA-00001: unique constraint (xxxxx.MV_BOT_IDX001) violated