|
【赛迪网－it技术报道】很多人都知道，在普通的函数中，使用dbms_output输出的信息，需要在服务器执行完整个函数后一次性的返回给客户端。但你如果需要在客户端实时的输出函数执行过程中的一些信息，在oracle 9i以后则可以使用管道函数(pipeline function)。
pipelined（关键字）表明这是一个管道函数，管道函数的返回值类型必须为集合，在函数中，pipe row语句被用来返回该集合的单个元素，函数则以一个空的return语句结束，以表明它已经完成。
create or replace type msgtype as table of varchar2(4000);
/
create or replace function f_pipeline_test
return msgtype
pipelined
as
begin
for i in 1 .. 10
loop
pipe row( 'iteration ' || i || ' at ' || systimestamp );
dbms_lock.sleep(1);
end loop;
pipe row( 'all done!' );
return;
end;
/
|
在sql*plus中执行该函数，大家需要首先设置arraysize为1，否则服务器会按照默认的15来向客户端返回信息，这会影响我们的测试效果。
sql> set arraysize 1
sql> select * from table( f_pipeline_test );
column_value
-----------------------------------------------------
iteration 1 at 14-feb-08 02.13.18.273988000 pm +08:00
iteration 2 at 14-feb-08 02.13.19.275988000 pm +08:00
iteration 3 at 14-feb-08 02.13.20.277767000 pm +08:00
iteration 4 at 14-feb-08 02.13.21.279591000 pm +08:00
iteration 5 at 14-feb-08 02.13.22.281366000 pm +08:00
iteration 6 at 14-feb-08 02.13.23.283189000 pm +08:00
iteration 7 at 14-feb-08 02.13.24.283965000 pm +08:00
iteration 8 at 14-feb-08 02.13.25.285785000 pm +08:00
iteration 9 at 14-feb-08 02.13.26.286570000 pm +08:00
iteration 10 at 14-feb-08 02.13.27.288387000 pm +08:00
all done!
|
11 rows selected.
如果要在pipeline中执行dml操作，则必须使用自治事务，否则会报ora-14551错误
create or replace function f_pipeline_testdml
return msgtype
pipelined
as
begin
for i in 1 .. 10
loop
insert into test values(1);
pipe row( 'insert into test values( ' || i || ') success at ' || systimestamp );
dbms_lock.sleep(1);
end loop;
pipe row( 'all done!' );
return;
end;
/
sql> select * from table( f_pipeline_testdml );
select * from table( f_pipeline_testdml )
*
error at line 1:
ora-14551: cannot perform a dml operation inside a query
ora-06512: at "ning.f_pipeline_testdml", line 8
create or replace function f_pipeline_testdml
return msgtype
pipelined
as
pragma autonomous_transaction;
begin
for i in 1 .. 10
loop
insert into test values(1);
commit;
pipe row( 'insert values ' || i || ' success at ' || systimestamp );
dbms_lock.sleep(1);
end loop;
pipe row( 'all done!' );
return;
end;
/
sql> select * from table( f_pipeline_testdml );
column_value
--------------------------------------------------------------------------------
insert values 1 success at 14-feb-08 02.16.47.855158000 pm +08:00
insert values 2 success at 14-feb-08 02.16.48.865559000 pm +08:00
insert values 3 success at 14-feb-08 02.16.49.867377000 pm +08:00
insert values 4 success at 14-feb-08 02.16.50.873154000 pm +08:00
insert values 5 success at 14-feb-08 02.16.51.874942000 pm +08:00
insert values 6 success at 14-feb-08 02.16.52.880781000 pm +08:00
insert values 7 success at 14-feb-08 02.16.53.882543000 pm +08:00
insert values 8 success at 14-feb-08 02.16.54.894348000 pm +08:00
insert values 9 success at 14-feb-08 02.16.55.896153000 pm +08:00
insert values 10 success at 14-feb-08 02.16.56.901904000 pm +08:00
all done!
11 rows selected.
|
在oracle 9205及其之后的版本中，在pipeline function中使用自治事务，则必须在pipe row之前提交或者回滚事务，否则会报ora-06519错误。
create or replace function f_pipeline_testdml
return msgtype
pipelined
as
pragma autonomous_transaction;
begin
for i in 1 .. 10
loop
insert into test values(1);
pipe row( 'insert values ' || i || ' success at ' || systimestamp );
dbms_lock.sleep(1);
end loop;
pipe row( 'all done!' );
commit;
return;
end;
/
sql> select * from table( f_pipeline_testdml );
select * from table( f_pipeline_testdml )
*
error at line 1:
ora-06519: active autonomous transaction detected and rolled back
ora-06512: at "ning.f_pipeline_testdml", line 10
|
此处是由于在9205中修复bug 2711518导致了自治事务的行为有所改变。如果系统从9205之前的版本升级到之后的版本，需要保证pipeline function的行为和以前版本一致，oracle提供了一个10946事件来设置和以前版本的兼容性，如果在管道函数中使用了select for update的cursor，则必须设置event回归以前的特性，否则即使在pipe row之前commit也会导致出现ora-1002错误。
alter system set event = "10946 trace name context forever, level 8" scope=spfile;
|
|
