Run a procedure ON-INSERT

Question

I am on an Oracle 10.2.0.4.0 database. I struggled if I should post the question here or in the General Questions forum, but decided to start here first.
I need to create a table that will be populated by another database probably 1000 times per a 24 hour day. The data will be populated via a database link via a SQL INSERT statement then committed. As far as the user of the other database is concerned, once they do the insert they are finished.
As the owner of the database that now has a new record in it, I need to process that new record. Basically, the table that was inserted into was a "staging" table. I want to pull the data from that table and populate various tables in my database with that information, then remove the record from the "staging" table and I would like to do that as close to "real-time" as possible.
The first thought that came in my mind was to create an ON INSERT trigger on the "staging" table that would call a procedure to move the data to the appropriate tables then remove the record from the "staging" table, but I figured that I would receive a "mutating table" error. I tried it and that is exactly what happened. My DBA recommended that I attempt to do the same thing with a Global Temporary table that removes the data on insert. It sounded good, but when I tried it, I received the same "mutating trigger" error.
Below are the statements to create the objects:
-- This will emulate the table that a user from another database will insert a record into.
create global temporary table temp_staging_table
(pk_id         number(5),
data_1        varchar2(10),
data_2        varchar2(10),
data_3        varchar2(10)) on commit delete rows;
--The first attempt I created a regular table:
create table temp_staging_table
(pk_id         number(5),
data_1        varchar2(10),
data_2        varchar2(10),
data_3        varchar2(10));
-- The next three tables are just tables used to simulate populating production tables from the staging table
create table temp_table_1
(pk_id         number(5),
data_1        varchar2(10));
create table temp_table_2
(pk_id         number(5),
data_1        varchar2(10));
create table temp_table_3
(pk_id         number(5),
data_1        varchar2(10));
-- The procedure to move the data from the staging table into the fake production tables
create or replace procedure tst_procedure (p_pk_id in number)
is
begin
  insert into temp_table_1 (pk_id, data_1)
  select pk_id, data_1
    from temp_staging_table
   where pk_id = p_pk_id;
  insert into temp_table_2 (pk_id, data_1)
  select pk_id, data_2
    from temp_staging_table
   where pk_id = p_pk_id;
  insert into temp_table_3 (pk_id, data_1)
  select pk_id, data_3
    from temp_staging_table
   where pk_id = p_pk_id;
end;
-- The trigger on the staging table that will call the procedure
create or replace TRIGGER tst_trg
  AFTER
  INSERT
  ON temp_staging_table
  REFERENCING OLD AS OLD NEW AS NEW
  FOR EACH ROW
DECLARE
BEGIN
  tst_procedure(:new.pk_id);
END;
/You might notice that I did not include any commits in the procedure or trigger hoping that I could avoid the "mutating" trigger error. The commit I included in a separate script where I performed the actual insert, but it never makes it that far because of the "mutating trigger" error. Below is that script to insert the data:
insert into temp_staging_table
values (500,'THIS','IS A','TEST');
commit;I don't have to do it this way, but I'm not sure what other options I have. I could create a DBMS_JOB that runs every minute of a 24 hour period, but that seems very inefficient and not quite "real-time". I thought about replacing my call to the procedure in the trigger to logic that would schedule a job that would run immediately that would call the procedure, but I don't have the right privileges so I haven't tried that yet, but once again is that the most efficient way of doing it?
I am hoping by posting this that someone might have a better option. I have Googled the problem but haven't found any good solutions.

Answer 1

>
I had never heard of "Oracle Streams" until you mentioned it.
>
Here is an Oracle-base article that has sample code
http://www.oracle-base.com/articles/9i/advanced-queuing-9i.php
You could also just run a scheduler job that continuously queries the staging table for data and processes whatever rows it finds.
Before you start implementing any solution I suggest you flesh out your requirements regarding restart/recovery. What happens if the source or target DB is unavailable for some period of time? Do the records need to be processed in any particular order?