Trigger to update field on a table with the sum of fields on another table

Question

My experience creating triggers and pl/sql in general can best be described in oracle terms as null. I've been practicing by creating tables and applications on my personal home server to help me with some of my work related tasks. Right now I'm trying to create a trigger that will, after insert, update, delete on the assignment_time_track table update the time_spent field on the assignments table with the sum of the time_spent fields on the assignment_time_track table. Hopefully that run on sentence there is clear to people other than myself. I've attempted to script this on my own using the trigger creation tool for Oracle Database Express Edition but I get the following error:
Trigger create was not successful for the following reason:
ORA-06552: PL/SQL: Compilation unit analysis terminated ORA-06553: PLS-320: the declaration of the type of this expression is incomplete or malformed
Here is my attempt at creating the trigger on my own.
create or replace trigger "ASSIGNMENT_TIME_TRACK_T1"
AFTER
insert or update or delete on "ASSIGNMENT_TIME_TRACK"
for each row
begin
update assignments
set time_spent = (select sum(time_spent)
from assignment_time_track
where assignment_time_track.name = assignments.name);
end;
If what I've posted isn't clear or more detail is needed, let me know and I'll respond with a complete description of both tables and my goals for each table. Thanks in advance for any help. I will also gladly accept links to tutorials or lessons that explain how to do this sort of thing.
Edited by: bobonthenet on Mar 9, 2009 2:01 PM

Answer 1

Hi,
If the assignments table has only one row per assignment, why is the primary key the combination of name and time_spent? If you have two two assignments called "Lab Report", isn't it possible that you would spend the same amount of time on each of them? I suggest using a sequence to assign an arbitrary id number to each assignment, and use that as the primary key.
What does each row in assuignment_time_track represent? It sounds like it is a chunk of time spent on one assignment (that is, you want to know that you spent 90 minutes on Tudesday morning working on some assignment, and that you spent another 30 minutes on Tuesday afternoon working on the same assignment). If so, then there should be a foreign key constraint in assignment_time_track referencing the primary key of assignemnt, and not the other way around.
Alex is right; you can get the total time spent on each project in a query or view; there is no need to replicate that data.
If you're new to Oracle and SQL, you should invest your time in getting more experience with the basics: everyday things like queries (using joins and GROUP BY) and views, and not spend much time on things that aren't used that much, like triggers.
If you really did have to copy the data, then you could have a trigger on assignemnt_time_track that kept the total in assignment up to date, like this:
UPDATE  assignment
SET     total_time_spent = total_time_spent
                + NVL (:NEW.time_spent, 0)
                         - NVL (:OLD.time_spent, 0);I suggest you name the column in assignment something different than the column in assignment_time_track, to reduce the risk of confusion. Also, since they represent different things, the same name can't be the most descripttive for each of them.
In case you're wondering about the use of NVL, above: It allows the same statement to take care of the situation when you INSERT, UPDATE or DELETE a row in assignment_time_track. That is, if you UPDATE a row in assignment_time_track, and change the time_spent from 60 to 90, then you want to add the new time (90) and subtract the old time (60) fro the total_time_spent in assignment: that is, total_time_spent would increase by 30. If you INSERT a new row into assignment_time_track with time_spent=30, you just need to add the new time_spent (30): there is nothing to subtract. But rather than write an IF statement and a second UPDATE for that situation, you can just rely on hte fact that all :OLD values are NULL iwhen INSERTing, and treat that NULL as a 0. Likewise, when DELETing, all :NEW values are NULL..