Isak Hansen wrote:

Hello Isak! I was speaking to you about this on comp.databases, glad to see you here on the Postgres group.

What you want to do is easy enough in the single case, but can get complicated if you do a lot of it.

<shameless plug>
We have a framework that is freely available that does exactly what you are trying to do, and works against PostgreSQL and is written in PHP. It writes all of the triggers for you to save the hassle and prevent mistakes. We even have on our website an example of the kind of constraint you are doing:

http://docs.secdat.com/index.php?gp_page=x_docview&gppn=Customer+Credit+Limit
</shamelss plug>

If you wish to code this by hand, here is what you must do:

1)  Add column "amount" to table A
2) Add insert, update, and delete triggers to B that increase and decrease the value of A.amount 3) Add an update trigger to A (insert and delete not necessary) that enforces your constraint, or just do it as a check constraint (i personally prefer triggers)

BTW, is table A supposed to be a GL batch summary table or something? Why must it always be zero? If it is a GL batch table, can it be out of balance while people are actually entering the data? Should the constraint only be enforced when the batch is closed?


create table a (
 id serial primary key,
);
create table b (
 id serial primary key,
 a_id int4 references a (id),
 amount decimal(16, 2)
);

and would like a constraint to guarantee that "sum(b.amount) = 0 group
by b.a_id".


From my testing so far, and this thread:
<http://groups.google.com/group/comp.databases/browse_thread/thread/c2ce2d61eecf5c6c/1e0fa71282aea7d8#1e0fa71282aea7d8>,
i think a trigger is the way to go.

Problem is, i have no idea where to go from here. Getting the model
nearly to to 3NF and writing some simple queries is about the extent
of my db-related skillset..

Anyone feel the calling..? ;)


Also, how would this kind of check affect performance? Table 'b' is
our ledger table, busiest one in the app i assume, while 'a' groups
related transactions and holds common info. We inherited the term
voucher for 'a', anyone know if that is(n't) appropriate?

If someone are interested, the actual tables are here:
a: http://trac.lodo.no/wiki/vouchers
b: http://trac.lodo.no/wiki/voucher_lines


Any feedback appreciated,
Isak

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
      choose an index scan if your joining column's datatypes do not
      match


begin:vcard
fn:Kenneth  Downs
n:Downs;Kenneth 
email;internet:[EMAIL PROTECTED]
tel;work:631-689-7200
tel;fax:631-689-0527
tel;cell:631-379-0010
x-mozilla-html:FALSE
version:2.1
end:vcard

---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

Reply via email to