Subject deleting and inserting records
Author Sergio H. Gonzalez
From: ""Adam""
> updates and deletes create record versions which must be garbage
> collected at a later time, so only update or delete records that you
> want to (there was a post a few days back where someone was deleting
> records only to reinsert them).

That "someone" was me, I'm affraid... Sorry for insisting with this, but I'm
really concerned about do things the way they should be done. I will explain a
real case and what are my reasons to do what I'm doing. I'm sure, that my
English doesn't help to make myself clear, but I will try. Of course I WOULD
LOVE to hear your opinions (being experimented people in DBs)

The case: a user makes an Invoice to a customer. For this example I will use
just 3 tables

INVOICE_MASTER
ID = 1 - Date = 2008/09/5 - Customer_ID = 200 - Payment_ID = 1

(Payment_ID = 1 means 3 payments at 30, 60, 90 days)

INVOICE_DETAIL
Master_ID = 1 - Article_ID = 10 - Quantity = 1 - Unit_Price = 10
Master_ID = 1 - Article_ID = 11 - Quantity = 1 - Unit_Price = 20
Master_ID = 1 - Article_ID = 43 - Quantity = 1 - Unit_Price = 20

INVOICE__PAYMENTS (populated by a trigger)
Due_Date = 2008/10/5 - Amount = 16.66
Due_Date = 2008/11/5 - Amount = 16.66
Due_Date = 2008/12/5 - Amount = 16.66

There are some triggers that would update stock quantyties, customer balance,
etc...

Now the user realizes he made a mistake. The article_id = 43 would not go in
the invoice and also the customer wants 5 of the article_ID = 11, not 1

This is only a case. Of course the combination of variations is huge: the
customer_id may change, the payment terms, the invoice date, etc...

So if I have all the actions in "after delete" and "after insert" triggers, I
think it's much more clear to delete the complete invoice and save it again with
all the posible changes the user may have introduced. If I dont do this, I would
have to write a lot of "after update" triggers to contemplate all the possible
cases of relevant fields that may have changed.

I also must say that I do all the invoice editing in local temp tables, and when
the user saves the invoice I construct all the SQL and send it to the server.
Before that, I worked directly into the database inserting and posting every
record and if the user exited the invoice without saving I did a rollback. But
later I came up with this idea of working in temp tables "outside" the database
and then send all the data in a small and very short transaction. I use this
approach ONLY for master/detail documents (as invoices, sales orders, etc). For
"simple" tables, of course, I use the known tools everybody uses (in my case IBX
and ClientDataSets). So... probably, mine is a very bad idea. And I'd LOVE to
hear other opinions

Thanks to you all, as usual !

-sergio