Comments on: Joining to one of two tables with one ID

By: Damien

Damien — Sat, 13 Aug 2011 08:09:56 +0000

I'm doing a similar thing – one table with all contacts.
I'm also doing this for the products and services tables.
All relationships between objects are linked through a separate master table 'idlinker'
In this table, any row can be related to any row in any table.. I'm not sure I explained that very well..
This means I can have one contact assigned to many locations, orders can be assigned here depending on the accent of the query, etc..

Im sure there are shortfalls to this method but it works for me..

By: Ryan

Ryan — Tue, 31 May 2011 01:03:22 +0000

Interesting ideas, thanks for the feedback guys.

By: Seth S.

Seth S. — Mon, 30 May 2011 18:56:33 +0000

My first thought would be to have a generic person table (e.g., "contact") for the basic and common attributes like name, etc.

Then have vendor and customer tables that are associated to contact by contactID.

Then you could use "contactID" as your creatorID.

Obviously that would be a major re-org, but just thinking theoretically here.

By: CephalidOne

CephalidOne — Mon, 30 May 2011 08:47:29 +0000

or you could do 2 select with cross join and a union

By: Ryan

Ryan — Mon, 23 May 2011 23:46:49 +0000

Roland, sorry I had a typo in my last example. I am doing a JOIN here without creating another ID field. So there is no orders.vendorID field, that was a typo. I am joining to either of one of two tables using a single ID field (creatorID).

Thanks for your comment.

By: Roland Collins

Roland Collins — Mon, 23 May 2011 22:55:23 +0000

You could also use a COALESCE, which is going to perform better and require no new fields at all!

SELECT orders.*,
COALESCE(customer.name, vendor.name) AS name
FROM orders
LEFT JOIN customers on orders.creator ID = customers.customerID
LEFT JOIN vendors on orders.vendorID = vendors.vendor ID