Question

I’m really not sure how to even phrase the question. So I’ll go by example. I currently use this SQL command to get a total of all sales:

SELECT sum(ot.value*o.currency_value)
  FROM orders_total ot
  LEFT JOIN orders o on o.orders_id = ot.orders_id
 WHERE ot.class = 'ot_total'
   AND o.date_purchased between '2010-01-01' and '2010-12-31'

And that works fine, but now I want to do a report that cross-references the above with a 3rd table (orders_products) to sum up only the orders containing a specific product id. So I tried this:

SELECT sum(ot.value*o.currency_value)
  FROM orders_total ot
 (LEFT JOIN orders o on o.orders_id = ot.orders_id)
  LEFT JOIN orders_products op on ot.orders_id = op.orders_id
 WHERE ot.class = 'ot_total'
   AND o.date_purchased between '2010-01-01' and '2010-12-31'
   AND op.products_id = 321

But that gives me a higher-than-expected total. Investigating manually, I discovered the obvious reason is that any given order can (of course) have more than one product.

I’d like to show an example but I can’t do tables here it seems.

Q: How do I sum up a total value without getting duplicate records from orders matching multiple entries in the op table?

Does that make any sense at all?

Edit:

I feel like I’m sort of onto something with this:

SELECT distinct o.orders_id
  FROM orders o
  JOIN orders_products op on o.orders_id = op.orders_id
 WHERE o.date_purchased between '2010-01-01' and '2010-12-31'
   AND op.products_id = 321

That results in a listing of all orders_id represented by the product in question. So now I need to sort of inject that result into another statement summing up the values? But how??

Edit:

Here’s an attempt to show my tables:

orders_total

orders_total_id orders_id class      value 
--------------- --------- ---------- -------
    1                 1   ot_sub       100
    2                 1   ot_shipping  10
    3                 1   ot_total     110
    4                 2   ot_sub       200
    5                 2   ot_shipping  10
    6                 2   ot_total     210
    7                 3   ot_sub       50
    8                 3   ot_shipping  5
    9                 3   ot_sub       55

orders

orders_id    currency_value    date_purchased
---------    --------------    --------------
   1              1.0000         2010-04-20
   2              1.0000         2010-05-05
   3              1.0000         2010-06-01

orders_products

orders_products_id    orders_id    products_id
------------------    ---------    -----------
    1                     1            321
    2                     2            555
    3                     2            132
    4                     2            321
    5                     3            132

So I want an SQL statement that will give me a result of 320 (total of all orders containing product ID 321, which is orders 1 and 2 but not 3; “value” of “ot_total” for 1 is 110 and for 2 is 210. 110 + 210 = 320).

EDIT/SOLUTION:

Thanks to JNK for turning me on to EXISTS. As it turns out, this did the job nicely:

SELECT sum(ot.value*o.currency_value) FROM orders_total ot LEFT JOIN orders o ON o.orders_id = ot.orders_id WHERE EXISTS (SELECT NULL FROM orders_products op WHERE op.products_id = 321 AND op.orders_id = o.orders_id) and o.date_purchased between '2010-01-01' and '2010-12-31' and ot.class = 'ot_total'

Answer 1

Use EXISTS – this is a perfect use case.

SELECT <all your fields>
FROM table
LEFT JOIN orders o
   ON table2.key = table.key
WHERE EXISTS (SELECT NULL
              FROM orders_products op
              WHERE op.products_id = xxx
              AND op.orderid = o.orderid)

This will do a short-circuit comparison on the subquery. If the row in the outer query matches, it gets included. If not, it’s not in the final result set.