Question

Firstly, I’m rather new to SQL and I’ve run into a roadblock. I’m using the Mimer SQL system.

I have three tables: “Transactions”, roughly equivalent with a receipt total, which I want to update with data from a selection and the tables “item”, containing prices and “sale”, containing number of items and item IDs for a given transaction, which I have to join in order to get the data to update Transaction with.

SELECT sale.T_ID, SUM(sale.quantity * item.price) as total
FROM sale
INNER JOIN item
ON sale.I_ID = item.I_ID
GROUP BY T_ID

Gives me the desired data selection with the transaction IDs and the sum total for that transaction:

T_ID Amount
1    100
2    150
etc...

I want to update the Transaction table, which contains columns “T_ID” and “Total”. I want to match the T_IDs and update the Total with the data from the corresponding Amount in the selection. The query:

UPDATE transaction SET total = (
    SELECT total FROM (
        SELECT sale.T_ID, SUM(sale.quantity * item.price) as total
            FROM sale
            INNER JOIN item
            ON sale.I_ID = item.I_ID
            GROUP BY T_ID)
    WHERE transaction.T_ID=T_ID);

I can sense that the above statement is faulty, but unable to discern the problem. How should I construct the query?

Answer 1

Only select the SUM(sale.quantity * item.price) in your subquery and remove the select total one.

update Transaction
    set total = (
        select SUM(sale.quantity * Item.price)
            from Sale
                inner join Item on Sale.I_ID = Item.I_ID
            where Sale.T_ID = Transaction.T_ID
    )

That is what I would try out in SQL Server.

I would have prefered to have some sample data so that I can test it against my local SQL Server database in order to verify whether this statement does what is expected, though you’re not using SQL Server, the idea is still the same.

EDIT #1

Though I do not fully understand why it does not overwrite values in the transaction.total column when a transaction is composed of several different items, your suggested query works.

The behaviour of the SUM function is to sum all targeted records resulting in only one scalar value for each Sale and Item rows. There can be only one sum, can it not?

That said, it multipies each resulting rows from the constraint, that is, for a particular transaction.

1. It selects both Sale and Item rows for this very transaction;
2. It then iterates through each of the resulting rows and multiply Sale.Quantity and Item.Price together, which is worth a value for each row;
3. Once a row is multiplied, its total is then added to a total which is stored somewhere in memory;
4. Once it has processed all the rows for that transaction, it comes out with a scalar value, which is the sum of all the rows in Sale and Item for that given transaction;
5. This SUM ends up to be this transaction’s total.

In other words, the subquery “knows” for which transaction to sum which is filtered in the subquery itself. The Transaction table is accessible in the subquery as part of the main SQL statement. So, putting it in the where clause filters the rows from Sale and Item that will be later multiplied and additioned for the total.

Does this help you better understand this update statement?

Please feel free to ask your questions. =)