Query Filtering Entries

Here's one for you guys:

I have a table with sales entries in it. To make things simple, I will put it this way:

There is an entry in the table for every item scanned. There may be some items that get voided out, but there will still be an entry in the table. No problem for the one transaction that was voided, because there is a negative dollar amount which I've already filtered out easily. However, there will still be an entry in the table for the same Item that got voided with a positive dollar amount. Make sense? I want to filter out the one entry with the positive dollar amount because it's not really a sale. It was voided in another entry in the table... So, If I come to the store, and buy a pack of gum for $0.99. The cashier scans it, then I decide I don't want it, so the cashier voids it off. I've got two entries in the table for that pack of gum. One for $0.99 where the cashier scanned, and one for -$0.99 where the cashier voided it off. I've got to get rid of the positive transaction.

Too confusing?Any chance you could add a column to use as a void flag? This would
make filtering this item much less cumbersome...|||nope. I can't make structural changes like that.|||Originally posted by AnSQLQuery
Here's one for you guys:

I have a table with sales entries in it. To make things simple, I will put it this way:

There is an entry in the table for every item scanned. There may be some items that get voided out, but there will still be an entry in the table. No problem for the one transaction that was voided, because there is a negative dollar amount which I've already filtered out easily. However, there will still be an entry in the table for the same Item that got voided with a positive dollar amount. Make sense? I want to filter out the one entry with the positive dollar amount because it's not really a sale. It was voided in another entry in the table... So, If I come to the store, and buy a pack of gum for $0.99. The cashier scans it, then I decide I don't want it, so the cashier voids it off. I've got two entries in the table for that pack of gum. One for $0.99 where the cashier scanned, and one for -$0.99 where the cashier voided it off. I've got to get rid of the positive transaction.

Too confusing?

Not too confusing...but it will be easier if you are running this through several tables and views.

Assuming that the transactions have a date time stamp and/or indexing column on it, and item number you could work it this way

Assuming that the sales_table is something like this

Index Item_Number Sales_Amt .....

Make a view called voided_sales which just sees the voided transactions and the sales_amt*-1

Insert into Voided_Indexes
Select max(index)
From sales_table
group by index
having index < (select index from sales_table, voided_sales
where sales_table.sales_Amt = voided_sales.Sales_Amt
and sales_table.Item_Number = voided_sales.Item_Number)

Once you get the voided_indexes list built then your query becomes
Select *
From sales_table
where index not in (select index from voided_sales)
and index not in (select index from voided_Indexes)

There are probably more efficient ways of doing it, but this is a quick scratch off the top of my head.|||select * from sales where saleAmt not in (select saleAmt * -1 from sales1)

The subquery may not make it efficient but this is probably the simplest
form of query...|||Originally posted by rocket39
select * from sales where saleAmt not in (select saleAmt * -1 from sales1)

The subquery may not make it efficient but this is probably the simplest
form of query...

The problem with that is that if the item subsequently sells it will be blocked out as well as the valid void.|||Originally posted by jimpen
The problem with that is that if the item subsequently sells it will be blocked out as well as the valid void.

Yep, I guess I was assuming that there would be some piece of information that would relate the void to the sale transaction.|||It just seems like there should be a more simple solution. (I haven't exactly figured out the creating a view suggestion just yet.)

It seems like i should be able to say if there is an entry that matches this TransactionID # and this dollar amt*-1 within the query, so that I'm using whatever transaction Id the query is on at that moment..don't count it. :) I know that is kind of funny sounding, but ...|||Originally posted by AnSQLQuery
It just seems like there should be a more simple solution. (I haven't exactly figured out the creating a view suggestion just yet.)

It seems like i should be able to say if there is an entry that matches this TransactionID # and this dollar amt*-1 within the query, so that I'm using whatever transaction Id the query is on at that moment..don't count it. :) I know that is kind of funny sounding, but ...

Create or Replace View voided_sales as Select Index, Item_Number, (Sales_Amt*-1) As Voided_Sales_Amt, ... or something similar...I've been doing more Oracle this week than MSSQL.

I rushed it should be
sales_table.sales_Amt = voided_sales.Voided_Sales_Amt

The reason for the view is trying to keep multiple recursive queries straight is a royal PITA. The reason for the max index is to get the most recent sale to match the return. That is not a guarentee match to the sale, but it will at least get it out of the totals.|||This just came to me, and I don't know why I didn't think of it before, but I would think that I could just get rid of the filter that takes out the voided transactions all together.

Since the voided transactions have a negative dollar amount, when I do a SUM() on the Dollars the negative amounts will cancel out, and The dollar figure will be correct. (I think anyway)

As for the number of sales, all I have to do is count the number of voided transactions, and subtract that amount from the total number of sales.

I'll let you know if this works...|||you can use self join for that, join the table with itself using the product id, price (with different signs) to get the records you want. thats the most simple - elegant way i can think of though im not sure how hot it is performance wise.

consider flaggnig voided records when inserting the negative valued record using either a trigger or (much better) in the sproc performing the insertion.

Good luck