mysql subquery understanding - mysql

Question

I am trying to find all sale_id's that have an entry in sales_item_taxes table, but do NOT have a corresponding entry in the sales_items table.
mysql> describe phppos_sales_items_taxes;
+------------+---------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+------------+---------------+------+-----+---------+-------+
| sale_id | int(10) | NO | PRI | NULL | |
| item_id | int(10) | NO | PRI | NULL | |
| line | int(3) | NO | PRI | 0 | |
| name | varchar(255) | NO | PRI | NULL | |
| percent | decimal(15,3) | NO | PRI | NULL | |
| cumulative | int(1) | NO | | 0 | |
+------------+---------------+------+-----+---------+-------+
6 rows in set (0.01 sec)
mysql> describe phppos_sales_items;
+--------------------+----------------+------+-----+--------------+-------+
| Field | Type | Null | Key | Default | Extra |
+--------------------+----------------+------+-----+--------------+-------+
| sale_id | int(10) | NO | PRI | 0 | |
| item_id | int(10) | NO | PRI | 0 | |
| description | varchar(255) | YES | | NULL | |
| serialnumber | varchar(255) | YES | | NULL | |
| line | int(3) | NO | PRI | 0 | |
| quantity_purchased | decimal(23,10) | NO | | 0.0000000000 | |
| item_cost_price | decimal(23,10) | NO | | NULL | |
| item_unit_price | decimal(23,10) | NO | | NULL | |
| discount_percent | int(11) | NO | | 0 | |
+--------------------+----------------+------+-----+--------------+-------+
9 rows in set (0.00 sec)
mysql>
Proposed Query:
SELECT DISTINCT sale_id
FROM phppos_sales_items_taxes
WHERE item_id NOT IN
(SELECT item_id FROM phppos_sales_items WHERE sale_id = phppos_sales_items_taxes.sale_id)
The part I am confused by is the subquery. The query seems to work as intended but I am not understanding the subquery part. How does it look for each sale?
For example if I have the following data:
mysql> select * from phppos_sales;
+---------------------+-------------+-------------+---------+-------------------------+---------+--------------------+-----------+-----------+------------+---------+-----------+-----------------------+-------------+---------+
| sale_time | customer_id | employee_id | comment | show_comment_on_receipt | sale_id | payment_type | cc_ref_no | auth_code | deleted_by | deleted | suspended | store_account_payment | location_id | tier_id |
+---------------------+-------------+-------------+---------+-------------------------+---------+--------------------+-----------+-----------+------------+---------+-----------+-----------------------+-------------+---------+
| 2014-08-09 17:53:38 | NULL | 1 | | 0 | 1 | Cash: $12.96<br /> | | | NULL | 0 | 0 | 0 | 1 | NULL |
| 2014-08-09 17:56:59 | NULL | 1 | | 0 | 2 | Cash: $12.96<br /> | | | NULL | 0 | 0 | 0 | 1 | NULL |
+---------------------+-------------+-------------+---------+-------------------------+---------+--------------------+-----------+-----------+------------+---------+-----------+-----------------------+-------------+---------+
mysql> select * from phppos_sales_items;
+---------+---------+-------------+--------------+------+--------------------+-----------------+-----------------+------------------+
| sale_id | item_id | description | serialnumber | line | quantity_purchased | item_cost_price | item_unit_price | discount_percent |
+---------+---------+-------------+--------------+------+--------------------+-----------------+-----------------+------------------+
| 2 | 1 | | | 1 | 1.0000000000 | 10.0000000000 | 12.0000000000 | 0 |
+---------+---------+-------------+--------------+------+--------------------+-----------------+-----------------+------------------+
1 row in set (0.00 sec)
mysql> select * from phppos_sales_items_taxes;
+---------+---------+------+-----------+---------+------------+
| sale_id | item_id | line | name | percent | cumulative |
+---------+---------+------+-----------+---------+------------+
| 1 | 1 | 1 | Sales Tax | 8.000 | 0 |
| 2 | 1 | 1 | Sales Tax | 8.000 | 0 |
+---------+---------+------+-----------+---------+------------+
2 rows in set (0.00 sec)
When I run the query below it does find sale_id 1. But how does the subquery know to filter correctly. I guess I am not understanding how the sub query works.
mysql> SELECT DISTINCT sale_id
-> FROM phppos_sales_items_taxes
-> WHERE item_id NOT IN
-> (SELECT item_id FROM phppos_sales_items WHERE sale_id = phppos_sales_items_taxes.sale_id)
-> ;
+---------+
| sale_id |
+---------+
| 1 |
+---------+
1 row in set (0.00 sec)

Answer 1

Duffy356 link to the SQL-Joins is good, but sometimes seeing with your own data might sometimes make more sense...
First, your query as written and obviously learning will be very expensive to the engine. How it knows what to include is because it is doing a correlated sub-query -- meaning that FOR every record IN the sales_items_taxes table it is running a query TO the sales_items table, which is returning every item possible for said sale_id. Then it comes back to the main query and compares it to the sales_items_taxes table. If it does NOT find it, it allows the sale_id to be included in the result set. Then it goes to the next record in the sales_items_taxes table.
(Your query reformatted for better readability)
SELECT DISTINCT
sale_id
FROM
phppos_sales_items_taxes
WHERE
item_id NOT IN ( SELECT item_id
FROM phppos_sales_items
WHERE sale_id = phppos_sales_items_taxes.sale_id)
Now, think about this. You have 1 sale with 100 items. It is running the correlated sub-query 100 times. Now do this with 1,000 sales id entries and each has however many items, gets expensive quickly.
A better alternative is to take advantage of databases and do a left-join. The indexes work directly with the LEFT JOIN (or inner join) and are optimized by the engine. Also, notice I am using "aliases" for the tables and qualifying the aliases for readability. By starting with your sales items taxes table (the one you are looking for extra entries) is the basis. Now, left-join this sales items table on the two key components of the sale_id and item_id. I would suggest that each table has an index ON (sale_id, item_id) to match the join condition here.
SELECT DISTINCT
sti.sale_id
FROM
phppos_sales_items_taxes sti
LEFT JOIN phppos_sales_items si
ON sti.sale_id = si.sale_id
AND sti.item_id = si.item_id
WHERE
si.sale_id IS NULL
So, from here, think of it that each table is lined-up side-by-side with each other and all you are getting are those on the left side (sale items taxes) that DO NOT have an entry on the right side (sales_items).

Answer 2

Your problem can be fixed by using joins.
Read the following article about SQL-Joins and think about your problem -> you will be able to fix it ;)
The IN-clause is not the best solution, because some databases have limits on the number of arguments contained in it.

Answer 3

what you really wanted here is:
SELECT DISTINCT sale_id
FROM phppos_sales_items_taxes
WHERE sale_id NOT IN
(SELECT sale_id FROM phppos_sales_items)
WHERE field NOT IN (SELECT field FROM anothertable WHERE ...) is a perfectly fine query construct.

Answer 4

Your original query:
SELECT DISTINCT sale_id
FROM phppos_sales_items_taxes
WHERE item_id NOT IN
(SELECT item_id FROM phppos_sales_items WHERE sale_id = phppos_sales_items_taxes.sale_id)
Here you are pulling all the item_ids from the phppos_sales_items table where sale_id matches with taxes table, and removing those item_ids from the final result.
You can do also get same results in couple other ways, which may be easy to understand.
Use IN query with multiple columns:
select distinct sales_id
from sales_item_taxes
where (sale_id, item_id) not in (select sale_id, item_id from phppos_sales_items)
-- This form of query is easy to read and understand. Performance may not be good for large tables.
Exists / not exists format:
select distinct sales_id
from sales_item_taxes t1
where not exists (select '1' from phppos_sales_items t2
where t2.sale_id = t1.sale_id
and t2.item_id = t1.item_id
)
I would have also suggested the same solution as 'bwperrin' did - not sure why you didn't get any output by running the query. If your criteria is to filter on sale_id - that is the best solution. But looks like you are using (sale_id, item_id) as a way to identify sales record. Make sure your table structure makes sense.