Mysql Select from table where a date is max for each unique location_id AND product

Question

I have a table - records - that looks like this

id | location_id | product | stock | date  
===|=============|=========|=======|======
1  | Bakery 1    |  cake   |  21   | 2
2  | Bakery 1    |  bread  |  23   | 2
3  | Bakery 2    |  cake   |  21   | 2
4  | Bakery 2    |  bread  |  21   | 2
5  | Bakery 1    |  cake   |  21   | 3
6  | Bakery 1    |  bread  |  23   | 3
7  | Bakery 2    |  cake   |  21   | 3
8  | Bakery 2    |  bread  |  21   | 3
9  | Bakery 1    |  cake   |  21   | 4
10 | Bakery 1    |  bread  |  23   | 4
11 | Bakery 2    |  bread  |  23   | 4

For each location and for each product, i want to pick the row that has the highest date value. Which will look like ths

id | location_id | product | stock | date  
===|=============|=========|=======|======
7  | Bakery 2    |  cake   |  21   | 3
9  | Bakery 1    |  cake   |  21   | 4
10 | Bakery 1    |  bread  |  23   | 4
11 | Bakery 2    |  bread  |  23   | 4

How do i execute this with one query? I could loop through all the locations and all the products and build the queries but its more time and memory consuming?

Answer 1

select r1.*
from records r1
join
(
  select location_id, product, max(date) as date
  from records
  group by location_id, product
) r2 on r1.location_id = r2.location_id 
    and r1.product = r2.product
    and r1.date = r2.date

Answer 2

If the following conditions are true you can do this job very efficiently using the id column

1. your id column is autoincrementing
2. the date column represents the current date: that is, you never insert a backdated record.
3. you don't UPDATE the date column

If these conditions are true, that means the unique identifier id can be used as a proxy for the non-unique value date for each record. A row with a higher date value is always guaranteed to have a higher id value.

If you can do this you can make your query perform very well.

First, you create a subquery to get the id of the latest row for each combination of location and product.

SELECT MAX(id)
  FROM records
 GROUP BY location_id, date

Then you use that set of id values to pull the correct records from your table

SELECT *
  FROM records
 WHERE id IN (
                SELECT MAX(id)
                 FROM records
                GROUP BY location_id, date
              )
ORDER BY location_id, product