MySQL 8.0 GROUP BY / FULL_GROUP_BY - mysql

Question

Since upgrading to MySQL 8 I have a lot of queries that dont comply to the new full-group-by setting of MySQL. Below is a simplified version of one of the queries. I have A LOT of them and before going through all the code, I want to fully understand the problem.
I have the following records in the database:
[prod_id] => 1
[prod_name] => Product 1
[prod_new] => 50.00
[prod_size] => L
[prod_id] => 2
[prod_name] => Product 1
[prod_new] => 45.00
[prod_size] => M
[prod_id] => 3
[prod_name] => Product 1
[prod_new] => 40.00
[prod_size] => S
[prod_id] => 4
[prod_name] => Product 4
[prod_new] => 100.00
[prod_size] => M
[prod_id] => 5
[prod_name] => Product 5
[prod_new] => 200.00
[prod_size] => M
When I ran the following query in MySQL 5.x, I got 3 results. Containing products 1, 4, 5. With the corresponding name, price and size.
SELECT prod_id, prod_name, prod_price, prod_size
FROM prod_product
GROUP BY prod_name
Since the upgrade I get the widely know error about nonaggregated columns. So I want to fix this, but in some cases this will get me unwanted results. Lets say, for some reason, I wanted the highest product-id.
SELECT MAX(prod_id), prod_name, ANY_VALUE(prod_price), ANY_VALUE(prod_size)
FROM prod_product
GROUP BY prod_name
This will give me product-id's 3, 4, 5. But with product-id 3 it gives me the price and size of product-id 1.
Obviously that is unwanted behaviour. I would assume that, since prod_id is the primary-key, the database knows which values to show with the corresponding id. When I say MAX(prod_id) this already pinpoints a single record in this group, why give me values of other records from this group?
I guess I am missing something important here. =)
Thanks!

Answer 1

I would assume that, since prod_id is the primary-key, the database knows which values to show with the corresponding id.
Why would it know that?
Consider the following queries:
SELECT prod_name, prod_id, MIN(prod_price), MAX(prod_price)
FROM prod_product
GROUP BY prod_name
Which value should it return for prod_id here? The product that corresponds to the minimum price? Or the product that corresponds to the maximum price?
Also, what if there are multiple products that tie for the minimum or maximum price? Which one should it return?
SELECT prod_name, prod_id, AVG(prod_price)
FROM prod_product
GROUP BY prod_name
Now which prod_id should it infer? The aggregate calculation AVG() is likely to return a value that doesn't correspond to any single product.
The same happens with the aggregate SUM().
The fact is, there is no implicit correlation between an aggregate function and a specific row in the group. You should not expect SQL to guess which row from the group you mean to reference when you use non-aggregated expressions.

Answer 2

If you want the first record in every group of rows having the same prod_name ordered by prod_id, you can use window function ROW_NUMBER(), which is available in MySQL 8 :
SELECT x.prod_id, x.prod_name, x.prod_new, x.prod_size
FROM (
SELECT
p.prod_id, p.prod_name, p.prod_new, p.prod_size,
ROW_NUMBER() OVER(PARTITION BY p.prod_name ORDER BY p.prod_id) rn
FROM prod_product p
) x WHERE x.rn = 1
The inner query assigns a number to each record within each group, and the outer query filters in the first record in each group.
Demo on DB Fiddle :
WITH prod_product AS (
SELECT 1 prod_id, 'Product 1' prod_name, 50 prod_new, 'L' prod_size
UNION ALL SELECT 2, 'Product 1', 45, 'M'
UNION ALL SELECT 3, 'Product 1', 40, 'S'
UNION ALL SELECT 4, 'Product 4', 100, 'M'
UNION ALL SELECT 5, 'Product 5', 200, 'M'
)
SELECT x.prod_id, x.prod_name, x.prod_new, x.prod_size
FROM (
SELECT
p.prod_id, p.prod_name, p.prod_new, p.prod_size,
ROW_NUMBER() OVER(PARTITION BY p.prod_name ORDER BY p.prod_id) rn
FROM prod_product p
) x WHERE x.rn = 1;
| prod_id | prod_name | prod_new | prod_size |
| ------- | --------- | -------- | --------- |
| 1 | Product 1 | 50 | L |
| 4 | Product 4 | 100 | M |
| 5 | Product 5 | 200 | M |