Question

i made a query for sql but it’s really heavy and i really don’t see how i can optimize it, the explanations comes after :

SELECT tyds.user_tyd, 
       tyds.product_tyd, 
       tyds.action_tyd, 
       products.price_product, 
       data_8, 
       data_9, 
       avg_8, 
       sum_9 
FROM   tyds 
       INNER JOIN products 
               ON tyds.product_tyd = products.id_product 
       INNER JOIN (SELECT product_tyd, 
                          Avg(data_tyd) AS avg_8 
                   FROM   tyds 
                   WHERE  action_tyd = 8 
                   GROUP  BY product_tyd) Agg_1 
               ON Agg_1.product_tyd = tyds.product_tyd 
       INNER JOIN (SELECT product_tyd, 
                          Sum(data_tyd) AS sum_9 
                   FROM   tyds 
                   WHERE  action_tyd = 9 
                   GROUP  BY product_tyd) Agg_2 
               ON Agg_2.product_tyd = tyds.product_tyd 
       INNER JOIN (SELECT product_tyd, 
                          data_tyd AS data_8 
                   FROM   tyds 
                   WHERE  user_tyd = 3 
                          AND action_tyd = 8) Agg_3 
               ON Agg_3.product_tyd = tyds.product_tyd 
       INNER JOIN (SELECT product_tyd, 
                          data_tyd AS data_9 
                   FROM   tyds 
                   WHERE  user_tyd = 3 
                          AND action_tyd = 9) Agg_4 
               ON Agg_4.product_tyd = tyds.product_tyd 
WHERE  tyds.user_tyd = 3 
       AND tyds.action_tyd = 1 
GROUP  BY tyds.product_tyd 

All this jointures because i need to get a lot of things :
I define an id_user in tyds.user_tyd = 3
tyds.action_tyd = 1 because i want to group by the tyds.product_tyd WHERE the previous id_user as an tyds.action_tyd = 1.

Then i want :
-data_8 is the value of tyds.data_tyd WHERE user_tyd = 3 AND action_tyd = 8
-same for data_9

-for AVG and COUNT i want to skip the tyds.user_tyd condition and only group by id_product.

Actually, this query is working but i think it’s really heavy and has too much SELECT…

I put a similar question a month ago but i had to review my architecture so i’m sorry…

Thanks a lot.

e1 : i use MySQL

Answer 1

You don’t actually need all of those joins and subqueries; you can eliminate most of them by moving logic from WHERE-clauses into the field-list, using CASE expressions:

SELECT 3 AS user_tyd, tyds.product_tyd, 1 AS action_tyd,
       products.price_product,
       MAX(CASE WHEN user_tyd = 3 AND action_tyd = 8 THEN data_tyd END) data_8,
       MAX(CASE WHEN user_tyd = 3 AND action_tyd = 9 THEN data_tyd END) data_9,
       AVG(CASE WHEN action_tyd = 8 THEN data_tyd END) AS avg_8,
       SUM(CASE WHEN action_tyd = 9 THEN data_tyd END) AS sum_9
  FROM tyds
  JOIN products
    ON tyds.product_tyd = products.id_product
 WHERE tyds.product_tyd IN
         ( SELECT product_tyd
             FROM tyds
            WHERE tyds.user_tyd = 3
              AND tyds.action_tyd = 1
         )
 GROUP
    BY tyds.product_tyd
;

(Note: the MAX(...) in the third and fourth lines are just a way of forcing a non-null value to be preferred over a null one.)

You’ll have to test to see if this is actually faster, but I’m betting that it is.