Using a sub-query in WP instead of Inner Join

Question

I've been investing tons of hours searching and applying many different tests to try and see if I could optimize this query but I was not successfull so far.

So this is based on a table with more than 6M entries and this query takes more than 2 seconds to perform:

SELECT DISTINCT b.meta_value as 'min'
FROM wp_postmeta as a
INNER JOIN wp_postmeta as b
ON a.post_id = b.post_id AND a.meta_key = 'Brand' AND b.meta_key = 'Yearofputtingintoproduction'
WHERE a.meta_value = 'BMW' 
ORDER BY b.meta_value DESC

And then I have this one that takes also more than 2 seconds to load:

SELECT DISTINCT b.meta_value
FROM wp_postmeta as a   
INNER JOIN wp_postmeta as b
ON a.post_id = b.post_id AND b.meta_key = 'OnlineModel'
INNER JOIN wp_postmeta as c
ON a.post_id = c.post_id AND c.meta_key = 'Brand' AND a.meta_key = 'Yearofputtingintoproduction'
INNER JOIN wp_postmeta as d
ON a.post_id = d.post_id AND d.meta_key = 'Yearofstoppingproduction'
WHERE  d.meta_key = 'Yearofstoppingproduction' AND d.meta_value >= '2020' AND a.meta_key = 'Yearofputtingintoproduction' AND a.meta_value <= '2021' AND c.meta_value = 'BMW'
ORDER BY b.meta_value ASC

What am I doing wrong here for this to take so long?

Any help will be much appreciated.

Answer 1

You need the value from one table copy only, and you need to use DISTINCT for to remove duplicates. This leads to replace JOIN with WHERE EXISTS.

SELECT [DISTINCT] b.meta_value as 'min'
FROM wp_postmeta as b
WHERE EXISTS ( SELECT NULL
               FROM wp_postmeta as a
               WHERE a.post_id = b.post_id 
                 AND a.meta_key = 'Brand' 
                 AND a.meta_value = 'BMW' )
  AND b.meta_key = 'Yearofputtingintoproduction'
ORDER BY b.meta_value DESC

Whether DISTINCT can be deleted or not depends on the data.

Answer 2

Have you installed https://wordpress.org/plugins/index-wp-mysql-for-speed/ ? That provides a generic fix for performance issues with "meta" tables in WP.

(This is an xample of where WP's EAV schema is especially ill-suited for the query in question. A simple table with brand , model , start_year , and stop_year would lead to a much simpler and intrinsically faster query.)