对 meta_query 使用 NOT EXISTS 时 Wordpress 慢查询

Question

I am running a query on a relatively small database (10,000 rows in wp_posts and 100,000 rows in wp_postmets) but for some reason the query takes around ~5 to 7 seconds to execute.

The args I'm passing into wp_query are;

Array(
   [post_type] => post
   [post_status] => publish
   [ignore_sticky_posts] => 0
   [posts_per_page] => 1
   [orderby] => rand
   [meta_key] => 
   [meta_value] => 
   [order] => DESC
   [offset] => 0
   [post__not_in] => Array
       (
       )

   [author__in] => Array
       (
       )

   [tax_query] => Array
       (
           [relation] => AND
           [0] => Array
               (
                   [taxonomy] => category
                   [field] => slug
                   [terms] => Array
                       (
                           [0] => beauty
                       )

                   [operator] => IN
                   [include_children] => 0
               )

       )

   [meta_query] => Array
       (
           [relation] => OR
           [0] => Array
               (
                   [key] => sponsored
                   [value] => 1
                   [compare] => !=
                   [type] => CHAR
               )

           [1] => Array
               (
                   [key] => sponsored
                   [value] => 
                   [compare] => NOT EXISTS
                   [type] => CHAR
               )

       )

)

And the SQL query it runs is;

SELECT SQL_CALC_FOUND_ROWS wp_posts.id
FROM   wp_posts
       LEFT JOIN wp_term_relationships
              ON ( wp_posts.id = wp_term_relationships.object_id )
       LEFT JOIN wp_postmeta
              ON ( wp_posts.id = wp_postmeta.post_id )
       LEFT JOIN wp_postmeta AS mt1
              ON ( wp_posts.id = mt1.post_id
                   AND mt1.meta_key = 'sponsored' )
WHERE  1 = 1
       AND ( wp_term_relationships.term_taxonomy_id IN ( 3 ) )
       AND ( ( wp_postmeta.meta_key = 'sponsored'
               AND wp_postmeta.meta_value != '1' )
              OR mt1.post_id IS NULL )
       AND wp_posts.post_type = 'post'
       AND (( wp_posts.post_status = 'publish' ))
GROUP  BY wp_posts.id
ORDER  BY Rand()
LIMIT  0, 1

I have noticed that the long-query issue resolves itself when I remove this part;

Array(
   [key] => sponsored
   [value] => 
   [compare] => NOT EXISTS
   [type] => CHAR
)

Which in the SQL query is;

OR mt1.post_id IS NULL

What I am trying to do is look for articles that are marked as sponsored == 0 or where they're not marked as both sponsored or not-sponsored (sponsored meta_key doesn't exist).

Can someone please point out to me/help me with how I can achieve the same results with a more efficient query/wp_query? Or if you can give me an idea around why OR mt1.post_id IS NULL makes the query run for ~5 to 7 seconds whereas without it the query takes ~20 milliseconds?

Answer 1

Rethink how you specify "is sponsored". You currently have

    LEFT JOIN  wp_postmeta  ON ( wp_posts.id = wp_postmeta.post_id )
    LEFT JOIN  wp_postmeta AS mt1  ON ( wp_posts.id = mt1.post_id
              AND  mt1.meta_key = 'sponsored' 
                          )
    WHERE  ( ( wp_postmeta.meta_key = 'sponsored'
                      AND  wp_postmeta.meta_value != '1' )
              OR  mt1.post_id IS NULL 
           )

This might be simpler to get rid of those 2 JOINs and change that part of WHERE to

    WHERE NOT EXISTS( SELECT 1 FROM wp_postmeta AS m
                WHERE m.post_id = wp_posts.id
                  AND m.meta_key = 'sponsored'
                  AND m.meta_value = '1' )

And, be sure to have WP Index Improvements . It will add an index that will help that subquery.

I can't be sure, but I think that this is unnecessary (and leads to an extra sort of the intermediate results):

    GROUP BY  wp_posts.id

RAND and SQL_CALC_FOUND_ROWS both necessitate finding all the rows with term taxonomy 3. Please provide EXPLAIN SELECT ... so we can discuss whether it is unnecessarily looking at all posts, not just the taxonomy 3 posts. (That may lead to an optimization.)

It's unfortunate that WP does not facilitate putting the sponsored and taxonomy flags in wp_posts. Then, a composite index would lead to a much faster query.