Query Optimization with Indexes - mysql

Question

I am needing some assistance on optimizing this WordPress/WooCommerce query:
SELECT
p.ID AS order_id
,DATE(p.post_date) AS order_date
,SUBSTR(comment_content,17) AS csr
,SUBSTR(p.post_status,4) AS order_status
,UCASE(CONCAT((SELECT wp_postmeta.meta_value FROM wp_postmeta WHERE meta_key = '_billing_first_name' and wp_postmeta.post_id = p.ID),' ',(SELECT wp_postmeta.meta_value FROM wp_postmeta WHERE meta_key = '_billing_last_name' and wp_postmeta.post_id = p.ID))) AS customer
,(SELECT GROUP_CONCAT(DISTINCT order_item_name ORDER BY order_item_name ASC SEPARATOR ', ') FROM wp_woocommerce_order_items WHERE order_id = p.ID AND order_item_type = 'line_item' GROUP BY order_id) AS products
,(SELECT GROUP_CONCAT(CONCAT(serial_number,'',serial_feature_code)) FROM wp_custom_serial WHERE wp_custom_serial.order_id = p.ID GROUP BY wp_custom_serial.order_id) AS serials
FROM
wp_posts AS p
INNER JOIN wp_comments AS c ON p.ID = c.comment_post_ID
INNER JOIN wp_postmeta AS pm ON p.ID = pm.post_id
WHERE
p.post_type = 'shop_order'
AND comment_content LIKE 'Order placed by%'
GROUP BY p.ID
ORDER BY SUBSTR(comment_content,17) ASC, p.post_date DESC;
I do not understand what EXPLAIN is telling me and need some guidance on how to speed it up. Can someone describe what, in the EXPLAIN response, indicates where my issue is and where to look for answers?
id
select_type
table
partitions
type
possible_keys
key
key_len
ref
rows
filtered
Extra
1
PRIMARY
c
NULL
ALL
comment_post_ID
NULL
NULL
NULL
20452
11.11
Using where; Using temporary; Using filesort
1
PRIMARY
p
NULL
eq_ref
PRIMARY,post_name,type_status_date,post_parent,post_author
PRIMARY
8
db.c.comment_post_ID
1
50.00
Using where
1
PRIMARY
pm
NULL
ref
post_id
post_id
8
db.c.comment_post_ID
33
100.00
Using index
2
DEPENDENT SUBQUERY
wp_postmeta
NULL
ref
post_id,meta_key
post_id
8
func
33
2.26
Using where
3
DEPENDENT SUBQUERY
wp_postmeta
NULL
ref
post_id,meta_key
post_id
8
func
33
2.30
Using where
4
DEPENDENT SUBQUERY
wp_woocommerce_order_items
NULL
ref
order_id
order_id
8
func
2
10.00
Using where
5
DEPENDENT SUBQUERY
wp_custom_serial
NULL
ALL
NULL
NULL
NULL
NULL
5160
10.00
Using where; Using filesort

Answer 1

Queries are processed in distinct stages. The first clauses processed are the FROM, then WHERE, and then the SELECT clause. Those dependent subqueries mean that for each row that you "have" after processing the FROM and WHERE clauses you are running separate, new subqueries for each row of those results. In your case you are doing that times four.
You can usually rework this to move these queries out of the SELECT clause and into the FROM clause.
Taking one column you have, the serials column, I think you would want to move that into the FROM clause in a way like this
SELECT p.ID AS order_id
, DATE(p.post_date) AS order_date
, SUBSTR(comment_content, 17) AS csr
, SUBSTR(p.post_status, 4) AS order_status
, UCASE(CONCAT((SELECT wp_postmeta.meta_value
FROM wp_postmeta
WHERE meta_key = '_billing_first_name' and wp_postmeta.post_id = p.ID), ' ',
(SELECT wp_postmeta.meta_value
FROM wp_postmeta
WHERE meta_key = '_billing_last_name' and wp_postmeta.post_id = p.ID))) AS customer
, (SELECT GROUP_CONCAT(DISTINCT order_item_name ORDER BY order_item_name ASC SEPARATOR ', ')
FROM wp_woocommerce_order_items
WHERE order_id = p.ID
AND order_item_type = 'line_item'
GROUP BY order_id) AS products
, serials_sub.serials
FROM wp_posts AS p
INNER JOIN wp_comments AS c ON p.ID = c.comment_post_ID
INNER JOIN wp_postmeta AS pm ON p.ID = pm.post_id
LEFT JOIN (
SELECT p.ID as post_id, GROUP_CONCAT(CONCAT(cs.serial_number, '', cs.serial_feature_code)) AS serials
FROM wp_custom_serial cs
JOIN wp_posts AS p ON cs.order_id = p.ID
WHERE p.post_type = 'shop_order'
AND comment_content LIKE 'Order placed by%'
GROUP BY cs.order_id
) as serials_sub ON serials_sub.post_id = p.ID
WHERE p.post_type = 'shop_order'
AND comment_content LIKE 'Order placed by%'
GROUP BY p.ID
ORDER BY SUBSTR(comment_content, 17) ASC, p.post_date DESC;
The difference here is that instead of separate queries being performed for each row, a single subquery is used in the initial FROM clause. So while perhaps looking more unwieldy, in fact this will give you much better performance.
Following this pattern for the other subqueries I think will resolve your issues.
If interested here is the documentation on the EXPLAIN.
https://dev.mysql.com/doc/refman/8.0/en/execution-plan-information.html
And I recommend the book High Performance MySQL.

Answer 2

The outer wp_postmeta does not seem to be used. Remove the JOIN if possible.
Hiding information in the middle of strings leads to SUBSTR() usage, which is inefficient.
The GROUP BY p.ID seems to be unnecessary.
The plugin WP Index Improvements would help with some parts.