如何优化具有多对多关系的 mysql COUNT() 查询（通过第 3 个表）

Question

Here is the query that i execute:

SELECT COUNT(*)
FROM (SELECT `p`.*
      FROM `shop_products` `p` LEFT JOIN
           `shop_tag_assignments`
           ON `p`.`id` = `shop_tag_assignments`.`product_id` LEFT JOIN
           `shop_tags`
           ON `shop_tag_assignments`.`tag_id` = `shop_tags`.`id`
      WHERE `p`.`status`=1
      GROUP BY `p`.`id`
     ) `c`

This query takes about 300 milliseconds(i think it's too long..)

EXPLAIN QUERY:

EXPLAIN QUERY IMAGE

DB tables dump:

1k records in shop_tags table

CREATE TABLE `shop_tags` (
  `id` int(11) NOT NULL,
  `name` varchar(255) COLLATE utf8_unicode_ci NOT NULL,
  `slug` varchar(255) COLLATE utf8_unicode_ci NOT NULL,
  `label` text COLLATE utf8_unicode_ci NOT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;

ALTER TABLE `shop_tags`
  ADD PRIMARY KEY (`id`),
  ADD UNIQUE KEY `idx-shop_tags-slug` (`slug`),
  ADD KEY `id` (`id`);

ALTER TABLE `shop_tags`
  MODIFY `id` int(11) NOT NULL AUTO_INCREMENT, AUTO_INCREMENT=1162;
COMMIT;

Table structure for shop_tag_assignments :

224k records in shop_tag_assignments table

CREATE TABLE `shop_tag_assignments` (
  `product_id` int(11) NOT NULL,
  `tag_id` int(11) NOT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

ALTER TABLE `shop_tag_assignments`
  ADD PRIMARY KEY (`product_id`,`tag_id`),
  ADD KEY `idx-shop_tag_assignments-product_id` (`product_id`),
  ADD KEY `idx-shop_tag_assignments-tag_id` (`tag_id`),
  ADD KEY `_index_name` (`product_id`,`tag_id`),
  ADD KEY `__index_name` (`tag_id`,`product_id`);

ALTER TABLE `shop_tag_assignments`
  ADD CONSTRAINT `fk-shop_tag_assignments-product_id` FOREIGN KEY (`product_id`) REFERENCES `shop_products` (`id`) ON DELETE CASCADE,
  ADD CONSTRAINT `fk-shop_tag_assignments-tag_id` FOREIGN KEY (`tag_id`) REFERENCES `shop_tags` (`id`) ON DELETE CASCADE;
COMMIT;

Mysql version : 5.7.16-10-log

Answer 1

Get rid of the subquery. And don't use backticks so much:

SELECT COUNT(DISTINCT p.id)
FROM shop_products p LEFT JOIN
     shop_tag_assignments sta
     ON p.id = sta.product_id LEFT JOIN
     shop_tags st
     ON sta.tag_id = st.id
WHERE p.status = 1;

For this query, you want indexes on shop_products(status, id) .

Next, your query is counting the number of rows returned by the inner subquery after aggregation by p.id . You are using LEFT JOIN , so nothing is being filtered out. That means that the joins are really redundant. I think this does the same thing:

SELECT COUNT(*)
FROM shop_products p 
WHERE p.status = 1;

This should be much faster than your version . . . however, your version is pretty fast so you might not notice a really big difference.