Question

Greetings fellow flowstackers!

Here’s the db schema (reduced to relevant fields) to help illustrate my conundrum:

• A webstore has items, and items have item_options.
• The same item can appear in multiple categories.
• categories, items and item_options can all be active or inactive (BOOL).

Categories look like this (notice the parent_id nesting, where Fruit Seeds are inside of Seeds):

id   parent_id   name              active
 1           0   Seeds                  1
 2           1   Vegetable Seeds        1
 3           1   Fruit Seeds            0
 4           0   Plants                 1
 5           4   Vegetable Plants       1
 6           4   Fruit Plants           1

What I want is a fast list of all active categories (id, parent_id and name), and the count of each category’s active items containing active item_options.

The query result would look like this:

id   parent_id   name              item_count
 1           0   Seeds                      0
 2           1   Vegetable Seeds           52
 4           0   Plants                     0
 5           4   Vegetable Plants         103
 6           4   Fruit Plants              79

This query works, but takes ~430ms:

    SELECT c.`id`, c.`parent_id`, c.`name`,
        (SELECT COUNT(*)
            FROM `item_categories` AS ic
            LEFT JOIN `items` AS i
                ON (i.`id` = ic.`item_id`)
            LEFT JOIN `item_options` AS io
                ON (i.`id` = io.`item_id`)
            WHERE c.`id` = ic.`category_id`
                AND i.`active` = 1
                AND io.`active` = 1
            ) AS `item_count`
    FROM `categories` AS c
    WHERE c.`active` = 1;

This next query takes only ~55ms, but fails to include the top-level categories (where parent_id = 0):

    SELECT c.`id`, c.`parent_id`, c.`name`,
        COUNT(ic.`item_id`) AS `item_count`
    FROM `categories` AS c
    LEFT JOIN `item_categories` AS ic
        ON (c.`id` = ic.`category_id`)
    LEFT JOIN `items` AS i
        ON (i.`id` = ic.`item_id`)
    LEFT JOIN `item_options` AS io
        ON (i.`id` = io.`item_id`)
    WHERE c.`active` = 1
      AND i.`active` = 1
      AND io.`active` = 1
    GROUP BY c.`id`;

Anyone see how to speed up the first query, or fix the second one?

Answer 1

Oh! This seemed to do it …

(
    SELECT c.`id`, c.`parent_id`, c.`name`,
        COUNT(ic.`item_id`) AS `item_count`
    FROM `categories` AS c
    LEFT JOIN `item_categories` AS ic
        ON (c.`id` = ic.`category_id`)
    LEFT JOIN `items` AS i
        ON (i.`id` = ic.`item_id`)
    LEFT JOIN `item_options` AS io
        ON (i.`id` = io.`item_id`)
    WHERE c.`active` = 1
      AND i.`active` = 1
      AND io.`active` = 1
    GROUP BY c.`id`
)
UNION
(
    SELECT c.`id`, c.`parent_id`, c.`name`, 0
    FROM `categories` AS c
    WHERE c.`parent_id` = 0
      AND c.`active` = 1
)
ORDER BY `parent_id`, `NAME`;

Anyone see a better way? 😀