Question

Confusing title…my apologies.

What I’ve got is a table with two related rows. I need to get the value of a column in one row based on the value of a column in another row.

Given the following postmeta table:

+----------+------------+---------------------------------------------------+--------------------+
| meta_id  | post_id    | meta_key                                          |  meta_value        |
+----------+------------+---------------------------------------------------+--------------------+
| 6917     | 661        | member_categories_0_member_categories_name        | 11                 |
+----------+------------+---------------------------------------------------+--------------------+
| 6918     | 661        | member_categories_0_member_categories_description | First description  |
+----------+------------+---------------------------------------------------+--------------------+
| 6919     | 661        | member_categories_1_member_categories_name        | 12                 |
+----------+------------+---------------------------------------------------+--------------------+
| 6920     | 661        | member_categories_1_member_categories_description | Second description |
+----------+------------+---------------------------------------------------+--------------------+

I need to get the meta_value category description based on the meta_value category ID and the post_id. For instance, if my category ID is 11 and my post_id is 661, I should get “First description”.

I thought about using a subquery to get the meta_key corresponding with a meta_value of ’11’, but I don’t know how to find the description based on the counter inside ‘member_categories_x_member_categories_name’.

Unfortunately, I don’t have control over the name of the meta_key. I got as far as this simple query, which returns ‘member_categories_0_member_categories_name’. How do I use that value to find ‘First description’?

    SELECT pm.meta_key 
    FROM postmeta pm
    WHERE pm.meta_value = "11"
    AND pm.post_id = 661

Here’s the SQL for the table:

CREATE TABLE `postmeta` (
    `meta_id` BIGINT(20) UNSIGNED NOT NULL AUTO_INCREMENT,
    `post_id` BIGINT(20) UNSIGNED NOT NULL DEFAULT '0',
    `meta_key` VARCHAR(255) NULL DEFAULT NULL,
    `meta_value` LONGTEXT NULL,
    PRIMARY KEY (`meta_id`),
    INDEX `post_id` (`post_id`),
    INDEX `meta_key` (`meta_key`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB
AUTO_INCREMENT=30814;

INSERT INTO `postmeta` (`meta_id`, `post_id`, `meta_key`, `meta_value`) VALUES (6917, 661, 'member_categories_0_member_categories_name', '11');
INSERT INTO `postmeta` (`meta_id`, `post_id`, `meta_key`, `meta_value`) VALUES (6918, 661, 'member_categories_0_member_categories_description', 'First description');
INSERT INTO `postmeta` (`meta_id`, `post_id`, `meta_key`, `meta_value`) VALUES (6919, 661, 'member_categories_1_member_categories_name', '12');
INSERT INTO `postmeta` (`meta_id`, `post_id`, `meta_key`, `meta_value`) VALUES (6920, 661, 'member_categories_1_member_categories_description', 'Second description');

Answer 1

This is ugly, but then, that table is so far from normalized that any answer short of scanning every row is going to be something like this:

SELECT pm2.meta_value 
  FROM postmeta pm1
  JOIN postmeta pm2
    ON pm1.post_id = pm2.post_id
   AND SUBSTRING(pm1.meta_key,1,LENGTH(pm1.meta_key)-5) = SUBSTRING(pm2.meta_key,1,LENGTH(pm2.meta_key)-12)
   AND pm1.meta_key like '%_name'
   AND pm2.meta_key like '%_description'
 WHERE pm1.meta_value = 11
   AND pm1.post_id = 661

The idea is to join the table to itself, linking rows that have the same post_id, and whose meta_key is ‘similar’ — it needs to be exactly the same, except that one ends with _name and one ends with _description.