Discovered something neat with the new version of MySQL and thought it warranted a mention. Storing tree structures in a relational database is a common use case across many different areas of tech. The problem comes when you need to construct a query based on a subset of that tree.
But MySQL 8 has some nice new features that makes doing this a breeze.
For example, let’s assume you have a set of tables that look like this:
CREATE TABLE `files` ( `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT, `name` varchar(255) NOT NULL, `parent_id` bigint(20) NOT NULL, `kind` enum('file','folder') NOT NULL, PRIMARY KEY (`id`), KEY `parent_id` (`parent_id`) ); CREATE TABLE `tags` ( `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT, `name` varchar(255) NOT NULL, PRIMARY KEY (`id`) ); CREATE TABLE `tag_file` ( `tag_id` bigint(20) NOT NULL, `file_id` bigint(20) NOT NULL, PRIMARY KEY (`tag_id`,`file_id`) );
This is a pretty standard setup for storing a tree setup in a relational table,
parent_id key referencing
id of the parent. And this all works
well … right up until you need to query a parent and all of it’s children.
So let’s say you want to find all children that have a tag of ‘Important’.
MySQL 8 includes support for recursive common table expressions. Using this, this becomes a pretty easy query. You can create a CTE query and recursively call it! You could so something like this:
with recursive cte (id, parent_id) as ( select id, parent_id from files where parent_id = ? and kind = 'folder' union all select p.id, p.parent_id from files p inner join cte on p.parent_id = cte.id where kind = 'folder' ) select * from files inner join tag_file on (tag_file.file_id = file.id) inner join tags on (tags.id = tag_file.tag_id) where kind = 'file' and tag = 'Important' and parent_id in (select id from cte);
(That is a prepared statement, replace the
? with the ID of the parent.)
What surprised me was how fast that query is with the right keys. On a table that has nearly 600,000 items in it, that query completes in about 0.3 seconds. Slow, but considering the number of rows in the table quite fast.
Thanks to this post on Stack Overflow for the heads up.