Comments on: Tags with MySQL fulltext

By: Ricson

Ricson — Mon, 06 Sep 2010 14:21:45 +0000

I initially user first solution without fulltext index, at the beginning no performance issue was found. But when I got 500,000 articles on my sites, it took several seconds to fetch the list of articles with specified tag. Then I added full text index on tag field, now it’s much faster on tag query.

By: S!ava

S!ava — Sat, 25 Jul 2009 15:18:21 +0000

Nice solution but limited to use one word tag.
I found another good solution to store tags in the same structure separated by comma. And you can use any character includin ‘-’ in tag name.

CREATE TABLE `delicious` (
`id` int(11) NOT NULL auto_increment,
`tags` text
PRIMARY KEY (`id`),
FULLTEXT KEY (`tags`)
) ENGINE=MyISAM

The sql will be

SELECT * FROM `delicious` WHERE concat(‘, ‘,tags,’,') like ‘%, two,%’

and if executed against a table with this data

ID, tags
—————————————
1, ‘one two three, numbers, consecutive
2, ‘two times, three times, another tag
3, ‘tags, two, five, seven’

will show the result:

ID, Tags
—————————————
3, ‘tags, two, five, seven’

By: The Experience Game Blog

The Experience Game Blog — Wed, 15 Oct 2008 14:23:12 +0000

Tags, search and a tagcloud…

I have been thinking about how to implement the Tags and which database structure would be best for it. In the end I settled for a method using fulltext search. Now that I had the tags working and knew how search worked I added the actual search functi…

By: html-utvikler » Blog Archive » Tags og MySQL - en perfekt kombinasjon

html-utvikler » Blog Archive » Tags og MySQL - en perfekt kombinasjon — Wed, 09 Apr 2008 19:23:37 +0000

[...] Tags with MySQL fulltext [...]

By: Egbert Wietses

Egbert Wietses — Mon, 09 Jul 2007 04:43:15 +0000

Why not have both, fulltext with mysql and a table wich keeps track off all different tags and adds everytime one tag is used.

- Egbert

By: Philipp Keller

Philipp Keller — Sun, 10 Jun 2007 13:34:31 +0000

Nickel: I did some perfomance measurements. It depends how your data looks and how big your database will become. If you "just" have a few thousand entries, then I would go for the "toxi schema" and would add a "fulltext table" with the columns "bookmark_id", "user_id" and "tags". This extra table holds redundant data that helps you with the boolean fulltext queries (much much faster than LIKE queries). If you have less complicated queries (just and), then I'd just stick with the toxi schema and wouldn't add the extra fulltext table. As for the tag cloud: If you want to have a "global tag cloud" (whatever this is for, I find them rather silly), you'd have to compute this from time to time. Doing tag clouds with the toxi solution is fairly simple (some group by clauses and you're in)

By: Nickel

Nickel — Thu, 07 Jun 2007 18:32:33 +0000

Hi guys,

Very informative article and comments!

I will appreciate if you can suggest the most scalable and performance-oriented solution in my scenario. I have a PHP/MySQL based application, with the following DB schema (simplified):

Table: Users
————
UserID
[other misc. fields]

Table: Bookmarks
—————-
BookmarkID
BookmarkURL
[other misc. fields]

Table: UserBookmarks
——————–
UBID
UserID
BookmarkID
BookmarkTags (Varchar)

This schema allows multiple users to manage multiple bookmarks. Now, what approach would be ideal in the following two scenarios:

1. If I want to search for bookmarks based on keywords. Eg: “mysql database” -postgres. Programatically, a boolean search parser can be implemented with either ‘LIKE’ or ‘Fulltext’, but I’m concerned about performance (say with a thousands of bookmarks and tags, and high frequency search queries from the users). I need to show unique bookmarks matching the search query and also the number/count of users who have bookmarked it? What do you guys reckon?

2. If I want to build a tag cloud, displaying all the unique tags and their count per user bookmarks.

Please share your thoughts. Thanks in advance!

By: Philipp Keller

Philipp Keller — Fri, 23 Mar 2007 12:45:02 +0000

Mark: You would do that in your program. It is not possible to do this on the database side, that’s actually a downside of this db schema.

By: Mark Carey

Mark Carey — Fri, 23 Mar 2007 12:10:32 +0000

I like the fulltext approach to tagging.

With this approach what is the best way to generate a list of unique tags, with counts for each?

Such as:

books (3)
programming (10)
work (1)

By: jon lee dot see eh » Blog Archive » mediatemple woes

jon lee dot see eh » Blog Archive » mediatemple woes — Fri, 02 Feb 2007 19:54:54 +0000

[...] However, perhaps more exciting is their promise of every user’s own MySQL container! I hope this means I can finally be able to tinker with my own MySQL server settings! What do I need this for you ask? Well for example for one of my projects I needed to use a MySQL fulltext implementation of a tagging system, and by default, MySQL has its minimum word length limit set to 4, meaning all tags of length 3 or less won’t be searched! So I asked mediatemple to set ft_min_word_length = 3 but of course they wouldn’t because that would affect everybody else on my server (grid?) [...]