Has anyone written or ported a tagging system for Code Igniter? I’m looking for a tagging library that I can plug into CI.

Freetag (http://code.google.com/p/freetag/) is a very good tagging library, but it is written for ADODB. I’d port it over myself, but that would take a few hours to do properly and I was hoping someone might have a library ready to go.

Thanks in advance!

I wrote some tagcloud code a while ago. It can generate folksonomy (no <font> tags, yay!) and goes off a single comma separated list stored in a database (one list per tagged item). All the files are at http://wgilk.com/code/ci/tagcloud/. It’s probably not as advanced as freetag, but it probably wouldn’t be hard to duplicate the API.

I’ve got a tagging system on Filepanda
Which uses 2 DB Fields… ‘tags’ and ‘files_tags’ to remove duplication.

So, in ‘tags’, you might have:

and in files_tags:

As a linking table.
It works great, and provides a much more normalized database structure.
I’ll post the actual code up as soon as I can.

Elliot

I’m doing a similar thing to Elliot for an email address management application i’ve built for a client.  i also added ‘Collections’ to the tags (which are basically just tags for tags) which have more or less the same db structure, a table for the collections and a joining table for the tags.

Ok, this is my tagging code:

It relies on 2 db tables: tags ( id [int, autoincrement], tag [varchar] ) and files_tags ( file_id [int, indexed], tag_id [int, indexed] ).

Then, to get a file’s tags you have to select from files_tags with a join on tags.

Hope this is helpful…
Elliot.

I considered using a table for tags, but I honestly think it’s less efficient to use a table for tags, unless you are trying to limit the name or number of tags. There’s a bit more processing involved with using a comma separated list, but imho, it’s more flexible.

all in all… saving in a normalized database method has many advantages.

Firstly, you can keep the tags indexed within the db… which is more efficient than constantly sorting arrays.
Secondly, you save space… potentially huge amounts of space.
Third, you can mass ‘rename’ a tag…. for example, ‘coool’ > ‘cool’.......just edit one record in the DB, and every instance of that tag changes.
Fourthly, you can use the database of tags for other things too…
I have ‘tags’ and ‘files_tags’, but you could also have ‘blog_tags’, ‘images_tags’, all as linking tables to the ‘main’ database tags table.

As a final example… let’s imagine you have 100,000 files stored… and each has an average of 3 tags at 8 characters each.

Thats’ 24 characters + 2 for commas, for each file.
which comes to 2,600,000 characters of text… which is roughly translated at 2.4MB.
using a db, with a linking table, each tag reference is just 2 integer numbers.

2 x 3 = 6 numbers stored.
6 X 100,000 = 600,000   which is 0.57MB

As far as flexibility goes… using a normalized database is still just as flexible, you’ve just got to get your head around the queries that you need to use to do what you want it to do.

Elliot

I appreciate the responses. Elliot, you make some great points. I’ve decided to atleast *attempt* to port Freetag over to CI. I’ve spent the last hour or so cleaning out all the ADODB functions and replacing them with the equivalent CI database functions.

Here’s what I’ve got so far. Keep in mind this is still a work in progress and I haven’t even begun to fully test it so I am sure there are plenty of bugs. If you attempt to use this class, please make sure you know what you’re doing and can help me fix the bugs and make improvements as needed. I know for a fact that there are things here that still need to be properly ported, but feel free to point out anything that may look amiss even if it is obvious.

Here it is so far:
http://codeigniter.com/wiki/Freetag/

How would you guys go about implementing a general tagging library that could be used with multiple linking tables for various different things eg: blogs, videos, photos, etc.

Should I just adopt a naming convention for the tables and pass a name on initialization of the class?

meaning tables names are Video_Tags, Blog_tags, Photo_Tags and they all contain objectID and tagID. Then it will take the name of the object id field too. eg:

Should this be the best way to go? I didnt like how freetag restricts you to the table names, seems hugely useless to me!

I know that a table based tagging system is the correct was to do these things, but you will get far better real world performance using your DB’s fulltext search function. For example, using a FULLTEXT index on a tags VARCHAR column in a mysql table allows you to write complex queries (”+cheese +mince -rice” means find all rows tagged cheese AND mince BUT NOT rice) that run very, very fast. Note from someone who has done this: My default mysql will not index words of 3 characters or less. Append underscores to tags shorter than this to make them 3 chars and strip them off later!

If you want REAL power you could use the sphinx fulltext search component (although its a pain to set up). But that would allow you to normalise terms (i.e. treat ‘machine’ and ‘machines’ as the same tag)....

Got bored and made a mini one. Seems to work fine for what i’ve tested.

Temp Link

Theifed a few minor bits from freetag but hopefully this should give some people an idea of where to start on their own.

Elliot thx for sharing. Always checking your blog for new entries.

Elliot,
Can you please repost you tagging system? Looks like it’s been deleted from filepanda.

Interested in knowing how it works.
Cheers