So, recently one of the object listings on one of my Rails sites crossed the dreaded 50+ mark and It was time to add searching. However, I didn't want to just scan all the fields of a model or make complicated queries, so I set out to write something more meaningful. How do you use it? Like this. First of all you add a column: ALTER TABLE posts ADD COLUMN searchable varchar(250); and then you tell something to your ActiveRecord class Post < ActiveRecord::Base indexes_columns :title, :tags, :body, :author end It will give you reasonably sane keywords for the columns. If the columns are associations (which is the idea of the plugin also) they will get indexed together with the main model. Of course your index can get out of sync when you delete associated objects, but the thing was meant to be simple, so don't be too harsh. Also, you can do: class Post < ActiveRecord::Base indexes_columns :title, :tags, :body, :author, :into=>:search_terms end which will index separate keywords into the table called "search_terms" and give you a model to work with this table (SearchTerm). After that you can query for stuff: Post.find_with_term("Windows sucks") # I hope everyone is offended which will issue a LIKE query for "windows" AND "sucks" if you use a column in the model OR "windows" and "sucks" from the "search_terms" table if you are using it instead. The "terms table" is clunky but effective (I've seen it used many times in other projects so I thought - why not). All the index maintenance is performed automagically (or at least is supposed to). The only thing missing for complete joy is the third indexer, for Ferret. Ferret is much more advanced and much more versatile than all this rough crap, so what I would like to have is a thin wrapper around it which "hooks" your ActiveRecords into ferret but still lets you fine-tune everything. All through the same syntax. The reason why I didn't START with Ferret in the first place was because I wanted to install it ASAP and TxD hasn't got ferret on it's shareds yet. This was also not my original intention - I just needed an elegant way for my clients to search their lists :-) The approach (in it's current form) seems quite extensible (you can make new Indexers yourself) and anyone who wants to tackle Ferret integration (there are starting points) is free to do so - I will have to learn the Ferret API first and I ain't got time now. If someone wants he can also implement fulltext searching for MySQL etc. but I didn't want to do it right away because I like the plugin being DB-agnostic. The plugin has some little dirty schticks to it that I implemented and I consider nice,. for example dates will be indexed as "day of week - month", and if you search for "thursday" you should, among others, get the blog entries written on, well, Thursdays. Which is sorta nice. Absolutely not academic but nice. So, the docs are here: http://www.julik.nl/code/active-search/ and the plugin itself can be pulled from my subversion http://julik.textdriven.com/svn/tools/rails_plugin... Any contributions are welcome. Note that the SVN URL is expected to be unavailable sometimes because it's on the server with a very bad track record and I currently can't do anything about it. It was called Julik::SimpleSearch but I just bit the bullet and renamed it.
on 2005-12-05 02:13