ActAsFuzzySearch
================
this plugin is inspired by the following blog entry
http://unirec.blogspot.com/2007/12/live-fuzzy-search-using-n-grams-in.html
it can add a fuzzy search to model. this search needs another table where all the
trigrams are stored but that table is hidden otherwise.
the whole plugins does not work with ActiveRecords, but for Datamapper !! to get it to work
for ActiveRecord only a few code places need to be changed, I marked the in the code #DM.
configuration
=============
add the following in the model class
include FuzzySearch
act_as_fuzzy_search :property1, :property1
where :property1, :property2 are the properties which are index for the fuzzy search. each property gets split into words (separator is the whitespaces)
if you define a methods returning a string
def self.normalize(word)
before include the FuzzySearch than this method gets used to normlize any word. default is to downcast
the string (which does only work properly within ascii)
trigrams table
==============
the naming follows the convention
class <model_classname>Trigram
property :id, Integer, :serial => true
property :<model_lowercase>_id, Integer, :index => true
property :token, String, :nullable => false, :length => 3, :index => true
end
example
=======
class UserTrigram
include DataMapper::Resource
property :id, Integer, :serial => true
property :user_id, Integer, :index => true
property :token, String, :nullable => false, :length => 3, :index => true
end
the model class looks like this
class User
include DataMapper::Resource
def self.normalize(word)
#TODO something
word.downcase
end
include FuzzySearch
act_as_fuzzy_search :firstname, :surname
property :id, Integer, :serial => true
property :surname, String, :nullable => false , :format => /^[^<'&">]*$/, :length => 32
property :firstname, String, :nullable => false , :format => /^[^<'&">]*$/, :length => 32
end
now you can create users and search them in a fuzzy manner like
User.fuzzy_find("Heywy")
enjoy !!
Copyright (c) 2008 kristian, released under the MIT license
