Sunspot::Rails
Sunspot::Rails is a Rails plugin that provides drop-in integration of the Sunspot Solr search library with Rails. It provides the following features:
-
Configure Sunspot using config/sunspot.yml
-
Extend ActiveRecord for easy index configuration, search, and indexing
-
Automatically index ActiveRecord objects when they are saved, and remove them from the index when they are destroyed (can be disabled)
-
Automatically commit Solr changes at the end of each request
-
Provide utility methods to find and fix orphaned documents and rebuild the Solr index for a given class
-
Provide rake tasks for starting and stopping the development Solr instance, using the configuration in sunspot.yml
Sunspot::Rails has been tested with Rails versions 2.1, 2.2, and 2.3
Installation
First, install the Sunspot and Sunspot::Rails gems:
sudo gem install outoftime-sunspot outoftime-sunspot_rails --source=http://gems.github.com
In your project’s config/environment.rb
, add the following gem dependencies:
config.gem 'outoftime-sunspot', :lib => 'sunspot'
config.gem 'outoftime-sunspot_rails', :lib => 'sunspot/rails'
If you are using an older version of Rails that doesn’t support plugins-as-gems, use:
script/plugin install git://github.com/outoftime/sunspot_rails.git
Create the file config/sunspot.yml
and set it up for your environments. Here is a sample:
common: &common
solr:
hostname: localhost
port: 8983
production:
<<: *common
solr:
path: /solr/myindex
development:
<<: *common
solr:
port: 8982
test:
<<: *common
solr:
port: 8981
Rails doesn’t automatically load rake tasks from plugins installed as gems (rails.lighthouseapp.com/projects/8994/tickets/59). If you installed Sunspot::Rails as a gem, add the following line to your project’s Rakefile:
require 'sunspot/rails/tasks'
If you wish to make modifications to the Solr schema, you can create a custom Solr home in your project directory. In order to do so, create the directory RAILS_ROOT/solr/conf
, and copy in the contents of the Solr gem’s solr/solr/conf
directory. If the files are in the right place, Sunspot::Rails will detect them and tell Solr to use your local configurations. Use caution when modifying schema.xml
- Sunspot relies on the field naming scheme in the packaged schema file.
To start up a Solr instance, issue the following:
rake sunspot:solr:start
Note that using the built-in Solr instance packaged with Sunspot is great for development, but is not recommended for production. See the Sunspot documentation for more information.
Usage
Setup
In order for an ActiveRecord model to be indexable and searchable, it must be configured for search. For example:
class Post < ActiveRecord::Base
searchable do
text :title, :body
integer :blog_id
time :updated_at
string :sort_title do
title.downcase.sub(/^(an?|the) /, '')
end
end
end
See the documentation for Sunspot.setup for full details on what can go in the configuration block.
Indexing
By default, models are indexed whenever they are saved, and removed from the index whenever they are destroyed. This behavior can be disabled:
class Post < ActiveRecord::Base
searchable :auto_index => false, :auto_remove => false do
# setup...
end
end
Note that using the :auto_remove
option is not recommended , as destroying an object without removing it from the index will create an orphaned document in the index, which is a Bad Thing. Turning off :auto_index
is perfectly safe if you prefer to manage indexing manually (perhaps using a background job).
If you have disabled lifecycle indexing hooks, you can invoke indexing operations directly on your model:
post = Post.create
post.index
post.remove_from_index
Committing
When data is changed in Solr, it is initially stored in memory and not made available to the currently running searcher instance. Issuing a commit
to Solr will cause it to write the changes to disk, and instantiate a new searcher instance. This operation is fairly expensive, so rather than issuing a commit every time a document is added or removed, Sunspot::Rails issues a commit at the end of any request where data has been added to or removed from Solr. If you need to immediately issue a commit, bang!-versions of the methods are available:
post = Post.create
post.index!
# this is the same as...
post.index
Sunspot.commit
When writing tests outside of the context of a controller request, you will want to use one of these two approaches.
Searching
Do it like this:
Post.search do
with :blog_id, 1
with(:updated_at).greater_than(Time.now - 2.weeks)
order :sort_title, :asc
paginate :page => 1, :per_page => 15
end
See the documentation for Sunspot.search
for all the options available in the search block, and the information available in the result block.
Searching for IDs
In some situations, you may want to get the IDs for models returned by a search without actually loading the models out of the database. For that, you can call search_ids
, using the same block format as #search. This will return an array of IDs.
Searching for multiple types
Sunspot is entirely agnostic about whether searches are for one or more types; the only restriction is that columns used for restriction, ordering, etc. are defined in the same way for all types being searched. Sunspot::Rails does not provide any additional support for this, since there is not anything useful to be added, so just use the interface provided by Sunspot:
Sunspot.search(Post, Comment) do
with :blog_id, 1
order :created_at, :asc
end
Be sure to check out the Sunspot documentation for all the details.
Adding search functionality in mixins
Sunspot does not require that search setup for a given class happen all in one place; it is perfectly acceptable to call the Sunspot.setup
method more than once. This capability is particularly useful for adding search functionality in mixins. For instance, if you have a Ratable
module, you may wish to add additional search fields for searchable classes that mix in that module. For example:
module Ratable
def self.included(base)
if base.searchable?
base.searchable do
float :average_rating do
.average(:value)
end
end
end
end
end
Note the use of base.searchable?
- this ensures that only classes that already have search enabled will have the additional configuration added. The above pattern requires that the class be declared searchable before the module is mixed in; other patterns (such as passing a :searchable option to an acts_as_-style declaration) may be more flexible.
Utility methods
If you need to completely reindex all of the instances of a given model class, you can issue:
Post.reindex
If for some reason models get deleted from the database, but not from the index, they will become index orphans - not a good situation. To get IDs that exist in the index but not the database, you can use the index_orphans
method; to remove those documents from the index, use clean_index_orphans
. Note that neither of these operations should be needed if Sunspot and Sunspot::Rails are used as intended.
Further Reading
Reading the Sunspot documentation is highly recommended. Sunspot::Rails exists to wrap Sunspot with a Rails-friendly API, but almost all of the functionality you use in Sunspot::Rails is implemented in Sunspot.
Posts about Sunspot on my blog are available at outofti.me/tagged/sunspot
Bugs
Please submit bug reports to outoftime.lighthouseapp.com/projects/20339-sunspot
Contributors
-
Mat Brown ([email protected])
-
Peer Allan ([email protected])
-
Michael Moen ([email protected])
-
Benjamin Krause ([email protected])
License
Sunspot::Rails is distributed under the MIT License, copyright © 2009 Mat Brown