原文出处:http://www.quarkruby.com/2007/8/12/acts_as_solr-for-search-and-faceting
Table of Contents
Solr is a search server based on lucene java search library with a HTTP/XML interface. Using Solr, large collections of documents can be indexed based on strongly typed field definitions, thereby taking advantage of Lucene's powerful full-text search features. is a ruby on rails plugin adding Solr capabilities to activerecord models. It hides all configuration and manual setting efforts with Solr and provides you with simple find_by...
methods. acts_as_solr can be used as a replacement to acts_as_ferret because of inbuilt full text search capabilities ;-) . The purpose of this article is to explain acts_as_solr with examples.
acts_as_solr
Installation: Installation is well explained on acts_as_solr homepage and getting started with acts_as_solr
Note: acts_as_solr requires jre1.5 on system. Before running any of the solr methods make sure you start solr server with rake solr:start
command.
Our example model for this tutorial will be DigitalCamera [classname: Camera] with following fields
- name (type:string)
- brand (type:string) [we want faceted browsing on this field]
- resolution (type:float) [we want faceted browsing on this field]
- other fields which we do not want to index
Basic Usage : for search
Lets start with basic search and then we will move on to faceted browsing. You need to specify which of the columns from your model file you want to be indexed for search (if no :fields
param is given, then all columns are indexed) :
1 2 3 | class Camera acts_as_solr :fields=>[:name,:brand,:resolution] end |
| @results = Camera.find_by_solr("canon powershot") |
:limit, :offset, :scores, :order, :operator
]
limit
: to limit the count of search resultsoffset
: starting index of the search results.scores
: solr scores for each of the result returnedorder
: by which field results should be ordered and in asc/desc order?operator
: words in query should be separated by "and" or "or" i.e. all words should exist in matching results or any of the words respectively
Note:
- If you do not want Camera models objects as results, you can use
find_id_by_solr
which will return just id's from the Camera table (score
option is not supported yet) - The result of
find_by_solr
is not an array of camera objects. Its a ActsAsSolr::SearchResults object.
You can get model objects and count of results by
1 2 | @products = @results.docs @total_hits = @results.total_hits |
Pagination
We will be using will_paginate for pagination (rails default pagination is buggy). One way of paginating returned results is explained nicely at Using will paginate with acts_as_solr. But we don't need an additional module for getting count of results returned. This module is an overhead because of extra Solr query. acts_as_solr returns the total number of results via
find_by_solr
and you don't need to call
count_by_solr
separately for getting result count. In your .rhtml file include this where you want pagination
1 2 | <%= will_paginate WillPaginate::Collection.new((params[:page]||1), (params[:products_per_page]||10),@total_hits) -%> |
Faceting
Presentation by Keith Instone is an excellent read on faceted browsing with examples.
Back to acts_as_solr ... open the model file. Using option :facets => ...
add the columns on which you want to allow faceted browsing.
1 2 3 4 | class Camera acts_as_solr :fields=>[:name,:brand,:resolution], :facets=>[:brand,:resolution] end |
resolution between 5 and 7
). Modify the model to look like:
1 2 3 4 | class Camera acts_as_solr :fields=>[:name,:brand,{:resolution=>:range_float}], :facets=>[:brand,:resolution] end |
1 2 | @results = Camera.find_by_solr("powershot",{:facets=> {:zeros=>false,:fields=>[:brand]}}) |
:zeros
parameter tells solr not to return the brand values whose facet count is zero.@results.facets["facet_fields"]["brand_facet"]
contains the names of brand with the corresponding counts. A sample result:{"Canon USA"=>1, "Canon PowerShot"=>1, "Canon"=>91}
But what about values for resolution facet as we have defined resolution to be varying float type (float_range)? Something like this does not work:
1 2 3 | # INCORRECT @results = Camera.find_by_solr("powershot",{:facets=> {:zeros=>false,:fields=>[:brand,:resolution]}}) |
facet_query
param of
find_by_solr
with ranges for resolution predefined i.e. Solr cannot calculate the range for the float/int fields itself and we need to specify the range of values while querying solr. For example if values in resolution column range from 0 to 20 and we want to have 4 facet ranges. Your query would be something like
1 2 3 4 5 6 7 | @results = Camera.find_by_solr("powershot", {:facets=>{:zeros=>false, :fields=>[:brand], :query=>["resolution:[0 TO 4]","resolution:[5 TO 9]", "resolution:[10 TO 14]","resolution:[15 TO 20]"] } }) |
@results.facets["facet_queries"]
If you want to display the results to the user as links along with counts (something like example images at the top), where user can make a further selection, you need to get results from solr using
1 2 3 | @results = Camera.find_by_solr("powershot"+" AND resolution:[0 TO 4]", {:facets=>{:zeros=>false, :fields=>[:brand]}}) ## resolution queries has been removed since you already made a selection in resolution itself. |
between 0 to 4
and
between 6 to 7
. So we can define query (first argument to find_by_solr) as
"powershot" + " AND (resolution:[0 TO 4] OR resolution:[6 TO 7])"
.
But unfortunately,
find_by_solr(query+" AND brand:Canon")
doesn't seems to work
By default all fields are of type string and faceting for these fields is done using
browse
parameter i.e.
1 2 3 4 5 6 | @results = Camera.find_by_solr(query,{:facets=> {:zeros=>false, :query=>["resolution:[0 TO 4]", "resolution:[5 TO 9]", "resolution:[10 TO 14]", "resolution:[15 TO 20]"], :browse=>["brand:Canon"]} }) |
Note: I don't know how to make this work with browse field because multiple options in browse are separated by AND and
:operator
option from
find_by_solr
works only with words in query.
A good alternative is to redefine your model fields which are of string type to field_type as
:facet
. Our acts_as_solr declaration becomes:
1 2 | acts_as_solr :fields=>[:name,{:brand=>:facet}, {:resolution=>:range_float}], :facets=>[:brand,:resolution] |
"powershot" + " AND (brand:Canon OR brand:Sony)"
Boosting
Using boost option you can give one field priority over the others. Just a small change in acts_as_solr declaration is enough.
| acts_as_solr :fields => [{:name=>{:boost=>2}},:brand,:resolution] |
Quick Tips
- How to search solr with no query or 'nil' query. i.e. Product pages with navigation without a search query. Use
query = "[* TO *]"
1 2 3 4 5 6 7
@results = Camera.find_by_solr("[* TO *]", {:facets=>{:zeros=>false, :fields=>[:brand], :query=>["resolution:[0 TO 4]", "resolution:[5 TO 9]", "resolution:[10 TO 14]", "resolution:[15 TO 20]"] } })
- Already have database, how to integrate solr now? Goto ./script/console and run
Camera.rebuild_solr_index
- Adding or updating rows to table automatically adds/modifies them in table.
- If you have already created index on solr (not acts_as_solr), it will not work with acts_as_solr. This is because, in solr configuration, schema.xml tells the solr exactly about the columns you are going to index, their data types. But, with acts_as_solr you never modify
schema.xml
. Actually, acts_as_solr uses dynamic fields to tell solr about the fields and their data types. So, acts_as_solr does not finds anything in your already created index because field names are no more same.
Testing Solr
Copy vendor/plugins/acts_as_solr/test/test_helper.rb
modified as shown below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 | class Test::Unit::TestCase begin Net::HTTP.get_response(URI.parse('http://localhost:8981/solr/')) rescue Errno::ECONNREFUSED raise "You forgot to 'rake solr:start RAILS_ENV=test', foo!" end def self.fixtures(*table_names) if block_given? Fixtures.create_fixtures(Test::Unit::TestCase.fixture_path, table_names) { yield } else Fixtures.create_fixtures(Test::Unit:: TestCase.fixture_path, table_names) end table_names.each do |table_name| clear_from_solr(table_name) klass = instance_eval table_name.to_s.capitalize.singularize.camelize klass.find (:all).each{|content| content.solr_save} end end private def self.clear_from_solr(table_name) ActsAsSolr::Post.execute(Solr::Request::Delete.new(:query => "type_t:#{table_name.to_s.capitalize.singularize.camelize}") end end |
Modifications in the original file
- removed top lines
- added camelize for removing "_" from model names
- added lines to test presence of test server running .. (from google groups)
Now include, this file in your unit/functional tests just like you include test_helper.rb
in all your test files. After you have saved above file as RAILS_APP/test/solr_test_helper.rb
do not forget to include solr_test_helper in your test files and start writing your tests. :)
Oops! Still not working
Did you make sure that config/solr.yml
has been configured for testing environment. (by default solr test server runs on port 8981)
References
- acts_as_solr benchmark results
- acts_as_solr performance on some live site
- http://www.webdesignpractices.com/navigation/facets.html
- http://cfis.savagexi.com/articles/2007/07/10/how-to-profile-your-rails-application
- Deploying acts_as_solr on tomcat