rlibsphinxclient¶ ↑

A Ruby wrapper for pure C searchd client API library. This is *highly experimental* library so use it at your own risk.

This gem can be more difficult to install than the typical Ruby extension. First you have to install Sphinx and Sphinx pure C searchd client API library.

Go to sphinxsearch.com/downloads.html and download the latest stable release. Then go to api/libsphinxclient directory and install client API to your preferred folder (I like /opt/sphinx):

cd api/libsphinxclient
./configure --prefix=/opt/sphinx
make
sudo make install

On Max OS X you may get the following error:

configure: error: C++ preprocessor "/lib/cpp" fails sanity check

In this case you should specify environment variable for ./configure script:

CXXCPP="gcc -E" ./configure --prefix=/opt/sphinx

If you have installed the Sphinx to /opt/sphinx, just run:

sudo gem install kpumuk-rlibsphinxclient --no-ri --no-rdoc

Otherwise, specify where sphinx has been installed to:

sudo gem install kpumuk-rlibsphinxclient --no-ri --no-rdoc -- --with-libsphinxclient-dir=/opt/sphinx-0.9.9

On Mac OS X with MacPorts you should specify ARCHFLAGS environment variable:

sudo env ARCHFLAGS="-arch i386" gem install kpumuk-rlibsphinxclient --no-rdoc --no-ri -- --with-libsphinxclient-dir=/opt/sphinx-0.9.9

If you are working on Ruby on Rails application, you can add gem dependency to your config/environment.rb:

config.gem 'kpumuk-rlibsphinxclient', :lib => 'sphinx'

Also don’t forget to remove the sphinx plugin, because it’s functionality is completely covered by this gem.

The gem includes two versions of the client API: pure Ruby and wrapper for pure C client API. They are 100% equivalent in use, so you can switch to any of them. To use pure Ruby client, instantiate the Sphinx::Client, for pure C wrapper use Sphinx::FastClient.

Important note: you should call destroy method when you do not need client API any more. The reason for that is the C wrapper saves all query results in memory, and frees them in the destroy method call. You can omit this call in pure Ruby library, but I’d like to do call in any case just for consistence (to be able to switch to another client).

Important note #2: to ensure that destroy method will be called, use ensure block:

begin
  @sphinx = Sphinx::FastClient.new
  @sphinx.Query('test')
ensure
  @sphinx.destroy
end

Ok, let’s take a look at the examples. First, here is the search example with all possible filters and options set:

require 'sphinx'
@sphinx = Sphinx::FastClient.new
@sphinx.SetServer('localhost', 3312)
@sphinx.SetLimits(1, 100, 20, 30)
@sphinx.SetMaxQueryTime(5)
@sphinx.SetMatchMode(Sphinx::Client::SPH_MATCH_EXTENDED2)
@sphinx.SetRankingMode(Sphinx::Client::SPH_RANK_BM25)
@sphinx.SetSortMode(Sphinx::Client::SPH_SORT_RELEVANCE)
@sphinx.SetFieldWeights('group_id' => 10, 'rating' => 20)
@sphinx.SetIndexWeights('test1' => 20, 'test2' => 30)
@sphinx.SetIDRange(1, 100)
@sphinx.SetFilter('group_id', [1], true)
@sphinx.SetFilterRange('group_id', 1, 2, true)
@sphinx.SetFilterFloatRange('rating', 1, 3, true)
@sphinx.SetGroupBy('created_at', Sphinx::Client::SPH_GROUPBY_DAY)
@sphinx.SetGroupDistinct('group_id')
@sphinx.SetRetries(5, 10)
results = @sphinx.Query('test')
@sphinx.destroy

BuildKeywords example:

require 'sphinx'
@sphinx = Sphinx::FastClient.new
results = @sphinx.BuildKeywords('wifi gprs', 'test1', true)
@sphinx.destroy

BuildExcerpts example:

require 'sphinx'
@sphinx = Sphinx::FastClient.new
results = @sphinx.BuildExcerpts(['what the world', 'London is the capital of Great Britain'], 'test1', 'the')
@sphinx.destroy

UpdateAttributes example:

require 'sphinx'
@sphinx = Sphinx::FastClient.new
results = @sphinx.UpdateAttributes('test1', ['group_id'], { 2 => [1] })
@sphinx.destroy

The reason to write this gem was to investigate why we keep getting timeout errors when using Sphinx (occur rarely, but they are annoying me.) But the side effect of this library was the slight search performance improvement: Ruby library is slower when generating Sphinx request and parsing its results.

require 'sphinx'
require 'benchmark'

def run_test(klass)
  sphinx = klass.new
  sphinx.Query('test hello')
ensure
  sphinx.destroy
end

Benchmark.bm do |x|
  x.report('pure ruby') { 1000.times { run_test(Sphinx::Client) } }
  x.report('c wrapper') { 1000.times { run_test(Sphinx::FastClient) } }
end

On my MBP I got the following results:

           user       system     total    real
pure ruby  0.420000   0.230000   0.650000 ( 14.721659)
c wrapper  0.060000   0.090000   0.150000 (  2.248645)

This plugin has been created in Scribd.com for our internal use and then the sources were opened for other people to use. All the code in this package has been developed by Dmytro Shteflyuk for Scribd.com and is released under the MIT license. For more details, see MIT-LICENSE file.