ExpressionLab

In one of my Ruby on Rails projects, I needed a full-text search engine to integrate. And after some research, I decided that Thinking Sphinx is the way to go. It seemed simple, fast and well documented. Unfortunately, I had some issues when tried to get it running on Windows.

The first step is installing Sphinx itself. While this looks simple, it took me five hours to move to the next step!

When I checked Sphinx documentation, it said that I needed Microsoft Visual C/C++ Studio to compile the source code! What? This is another project. After some (a lot of) research, some people suggested that I just needed to download the binaries, not the source code, and add the bin folder to the Windows Path environemnt variable. I tried that, but got errors later when tried to index data. So here is what works:

1. Go to Sphinx downloads page and download the Win32 release binaries with MySQL support. At the time of writing it is sphinx-0.9.8-win32.zip. Of course, you should get the one with PostgreSQL support if you need it.
2. Unzip the downloaded file and open the bin folder. Select and copy all files and paste them in your ruby/bin folder. If some files are already exist in your ruby/bin folder, don't replace them.
3. Install the Thinking Sphinx plugin:

ruby script/plugin install git://github.com/freelancing-god/thinking-sphinx.git

4. To set a model to be indexed, add some fields and attributes (refer to the primer):

   class Post < ActiveRecord::Base  
     define_index do   
       #sphinx fields
       indexes title
     end  
   end

5. Tell Sphinx to index the data:

   rake thinking_sphinx:index

   Now, you may get something like this in your console:

   rake thinking_sphinx:index
   Generating Configuration to F:/InstantRails/rails_apps/community/config/development.sphinx.conf
   indexer --config F:/InstantRails/rails_apps/community/config/development.sphinx.conf --all

   It seems that it couldn't continue. To solve this, make sure you copied all the .exe files from sphinx-0.9.8-win32/bin to ruby/bin. If you keep getting the same result, you may need to copy the libmySQL.dll file from mysql/bin to ruby/bin (check this post for details). Now, you should get something like this:

   Generating Configuration to F:/InstantRails/rails_apps/community/config/development.sphinx.conf
   indexer --config F:/InstantRails/rails_apps/community/config/development.sphinx.conf --all
   Sphinx 0.9.8-release (r1371)
   Copyright (c) 2001-2008, Andrew Aksyonoff
   
   using config file 'F:/InstantRails/rails_apps/community/config/development.sphinx.conf'...
   indexing index 'post_core'...
   collected 2 docs, 0.0 MB
   collected 0 attr values
   sorted 0.0 Mvalues, 100.0% done
   sorted 0.0 Mhits, 100.0% done
   total 2 docs, 27 bytes
   total 0.030 sec, 904.37 bytes/sec, 66.99 docs/sec
   distributed index 'post' can not be directly indexed; skipping.

   Also, note the newly generated files:
   • config/development.sphinx.conf
   • db/sphinx/development/post_core.spa
   • db/sphinx/development/post_core.spd
   • db/sphinx/development/post_core.sph
   • db/sphinx/development/post_core.spi
   • db/sphinx/development/post_core.spm
   • db/sphinx/development/post_core.spp
   Great! But what was that last line about? Don't worry, Pat says it's OK:

   A distributed index is made up of other indexes. It doesn't need to be indexed, but Sphinx tries to anyway, and then complains when it can't. I'm not sure about the reasoning for this, but it's not important: You can ignore the error, it doesn't stop anything working, searching and indexing will happen as expected.

6. Start a Sphinx searchd daemon:

   rake thinking_sphinx:start

   You should get something like this:

   searchd --config F:/InstantRails/rails_apps/community/config/development.sphinx.conf
   [Wed Oct 29 14:26:44.390 2008] [ 2796] WARNING: forcing --console mode on Windows
   [Wed Oct 29 14:26:44.390 2008] [ 2796] using config file 'F:/InstantRails/rails_apps/community/config/development.sphinx.conf'...
   [Wed Oct 29 14:26:44.406 2008] [ 2796] creating server socket on 127.0.0.1:3312
   [Wed Oct 29 14:26:44.406 2008] [ 2796] accepting connections

Now, you are ready to search. To test it, you can modify the index action as follows:

class PostsController < BaseController
  def index
    @posts = Post.search params[:search]
  end  
end

And your view may look like this:

<% for post in @posts %>
  <h2><%=h post.subject %></h2>
  <%=h post.body %>
<% end %>

Now, navigate to /posts?search=myword where myword is a string that can be found in the title of some of your posts. You should see that the results are filtered according to the keyword sent.

Congratulations! You are thinking sphinx on Windows.

Did you like this article? Bookmark it: