Testing with Thinking Sphinx

Before you get caught up in the specifics of testing Thinking Sphinx using certain tools, it’s worth noting that no matter what the approach, you’ll need to turn off transactional fixtures and index your data after creating the appropriate records – otherwise you won’t get any search results.

Also: make sure you have your test environment using a different port number in config/sphinx.yml (which you may need to create if you haven’t already). If this isn’t done, then you won’t be able to run Sphinx in your development environment and your tests at the same time (as they’ll both want to use the same port for Sphinx).

  port: 9313

Unit Tests and Specs

It’s recommended you stub out any search calls, as Thinking Sphinx should ideally only be used in integration testing (whether that be via Cucumber or other methods).


As of version 1.3.2, Thinking Sphinx has a helper object to make combining Thinking Sphinx and Cucumber quite easy. You’ll need to add the following two lines to your features/support/env.rb file:

require 'cucumber/thinking_sphinx/external_world'

Don’t forget, you also need to turn transactional fixtures off. This can be done on a global level in your features/support/env.rb file:

Cucumber::Rails::World.use_transactional_fixtures = false

Or, you can tag either an entire feature or single scenarios with the @no-txn tag:

Feature: Searching for articles

The reason for this is that while ActiveRecord can run all its operations within a single transaction, Sphinx doesn’t have access to that, and so indexing will not include your transaction’s changes.

The added complication to this is that you’ll probably want to clear all the data from your database between scenarios. This can be done within the Before block, in one of your steps files (see below). Another option is Ben Mabey’s Database Cleaner library – and make sure you use the truncation strategy.

Before do
  # Add your own models here instead.
  [Article, User].each do |model|

Once this is all set up, then Sphinx will automatically index and start the daemon when you run your features – but only once at the very beginning, not for every scenario (as that could be quite slow).

To re-index during specific scenarios, I recommend adding steps something like the following (to be called after preparing your model data, but before Webrat browses the application):

Given 'the Sphinx indexes are updated' do
  # Update all indexes
  sleep(0.25) # Wait for Sphinx to catch up

Given 'the Sphinx indexes for articles are updated' do
  # Update specific indexes
  ThinkingSphinx::Test.index 'article_core', 'article_delta'
  sleep(0.25) # Wait for Sphinx to catch up

Delta indexes (if you’re using the default approach) will automatically update just like they do in a normal application environment.

Any suggestions to improve this workflow are very much welcome.

Rails Functional and Integration Tests

In much the same way, Thinking Sphinx can be used in traditional functional and integration tests. You’ll want to add the following lines to your test_helper.rb file:

require 'thinking_sphinx/test'

You can turn off transactional features on a per-test basis within the test class definition:

class SearchControllerTest
  self.use_transactional_fixtures = false

  # ...

To actually have Sphinx running, you have a few options…

If you want it running constantly for all of your tests, you can call start_with_autostop in your test_helper.rb file:


However, you probably don’t want Sphinx running for your unit tests, and so it’s recommended you just start and stop Sphinx as required. ThinkingSphinx::Test has methods named start and stop for that very purpose:

test "Searching for Articles" do

  get :index
  assert [@article], assigns[:articles]


You can also wrap the code that needs Sphinx in a block called by ThinkingSphinx::Test.run, which will start up and stop Sphinx either side of the block:

test "Searching for Articles" do
  ThinkingSphinx::Test.run do
    get :index
    assert [@article], assigns[:articles]

If you need to manually process indexes, just use the index method, which defaults to all indexes unless you pass in specific names.

ThinkingSphinx::Test.index # all indexes
ThinkingSphinx::Test.index 'article_core', 'article_delta'