edge badge

Fixtures are a way of organizing data that you want to test against; in short, sample data.

They are stored in YAML files, one file per model, which are placed in the directory appointed by ActiveSupport::TestCase.fixture_path=(path) (this is automatically configured for Rails, so you can just put your files in <your-rails-app>/test/fixtures/). The fixture file ends with the .yml file extension (Rails example: <your-rails-app>/test/fixtures/web_sites.yml). The format of a fixture file looks like this:

rubyonrails:
  id: 1
  name: Ruby on Rails
  url: http://www.rubyonrails.org

google:
  id: 2
  name: Google
  url: http://www.google.com

This fixture file includes two fixtures. Each YAML fixture (ie. record) is given a name and is followed by an indented list of key/value pairs in the “key: value” format. Records are separated by a blank line for your viewing pleasure.

Note that fixtures are unordered. If you want ordered fixtures, use the omap YAML type. See yaml.org/type/omap.html for the specification. You will need ordered fixtures when you have foreign key constraints on keys in the same table. This is commonly needed for tree structures. Example:

--- !omap
- parent:
    id:         1
    parent_id:  NULL
    title:      Parent
- child:
    id:         2
    parent_id:  1
    title:      Child

Using Fixtures in Test Cases

Since fixtures are a testing construct, we use them in our unit and functional tests. There are two ways to use the fixtures, but first let's take a look at a sample unit test:

require 'test_helper'

class WebSiteTest < ActiveSupport::TestCase
  test "web_site_count" do
    assert_equal 2, WebSite.count
  end
end

By default, test_helper.rb will load all of your fixtures into your test database, so this test will succeed.

The testing environment will automatically load the all fixtures into the database before each test. To ensure consistent data, the environment deletes the fixtures before running the load.

In addition to being available in the database, the fixture's data may also be accessed by using a special dynamic method, which has the same name as the model, and accepts the name of the fixture to instantiate:

test "find" do
  assert_equal "Ruby on Rails", web_sites(:rubyonrails).name
end

Alternatively, you may enable auto-instantiation of the fixture data. For instance, take the following tests:

test "find_alt_method_1" do
  assert_equal "Ruby on Rails", @web_sites['rubyonrails']['name']
end

test "find_alt_method_2" do
  assert_equal "Ruby on Rails", @rubyonrails.name
end

In order to use these methods to access fixtured data within your testcases, you must specify one of the following in your ActiveSupport::TestCase-derived class:

  • to fully enable instantiated fixtures (enable alternate methods #1 and #2 above)

    self.use_instantiated_fixtures = true
  • create only the hash for the fixtures, do not 'find' each instance (enable alternate method #1 only)

    self.use_instantiated_fixtures = :no_instances

Using either of these alternate methods incurs a performance hit, as the fixtured data must be fully traversed in the database to create the fixture hash and/or instance variables. This is expensive for large sets of fixtured data.

Dynamic fixtures with ERB

Some times you don't care about the content of the fixtures as much as you care about the volume. In these cases, you can mix ERB in with your YAML fixtures to create a bunch of fixtures for load testing, like:

<% 1.upto(1000) do |i| %>
fix_<%= i %>:
  id: <%= i %>
  name: guy_<%= 1 %>
<% end %>

This will create 1000 very simple fixtures.

Using ERB, you can also inject dynamic values into your fixtures with inserts like <%= Date.today.strftime("%Y-%m-%d") %>. This is however a feature to be used with some caution. The point of fixtures are that they're stable units of predictable sample data. If you feel that you need to inject dynamic values, then perhaps you should reexamine whether your application is properly testable. Hence, dynamic values in fixtures are to be considered a code smell.

Helper methods defined in a fixture will not be available in other fixtures, to prevent against unwanted inter-test dependencies. Methods used by multiple fixtures should be defined in a module that is included in ActiveRecord::FixtureSet.context_class.

  • define a helper method in `test_helper.rb`

    class FixtureFileHelpers
      def file_sha(path)
        Digest::SHA2.hexdigest(File.read(Rails.root.join('test/fixtures', path)))
      end
    end
    ActiveRecord::FixtureSet.context_class.send :include, FixtureFileHelpers
    
  • use the helper method in a fixture

    photo:
      name: kitten.png
      sha: <%= file_sha 'files/kitten.png' %>

Transactional Fixtures

Test cases can use begin+rollback to isolate their changes to the database instead of having to delete+insert for every test case.

class FooTest < ActiveSupport::TestCase
  self.use_transactional_fixtures = true

  test "godzilla" do
    assert !Foo.all.empty?
    Foo.destroy_all
    assert Foo.all.empty?
  end

  test "godzilla aftermath" do
    assert !Foo.all.empty?
  end
end

If you preload your test database with all fixture data (probably in the rake task) and use transactional fixtures, then you may omit all fixtures declarations in your test cases since all the data's already there and every case rolls back its changes.

In order to use instantiated fixtures with preloaded data, set self.pre_loaded_fixtures to true. This will provide access to fixture data for every table that has been loaded through fixtures (depending on the value of use_instantiated_fixtures).

When not to use transactional fixtures:

  1. You're testing whether a transaction works correctly. Nested transactions don't commit until all parent transactions commit, particularly, the fixtures transaction which is begun in setup and rolled back in teardown. Thus, you won't be able to verify the results of your transaction until Active Record supports nested transactions or savepoints (in progress).

  2. Your database does not support transactions. Every Active Record database supports transactions except MySQL MyISAM. Use InnoDB, MaxDB, or NDB instead.

Advanced Fixtures

Fixtures that don't specify an ID get some extra features:

  • Stable, autogenerated IDs

  • Label references for associations (belongs_to, has_one, has_many)

  • HABTM associations as inline lists

  • Autofilled timestamp columns

  • Fixture label interpolation

  • Support for YAML defaults

Stable, Autogenerated IDs

Here, have a monkey fixture:

george:
  id: 1
  name: George the Monkey

reginald:
  id: 2
  name: Reginald the Pirate

Each of these fixtures has two unique identifiers: one for the database and one for the humans. Why don't we generate the primary key instead? Hashing each fixture's label yields a consistent ID:

george: # generated id: 503576764
  name: George the Monkey

reginald: # generated id: 324201669
  name: Reginald the Pirate

Active Record looks at the fixture's model class, discovers the correct primary key, and generates it right before inserting the fixture into the database.

The generated ID for a given label is constant, so we can discover any fixture's ID without loading anything, as long as we know the label.

Label references for associations (belongs_to, has_one, has_many)

Specifying foreign keys in fixtures can be very fragile, not to mention difficult to read. Since Active Record can figure out the ID of any fixture from its label, you can specify FK's by label instead of ID.

belongs_to

Let's break out some more monkeys and pirates.

### in pirates.yml

reginald:
  id: 1
  name: Reginald the Pirate
  monkey_id: 1

### in monkeys.yml

george:
  id: 1
  name: George the Monkey
  pirate_id: 1

Add a few more monkeys and pirates and break this into multiple files, and it gets pretty hard to keep track of what's going on. Let's use labels instead of IDs:

### in pirates.yml

reginald:
  name: Reginald the Pirate
  monkey: george

### in monkeys.yml

george:
  name: George the Monkey
  pirate: reginald

Pow! All is made clear. Active Record reflects on the fixture's model class, finds all the belongs_to associations, and allows you to specify a target label for the association (monkey: george) rather than a target id for the FK (monkey_id: 1).

Polymorphic belongs_to

Supporting polymorphic relationships is a little bit more complicated, since Active Record needs to know what type your association is pointing at. Something like this should look familiar:

### in fruit.rb

belongs_to :eater, polymorphic: true

### in fruits.yml

apple:
  id: 1
  name: apple
  eater_id: 1
  eater_type: Monkey

Can we do better? You bet!

apple:
  eater: george (Monkey)

Just provide the polymorphic target type and Active Record will take care of the rest.

has_and_belongs_to_many

Time to give our monkey some fruit.

### in monkeys.yml

george:
  id: 1
  name: George the Monkey

### in fruits.yml

apple:
  id: 1
  name: apple

orange:
  id: 2
  name: orange

grape:
  id: 3
  name: grape

### in fruits_monkeys.yml

apple_george:
  fruit_id: 1
  monkey_id: 1

orange_george:
  fruit_id: 2
  monkey_id: 1

grape_george:
  fruit_id: 3
  monkey_id: 1

Let's make the HABTM fixture go away.

### in monkeys.yml

george:
  id: 1
  name: George the Monkey
  fruits: apple, orange, grape

### in fruits.yml

apple:
  name: apple

orange:
  name: orange

grape:
  name: grape

Zap! No more fruits_monkeys.yml file. We've specified the list of fruits on George's fixture, but we could've just as easily specified a list of monkeys on each fruit. As with belongs_to, Active Record reflects on the fixture's model class and discovers the has_and_belongs_to_many associations.

Autofilled Timestamp Columns

If your table/model specifies any of Active Record's standard timestamp columns (created_at, created_on, updated_at, updated_on), they will automatically be set to Time.now.

If you've set specific values, they'll be left alone.

Fixture label interpolation

The label of the current fixture is always available as a column value:

geeksomnia:
  name: Geeksomnia's Account
  subdomain: $LABEL
  email: $LABEL@email.com

Also, sometimes (like when porting older join table fixtures) you'll need to be able to get a hold of the identifier for a given label. ERB to the rescue:

george_reginald:
  monkey_id: <%= ActiveRecord::FixtureSet.identify(:reginald) %>
  pirate_id: <%= ActiveRecord::FixtureSet.identify(:george) %>

Support for YAML defaults

You probably already know how to use YAML to set and reuse defaults in your database.yml file. You can use the same technique in your fixtures:

DEFAULTS: &DEFAULTS
  created_on: <%= 3.weeks.ago.to_s(:db) %>

first:
  name: Smurf
  <<: *DEFAULTS

second:
  name: Fraggle
  <<: *DEFAULTS

Any fixture labeled “DEFAULTS” is safely ignored.

Namespace
Methods
#
C
E
F
I
N
R
S
T
Constants
MAX_ID = 2 ** 30 - 1
 
Attributes
[R] config
[R] fixtures
[R] model_class
[R] name
[R] table_name
Class Public methods
cache_fixtures(connection, fixtures_map)
# File activerecord/lib/active_record/fixtures.rb, line 433
def self.cache_fixtures(connection, fixtures_map)
  cache_for_connection(connection).update(fixtures_map)
end
cache_for_connection(connection)
# File activerecord/lib/active_record/fixtures.rb, line 417
def self.cache_for_connection(connection)
  @@all_cached_fixtures[connection]
end
cached_fixtures(connection, keys_to_fetch = nil)
# File activerecord/lib/active_record/fixtures.rb, line 425
def self.cached_fixtures(connection, keys_to_fetch = nil)
  if keys_to_fetch
    cache_for_connection(connection).values_at(*keys_to_fetch)
  else
    cache_for_connection(connection).values
  end
end
context_class()

Superclass for the evaluation contexts used by ERB fixtures.

# File activerecord/lib/active_record/fixtures.rb, line 564
def self.context_class
  @context_class ||= Class.new
end
create_fixtures(fixtures_directory, fixture_set_names, class_names = {}, config = ActiveRecord::Base)
# File activerecord/lib/active_record/fixtures.rb, line 496
def self.create_fixtures(fixtures_directory, fixture_set_names, class_names = {}, config = ActiveRecord::Base)
  fixture_set_names = Array(fixture_set_names).map(&:to_s)
  class_names = ClassCache.new class_names, config

  # FIXME: Apparently JK uses this.
  connection = block_given? ? yield : ActiveRecord::Base.connection

  files_to_read = fixture_set_names.reject { |fs_name|
    fixture_is_cached?(connection, fs_name)
  }

  unless files_to_read.empty?
    connection.disable_referential_integrity do
      fixtures_map = {}

      fixture_sets = files_to_read.map do |fs_name|
        klass = class_names[fs_name]
        conn = klass ? klass.connection : connection
        fixtures_map[fs_name] = new( # ActiveRecord::FixtureSet.new
          conn,
          fs_name,
          klass,
          ::File.join(fixtures_directory, fs_name))
      end

      all_loaded_fixtures.update(fixtures_map)

      connection.transaction(:requires_new => true) do
        fixture_sets.each do |fs|
          conn = fs.model_class.respond_to?(:connection) ? fs.model_class.connection : connection
          table_rows = fs.table_rows

          table_rows.keys.each do |table|
            conn.delete "DELETE FROM #{conn.quote_table_name(table)}", 'Fixture Delete'
          end

          table_rows.each do |fixture_set_name, rows|
            rows.each do |row|
              conn.insert_fixture(row, fixture_set_name)
            end
          end
        end

        # Cap primary key sequences to max(pk).
        if connection.respond_to?(:reset_pk_sequence!)
          fixture_sets.each do |fs|
            connection.reset_pk_sequence!(fs.table_name)
          end
        end
      end

      cache_fixtures(connection, fixtures_map)
    end
  end
  cached_fixtures(connection, fixture_set_names)
end
fixture_is_cached?(connection, table_name)
# File activerecord/lib/active_record/fixtures.rb, line 421
def self.fixture_is_cached?(connection, table_name)
  cache_for_connection(connection)[table_name]
end
identify(label, column_type = :integer)

Returns a consistent, platform-independent identifier for label. Integer identifiers are values less than 2^30. UUIDs are RFC 4122 version 5 SHA-1 hashes.

# File activerecord/lib/active_record/fixtures.rb, line 555
def self.identify(label, column_type = :integer)
  if column_type == :uuid
    SecureRandom.uuid_v5(SecureRandom::UUID_OID_NAMESPACE, label.to_s)
  else
    Zlib.crc32(label.to_s) % MAX_ID
  end
end
instantiate_all_loaded_fixtures(object, load_instances = true)
# File activerecord/lib/active_record/fixtures.rb, line 449
def self.instantiate_all_loaded_fixtures(object, load_instances = true)
  all_loaded_fixtures.each_value do |fixture_set|
    instantiate_fixtures(object, fixture_set, load_instances)
  end
end
instantiate_fixtures(object, fixture_set, load_instances = true)
# File activerecord/lib/active_record/fixtures.rb, line 437
def self.instantiate_fixtures(object, fixture_set, load_instances = true)
  if load_instances
    fixture_set.each do |fixture_name, fixture|
      begin
        object.instance_variable_set "@#{fixture_name}", fixture.find
      rescue FixtureClassNotFound
        nil
      end
    end
  end
end
new(connection, name, class_name, path, config = ActiveRecord::Base)
# File activerecord/lib/active_record/fixtures.rb, line 570
def initialize(connection, name, class_name, path, config = ActiveRecord::Base)
  @name     = name
  @path     = path
  @config   = config
  @model_class = nil

  if class_name.is_a?(String)
    ActiveSupport::Deprecation.warn("The ability to pass in strings as a class name to `FixtureSet.new` will be removed in Rails 4.2. Use the class itself instead.")
  end

  if class_name.is_a?(Class) # TODO: Should be an AR::Base type class, or any?
    @model_class = class_name
  else
    @model_class = class_name.safe_constantize if class_name
  end

  @connection  = connection

  @table_name = ( model_class.respond_to?(:table_name) ?
                  model_class.table_name :
                  self.class.default_fixture_table_name(name, config) )

  @fixtures = read_fixture_files path, @model_class
end
reset_cache()
# File activerecord/lib/active_record/fixtures.rb, line 413
def self.reset_cache
  @@all_cached_fixtures.clear
end
Instance Public methods
[](x)
# File activerecord/lib/active_record/fixtures.rb, line 595
def [](x)
  fixtures[x]
end
[]=(k,v)
# File activerecord/lib/active_record/fixtures.rb, line 599
def []=(k,v)
  fixtures[k] = v
end
each(&block)
# File activerecord/lib/active_record/fixtures.rb, line 603
def each(&block)
  fixtures.each(&block)
end
size()
# File activerecord/lib/active_record/fixtures.rb, line 607
def size
  fixtures.size
end
table_rows()

Returns a hash of rows to be inserted. The key is the table, the value is a list of rows to insert to that table.

# File activerecord/lib/active_record/fixtures.rb, line 613
def table_rows
  now = config.default_timezone == :utc ? Time.now.utc : Time.now
  now = now.to_s(:db)

  # allow a standard key to be used for doing defaults in YAML
  fixtures.delete('DEFAULTS')

  # track any join tables we need to insert later
  rows = Hash.new { |h,table| h[table] = [] }

  rows[table_name] = fixtures.map do |label, fixture|
    row = fixture.to_hash

    if model_class
      # fill in timestamp columns if they aren't specified and the model is set to record_timestamps
      if model_class.record_timestamps
        timestamp_column_names.each do |c_name|
          row[c_name] = now unless row.key?(c_name)
        end
      end

      # interpolate the fixture label
      row.each do |key, value|
        row[key] = value.gsub("$LABEL", label) if value.is_a?(String)
      end

      # generate a primary key if necessary
      if has_primary_key_column? && !row.include?(primary_key_name)
        row[primary_key_name] = ActiveRecord::FixtureSet.identify(label, primary_key_type)
      end

      # If STI is used, find the correct subclass for association reflection
      reflection_class =
        if row.include?(inheritance_column_name)
          row[inheritance_column_name].constantize rescue model_class
        else
          model_class
        end

      reflection_class.reflect_on_all_associations.each do |association|
        case association.macro
        when :belongs_to
          # Do not replace association name with association foreign key if they are named the same
          fk_name = (association.options[:foreign_key] || "#{association.name}_id").to_s

          if association.name.to_s != fk_name && value = row.delete(association.name.to_s)
            if association.options[:polymorphic] && value.sub!(/\s*\(([^\)]*)\)\s*$/, "")
              # support polymorphic belongs_to as "label (Type)"
              row[association.foreign_type] = $1
            end

            fk_type = association.active_record.columns_hash[association.foreign_key].type
            row[fk_name] = ActiveRecord::FixtureSet.identify(value, fk_type)
          end
        when :has_many
          if association.options[:through]
            add_join_records(rows, row, HasManyThroughProxy.new(association))
          end
        end
      end
    end

    row
  end
  rows
end