Database Cleaner is a set of strategies for cleaning your database in Ruby.
The original use case was to ensure a clean state during tests. Each strategy is a small amount of code but is code that is usually needed in any ruby app that is testing with a database.
ActiveRecord, DataMapper, Sequel, MongoMapper, Mongoid, CouchPotato, Ohm and Redis are supported.
Here is an overview of the strategies supported for each library:
ORM | Truncation | Transaction | Deletion |
---|---|---|---|
ActiveRecord | Yes | Yes | Yes |
DataMapper | Yes | Yes | No |
CouchPotato | Yes | No | No |
MongoMapper | Yes | No | No |
Mongoid | Yes | No | No |
Sequel | Yes | Yes | No |
Redis | Yes | No | No |
Ohm | Yes | No | No |
Neo4j | Yes | Yes* | Yes* |
- Truncation and Deletion strategies for Neo4j will just delete all nodes and relationships from the database.
Driver | Truncation | Transaction | Deletion |
---|---|---|---|
Mongo | Yes | No | No |
Moped | Yes | No | No |
(Default strategy for each library is denoted in bold)
Database Cleaner also includes a null
strategy (that does no cleaning at all) which can be used with any ORM library.
You can also explicitly use it by setting your strategy to nil
.
For support or to discuss development please use the Google Group.
For the SQL libraries the fastest option will be to use :transaction
as transactions are simply rolled back. If you can use this strategy you should. However, if you wind up needing to use multiple database connections in your tests (i.e. your tests run in a different process than your application) then using this strategy becomes a bit more difficult. You can get around the problem a number of ways.
One common approach is to force all processes to use the same database connection (common ActiveRecord hack) however this approach has been reported to result in non-deterministic failures.
Another approach is to have the transactions rolled back in the application's process and relax the isolation level of the database (so the tests can read the uncommitted transactions).
An easier, but slower, solution is to use the :truncation
or :deletion
strategy.
So what is fastest out of :deletion
and :truncation
? Well, it depends on your table structure and what percentage of tables you populate in an average test. The reasoning is out of the scope of this README but here is a good SO answer on this topic for Postgres.
Some people report much faster speeds with :deletion
while others say :truncation
is faster for them. The best approach therefore is it try all options on your test suite and see what is faster.
If you are using ActiveRecord then take a look at the additional options available for :truncation
.
Because database_cleaner supports multiple ORMs, it doesn't make sense to include all the dependencies for each one in the gemspec. However, the DataMapper adapter does depend on dm-transactions. Therefore, if you use DataMapper, you must include dm-transactions in your Gemfile/bundle/gemset manually.
require 'database_cleaner'
DatabaseCleaner.strategy = :truncation
# then, whenever you need to clean the DB
DatabaseCleaner.clean
With the :truncation
strategy you can also pass in options, for example:
DatabaseCleaner.strategy = :truncation, {:only => %w[widgets dogs some_other_table]}
DatabaseCleaner.strategy = :truncation, {:except => %w[widgets]}
With Ohm and Redis, :only
and :except
take a list of strings to be
passed to keys
).
(I should point out the truncation strategy will never truncate your schema_migrations table.)
Some strategies need to be started before tests are run (for example the :transaction
strategy needs to know to open up a transaction). This can be accomplished by calling DatabaseCleaner.start
at the beginning of the run, or by running the tests inside a block to Database.cleaning
. So you would have:
require 'database_cleaner'
DatabaseCleaner.strategy = :transaction
DatabaseCleaner.start # usually this is called in setup of a test
dirty_the_db
DatabaseCleaner.clean # cleanup of the test
# OR
DatabaseCleaner.cleaning do
dirty_the_db
end
At times you may want to do a single clean with one strategy.
For example, you may want to start the process by truncating all the tables, but then use the faster transaction strategy the remaining time. To accomplish this you can say:
require 'database_cleaner'
DatabaseCleaner.clean_with :truncation
DatabaseCleaner.strategy = :transaction
# then make the DatabaseCleaner.start and DatabaseCleaner.clean calls appropriately
The following options are available for ActiveRecord's :truncation
strategy only for MySQL and Postgres.
:pre_count
- When set totrue
this will check each table for existing rows before truncating it. This can speed up test suites when many of the tables to be truncated are never populated. Defaults to:false
. (Also, see the section on What strategy is fastest?):reset_ids
- This only matters when:pre_count
is used, and it will make sure that a tables auto-incrementing id is reset even if there are no rows in the table (e.g. records were created in the test but also removed before DatabaseCleaner gets to it). Defaults totrue
.
The following option is available for ActiveRecord's :truncation
and :deletion
strategy for any DB.
:cache_tables
- When set totrue
the list of tables to truncate or delete from will only be read from the DB once, otherwise it will be read before each cleanup run. Set this tofalse
if (1) you create and drop tables in your tests, or (2) you change Postgres schemas (ActiveRecord::Base.connection.schema_search_path
) in your tests (for example, in a multitenancy setup with each tenant in a different Postgres schema). Defaults totrue
.
RSpec.configure do |config|
config.before(:suite) do
DatabaseCleaner.strategy = :transaction
DatabaseCleaner.clean_with(:truncation)
end
config.around(:each) do |example|
DatabaseCleaner.cleaning do
example.run
end
end
end
You'll typically discover a feature spec is incorrectly using transaction instead of truncation strategy when the data created in the spec is not visible in the app-under-test.
A frequently occurring example of this is when, after creating a user in a spec, the spec mysteriously fails to login with the user. This happens because the user is created inside of an uncommitted transaction on one database connection, while the login attempt is made using a separate database connection. This separate database connection cannot access the uncommitted user data created over the first database connection due to transaction isolation.
For feature specs using a Capybara driver for an external
JavaScript-capable browser (in practice this is all drivers except
:rack_test
), the Rack app under test and the specs do not share a
database connection.
When a spec and app-under-test do not share a database connection, you'll likely need to use the truncation strategy instead of the transaction strategy.
See the suggested config below to temporarily enable truncation strategy for affected feature specs only. This config continues to use transaction strategy for all other specs.
RSpec.configure do |config|
config.use_transactional_fixtures = false
config.before(:suite) do
if config.use_transactional_fixtures?
raise(<<-MSG)
Delete line `config.use_transactional_fixtures = true` from rails_helper.rb
(or set it to false) to prevent uncommitted transactions being used in
JavaScript-dependent specs.
During testing, the app-under-test that the browser driver connects to
uses a different database connection to the database connection used by
the spec. The app's database connection would not be able to access
uncommitted transaction data setup over the spec's database connection.
MSG
end
DatabaseCleaner.clean_with(:truncation)
end
config.before(:each) do
DatabaseCleaner.strategy = :transaction
end
config.before(:each, type: :feature) do
# :rack_test driver's Rack app under test shares database connection
# with the specs, so continue to use transaction strategy for speed.
driver_shares_db_connection_with_specs = Capybara.current_driver == :rack_test
if !driver_shares_db_connection_with_specs
# Driver is probably for an external browser with an app
# under test that does *not* share a database connection with the
# specs, so use truncation strategy.
DatabaseCleaner.strategy = :truncation
end
end
config.before(:each) do
DatabaseCleaner.start
end
config.after(:each) do
DatabaseCleaner.clean
end
end
DatabaseCleaner.strategy = :transaction
class Minitest::Spec
before :each do
DatabaseCleaner.start
end
after :each do
DatabaseCleaner.clean
end
end
# with the minitest-around gem, this may be used instead:
class Minitest::Spec
around do |tests|
DatabaseCleaner.cleaning(&tests)
end
end
If you're using Cucumber with Rails, just use the generator that ships with cucumber-rails, and that will create all the code you need to integrate DatabaseCleaner into your Rails project.
Otherwise, to add DatabaseCleaner to your project by hand, create a file features/support/database_cleaner.rb
that looks like this:
begin
require 'database_cleaner'
require 'database_cleaner/cucumber'
DatabaseCleaner.strategy = :truncation
rescue NameError
raise "You need to add database_cleaner to your Gemfile (in the :test group) if you wish to use it."
end
Around do |scenario, block|
DatabaseCleaner.cleaning(&block)
end
This should cover the basics of tear down between scenarios and keeping your database clean.
For more examples see the section "Why?".
Sometimes you need to use multiple ORMs in your application.
You can use DatabaseCleaner to clean multiple ORMs, and multiple connections for those ORMs.
#How to specify particular orms
DatabaseCleaner[:active_record].strategy = :transaction
DatabaseCleaner[:mongo_mapper].strategy = :truncation
#How to specify particular connections
DatabaseCleaner[:active_record, { :connection => :two }]
# You may also pass in the model directly:
DatabaseCleaner[:active_record, { :model => ModelWithDifferentConnection }]
Usage beyond that remains the same with DatabaseCleaner.start
calling any setup on the different configured connections, and DatabaseCleaner.clean
executing afterwards.
ORM | How to access | Notes |
---|---|---|
Active Record | DatabaseCleaner[:active_record] |
Connection specified as :symbol keys, loaded from config/database.yml . You may also pass in the ActiveRecord model under the :model key. |
Data Mapper | DatabaseCleaner[:data_mapper] |
Connection specified as :symbol keys, loaded via Datamapper repositories |
Mongo Mapper | DatabaseCleaner[:mongo_mapper] |
Multiple connections not yet supported |
Mongoid | DatabaseCleaner[:mongoid] |
Multiple databases supported for Mongoid 3. Specify DatabaseCleaner[:mongoid, {:connection => :db_name}] |
Moped | DatabaseCleaner[:moped] |
It is necessary to configure database name with DatabaseCleaner[:moped].db = db_name otherwise name `default` will be used. |
Couch Potato | DatabaseCleaner[:couch_potato] |
Multiple connections not yet supported |
Sequel | DatabaseCleaner[:sequel] |
Multiple databases supported; specify DatabaseCleaner[:sequel, {:connection => Sequel.connect(uri)}] |
Redis | DatabaseCleaner[:redis] |
Connection specified as Redis URI |
Ohm | DatabaseCleaner[:ohm] |
Connection specified as Redis URI |
Neo4j | DatabaseCleaner[:neo4j] |
Database type and path(URI) DatabaseCleaner[:neo4j, connection: {type: :server_db, path: 'http://localhost:7475'}]. |
One of my motivations for writing this library was to have an easy way to turn on what Rails calls "transactional_fixtures" in my non-rails ActiveRecord projects.
After copying and pasting code to do this several times I decided to package it up as a gem and save everyone a bit of time.
DatabaseCleaner has an autodetect mechanism where if you do not explicitly define your ORM it will use the first ORM it can detect that is loaded.
Since ActiveRecord is the most common ORM used that is the first one checked for.
Sometimes other libraries (e.g. ActiveAdmin) will load other ORMs (e.g. ActiveRecord) even though you are using a different ORM. This will result in DatabaseCleaner trying to use the wrong ORM (e.g. ActiveRecord) unless you explicitly define your ORM like so:
# How to setup your ORM explicitly
DatabaseCleaner[:mongoid].strategy = :truncation
If you are using Postgres and have foreign key constraints, the truncation strategy will cause a lot of extra noise to appear on STDERR (in the form of "NOTICE truncate cascades" messages).
To silence these warnings set the following log level in your postgresql.conf
file:
client_min_messages = warning
For ActiveRecord, you add the following parameter in your database.yml file:
test: adapter: postgresql # ... min_messages: WARNING
Due to an inconsistency in JRuby's implementation of Fibers, Sequel gives a different connection to DatabaseCleaner.start
than is used for tests run between .start
and .clean
. This can be worked around by running your tests in a block like DatabaseCleaner.cleaning { run_my_tests }
instead, which does not use Fibers.
When you are using neo4j gem it creates schema and reads indexes upon loading models. These operations can't be done during a transaction. You have to preload your models before DatabaseCleaner starts a transaction.
Add to your rails_helper or spec_helper after requiring database_cleaner:
require 'database_cleaner'
Dir["#{Rails.root}/app/models/**/*.rb"].each do |model|
load model
end
In rare cases DatabaseCleaner will encounter errors that it will log. By default it uses STDOUT set to the ERROR level but you can configure this to use whatever Logger you desire.
Here's an example of using the Rails.logger
in env.rb
:
DatabaseCleaner.logger = Rails.logger
Copyright (c) 2014 Ben Mabey. See LICENSE for details.