-
Star
(220)
You must be signed in to star a gist -
Fork
(36)
You must be signed in to fork a gist
-
-
Save coreyhaines/2068977 to your computer and use it in GitHub Desktop.
--colour | |
-I app |
require 'active_record' | |
connection_info = YAML.load_file("config/database.yml")["test"] | |
ActiveRecord::Base.establish_connection(connection_info) | |
RSpec.configure do |config| | |
config.around do |example| | |
ActiveRecord::Base.transaction do | |
example.run | |
raise ActiveRecord::Rollback | |
end | |
end | |
end |
class Coderetreat < ActiveRecord::Base | |
def self.running_today | |
where(scheduled_on: Date.today) | |
end | |
end |
require 'active_record_spec_helper' | |
require 'models/coderetreat' | |
describe Coderetreat do | |
describe ".running_today" do | |
it "returns a coderetreat scheduled for today" do | |
coderetreat = Coderetreat.create! city: "Chicago", scheduled_on: Date.today | |
Coderetreat.running_today.all.should =~ [coderetreat] | |
end | |
it "does not return a coderetreat not scheduled for today" do | |
coderetreat = Coderetreat.create! city: "Chicago", scheduled_on: Date.today.advance(:days => -1) | |
Coderetreat.running_today.should be_empty | |
end | |
end | |
end |
#Some databases get upset if you try to start a new transaction while a transaction is already in play, so running the whole spec suite chokes when rspec is trying to start a transaction. You need to update your spec_helper to rely on active_record_spec_helper to do this for you. | |
#Replace this line | |
config.use_transactional_fixtures = true | |
#With this | |
require 'active_record_spec_helper' | |
#TADA |
Thanks. It is a very handy addition.
Thanks! I wasn't sure if you were doing something to mock AR, or just isolating AR stuff to one spec helper.
No. I don't agree with mocking AR. This is really about a step forward in making progress towards a realistic approach to keeping things fast.
I have started wrapping my active_record_spec_helper in an "unless defined?(Rails)" block so that I can run the model specs alone or as part of the entire suite.
@rubysolo I've not found the need to do that in order to have them run in both situations. What error are you seeing that causes you to need this?
Interesting. If I run "rake", I get the following error:
activerecord-3.2.2/lib/active_record/connection_adapters/postgresql_adapter.rb:645:in `async_exec': PG::Error: connection is closed: SELECT tablename (ActiveRecord::StatementInvalid)
FROM pg_tables
WHERE schemaname = ANY (current_schemas(false))
which traces back to:
DatabaseCleaner.clean_with(:truncation)
in the active_record_spec_helper.
Does your spec_helper set the config.use_transactional_fixtures? Take that out, replace by requiring this spec helper in your regular spec helper
@coreyhaines Indeed it did, and this fixed the problem. :) Thanks!
@rubysolo Thanks so much for bringing this up. Glad it worked. I forgot to put that in here, so now we have it documented in the comments. :) I'll try to get it into the gist, itself, as well.
@coreyhaines is there a need for the double requires of "database_cleaner"? Isn't it unnecessary in the spec_helper?
@rmcstil A good point.
@coreyhaines Thanks for posting this! Although I'm starting to get spoiled by the near instantaneous feedback and am considering dropping model specs all together. Right now I've got it costing me about 6 seconds to load active record. Is that what you do with mercuryapp?
@rmcastil Yeah, that's the eventual goal. I tend to only want to load active record when I'm dealing with scopes. Those make sense to have the database loaded up.
This technique is very useful to shrink the feedback loop, so you can start effectively making changes to your design.
@coreyhaines Instead of DatabaseCleaner you could do something like this: ActiveRecord::Base.subclasses.each(&:delete_all) DatabaseCleaner is such a beast to load.
Thanks @revans. I'll try that.
Very glad that you gist'd this. I had similar thoughts as @steveklabnik. I appreciate the insight!
Not using database_cleaner
and using the below code cut off 0.2-0.3 sec.
config.around do |example|
ActiveRecord::Base.transaction do
example.run
raise ActiveRecord::Rollback
end
end
@brandonhilkert Thanks. I'm going to try that.
What happens with a class with a lot of relations? We may end up manually requiring a lot of things, won't we? Or is there a catch that I'm not seeing?
@brandonhilkert that's cool! But if you use capybara-webkit
for example, this doesn't work :(
@brandonhilkert I mean, unless you workaround https://gist.github.com/4469873
Two things I've noticed:
- I had to require the model like this
require_relative '../../app/models/coderetreat
- When loading the full stack I had an
ActiveModel::MassAssignmentSecurity::Error
which passed with this approach
Not sure if I'm doing anything wrong. The idea is awesome though.
@zamith, I had the same problem. Were you sure to add the -I app
line to your .rspec
file? Then you require 'models/coderetreat'
. That fixed it for me.
If you are using Rails 4 with SQLite and database_cleaner 1.1.1 you get this error
undefined local variable or method `postgresql_version'
to solve the problem downgrade to database_cleaner 1.0.1
Thank you @coreyhaines for sharing this.
I had to add require "yaml"
to the top of active_record_spec_helper
to get this to work. RSpec 3.1.7, Rails 4.1.8.
I love this idea, and I really want to implement it more in my test suites. But I'm having the problem that my ActiveRecord models just have too damn many dependencies, and I have to add so many "requires" at the top of my spec file that I might as well just require spec_helper
.
Of course, the obvious solution here is to reduce the number of dependencies. And in some cases this is pretty easy, but it's often not, because a lot of the dependencies are due to my associations.
E.g. say I have an model User
that looks like this:
class Comment < ActiveRecord::Base
belongs_to :design
end
If I set up my spec like this:
require 'active_record_spec_helper'
require "comment"
describe Comment do
... whatever
end
... then it will fail, because when it tries to load "comment.rb", it hits the line belongs_to :design
, and starts looking for a Design
class - which hasn't been required, so the whole thing crashes with uninitialized constant Comment::Design
.
So I add require 'design'
to the top of my spec - but within my Design
class I have the line belongs_to :user
, which crashes with uninitialized constant Design::User
... so I add require 'user'
to the top of my spec, which hits the line has_many :posts
and raises uninitialized constant User::Post
... and so on until I've required almost my entire app/models
directory.
Is there something conceptual I'm missing? How do I get around this? Basic Rails associations don't seem like something I should be refactoring out into a separate class or module, and stubbing out the associations feels a bit icky too (plus I'm not entirely sure how to do it anyway.)
Anyone have any pointers?
Sorry, me again. StackOverflow bounty is available to anyone who can help me out with the above question:
Did anyone came up with a solution to the question of georgemillo? I also struggle with the idea of loading all dependent models, which is undesireable to me.
Very nice.