Define a pact between service consumers and providers, enabling "consumer driven contract" testing.
Pact provides an RSpec DSL for service consumers to define the requests they will make to a service provider and the responses they expect back. These expectations are used in the consumers specs to provide a mock service provider. The interactions are recorded, and played back in the service provider specs to ensure the service provider actually does provide the response the consumer expects.
This allows testing of both sides of an integration point using fast unit tests.
This gem is inspired by the concept of "Consumer driven contracts". See http://martinfowler.com/articles/consumerDrivenContracts.html for more information.
Travis CI Status:
- A service is mocked using an actual process running on a specified port, so javascript clients can be tested as easily as backend clients.
- "Provider states" (similar to fixtures) allow the same request to be made with a different expected response.
- Consumers specify only the fields they are interested in, allowing a provider to return more fields without breaking the pact. This allows a provider to have a different pact with a different consumer, and know which fields each cares about in a given response.
- Expected interactions are verified to have actually occurred in the consumer specs.
- The mocked responses are verified to be valid by replaying the interactions against the provider codebase.
- Rake verification tasks allow a pacts at one or more URIs to be checked against a given service provider codebase.
- Different versions of a consumer/provider pair can be easily tested against each other, allowing confidence when deploying new versions of each (see the pact_broker and pact_broker-client gems).
Put it in your Gemfile. You know how.
Imagine a model class that looks something like this. The attributes for a SomethingModel live on a remote server, and will need to be retrieved by an HTTP call.
class SomethingModel
attr_reader :name
def intialize name
@name = name
end
def == other
other.is_a?(Something) && other.name == name
end
endImagine a service provider client class that looks something like this.
class MyServiceProviderClient
include HTTParty
base_uri 'http://my-service'
def get_something
# Yet to be implemented because we're doing Test First Development...
end
endThe following code will create a mock service on localhost:1234 which will respond to your application's queries over HTTP as if it were the real "My Service Provider" app. It also creats a mock service provider object which you will use to set up your expectations. The method name to access the mock service provider will be what ever name you give as the service argument - in this case "my_service_provider"
# In /spec/service_providers/pact_helper.rb
require 'pact/consumer/rspec'
Pact.service_consumer "My Service Consumer" do
has_pact_with "My Service Provider" do
mock_service :my_service_provider do
port 1234
end
end
end# In /spec/service_providers/my_service_provider_client_spec.rb
# Use the :pact => true describe metadata to include all the pact generation functionality in your spec.
describe MyServiceProviderClient, :pact => true do
before do
# Configure your client to point to the stub service on localhost using the port you have specified
MyServiceProviderClient.base_uri 'localhost:1234'
end
describe "get_something" do
before do
my_service_provider.
.given("something exists")
.upon_receiving("a request for something")
.with( method: :get, path: '/something' )
.will_respond_with(
status: 200,
headers: { 'Content-Type' => 'application/json' },
body: {name: 'A small something'}
)
end
it "returns a Something" do
expect(MyServiceProviderClient.get_something).to eq(SomethingModel.new('A small something'))
end
end
endRunning the consumer spec will generate a pact file in the configured pact dir (spec/pacts by default). Logs will be output to the configured log dir that can be useful when diagnosing problems.
Of course, the above specs will fail because the client method is not implemented, so next, implement your client methods.
class MyServiceProviderClient
include HTTParty
base_uri 'http://my-service'
def get_something
name = JSON.parse(self.class.get("/something").body)['name']
SomethingModel.new(name)
end
endGreen! You now have a pact file that can be used to verify your expectations in the provider project. Now, rinse and repeat for ALL the likely status codes that may be returned (recommend 400, 404, 500 and 401/403 if there is authorisation.)
Create your API class using the framework of your choice (e.g. Sinatra, Grape) - leave the methods unimplemented, we're doing Test First Develoment, remember?
Create a pact_helper.rb in your service provider project. The file must be called pact_helper.rb, however there is some flexibility in where it can be stored. The recommended place is specs/service_consumers/pact_helper.rb.
require 'pact/provider/rspec'
# If you wish to use the same spec_helper file as your unit tests, require it here.
# Otherwise, you can set up a separate RSpec configuration in this file just for pact:verify.
require './spec_helper'
Pact.service_provider "My Service Provider" do
# If you have a config.ru file, the app will be loaded from it automatically and you can skip the app config.
# Otherwise, set your rack app here as you would when using Rack::Test::Methods
app { MyApp.new }
honours_pact_with 'My Service Consumer' do
# This example points to a local file, however, on a real project with a continuous
# integration box, you would publish your pacts as artifacts,
# and point the pact_uri to the pact published by the last successful build.
pact_uri '../path-to-your-consumer-project/specs/pacts/my_consumer-my_provider.json'
end
endRequire "pact/tasks" in your Rakefile. If the pact gem is in the test/development section of your Gemfile, you may want to put an env check around this so it doesn't load the pact tasks in prod.
# In Rakefile
require 'pact/tasks'$ rake pact:verify
Congratulations! You now have a failing spec to develop against.
At this stage, you'll probably want to be able to run your specs one at a time while you implement. Define the environment variables PACT_DESCRIPTION and/or PACT_PROVIDER_STATE as so:
$ PACT_DESCRIPTION="a request for something" PACT_PROVIDER_STATE="something exists" rake pact:verify
Yay! Your provider now honours the pact it has with your consumer. You can now have confidence that your consumer and provider will play nicely together.
Having different service provider states allows you to test the same request with different expected responses.
For example, some code that creates the pact in a consumer project might look like this:
my_service.
given("a thing exists").
upon_receiving("a request for a thing").
with(method: 'get', path: '/thing').
will_respond_with(status: 200, :body => {thing: "yay!"} )
...
my_service.
given("a thing does not exist").
upon_receiving("a request for a thing").
with(method: 'get', path: '/thing').
will_respond_with(status: 404, :body => {error: "There is no thing :("} )To define service provider states that create the right data for "a thing exists" and "a thing does not exist", write the following in the service provider project. (The consumer name here must match the name of the consumer configured in your consumer project for it to correctly find these provider states.)
# In /spec/service_consumers/provider_states_for_my_service_consumer.rb
Pact.provider_states_for 'My Service Consumer' do
provider_state "a thing exists" do
set_up do
# Create a thing here using your factory of choice
end
tear_down do
# Any tear down steps to clean up your code (or use RSpec.after(:each))
end
end
provider_state "a thing does not exist" do
no_op # If there's nothing to do because the state name is more for documentation purposes, you can use no_op to imply this.
end
end# In /spec/service_consumers/pact_helper.rb
require_relative 'provider_states_for_my_service_consumer.rb'If a state should be used for all consumers, the top level Pact.with_consumer can be skipped, and a global Pact.provider_state can be defined on its own.
You can verify a pact at an arbitrary local or remote URL
$ rake pact:verify:at[../path-to-your-consumer-project/specs/pacts/my_consumer-my_provider.json]
$ rake pact:verify:at[http://build-box/MyConsumerBuild/latestSuccessful/artifacts/my_consumer-my_provider.json]
To make a shortcut task for pact at an arbitrary URL, add the following to your Rakefile. The pact.uri may be a local file system path or a remote URL.
# In Rakefile or /tasks/pact.rake
# This creates a rake task that can be executed by running
# $ rake pact:verify:dev
Pact::VerificationTask.new(:dev) do | pact |
pact.uri '../path-to-your-consumer-project/specs/pacts/my_consumer-my_provider.json'
endPact.configure do | config |
config.pact_dir = "???" # Optional, default is ./spec/pacts
config.log_dir = "???" # Optional, default is ./log
config.logger = "??" # Optional, defaults to a file logger to the configured log_dir.
config.logger.level = Logger::DEBUG #By default this is INFO, bump this up to debug for more detailed logs
config.pactfile_write_mode = :ovewrite / :update / :smart # Optional. The default pactfile_write_mode is :overwrite. See notes in Advanced section for further information.
endThis makes the pact available via URL, which your provider build can then use when it runs pact:verify.
Do not hand create any HTTP requests in your consumer app or specs. Testing through your provider client class gives you the assurance that your consumer app will be creating exactly the HTTP requests that you think it should.
Sure, you've checked that your client deserialises the HTTP response into the object you expect, but then you need to make sure in your other tests where you stub your client that you're stubbing it with a valid object. The best way to do this is to use factories for all your tests.
Configure the pact_uri in the Pact.service_provider block with the pact artifact URL of your last successful build. This way you're only verifying green builds. No point verifying a broken one. (Watch this space - pact-broker coming soon, so we don't have to use messy build box artifact URLs)
It should run with all your other tests. If an integration is broken, you want to know about it before you check in.
Do not stub your database calls for pact:verify. This is the best time for you to test your database integration. If you stub your database calls, you are getting little more assurance that the real end-to-end will work than if you'd used a unit test. It's the appropriate time to incur the overhead of a database call.
- Be aware when using the app from the config.ru file is used (the default option) that the Rack::Builder.parse_file seems to require files even if they have already been required, so make sure your boot files are idempotent.
To execute a subset of the specs when running any of the pact verification tasks, define the environment variables PACT_DESCRIPTION and/or PACT_PROVIDER_STATE.
$ PACT_DESCRIPTION="a request for something" PACT_PROVIDER_STATE="something exists" rake pact:verify
A pact service can be run locally and is really useful for debugging purposes.
$ bundle exec pact service -p <port-num>
The service prints messages it recieves to stdout which can be really useful when diagnosing issues with pacts.
Eg. for Capybara tests
Pact.service_consumer "My Consumer" do
app my_consumer_rack_app
port 4321
endRFC 2616 states that two headers with the same name can interpreted as a single header with two comma-separated values. This is the safest way to specify multiple headers with the same name, as Rack will only pass the last value through when they are defined separately (see rack/rack#436).
my_service_provider.
.given("it is RFC 2616 compliant")
.upon_receiving("a request with a header with commas separated values")
.with( method: :get, path: '/', headers: {'X-Request-Multival' => "A, B"} )
.will_respond_with(
status: 200, headers: {'X-Response-Multival' => "C, D"}
)By default, the pact file will be overwritten (started from scratch) every time any rspec runs any spec using pacts. This means that if there are interactions that haven't been executed in the most recent rspec run, they are effectively removed from the pact file. If you have long running pact specs (e.g. they are generated using the browser with Capybara) and you are developing both consumer and provider in parallel, or trying to fix a broken interaction, it can be tedius to run all the specs at once. In this scenario, you can set the pactfile_write_mode to :update. This will keep all existing interactions, and update only the changed ones, identified by description and provider state. The down side of this is that if either of those fields change, the old interactions will not be removed from the pact file. As a middle path, you can set pactfile_write_mode to :smart. This will use :overwrite mode when running rake (as determined by a call to system using 'ps') and :update when running an individual spec.
The pact:verify RSpec examples have the metadata {:pact => :verify} defined. You can add RSpec hooks using a filter as shown here:
RSpec.configure do | config |
config.before :each, :pact => :verify do
# Your code here
end
endSee https://www.relishapp.com/rspec/rspec-core/docs/hooks/filters for more information.
Short term:
- FIX EXAMPLE!!!
Long term:
- Provide more flexible matching (eg the keys should match, and the classes of the values should match, but the values of each key do not need to be equal). This is to make the pact verification less brittle.
- Decouple Rspec from Pact and make rspec-pact gem for easy integration
- Fork it
- Create your feature branch (
git checkout -b my-new-feature) - Commit your changes (
git commit -am 'Add some feature') - Push to the branch (
git push origin my-new-feature) - Create new Pull Request