parisyee / puffing-billy

A rewriting web proxy for testing interactions between your browser and external sites. Works with ruby + rspec.

Puffing Billy

A rewriting web proxy for testing interactions between your browser and external sites. Works with ruby + rspec.

Puffing Billy is like webmock or VCR, but for your browser.

Overview

Billy spawns an EventMachine-based proxy server, which it uses to intercept requests sent by your browser. It has a simple API for configuring which requests need stubbing and what they should return.

Billy lets you test against known, repeatable data. It also allows you to test for failure cases. Does your twitter (or facebook/google/etc) integration degrade gracefully when the API starts returning 500s? Well now you can test it!

it 'should stub google' do
  proxy.stub('http://www.google.com/').and_return(:text => "I'm not Google!")
  visit 'http://www.google.com/'
  expect(page).to have_content("I'm not Google!")
end

You can also record HTTP interactions and replay them later. See caching below.

Installation

Add this line to your application's Gemfile:

gem 'puffing-billy'

And then execute:

$ bundle

Or install it yourself as:

$ gem install puffing-billy

RSpec Usage

Setup for Capybara

In your rails_helper.rb:

require 'billy/capybara/rspec'

# select a driver for your chosen browser environment
Capybara.javascript_driver = :selenium_billy # Uses Firefox
# Capybara.javascript_driver = :selenium_chrome_billy
# Capybara.javascript_driver = :webkit_billy
# Capybara.javascript_driver = :poltergeist_billy

Note: :poltergeist_billy doesn't support proxying any localhosts, so you must use :webkit_billy for headless specs when using puffing-billy for other local rack apps. See this phantomjs issue for any updates.

Setup for Watir

In your rails_helper.rb:

require 'billy/watir/rspec'

# select a driver for your chosen browser environment
@browser = Billy::Browsers::Watir.new :firefox
# @browser = Billy::Browsers::Watir.new = :chrome
# @browser = Billy::Browsers::Watir.new = :phantomjs

In your tests (Capybara/Watir)

# Stub and return text, json, jsonp (or anything else)
proxy.stub('http://example.com/text/').and_return(:text => 'Foobar')
proxy.stub('http://example.com/json/').and_return(:json => { :foo => 'bar' })
proxy.stub('http://example.com/jsonp/').and_return(:jsonp => { :foo => 'bar' })
proxy.stub('http://example.com/headers/').and_return({
  :headers => { 'Access-Control-Allow-Origin' => '*' },
  :json    => { :foo => 'bar' }
})
proxy.stub('http://example.com/wtf/').and_return(:body => 'WTF!?', :content_type => 'text/wtf')

# Stub redirections and other return codes
proxy.stub('http://example.com/redirect/').and_return(:redirect_to => 'http://example.com/other')
proxy.stub('http://example.com/missing/').and_return(:code => 404, :body => 'Not found')

# Even stub HTTPS!
proxy.stub('https://example.com:443/secure/').and_return(:text => 'secrets!!1!')

# Pass a Proc (or Proc-style object) to create dynamic responses.
#
# The proc will be called with the following arguments:
#   params:  Query string parameters hash, CGI::escape-style
#   headers: Headers hash
#   body:    Request body string
#   url:     The actual URL which was requested
#   method:  The HTTP verb which was requested
proxy.stub('https://example.com/proc/').and_return(Proc.new { |params, headers, body, url, method|
  {
    :code => 200,
    :text => "Hello, #{params['name'][0]}"
  }
})

# You can also use Puffing Billy to intercept requests and responses. Just pass
# a Proc and use the pass_request method. You can manipulate the request
# (headers, URL, HTTP method, etc) and also the response from the upstream
# server. The scope of the delivered callable is the user scope where
# it was defined.
proxy.stub('http://example.com/').and_return(Proc.new { |*args|
  response = Billy.pass_request(*args)
  response[:headers]['Content-Type'] = 'text/plain'
  response[:body] = 'Hello World!'
  response[:code] = 200
  response
})

# Stub out a POST. Don't forget to allow a CORS request and set the method to 'post'
proxy.stub('http://example.com/api', :method => 'post').and_return(
  :headers => { 'Access-Control-Allow-Origin' => '*' },
  :code => 201
)

# Stub out an OPTIONS request. Set the headers to the values you require.
proxy.stub('http://example.com/api', :method => :options).and_return(
  :headers => {
    'Access-Control-Allow-Methods' => 'GET, PATCH, POST, PUT, OPTIONS',
    'Access-Control-Allow-Headers' => 'X-Requested-With, X-Prototype-Version, Content-Type',
    'Access-Control-Allow-Origin'  => '*'
  },
  :code => 200
)

Stubs are reset between tests. Any requests that are not stubbed will be proxied to the remote server.

If for any reason you'd need to reset stubs manually you can do it in two ways:

# reset a single stub
example_stub = proxy.stub('http://example.com/text/').and_return(:text => 'Foobar')
proxy.unstub example_stub

# reset all stubs
proxy.reset

Cucumber Usage

An example feature:

Feature: Stubbing via billy

  @javascript @billy
  Scenario: Test billy
    And a stub for google

Capybara

In your features/support/env.rb:

require 'billy/capybara/cucumber'

After do
  Capybara.use_default_driver
end

And in steps:

Before('@billy') do
  Capybara.current_driver = :poltergeist_billy
end

And /^a stub for google$/ do
  proxy.stub('http://www.google.com/').and_return(:text => "I'm not Google!")
  visit 'http://www.google.com/'
  expect(page).to have_content("I'm not Google!")
end

It's good practice to reset the driver after each scenario, so having an @billy tag switches the drivers on for a given scenario. Also note that stubs are reset after each step, so any usage of a stub should be in the same step that it was created in.

Watir

In your features/support/env.rb:

require 'billy/watir/cucumber'

After do
  @browser.close
end

And in steps:

Before('@billy') do
  @browser = Billy::Browsers::Watir.new :firefox
end

And /^a stub for google$/ do
  proxy.stub('http://www.google.com/').and_return(:text => "I'm not Google!")
  @browser.goto 'http://www.google.com/'
  expect(@browser.text).to eq("I'm not Google!")
end

Minitest Usage

Please see this link for details and report back to Issue #49 if you get it fully working.

Caching

Requests routed through the external proxy are cached.

By default, all requests to localhost or 127.0.0.1 will not be cached. If you're running your test server with a different hostname, you'll need to add that host to puffing-billy's whitelist.

In your rails_helper.rb:

Billy.configure do |c|
  c.whitelist = ['test.host', 'localhost', '127.0.0.1']
end

If you would like to cache other local rack apps, you must whitelist only the specific port for the application that is executing tests. If you are using Capybara, this can be accomplished by adding this in your rails_helper.rb:

server = Capybara.current_session.server
Billy.config.whitelist = ["#{server.host}:#{server.port}"]

If you want to use puffing-billy like you would VCR you can turn on cache persistence. This way you don't have to manually mock out everything as requests are automatically recorded and played back. With cache persistence you can take tests completely offline.

Billy.configure do |c|
  c.cache = true
  c.cache_request_headers = false
  c.ignore_params = ["http://www.google-analytics.com/__utm.gif",
                     "https://r.twimg.com/jot",
                     "http://p.twitter.com/t.gif",
                     "http://p.twitter.com/f.gif",
                     "http://www.facebook.com/plugins/like.php",
                     "https://www.facebook.com/dialog/oauth",
                     "http://cdn.api.twitter.com/1/urls/count.json"]
  c.path_blacklist = []
  c.merge_cached_responses_whitelist = []
  c.persist_cache = true
  c.ignore_cache_port = true # defaults to true
  c.non_successful_cache_disabled = false
  c.non_successful_error_level = :warn
  c.non_whitelisted_requests_disabled = false
  c.cache_path = 'spec/req_cache/'
  c.certs_path = 'spec/req_certs/'
  c.proxy_host = 'example.com' # defaults to localhost
  c.proxy_port = 12345 # defaults to random
  c.proxied_request_host = nil
  c.proxied_request_port = 80
  c.record_requests = true # defaults to false
  c.cache_request_body_methods = ['post', 'patch', 'put'] # defaults to ['post']
end

The cache works with all types of requests and will distinguish between different POST requests to the same URL.

c.cache_request_headers is used to store the outgoing request headers in the cache. It is also saved to yml if persist_cache is enabled. This additional information is useful for debugging (for example: viewing the referer of the request).

c.ignore_params is used to ignore parameters of certain requests when caching. You should mostly use this for analytics and various social buttons as they use cache avoidance techniques, but return practically the same response that most often does not affect your test results.

c.allow_params is used to allow parameters of certain requests when caching. This is best used when a site has a large number of analytics and social buttons. c.allow_params is the opposite of c.ignore_params, a whitelist to a blacklist. In order to toggle between using one or the other, use c.use_ignore_params.

c.strip_query_params is used to strip query parameters when you stub some requests with query parameters. Default value is true. For example, proxy.stub('http://myapi.com/user/?country=FOO') is considered the same as: proxy.stub('http://myapi.com/user/?anything=FOO') and generally the same as: proxy.stub('http://myapi.com/user/'). When you need to distinguish between all these requests, you may set this config value to false.

c.dynamic_jsonp is used to rewrite the body of JSONP responses based on the callback parameter. For example, if a request to http://example.com/foo?callback=bar returns bar({"some": "json"}); and is recorded, then a later request to http://example.com/foo?callback=baz will be a cache hit and respond with baz({"some": "json"}); This is useful because most JSONP implementations base the callback name off of a timestamp or something else dynamic.

c.dynamic_jsonp_keys is used to configure which parameters to ignore when using c.dynamic_jsonp. This is helpful when JSONP APIs use cache-busting parameters. For example, if you want http://example.com/foo?callback=bar&id=1&cache_bust=12345 and http://example.com/foo?callback=baz&id=1&cache_bust=98765 to be cache hits for each other, you would set c.dynamic_jsonp_keys = ['callback', 'cache_bust'] to ignore both params. Note that in this example the id param would still be considered important.

c.dynamic_jsonp_callback_name is used to configure the name of the JSONP callback parameter. The default is callback.

c.path_blacklist = [] is used to always cache specific paths on any hostnames, including whitelisted ones. This is useful if your AUT has routes that get data from external services, such as /api where the ajax request is a local URL but the actual data is coming from a different application that you want to cache.

c.merge_cached_responses_whitelist = [] is used to group together the cached responses for specific uri regexes that match any part of the url. This is useful for ensuring that any kind of analytics and various social buttons that have slightly different urls each time can be recorded once and reused nicely. Note that the request body is ignored for requests that contain a body.

c.ignore_cache_port is used to strip the port from the URL if it exists. This is useful when caching local paths (via path_blacklist) or other local rack apps that are running on random ports.

c.non_successful_cache_disabled is used to not cache responses without 200-series or 304 status codes. This prevents unauthorized or internal server errors from being cached and used for future test runs.

c.non_successful_error_level is used to log when non-successful responses are received. By default, it just writes to the log file, but when set to :error it throws an error with the URL and status code received for easier debugging.

c.non_whitelisted_requests_disabled is used to disable hitting new URLs when no cache file exists. Only whitelisted URLs (on non-blacklisted paths) are allowed, all others will throw an error with the URL attempted to be accessed. This is useful for debugging issues in isolated environments (ie. continuous integration).

c.cache_path can be used to locate the cache directory to a different place other than system temp directory/puffing-billy.

c.certs_path can be used to locate the directory for dynamically generated SSL certificates to a different place other than system temp directory/puffing-billy/certs.

c.proxy_host and c.proxy_port are used for the Billy proxy itself which runs locally.

c.proxied_request_host and c.proxied_request_port are used if an internal proxy server is required to access the internet. Most common in larger companies.

c.record_requests can be used to record all requests that puffing billy proxied. This can be useful for debugging purposes, for instance if you are unsure why your stubbed requests are not being successfully proxied.

Example usage:

require 'table_print' # Add this dependency to your gemfile

Billy.configure do |c|
  c.record_requests = true
end

RSpec.configure do |config|
  config.prepend_after(:example, type: :feature) do
    puts "Requests received via Puffing Billy Proxy:"

    puts TablePrint::Printer.table_print(Billy.proxy.requests, [
      :status,
      :handler,
      :method,
      { url: { width: 100 } },
      :headers,
      :body
    ])
  end
end

This will generate a human readable list of all requests after each test run:

Requests received via Puffing Billy Proxy:
STATUS   | HANDLER | METHOD  | URL                                     | HEADERS                        | BODY
---------|---------|---------|-----------------------------------------|--------------------------------|-----
complete | proxy   | GET     | http://127.0.0.1:56692/                 | {"Accept"=>"text/html,appli... |
complete | proxy   | GET     | http://127.0.0.1:56692/assets/appl...   | {"Accept"=>"text/css,*/*;q=... |
complete | proxy   | GET     | http://127.0.0.1:56692/assets/app...    | {"Accept"=>"*/*", "Referer"... |
complete | proxy   | GET     | http://127.0.0.1:56692/javascript/index | {"Accept"=>"text/html,appli... |
complete | stubs   | OPTIONS | https://api.github.com:443/             | {"Access-Control-Request-Me... |
complete | stubs   | GET     | https://api.github.com:443/             | {"Accept"=>"*/*", "Referer"... |
inflight |         | GET     | http://127.0.0.1:56692/example          | {"Referer"=>"http://127.0.0... |
.

Finished in 1.98 seconds (files took 2.11 seconds to load)
1 example, 0 failures

The handler column indicates how Puffing Billy handled your request:

• proxy: This request was successfully routed to the original target
• stubs: This was handled via a stub
• error: This request was not handled by a stub, and was not successfully handled
• cache: This response was handled by a previous cache

If your status is set to in_flight this request has not yet been handled fully. Either puffing billy crashed internally on this request, or your test ended before it could complete successfully.

c.cache_request_body_methods is used to specify HTTP methods of requests that you would like to cache separately based on the contents of the request body. The default is ['post'].

c.after_cache_handles_request is used to configure a callback that can operate on the response after it has been retrieved from the cache but before it is returned. The callback receives the request and response as arguments, with a request object like: { method: method, url: url, headers: headers, body: body }. An example usage would be manipulating the Access-Control-Allow-Origin header so that your test server doesn't always have to run on the same port in order to accept cached responses to CORS requests:

c.use_ignore_params is used to choose whether to use the ignore_params blacklist or the allow_params whitelist. Set to true to use c.ignore_params, false to use c.allow_params

Billy.configure do |c|
  ...
  fix_cors_header = proc do |_request, response|
    allowed_origins = response[:headers]['Access-Control-Allow-Origin']
    if allowed_origins.present?
      localhost_port_pattern = %r{(?<=http://127\.0\.0\.1:)(\d+)}
      allowed_origins.sub!(
        localhost_port_pattern, Capybara.current_session.server.port.to_s
      )
    end
  end
  c.after_cache_handles_request = fix_cors_header
  ...
end

c.cache_simulates_network_delays is used to add some delay before cache returns response. When set to true, cached requests will wait from configured delay time before responding. This allows to catch various race conditions in asynchronous front-end requests. The default is false.

c.cache_simulates_network_delay_time is used to configure time (in seconds) to wait until responding from cache. The default is 0.1.

Cache Scopes

If you need to cache different responses to the same HTTP request, you can use cache scoping.

For example, an index page may return zero or more items in a list, with or without pagination, depending on the number of entries in a database.

There are a few different ways to use cache scopes:

# If you do nothing, it uses the default cache scope:
it 'defaults to nil scope' do
  expect(proxy.cache.scope).to be_nil
end

# You can change context indefinitely to a specific cache scope:
context 'with a cache scope' do
  before do
    proxy.cache.scope_to "my_cache"
  end

  # Remember to set the cache scope back to the default in an after block
  # within the context it is used, and/or at the global rails_helper level!
  after do
    proxy.cache.use_default_scope
  end

  it 'uses the cache scope' do
    expect(proxy.cache.scope).to eq("my_cache")
  end

  it 'can be reset to the default scope' do
    proxy.cache.use_default_scope
    expect(proxy.cache.scope).to be_nil
  end

  # Or you can run a block within the context of a cache scope:
  # Note: When using scope blocks, be sure that both the action that triggers a
  #       request and the assertion that a response has been received are within the block
  it 'can execute a block against a named cache' do
    expect(proxy.cache.scope).to eq("my_cache")
    proxy.cache.with_scope "another_cache" do
      expect(proxy.cache.scope).to eq "another_cache"
    end
    # It
    expect(proxy.cache.scope).to eq("my_cache")
  end
end

If you use named caches it is highly recommend that you use a global hook to set the cache back to the default before or after each test.

In Rspec:

RSpec.configure do |config|
  config.before :each { proxy.cache.use_default_scope }
end

Separate Cache Directory for Each Test (in Cucumber)

If you want the cache for each test to be independent, i.e. have it's own directory where the cache files are stored, you can use a Before tag like so:

Before('@javascript') do |scenario, block|
  Billy.configure do |c|
    feature_name = scenario.feature.name.underscore
    scenario_name = scenario.name.underscore
    c.cache_path = "features/support/fixtures/req_cache/#{feature_name}/#{scenario_name}/"
    Dir.mkdir_p(Billy.config.cache_path) unless File.exist?(Billy.config.cache_path)
  end
end

Stub requests recording

If you want to record requests to stubbed URIs, set the following configuration option:

Billy.configure do |c|
  c.record_stub_requests = true
end

Example usage:

it 'should intercept a GET request' do
  stub = proxy.stub('http://example.com/')
  visit 'http://example.com/'
  expect(stub.has_requests?).to be true
  expect(stub.requests).not_to be_empty
end

Proxy timeouts

By default, the Puffing Billy proxy will use the EventMachine:HttpRequest timeouts of 5 seconds for connect and 10 seconds for inactivity when talking to downstream servers.

These can be configured as follows:

Billy.configure do |c|
  c.proxied_request_connect_timeout = 20
  c.proxied_request_inactivity_timeout = 20
end

Customising the javascript driver

If you use a customised Capybara driver, remember to set the proxy address and tell it to ignore SSL certificate warnings. See lib/billy.rb to see how Billy's default drivers are configured.

Working with VCR and Webmock

If you use VCR and Webmock elsewhere in your specs, you may need to disable them for your specs utilizing Puffing Billy. To do so, you can configure your rails_helper.rb as shown below:

RSpec.configure do |config|
  config.around(:each, type: :feature) do |example|
    WebMock.allow_net_connect!
    VCR.turned_off { example.run }
    WebMock.disable_net_connect!
  end
end

As an alternative if you're using VCR, you can ignore requests coming from the browser. One way of doing that is by adding to your rails_helper.rb the excerpt below:

VCR.configure do |config|
  config.ignore_request do |request|
    request.headers.include?('Referer')
  end
end

Note that this approach may cause unexpected behavior if your backend sends the Referer HTTP header (which is unlikely).

Raising errors from stubs

By default PuffingBilly suppress errors from stub-blocks. To make it raise errors instead, add this test initializers:

EM.error_handler { |e| raise e }

SSL usage

Unfortunately we cannot setup the runtime certificate authority on your browser at time of configuring the Capybara driver. So you need to take care of this step yourself as a prepartion. A good point would be directly after configuring this gem.

Google Chrome Headless example

Google Chrome/Chromium is capable to run as a test browser with the new headless mode which is not able to handle the deprecated --ignore-certificate-errors flag. But the headless mode is capable of handling the user PKI certificate store. So you just need to import the runtime Puffing Billy certificate authority on your system store, or generate a new store for your current session. The following examples demonstrates the former variant:

# Install the fabulous `os` gem first
# See: https://rubygems.org/gems/os
# gem install os
#
# --

# Overwrite the local home directory for chrome. We use this
# to setup a custom SSL certificate store.
ENV['HOME'] = "#{Dir.tmpdir}/chrome-home-#{Time.now.to_i}"

# Clear and recreate the Chrome home directory.
FileUtils.rm_rf(ENV['HOME'])
FileUtils.mkdir_p(ENV['HOME'])

# Setup a new pki certificate database for Chrome
if OS.linux?
  system <<~SCRIPT
    cd "#{ENV['HOME']}"
    curl -s -k -o "cacert-root.crt" "http://www.cacert.org/certs/root.crt"
    curl -s -k -o "cacert-class3.crt" "http://www.cacert.org/certs/class3.crt"
    echo > .password
    mkdir -p .pki/nssdb
    CERT_DIR=sql:$HOME/.pki/nssdb
    certutil -N -d .pki/nssdb -f .password
    certutil -d ${CERT_DIR}  -A -t TC \
      -n "CAcert.org" -i cacert-root.crt
    certutil -d ${CERT_DIR} -A -t TC \
      -n "CAcert.org Class 3" -i cacert-class3.crt
    certutil -d sql:$HOME/.pki/nssdb -A \
      -n puffing-billy -t "CT,C,C" -i #{Billy.certificate_authority.cert_file}
  SCRIPT
end

# Setup the macOS certificate store
if OS.mac?
  prompt = 'Add Puffing Billy root certificate authority ' \
           'to system certificate store'
  system <<~SCRIPT
    sudo -p "# #{prompt}`echo $'\nPassword: '`" \
      security find-certificate -a -Z -c 'Puffing Billy' \
      | grep 'SHA-1 hash' | cut -d ':' -f2 | xargs -n1 \
      sudo security delete-certificate -Z >/dev/null 2>&1 || true
    sudo security add-trusted-cert -d -r trustRoot \
      -k /Library/Keychains/System.keychain \
      #{Billy.certificate_authority.cert_file}
  SCRIPT
end

Mind the reset of the HOME environment variable. Fortunately Chrome takes care of the users home, so we can setup a new temporary directory for the test run, without messing with potential user configurations.

The macOS support requires the input of your password to manipulate the system certificate store. If you are lazy you can turn off sudo password prompt for the security command, but it's strongly advised against. (You know passwordless security, is no security in this case) Further, the macOS handling here cleans up old Puffing Billy root certificate authorities and put the current one into the system store. So after a run of your the suite only one certificate will be left over. If this is not enough you can handling the cleanup again with a custom on-after hook.

Resources

• Bring Ruby VCR to Javascript testing with Capybara and puffing-billy
• Integration Testing Stripe.js With Mocked Network Requests
• Clean-up unused cache files periodically with this config

FAQ

1. Why name it after a train?

   Trains are cool.

Contributing

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Added some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request

TODO

1. Integration for test frameworks other than rspec.
2. Show errors from the EventMachine reactor loop in the test output.