Category Archive for Rails

Rack::Bug debugging toolbar in four minutes

April 22, 2009

Here’s a quick screencast demonstrating Rack::Bug, a new Rack middleware for inspecting requests to any Rack-compatible web application.

Download the screencast (12 MB, 4:00)

The source is on GitHub. It ships with nine panels:

  • Rails Info
  • Resource Usage
  • Rack Env
  • SQL
  • AR Objects
  • Memcached
  • Templates
  • Log
  • Memory Usage
./script/plugin install git://github.com/brynary/rack-bug.git
1
2
3
4
5
6
7
# config/initializers/middleware.rb
require "rack/bug"

ActionController::Dispatcher.middleware.use Rack::Bug,
  :ip_masks   => [IPAddr.new("127.0.0.1")],
  :secret_key => "epT5uCIchlsHCeR9dloOeAPG66PtHd9K8l0q9avitiaA/KUrY7DE52hD4yWY+8z1",
  :password   => "rack-bug-secret"

Switching Webrat to Selenium mode

April 06, 2009

Ryan Bates did a great job covering Webrat in Railscast #156, and I’d recommend anyone who hasn’t used Webrat check it out. It’s a clear demonstration of how you can start using Webrat in a few minutes today with Rails built-in integration testing framework. Webrat works with RSpec, Cucumber, Shoulda, etc. too, but they aren’t required.

Unfortunately, during the screencast some Webrat bugs prevented running the tests in Selenium mode. Webrat has worked with Selenium for some time, but there have been some lingering rough edges around getting it up and running that I’ve been meaning to resolve. Today I’m happy to report those issues are now fixed, and getting started using Selenium with Webrat should now be drop-dead simple.

Give it a try by checking out the full Episode #156 source code, then make the following changes:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
# config/environments/test.rb
config.gem "webrat", :version => ">=0.4.4" # was 0.4.3

# test/test_helper.rb
class ActiveSupport::TestCase
  # ...
  self.use_transactional_fixtures = false # was true

  # ...

  # Add these lines:
  setup do |session|
    session.host! "localhost:3001"
  end
end
  
Webrat.configure do |config|
  config.mode = :selenium # was :rails
end

Running the test again should pop open a Firefox window for the running tests and print output like this:

$ rake test:integration
(in /Users/bhelmkamp/p/railscasts-episodes/episode-156/authenticator)
Loaded suite /Library/Ruby/Gems/1.8/gems/rake-0.8.4/lib/rake/rake_test_loader
Started
==> Waiting for Selenium RC server on port 4444... Ready!
==> Waiting for rails application server on port 3001... Ready!
..
Finished in 11.679288 seconds.

2 tests, 2 assertions, 0 failures, 0 errors

Success!

Note: I just pushed Webrat 0.4.4, so it might not be available on all the gem sever mirrors yet. If not, you can install it with the following commands:

git clone git://github.com/brynary/webrat.git
cd webrat
rake install_gem

Speaking about Webrat at GoGaRuCo

March 30, 2009

Thanks to all those kind attendees who voted for my talk proposal, I’ve been invited to speak at GoGaRuCo in San Francisco on April 17-18th. My talk is titled Webrat: Rails Acceptance Testing Evolved, and I’m especially excited for two reasons:

  • The conference looks like it will be excellent. There are lots of talks I’m looking forward to seeing, Josh and Leah are doing a great job as organizers (harder than it looks!), and it’s in a cool city I haven’t visited before.
  • This is the first time I’ll be giving a talk about Webrat specifically, rather than covering Webrat as part of a broader topic. The project is almost a year and a half old now, and there’s plenty of interesting bits to talk about these days.

The conference is sold out, but if you’re going you should say hello. I’m always up for chatting about Ruby, scaling or testing, and grabbing a beer. Should be easy to find, and you can message me as @brynary on Twitter.

Rack::Test released: Simply test any Rack-compatible app

March 04, 2009

Today I’m happy to announce the first release of a new library I’ve been cooking up: Rack::Test. It’s a simple API for issuing fake requests to any Rack app (e.g. Rails, Sinatra, Merb, etc.) similar to what you might be familiar with in Rails’ integration tests or Merb’s request specs. Basically, I extracted Merb’s request helper code into a small, reusable, framework agnostic library. Here’s an example:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
require "rack/test"

class HomepageTest < Test::Unit::TestCase
  include Rack::Test::Methods
  
  def app
    MyApp.new
  end
  
  def test_redirect_logged_in_users_to_dashboard
    authorize "bryan", "secret"
    get "/"
    follow_redirect!
    
    assert_equal "http://example.org/redirected", last_request.url
    assert last_response.ok?
  end
end

While developing Webrat, I started to realize all of the Ruby web frameworks were reinventing boilerplate code to create a test environment for issuing fake requests. They all have to:

  • Convert params hashes into encoded strings (something the browser handles for normal requests)
  • Maintain a simulated cooke jar to enable sessions
  • Generate multipart requests (to support test file uploads)

Through discussions with Josh Knowles and Yehuda Katz, it became apparent that an abstracted layer for these concerns would be valuable.

With Rack::Test, we hope to make it easy for frameworks to encourage their users to write tests by making it trivial to provide a testing environment. We’d like to foster compatibility between Ruby web app testing environments (especially important as ideas like multi-framework apps become more prominent). The philosophy is the library should stay small and extendable so frameworks can layer on additional functionality they want to offer without modifying Rack::Test’s core behavior or resorting to monkeypatching.

The Rack::Test API

The full RDoc is available on gitrdoc, but I’ve reproduced some of it below. It should feel familiar if you’re used to Rails integration testing or Merb request specs:

Rack::Test.initialize(app) — Initialize a new session for the given Rack app

#get(uri, params = {}, env = {}) — Issue a GET request for the given URI with the given params and Rack environment. Stores the issues request object in #last_request and the app’s response in #last_response. Yield #last_response to a block if given.

#post, #put, #delete, and #head — Behave like #get but specifying a different HTTP method

#request(uri, env = {}) — Issue a request to the Rack app for the given URI and optional Rack environment. Stores the issues request object in #last_request and the app’s response in #last_response. Yield #last_response to a block if given.

#follow_redirect! — Rack::Test does not follow any redirects automatically. This method will follow the redirect returned in the last response. If the last response was not a redirect, an error will be raised.

#header(name, value) — Set a header to be included on all subsequent requests through the session. Use a value of nil to remove a previously configured header.

#authorize(username, password)— Set the username and password for HTTP Basic authorization, to be included in subsequent requests in the HTTP_AUTHORIZATION header.

#last_request — Return the last request issued in the session. Raises an error if no requests have been sent yet.

#last_response — Return the last response received in the session. Raises an error if no requests have been sent yet.

Note: Thanks to the people who helped get Rack::Test off the ground: The Merb Team for their request helper code; Yehuda Katz and Josh Knowles for guidance, encouragement and support; Simon Rozet and Pat Nakajima for early code contributions.

Our Git deployment workflow

August 03, 2008

At weplay, we recently switched to pure git version control from git-svn. Now that we've had a couple weeks for the dust to settle, I'd like to share our workflow for managing deployments to our staging and production clusters.

We started by outlining the goals of our system:

  • All code that's pushed to our staging and production environments must be in GitHub. Nothing goes straight from a local repository to our servers.

  • Any developer can deploy our most recent work to staging.

  • Any developer can deploy the code on staging into production. We (try to) avoid deploying anything to production that hasn't been pushed to staging first.

  • Any developer can see a diff between "What is deployed" and "What I'm about to deploy."

  • Any developer can branch from the production codebase for time sensitive tweaks and fixes. These need to be staged before they're deployed to production too.

  • That functionality is available in an automated, safe, easy to use form. (Hint: rake)

We ruled out having "production" and "staging" tags because updating tags across a tree of remotes doesn't seem to work smoothly. We also ruled out doing git merges from master into our staging branch. Our staging server/codebase jumps around from the latest changes to the production code (up to a week old) based on what we need to test, so we really just want to replace it.

We settled on using remote git branches for production and staging on origin and hard resetting them to other branches to simulate a copy. We treat these branches more like tags and never commit to them directly. This has worked great so far.

Here's what our rake tasks look like:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
class GitCommands

  def diff_staging
    `git fetch`
    puts `git diff origin/production origin/staging`
  end
  
  def tag_staging(branch_name)
    verify_working_directory_clean
    
    `git fetch`
    `git branch -f staging origin/staging`
    `git checkout staging`
    `git reset --hard origin/#{branch_name}`
    `git push -f origin staging`
    `git checkout master`
    `git branch -D staging`
  end
  
  def tag_production
    verify_working_directory_clean
    
    `git fetch`
    `git branch -f production origin/production`
    `git checkout production`
    `git reset --hard origin/staging`
    `git push -f origin production`
    `git checkout master`
    `git branch -D production`
  end

  def branch_production(branch_name)
    verify_working_directory_clean
    
    `git fetch`
    `git branch -f production origin/production`
    `git checkout production`
    `git branch #{branch_name}`
    `git checkout #{branch_name}`
    `git push origin #{branch_name}`
  end
  
protected

  def verify_working_directory_clean
    return if `git status` =~ /working directory clean/
    raise "Must have clean working directory"
  end
end


namespace :tag do
  desc <<-DESC
    Update the staging branch to prepare for a staging deploy.
    Defaults to master. Optionally specify a BRANCH=name
  DESC
  
  task :staging do
    branch_name = ENV['BRANCH'] || "master"
    GitCommands.new.tag_staging(branch_name)
  end

  desc "Update the remove production branch to prepare for a release"
  task :production => ['diff:staging'] do
    GitCommands.new.tag_production
  end
end

namespace :diff do
  desc "Show the differences between the staging branch and the production branch"
  task :staging do
    GitCommands.new.diff_staging
  end
end

namespace :branch do
  desc "Branch from production for tweaks or bug fixs. Specify BRANCH=name"
  task :production do
    branch_name = ENV['BRANCH']
    raise "You must specify a branch name using BRANCH=name" unless branch_name
    GitCommands.new.branch_production
  end
end

namespace :deploy do
  desc "Tag and deploy staging"
  task :staging => "tag:staging" do
    `cap staging deploy:long`
  end
end

The last one (rake deploy:staging) simply wraps up the common task of tagging our latest code to be pushed to staging and initiating a staging deploy.

Note: To use most of these commands, your local working directory must be clean. If we have outstanding changes in our tree when we need to run them, we use git-stash to temporarily move them out of the way.

Thanks to Scott Chacon for helping us work this out. Be sure to check out his GitCasts and Git Internals PDF.

Story Driven Development slides posted

April 26, 2008

Just wrapped up my Story Driven Development talk at GoRuCo 2008. There were some great questions at the end, and I'm looking forward to more hallway track conversations about SDD.

Download the slides as a PDF (1.6 MB)

Note: Confreaks is recording the talks today, so a full video of my presentation should be online soon.

weplay launches

March 26, 2008

The online community for youth sports I’ve been working on for the last month or so, weplay, is now in public beta. It’s been a roller coaster, but entirely worth it.

As part of this launch, we’re set to appear on the front page of the New York Times business section in tomorrow’s paper. The online version of the article is already available, and we’ve been busy all night responding to our first production load…. ever.

Check it out, don’t mind a few bumps and bruises as we settle in, and expect more to come soon. Oh, and be sure to checkout my weplay profile.

Presenting at GoRuCo in NYC

February 09, 2008

I just received word that my proposal to speak at NYC’s very own second annual GoRuCo has been accepted. The conference is set for Saturday, April 26th at Pace University downtown. My topic will be the same as my Scotland on Rails presentation.

I’m really excited to have the opportunity to present at home, in front of many of my friends and colleagues. Now I just need to make sure I don’t suck. Better get back to working on my slide deck…

Presenting at Scotland on Rails

January 31, 2008

I’m going to be presenting on Story Driven Development at the first Scottish Ruby on Rails conference, Scotland on Rails. I’ll be exploring the concept of Story-first development from both an agile process and implementation standpoint.

Topics will include:

  • Collaborating with the product owner on stories
  • What makes good stories and scenarios?
  • Driving stories with a browser simulator like Webrat
  • Potential pitfalls and ways to avoid them

If you’re going to be in Edinburgh for the conference, let me know.

YUI Pages plugin 0.1.0 released

December 15, 2007

Tonight I’m releasing version 0.1.0 of my new YUI Pages Rails plugin. Here’s a quick example of what it can do (from the README):

<% yui_page :width => YUIPages::PAGE_750, :secondary => YUIPages::RIGHT_180 do %>
  <% header do %>
    Page title
  <% end %>

  <% body do %>
    <% main do %>
      <% grid :columns => YUIPages::GRID_75_25 do %>
        <% unit do %>
          Content for the 75% width first column.
        <% end %>
        <% unit do %>
          Content for the 25% width second column.
        <% end %>
      <% end %>
    <% end %>

    <% secondary do %>
      Content for the 180px right sidebar.
    <% end %>      
  <% end %>

  <% footer do %>
    Footer links here
  <% end %>
<% end %>

This light (less than 90 lines of code) plugin saves me from having to remember or care about the YUI Grids CSS IDs and classes, and handles preparing correct markup regardless of my grid arrangement.

No more looking up that “yui-t3” means a 300px sidebar on the left or forgetting to class the first unit in a grid with “first” because I moved things around in my layout file. A small victory, but a victory nonetheless.

SVN is at http://svn.eastmedia.net/public/plugins/yui_pages/. Check out the full README.

"What's new in Rails 2?" slides posted

December 12, 2007

Thanks to everyone who attended my “What’s new in Rails 2?” talk last night at NYC.rb. The questions were great, and I picked up a few tips myself from the discussions we had.

Download the slides as a PDF (1.4 MB).

Speaking at NYC.rb on Tuesday

December 08, 2007

On Tuesday, December 11th, I’ll be speaking at NYC.rb about “What’s new in Rails 2?”. Come by and watch or check back after then and I’ll be sure to have my slides up.

Webrat 0.1.0 released

December 08, 2007

Last week I pushed Webrat 0.1.0 out the door. So far the response has been great, and I’ve already received a couple patched (Thanks, David!). Here’s a quick usage example (from the README):

1
2
3
4
5
6
7
8
def test_sign_up
    visits "/"
    clicks_link "Sign up"
    fills_in "Email", :with => "good@example.com"
    select "Free account"
    clicks_button "Register"
    ...
  end

Behind the scenes, this will perform the following work:

  1. Verify that loading the home page is successful
  2. Verify that a “Sign up” link exists on the home page
  3. Verify that loading the URL pointed to by the “Sign up” link leads to a successful page
  4. Verify that there is an “Email” input field on the Sign Up page
  5. Verify that there is an select field on the Sign Up page with an option for “Free account”
  6. Verify that there is a “Register” submit button on the page
  7. Verify that submitting the Sign Up form with the values “good@example.com” and “Free account” leads to a successful page

Take special note of the things not specified in that test, that might cause tests to break unnecessarily as your application evolves:

  • The input field IDs or names (e.g. “user_email” or “user[email]”), which could change if you rename a model
  • The ID of the form element (Webrat can do a good job of guessing, even if there are multiple forms on the page.)
  • The URLs of links followed
  • The URL the form submission should be sent to, which could change if you adjust your routes or controllers
  • The HTTP method for the login request

A test written with Webrat can handle these changes smoothly.

SVN is at http://svn.eastmedia.net/public/plugins/webrat/. Check out the full README.

Scourging your Ruby code with Flog

September 13, 2007

Flog is a tool build by Ryan Davis to analyze Ruby code complexity. It’s dead simple to run, and immediately provides useful metrics in the form of a “Flog score” per method. Don’t get too hung up on the actual values, but high outliers are prime candidates for refactoring.

I ran this on a couple recent projects I worked on, and it proved quite accurate in identifying the areas of code that seem to be in the most pain. Flog doesn’t replace hands-on code review, but it is still a helpful aid.

Below is a rake task I threw together for running Flog on a Rails project. Seems like a great candidate to install via Chris Wanstrath’s Sake, a system for handling Rake tasks that you want to be available from anywhere.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
def flog(output, *directories)
  `find #{directories.join(" ")} -name \\*.rb|xargs flog > #{RAILS_ROOT}/tmp/flog/#{output}.txt`
end

desc "Flog models, controller, helpers and lib"
task :flog do
  flog "all", *%w[app/models app/controllers app/helpers lib]
end

namespace :flog do
  desc "Flog code in app/models"
  task :models do
    flog "models", "app/models"
  end
  
  desc "Flog code in app/controllers"  
  task :controllers do
    flog "controllers", "app/controllers"
  end
  
  desc "Flog code in app/helpers"  
  task :helpers do
    flog "helpers", "app/helpers"
  end
  
  desc "Flog code in lib"  
  task :lib do
    flog "lib", "lib"
  end
end

Modeling tip: Replace booleans with timestamps

September 11, 2007

Let’s say you’re building a blogging application and you’ve got a Post model. Posts should not show up on the public site until they are approved. Approval is done by clicking an “Approve” button next to the post in the admin area.

Your first inclination (and mine, up until recently) might be to implement this with a boolean column:

1
2
3
4
5
6
7
8
class Post < ActiveRecord::Base
  # assuming posts.approved_at is a boolean column

  def approve!
    self.approved = true
  end

end

That works just fine, and may be all you need. On the other hand, my experience has shown that storing a little bit more data is generally worth the tiny incurred complexity. For example:

1
2
3
4
5
6
7
8
9
10
11
12
class Post < ActiveRecord::Base
  # assuming posts.approved_at is a datetime column

  def approve!
    self.approved_at = Time.now
  end

  def approved?
    !approved_at.nil?
  end

end

This has proven extremely valuable for debugging and audit trail purposes, even if the extra timestamp data is never exposed in a user interface. Approved posts can still be easily queried using SQL by using a WHERE approved_at IS NOT NULL condition.