Scott Sims

I have been working on converting our page object factory to open source for a few weeks. Now it is time to launch the  HomeAway sponsored open source project under the Apache 2.0 license.  It is a furiously quick way to implement test automation.  I am planning to add more configuration options in the future.

This project started when I had to test a page with 300+ check boxes and I did not want to enter them by hand.  I used the page object generator to build a page of ruby variables with Selenium locators. Everything was great until some number of the check boxes changed their ids and my tests started failing.  I needed  a quick way to find out how many changed so I could update the locators by hand or regenerate the page.  This is where the validators came in.  I used Ruby’s support of reflection to open a class, navigate to the url of the page and use Selenium to validate the locators and return a list of missing locators on the page.  It worked perfectly for my page of 300+ check boxes.  I had over 40 that changed I quickly regenerated the page.

Install with:

• gem install selenium_fury

Checkout the home page and examples at https://github.com/scottcsims/SeleniumFury

Thanks to HomeAway for sponsoring this project.

I wanted to test an error message in my class that processes test results and sends them to Rally.  The problem was that I did not want to actually create a rally object and run the method to parse the results, just to test the error message.
This is what my class looks like.

class AutomationRun
  def send_results_to_rally
    @rally = RallyUtils.new(ENV['RALLY_WORKSPACE'])
    push_test_case_results
    raise("Test Case Results Were Not Parsed Correctly") if test_case_results.empty?
    test_case_results.each do |result|
      @rally.update_test_case_result(:tc_id =>result.test_case_number_from_spec_report, :build =>result.build_number, :verdict =>result.verdict, :notes => result.note.format_note)
    end
  end
end

I want to skip these steps to test the error message:

• don’t instantiate RallyUtils.new  we don’t need @rally for the test
• don’t parse the results so don’t call parse_test_case_results

This is what my test code looks like:

 it "should raise an exception if there are not test case results" do
    RallyUtils.stub!(:new).and_return(nil)
    automation_run=AutomationRun.new
    automation_run.should_receive(:push_test_case_results).and_return([])
    begin
    automation_run.send_results_to_rally
    rescue Exception=>e
    end
    e.message.should == "Test Case Results Were Not Parsed Correctly"
  end

1. We use a rspec stub to intercept the new call to the RallyUtils class and return nil.  We could have also returned a mock object if we wanted to execute a call on the rally class. Now we are not creating a connection to Rally
2. Now it is time to new up or AutomationRun class and setup our mock expectation to skip running the real push_test_case_results method.
3. automation_run.should_receive tells us to intercept the symbol or method push_test_case_results
4. and_return([]) will just return an empty array
5. We setup a begin rescue block to capture the Exception object returned by our expected error message
6. Now we call send_results_to_rally to get our error message
7. Lastly we verify that the message attribute for our error is Test Case Results Were Not Parsed Correctly

So that is it, I created a partial mock  in that we are using a real automation run object but we stubbed out one of the methods.  We also stubbed out the new call to rally so we don’t create a connection.  Now I can focus my test on exactly the function I wanted to test, the error message.

I am looking forward to my first trip to Las Vegas http://www.stpcon.com/. I will have a chance to meet with other selenium grid users and test out my strategies in test automation patterns. I will be attending all of the topics on agile automation

Deep test allows you to run ruby rspec tests in parallel by providing a spec task that you include in your rake file. The 2.0 prerelease version of deep tests offers support for rspec 1.1.12. You can install it buy running: gem install deep_test_pre. Require it in your rake file like this:

gem “deep_test_pre”, “=2.0”
require “deep_test”

There currently is a bug that occurs if you run Test::Unit tests with the deep test spec runner.  The tests will all run twice but are reported once.

What I have already done to update the ruby example in the Selenium Grid repo. Code Link

• Updated rspec from 1.1.8 to 1.1.12. I accomplished this by getting the most recent deep-test code  and adding it to the /lib directory. The new version of the deep test code supports rspec 1.1.12. I then just require it in in the rake file. require File.expand_path(File.dirname(__FILE__) + ‘/lib/deep_test/rake_tasks’)
• The rake file does not need selenium-client 1.2.7. I removed the dependency by  adding the reporting files it was referencing into the lib/reporting  directory. report_formatter_path = “lib/reporting/selenium_test_report_formatter”
• Updated selenium-client from 1.2.7 to 1.2.18. With the change to the rake file, I am now able to use the most recent version of selenium-client in my specs.

Other helpful tools/tricks I have added to selenium grid (Maybe useful to other Ruby/Grid users)

• I start and stop the grid from the same rake file that I execute the tests from. Examples are grid:all_start and spec:run_in_parallel
• Rake tasks can just run and create a HTML report or run and send test results to Rally
• The tests and rake file runs on OS X, Cent OS, and Windows. (no parallelizaion on windows)
• Specify command options to the rake tasks to switch between hostnames I am running against. (This is useful when you have more than one instance like test.yourdomain.com and stage.yourdomain.com)
• Specify to only run tests for specific sites in the test suit. For example if you wrote tests for www.yourdomain.com and www.yourdomain.es.  You can specify which sites to run against
• I use gem bundler with all grid projects to help control gem installation and versions.  (My Rakefile automatically installs needed gems if you don’t have them)
• I have a framework for creating ruby classes with selenium locators automatically. I use nokogiri to help with this.
• Spec test to open ruby classes with locators and create a report of all of the changed selenium locators.

Future Goals

• Update Deep Test so that it uses the same version(1.2.8) of rspec as the latest version of Selenium-Client. Right now we are at rspec 1.1.12 and selenium client 1.2.18. Selenium-Client 1.2.18 can use rspec 1.2.8.
• When rspec is updated, I will be able to use the selenium client to provide screen shots if needed
• Figure out how to create a formatter that allows the Team City and Rubymine spec runners to provide the same test runner template as it does for test in sequence when you run test in parallel in deep test.
  

  Rubymine Spec Formatter

  

Some of the pages that I test have hundreds of check boxes. I needed a way to generate the locators to use in selenium and create classes to use those locators.
This is an example of a class that contains selenium css locators for navigation links

class Navigation
  def initialize * browser
    @location =  "css=a:contains(\"Location\")"
    @amenities =  "css=a:contains(\"Amenities\")"
    @photos =  "css=a:contains(\"Photos\")"
    @contact =  "css=a:contains(\"Contact\")"  
    @further_details =  "css=a:contains(\"Further Details\")"
  end
  attr_accessor :location, :amenities, :photos, :contact, :further_details
end

To use this class I would new it up and use standard selenium commands. This is how I would navigate to the amenities page

navigation = Navigation.new
browser.click amenities

The spec below demonstrates how to get data you need using nokogiri. I want to click an anmenity check box with the label beach whose input name is “amenity_1_1_6”. I could click this checkbox with this command browser.click “amenity_1_1_6”, but I want to have a way to store all of the amenities on the page with the variable named by their label.

require "spec"
require "nokogiri"
describe "Find elements to use as locators in a web test" do

  it "should find the value of name for the input element and the text in span" do
    sample_html = '
                        Beach
                        
                    '
    attribute_name=""
    attribute_value=""
    doc = Nokogiri::HTML(sample_html)
    doc.css("label").each do |check_box|
      attribute_name =  check_box.text
      attribute_value = check_box.children.css("input")[0]["name"]
      attribute_name = attribute_name.strip
      attribute_name = attribute_name.to_s.downcase

    end
    attribute_name.should=="beach"
    attribute_value.should=="amenity_1_1_6"
  end
end

To take this further I can print the contents of my class file to generate my locators from this spec.

puts "@#{attribute_name} = \"#{attribute_value}\""

I use bundler 0.9.6 with all of my projects. Checkout my earlier post on how to add bundled gems to your project.  I am excited about the updated bundler support in Rubymine 2.0.2 for a few reasons.

Rubymine can easily find my bundled gems in my Gemfile and attach gems to my project with suggestions. Now I can Go To Declaration for all of my attached gems.

Attech gems from the Gemfile

I don’t have to navigate to my project through the terminal to unlock the Gemfile and add a new gem. I can use the bundler menu for all bundler functions.

Bundler 9.0 options

I can get feedback from my bundler commands inside the Rubymine run window.

Unlocking through rubymine

I am impressed with the new logging and stability features of  Selenium-Client 1.2.18. One of the most useful improvements is the automatic failure when a page is showing service unavailable.

I used to put a line like this after every browser.open command.

browser.text?("Http/1.1 Service Unavailable").should(be_false)

Now I get a nice error message from the Selenium-Client.
Selenium::CommandError: Response_Code = 503 Error_Message = Service Unavailable

I have been using the gem bundler lately with my ruby projects to deploy to team city build agents.  My project needed to run on multiple OS X and Cent OS systems without having to manage ruby gems manually.  The bundler gem repository is located on github at http://github.com/carlhuda/bundler.  There are a few different ways to use it so I decided to write about how use it.  The short answer is that I lock and pack the gems I need, and fall back to sudo installed gems when needed. I am using bundler 0.9.6. The prerequisite is that you gem install bundler.

Use a begin / rescue to try to require the environment file.
If the file is not found then execute the bundle install command

begin
  # Try to require the preresolved locked set of gems.
  require File.dirname(__FILE__) + "/../.bundle/environment"
rescue Exception=>e
  # Fall back on doing an unlocked resolve at runtime.
  if (!system("bundle install"))
        puts $?
  end
end
require File.dirname(__FILE__) + "/../.bundle/environment"

On the first run the bundle install command will execute
This operation will create a .bundle directory in the root of your project that will contain an environment.rb file that will contain all of your gem configuration information.
All your files need to do, is include this environment file and they will have access to your bundled gems. If you run the bundle install command twice in a row it will not install again if your gems have already been installed.

Three step process to setup your bundler.
Create a Gemfile that will list what gems to install. Lock your gem file and then pack your gems.
Check out the information on github to see the format of the gemfile. You can use the bundle init command to create the Gemfile for you in the base of your project. Now run the bundle lock command. A Gemfile.lock file will be created you will check this in to your version control system. Next you need to pack the gems. Run the command bundle pack. You will see that a vendor/cache directory will be created. You will need to check in the vendor cache directory into your VCS. Now when your deploy your code will self execute the bundle install command making all of the gems you need available too your project.

Working with compiled gems
I really like using the Nokogri gem but I foud out that when I execute my project with in the Team City execution environment with a Cent OS build agent, that Nokogiri cannot build correctly because it cannot access the system files it needs. The work around is to list nokogiri in your Gemfile, but don’t check in the gem into the gem cache. Install it as the user that will be executing your scripts so it will be available to your execution environment. The bundler will resolve the dependency to your system installed gem.

That was a quick rundown of what you can get using the bundler and how to set it up. I will continue to post updates as I come across more challenging situations. For now I still have to touch every machine that needs odbc ruby configuration and nokogiri.

I like to navigate through Rubymine using only keyboard shortcuts if I can. One of the most useful techniques I have found is mapping a keyboard shortcut to enable you to run the test that your cursor is on. This way you can navigate up and down tests using ctl + up or down and then run the test your interested in with your key stroke. Even though I am on a Mac I prefer running a test to be mapped to control + t. The default configuration for this is command + alt + F8. The keymap configuration is under the label Run context configuration, which is under the All Actions -> Other section in the Keymap configuration menu.

The other shortcut that goes well with this one is Go To Test. I can put my cursor on a class name and press alt+shift+t and I can view the tests implementation for this class. Then I can use the short cut I created ctrl + t to run a test I am interested in.