IT Really Matters

Ruby is a great programming language, but unfortunately it does have some problems when using on Windows. One of it’s biggest drawbacks is it’s slowness when loading files. This is also slower than it ought to be on Unix platforms, but not as slow as on Windows. Thankfully there is some work going on to make it faster in future versions. It’s already possible to make it faster yourself!

Benchmarks

I will conduct two types of benchmarks. Firstly i’m gonna create an empty Rails 3 project and see how much time would it boot up. Secondly there is a project called measurements which allows to perform some operations on your Ruby and one of these are benchmarks for loading files. Let’s see how it goes for Ruby 1.9.2, 1.9.3 from RubyInstaller and patched 1.9.3.

These benchmarks show that Ruby 1.9.2 is really slow and patched 1.9.3 got about 50% performance boost compared to regular 1.9.3. That’s something to be happy about! I’m using average laptop PC which means that if you have a more decent hardware then the results might be very different from me.

Prerequisites

The following components are needed to speedup your Ruby:

• Ruby 1.9.3 from RubyInstaller
• Devkit
• Git

If you’d like to skip all that hassle of building all the things yourself then you can download already prebuilt patched Ruby versions too! Read the “Faster Way” part about that below.

Get Fenix

Fenix is a Ruby extension written by Luis Lavena. Luis is a really helpful and great guy, at least when it comes to Ruby on Windows. He is part of the Ruby core team, is the main man behind RubyInstaller and has created many nice libraries like sqlite3-ruby, rake-compiler and many others.

Back to Fenix. It changes File.expand_path method to be faster. All the 50% speedup seen from the benchmarks were coming from this change since this method is called more than once when loading files. Pretty impressive or rather sad bottleneck in Ruby. Get the Fenix extension, compile and benchmark it:

Not bad benchmark results.

Patching & Compiling Ruby

Unfortunately only File.expand_path method is faster when using this extension, but there’s Ruby’s require and load methods, which also execute expand_path, but they will do so by using internal C expand_path function instead. Solution for that problem is to patch Ruby code to use Fenix.expand_path internally also! First step would be to clone Ruby itself (make sure that line endings are not converted by Git):

Apply the patch created by Luis:

`git status` shows us that file.c and load.c files are modified. It’s time to start compiling Ruby itself. Best way to do that would be to use RubyInstaller itself (how ironic, we’re using Ruby to build Ruby). Make sure to disable ANSICON too since it is known to cause problems during the compilation process:

This takes some time and if the process raises any errors, just try again. If there’s still errors don’t hesitate to write to RubyInstaller mailing list or contact Luis directly. He’s very helpful, as i already stated above.

After the process is done “install” new Ruby with Fenix and make sure that it’s used by setting environment variables too:

Make sure that the PATH and RUBYOPT environment variables are set permanently. And try it out if it’s faster for you too!

Faster Way

In case you don’t want to spend your time cloning, compiling and testing then there is a faster way to boost your Ruby. There exists a project called The Code Shop, which tries to solve the Ruby performance problem on Windows. It makes different builds available to download, each experimenting with different set of patches. I can recommend tcs-ruby193_require_winio_fenix-20111113 because it seems to be the fastest. Here are some benchmark results:

Fenix is already precompiled into these builds. You still need to compile and install Fenix and set PATH and RUBYOPT environments as described above.

In Conclusion

The main point of this post is that things aren’t always as rosy as they could be, but it’s possible to make them better. It is really valuable that you give feedback to projects like The Code Shop (e.g. what are your benchmark results) so the final result could get better in the future without any additional hassle for everyone. I hope that you can now start your engines much faster!

I mentioned page objects in the post about WatirSplash 2. In this post i’m gonna explain what these things exactly are and how to use them.

What Is A Page?

When it comes to automated testing then the page is really nothing else than the page user sees when visiting the website. Any meaningful page consists of different components some of which just show information (e.g. text) others allow interaction (e.g. text fields and buttons). Different pages might also have some shared content like header, search field and footer. One important thing about components is that the result of some interactions is another page.

Let me bring you a concrete example. Let’s look at the Bing’s main page (enough of Google in all examples, alright). You can think of the possible sections (Web, Images, Videos, News and More) as a header. There is also the search field and button. Search button is special since clicking on it changes the current page to results page. Makes sense? Let’s see how to convert that information into WatirSplash Page Objects.

Point Of Having Page Objects

The point of having Page Objects is to have better structured code and keep it DRY in your tests too. This all makes your tests easier to maintain. If you’re a software developer then you care about these things when it comes to production code. Why not to use these good practices in tests also? Page Objects make it all possible. Having Page Objects means that it should be easier to change your tests if the application under test (AUT) changes since you need only one place (ideally) to change. Page Objects also make it very clear where to find existing and put the new code.

Let’s Get Rockin’ With Pages And Objects

Let’s get more concrete. I assume that you have created a WatirSplash 2 project already. If not then follow the instructions in the post about WatirSplash 2 or read everything you need to know straight from the README. Let’s create the Main page first:

This command created us the Main page object and a test for it. The generated file looks like this:

And the test itself is nothing fancy yet:

Let’s add something meaningful to our Main page object like search field and button:

You can use all the familiar Watir methods directly in your Page objects! For the sake of clarity let’s also add Header functionality. To do that we’d just use Ruby’s mixin functionality - the Header module:

And make sure to include it into your page object:

We can now update our main_spec to actually test something:

I made two tests - one for the header and another for the search functionality. Since header should be on every page, i could have taken advantage of the RSpec’s Shared Examples feature, but for the sake of this post’s clarity i didn’t.

The second test is a failing test since there’s no way that Watir’s Button#click will return us a results page object. But hey, i’ve just done a little TDD here! Let’s make it work like we want it to work:

I’ve used the WatirSplash::Page::Base#modify method which takes some object as it’s first argument (here button) and a Hash as a second argument. Hash consists of method name as a key and value as Proc object. In this particular case the code above overrides this particular button’s #click method with my #click, which returns Results page object due to the usage of #redirect_to method. Since #click is already existing method on that button, then it will call the original method (e.g. performs the actual clicking) before calling the defined block. The #modify method does all that by using Ruby’s meta-programming. You can always look at the source itself if you’re more interested in the nitty-gritty details.

Let’s create ourselves a Results page object with the results method. This time, let’s be even cooler and generate more code than we did with the Main page object:

This generated us the Results class with the results method:

Let’s add the #count and #[] methods to our #results object by using #modify again so we could chain method invocations naturally as i’ve written in the search functionality test:

I’m invoking the #results method again to get the Watir::Ul element to execute regular Watir methods on it - in these cases #li and #lis. This neat trick hides the implementation details from the tests and makes the tests even more natural to read.

In Conclusion

That’s it - we have created a few tests using Page Objects functionality in WatirSplash 2! Using Page Objects makes your tests (hopefully) more readable and easier to maintain. I like my implementation when compared to others because it makes your page objects to look really succinct and doesn’t add any additional complex API to learn - you can use already learnt Watir methods as you did before. And the code written to support creating and using these Page Objects is also quite clever and short. I love it! I hope you will too.