Code Count: Ruby on Rails vs. C#/ASP.NET
May 28, 2009 11:24

I'm going to start with this table so it's all up front.

(Widen your window if the CSS float sidebar is overlapping the content)

  C#/ASP.NETRuby on RailsRatioC#/ASP.NETRuby on RailsRatio

When I started working in Ruby and Rails back in 2005, I was mostly a Java and Struts developer. I had spent years messing with XML and type-casting and fighting logic:equal tags and building Iterators - not to mention hand-crafting my own pre-Hibernate object-relational database layers. Using Ruby on Rails felt like someone had opened a window and let in sunlight and fresh air. Things that would take dozens of lines of Java were reduced to a single line of Ruby, and entire application layers just disappeared.

I had always wondered exactly how much more efficient Ruby on Rails could be over Java and Struts, but I never had a direct translation of an application to work with.

Last fall I worked on just such a project, converting ClearFit's application from a crufty old .NET/C# application to Ruby on Rails. We kept the same screens and the same functionality - just cleaned things up and translated everything into Rails. This was a perfect opportunity to get some real numbers on how much more efficient Ruby on Rails can be compared to a Java / C# application.

I gave a presentation in early May and this was one of the more popular slides:

So, in general, a big Rails application will get the same thing done as a big C# app in roughly one fifth of the amount of code.

Since there are lies, damned lies, and statistics, here are some mitigating factors (in both directions):

  • 'Display' only counted the contents of .aspx or .html.erb pages
  • The 'display' amounts for Rails would have been less if we had used Haml, but we decided not to use it, since it would be easier to just copy-and-paste the HTML for the pages
  • The C#/ASP.NET application was fairly old and didn't use an object-relational library like NHibernate
  • Likewise, we didn't do very much 'leading edge' Rails for this project either - most of it could have been written in 2006
  • The 'Non-Display' code includes all .cs and .rb files in the projects, but doesn't include libraries, such as .dlls or items in the vendor/plugin directories, or migrations
  • Some of the C# code was very badly written, and some functions were much larger than they probably needed to be.
  • On the other hand, we didn't spend a lot of time making the Ruby code cleaner or more efficient than it was already
  • There were very few tests in the C# code, and limited automatic tests in the Ruby code (mea culpa mea maxima culpa), however, the Ruby total includes admin tools and an interactive 'testing' interface.
  • The Ruby version of the application has 56 tables in the database, while the C# version had 250 tables - but most of these 250 tables weren't referenced in the code.
If anyone is curious, the complete translation project took roughly 350 hours of development time and took from mid-October 2008 to early January 2009. It was mostly done by me, with some key help from Paul Doerwald. For more details, see the detailed presentation notes, including slides and some video clips.

A Brief Note on Twitter
May 26, 2009 08:53