Last summer, I took over a troubled Ruby on Rails project; the code was a mess, overly complex and bug-ridden, lack of meaningful unit tests, and an unhappy client.
I needed integration tests. After trying and failing with Cucumber, I switched to Selenium IDE.
A year later, I'm very happy with Selenium. The project is turning around and I sense the client is warming up to me.
The existing codebase, and the fact that I'm charging by the hour on a project who's budget is 3x initial estimates.
The site was, and still is, stuck at Rails 2.1.1. And, at the time, I was not using RVM or Bundler. Meanwhile, I was putting out fires on a daily base on a live system.
With Cucumber, I was unable to run even a simple feature against the codebase. I'd get an exception, fix it somehow, then get another. After 20 minutes of that, I started looking for other options.
And on top of that, I did not know the codebase: brand new project.
The quickest path to stable ground was Selenium IDE. Just hit the record button and start writing integration tests. I quickly gained enough confidence to change the underlying code. And I was working within a budget of only a couple thousand dollars a month.
What I've grown to love about Selenium, as opposed to Cucumber, is its complete separation from the site it's testing. An integration testing framework can take a long time to master; I'd rather 1) minimize that time, 2) learn a tool that can be applied to as wide a variety of projects as possible.
