“Guest” commented about my Agile 2009 tutorial, Integration Tests Are A Scam. “Guest” wrote this:
Thank you for posting this! This will help me sell Collaboration and Contract tests to my organization.
I prefer the format of the most recent post. In the older post I get the sense the solution described won't scale.
Your comment intrigues me. Can you articulate the difference so that I can understand it?
The Mars Rover is meant to be deployed on Mars. The variables which ought to be considered in an integration test may include atmosphere of mainly carbon dioxide at 1/100th the pressure of Earth's, temperature ranges from -80C to +20C, and last but not least a gravity of 0.37g.
An argument against integration testing a Mars rover may include the fact that *it couldn't be meaningfully done on this freaking planet*.
A very interesting point. Even so, we could simulate all those factors on Earth, if the physicists could be sure that we could safely ignore everything else.
Post Script: Gary Bernhardt moved Ruby one step closer to supporting the syntax I used in this article for declaring object collaborators. https://github.com/garybern...
The link to Ash Moran's blog entry is incorrect. The correct link is: http://blog.patchspace.co.u...
I would probably introduce just one or two integrated tests to ensure the whole system is properly initialized (e.g. configuration, IoC containers, etc.); maybe some simple ones, which satisfy some acceptance criteria. I think this is also in line with "Integrated Tests are a Scam", but I haven't watched it for a while now.
What do you think, @jbrains?
Yes. I would like at least one Smoke Test that verifies that a container can inject every interface that it tries to. I would probably also want some way to check that any configuration is at least syntactically valid, much like `httpd -t` does for Apache. Of course, I'll have some integrated tests, but when I wrote this article, I wanted to explore the claim "Only an integrated test would have found the problem."