RANDALL DEGGES


postgression - A PostgreSQL Database for Every Test Case

Monkey Sketch

The following are some facts about me:

  • I write a lot of software.
  • In order for me to write good software, I often write tests for my software. Code that makes sure my business logic works on data from my database. Code that makes sure my web requests return the right thing. You name it, I’ve probably tested it.
  • I do 100% of my development on a Linux laptop.
  • I usually run my tests on my laptop while I’m coding, as well as a remote Jenkins server (and sometimes Travis CI instance) that run all my tests to make sure I didn’t forget to do so locally.
  • I hate running a database server on my laptop, I hate running a database server on my Jenkins instance, and I hate telling my test code how to run tests against my testing databases on all the different types of machines I use.

Over the past year or so, I’ve become really annoyed at having to configure my database in all my environments. I love PostgreSQL (it’s an awesome database), but I can’t stand the idea of running it locally on my laptop, just so I can make some tests work. I also can’t stand the annoyance of having to SSH into my Jenkins server, configure PostgreSQL, and then write code which tells my tests to distinguish between my local PostgreSQL stuff and my Jenkins PostgreSQL stuff–and don’t even get me started on configuring it to work in all 3 environments: locally, on Jenkins, and on Travis. Ugh.

I think what annoys me about this is that setting up a database isn’t hard, I just think it’s stupid to have to remember to do it for each new project. It feels like I’m repeating the same thing over and over again, and each time I do it, I become slightly more annoyed.

So this past week, my buddy Alven and I teamed up to solve this mini-problem for ourselves. The result is our new service, postgression.

postgression is a simple web service, built on top of Heroku’s platform (Don’t know about Heroku yet? Read my book.), that instantly provisions a new PostgreSQL server (PostgreSQL 9.2.3, to be precise) for you to use in your tests.

Here’s how it works: you hit our public facing API (no account required), and we give you back a PostgreSQL database URL that you can use in your application. For example:

$ curl 'http://api.postgression.com'
postgres://username:password@hostname:port/dbname

Simple, right? So now that you’ve got the database, you have your tests run against this database (which is available in Amazon’s US East region, in case you’re curious), and that’s it!

Finally, after 30 minutes, this database will magically disappear.

So, why is this useful? Well, using postgression to generate a database for your tests means that:

  • You can run your tests locally without needing to install PostgreSQL server.
  • You can run your tests locally, remotely (on Jenkins / Travis / etc.) using all the same configuration–no need to do any custom scripting or environment checking.
  • It costs you nothing.

Is this the most useful service in the world? Nope. But I love it, I’ve been using it, and it’s made my testing quite a bit simpler.

Today I’m really happy to open postgression up to the public. Alven and I have created some handy tools on our GitHub page, which make using postgression easier, and we’ve written some basic documentation on the postgression website to help you get started.

If you’ve got any questions, comments, concerns, or otherwise, I’d absolutely love to hear them. I hope you’ll give postgression a try!

Check out postgression here.