We’re constantly looking for new technologies to ease the pain of our developers. Our heritage is in Java+Spring with small exceptions. Java8 and especially Spring Boot bring some most needed updates and changes to typically monolithic Java applications. And when you have proper api’s you just need a nice frontend framework to replace jsp’s and tons of jQuery: in our case we chose AngularJS. The first project using Angular was done almost two years ago and currently all projects are done using it.
Over 10 years of Java leaves a mark on your soul
For couple a years I have been looking for something better. Best thing in this profession is that you have great variety of choices. We have built couple projects with NodeJS and have learned how horrible Ruby ecosystem can be with Chef. Of course we have some Scala, but with other languages we have done only some experiments: Yes, I’m talking about you Clojure, Haskell and Rust. Then we have Go (golang). Which we’ve used only on couple of smallish services, but I’m really impressed about language, standard library, tooling and community around it. There’s a great deal of blog posts explaining why different companies have chosen Golang, this isn’t one of those posts, I might write that some day. Meanwhile get familiar with Golang with the interactive A tour of Go if you learn by coding, Effective Go if you prefer reading or A tour of Go 34:40 if video’s is you thing.
That was quite long intro to the actual topic. With all programming languages some unit testing for coding is needed, others go all the way with TDD and target for 100% test coverage. Dynamic languages require more testing for types. When your application is a decent sized you end up with hundreds of tests. Then comes the pain, depending on the language, your test runs start to take more time: What used to be couple of seconds will become minutes and in worst case tens of minutes. Team starts to run only unit tests and let CI handle integration tests. So you start mocking repository (database) calls. And build preload & cleanup methods for development database for integration tests. From time to time builds might fail for integration tests due to timeouts or just because two builds are running parallel with the same database.
Testing with Golang and Docker
Golang will not make exception with this, but with support of Golang’s exceptionally fast build & test cycles and some Docker magic you can start MongoDB Docker container and run all tests with it in seconds! That really is seconds from start to finish, with exception for first run that will download and provision MongoDB Docker container.
I got actual inspiration from this tweet and ever since have been looking for excuse to check if this is true:
Let’s build something nice where we can experiment with Docker
I’ve been playing with my spare time toy Golang+AngularJS project Inkblot for some time and now was perfect time to checkout if Docker is as magical as it has been hyped to be. There’s a small annoyance for OS X user when it comes to Docker: it runs on Linux only. Yes, you can install it to OS X with Boot2Docker that actually runs Docker daemon on virtualized (Virtualbox) Linux. I was using Vagrant with Ubuntu already as development environment for Inkblot so I just installed Docker on that.
First I got familiar with Camlistore implementation and copied it with pride, special thanks goes to Brad Fitzpatrick, you have done exceptional work with Camlistore and Golang standard library. Thanks!
The actual test can be found here story_test.go. For those who can’t read Golang, I added extra comments to most intresting parts of the code.
Seeing is believing
So it starts MongoDB Docker container, configures it to application then creates http server with built in testing support. Then we setup same routes to server that actual server has and run two request against test server, first creates story comment and another tries to get it. All data is stored and fetched from MongoDB. And how long all of this takes time?
Docker is for all developers not just Golang users
And for those that are not that fortune that they can use Golang Docker can help all you too. It wouldn’t be so super fast as with Golang but as fast as using external MongoDB server without extra cleanup hassle. No doubt that Docker is game changer in virtualization business and all hype that it has is well earned. There’s just no excuses to write any mock tests for MongoDB functions.