Testing has traditionally been the gateway process for Scala development at a number of Java shops looking to evolve their tech stack with minimal disruption or committment. This posts hopes to cover three of the main testing frameworks in the Scala landscape, (Specs2, Scalatest and Scalacheck) with an example of the classic FizzBuzz test, for how they can be used for fun and profit.
So, the brief outline of the FizzBuzz requirement are that given a sequence of numbers:
- Any number divisible by 3 should return the String “Fizz”
- Any number divisible by 5 should return the String “Buzz”
- Any number divisible by both 3 and 5 should return the String “FizzBuzz”
- Otherwise the number should be returned as a String
As a comparison, I’ve also included a sample Java JUnit test for this. The actual implementation code (included here) for this is trivial, (partly by design) as it is just to show illustrate the libraries being used.
My setup for this was using Eclipse v3.7.2 on Windows 7 with the Scala IDE v2.0.1.v-2_09 plugin and Scala version 126.96.36.199.
In a nutshell
- A meta framework for writing unit and acceptance tests by specification. Supports concurrent execution, mocking, reporting, scalatest and scalacheck feature support, datatable style testing and form based specification tests. The propject emerged from the original, (and now deprecated) Scala Specs BDD testing framework.
Yes. A project initiated as Specs (v1) in mid 2007 and later evolving to a more general framework as Specs2 in 2010. The project has a number of comitters (though primarily driven by Eric Torreborre) and supporting resources such as a google group, blog and corporate backing for the project.
Generally the pain points I encountered were with Scala IDE rather than Specs2. Specifically, I had originally wanted to include the different types of tests in the same source file, but differentiated by different subpackages. Unfortunately, Scala IDE didn’t pickup the sub packages so this was a non starter. One minor weirdness was the inability to reorder my arguments for the DataTable spec test. Update: Thanks to Eric Torreborre this was solved using !! instead of ! when a String type is used for the first column in the DataTable !! It was a minor annoyance though.
Otherwise, Specs 2 looks mature, with good tooling support and would likely be me go to choice as a ‘wrapper’ framework for testing. I found the default API for specifications clear and easy to use, and liked the seperation of tests to the expressions evaluated to support or refute them. Nice.
In a nutshell
- A generalised framework for Java and Scala which uses a variety of traits to mixin a number of different test styles and strategies. Out of the box, TDD, BDD, functional, integration TestNG and JUnit tests are supported.
Yes. This is a second generation framework spawned from a project called SuiteRunner by Bill Venners in 2001. There are actively maintained forum groups accessible via: http://www.scalatest.org/community
I only used one spec type and it seemed both clear and concise. The language dictated by this API (describe.. it..) was straightforward it not slightly awkward at first site. I also prefer a clearer distinction between my tests and assertions, (which is a matter of taste). Still, it’d be hard to find a more straightforward and ‘friendly’ framework to get a Java team up and running with first time around.
In a nutshell
- A specification based test generation framework for Scala and Java. The library was originally inspired by QuickCheck in Haskell.
Yes, but primarily stemming from Rickard Nilsson. There doesn’t appear to be any forum support for this project. Update: Thanks to @Daniel Sobral there is (in fact) a google group in support of ScalaCheck here !
Compared to the other frameworks, this one took me a while to get running with, (bear in mind, I was running with the others in minutes though !). The main sources of my pain were:
- No obvious out of the box JUnit support
- I originally ran my tests as a Scala App (i.e. something that extended App), though kept getting
from the ConsoleReporter whenever I tried to run my tests.
- Whenever I tried to add too many conditions to my implication operator, I found not enough tests were generated to validate my property, (hence the somewhat cludgy example I present).
Having said all that, I still find this framework unique and really valuable. On a few test runs I logged the inputs generated and Scalacheck really is great, (a fair span of inputs generated and passed through). This was probably the most challenging framework of the three, but also offered the sort of features I’d certainly want to use again and again. Likely next time I’d run this via Specs though !
In conclusion all three frameworks are much clearer, (or expressive) than comparative frameworks that I have personally used in Java previously. If I was trying to get a team to ‘evovle’ to Scala, I’d choose ScalaTest. If I was working with a team who were slightly more comfortable with Scala I’d go the Specs2 route. In any event I’d try to use ScalaCheck for test generation, (ideally abstracted away through one of the other frameworks though). Either way I can see why using Scala test frameworks helps to sway opinion towards teams progressing towards Scala in the enterprise.
I hope this feedback was useful and not too dry. As ever, the best ‘validation’ is in walking the path yourself (unless ScalaCheck will do it for you !).
Good luck with your TDD and Happy hacking !