End of Line
(yet another) programming blog

Specnaz 1.3 released!

At what point I'll get sick of writing about Specnaz? Hard to say, but this third post in a row will be the last one in the series - the next one won't be about Specnaz, I promise.

Version 1.3 is a major release, bringing with it two long-awaited features to the library: JUnit Rules support, and the capability to define parametrized (sometimes also called data-driven) tests. Let's dive right in.

JUnit Rules support

JUnit Rules is a way of wrapping the execution of JUnit tests with your own code. You can add custom code before and/or after the test executes, fail tests that would otherwise pass, or the opposite - make failing ones pass under some conditions. The API allows you to package your logic into a class that can then be re-used between many tests.

Over the years, a large ecosystem of third-party Rules has emerged that allows you to easily integrate your tests with libraries like Mockito, Spring or Dropwizard. Because using these third-party Rules is so convenient, it was important that Specnaz support them as well.

Well, I'm happy to say that, with the release of version 1.3, it now does. Here's an example of integrating with Mockito:

public class MockitoExampleSpec extends SpecnazJUnit {
    public Rule<MockitoRule> mockitoRule = Rule.of(
    	() -> MockitoJUnit.rule());

    private List<Integer> listMock;

        describes("Using the JUnit Mockito Rule in Specnaz", it -> {
            it.should("initialize fields annotated with @Mock", () -> {
                when(listMock.get(0)).thenReturn(400 + 56);

                assertEquals(456, (int) listMock.get(0));

Check out the documentation for that feature here, as there are some subtle differences in the way the Rules work in 'vanilla' JUnit and in Specnaz (mainly related to the different object lifecycle in the two libraries).

Parametrized tests

Parametrized (sometimes also called data-driven) tests are an important way of reducing duplication in tests. You define the test body once, but instead of hard-coding the inputs and outputs, you make them parameters of the test. Afterwards, you specify with what parameters you want that test to execute, and that will result in a separate test case for every set of parameters you provided.

JUnit ships with a Parametrized Runner that allows you to define data-driven tests. However, the experience that that class offers leaves a few things to be desired - so many, in fact, that there exists a separate library just for that functionality, JUnitParams (created, incidentally, by the CEO of the company I worked at before Amazon, Pragmatists).

The lambda-based structure of tests that Specnaz uses lends itself perfectly to defining parametrized tests. Instead of passing a no-argument lambda to the should method, you can now pass a lambda expecting between one and nine arguments to should, and use those arguments as parameters inside the test body. To specify the values for the parameters, you need to call the provided method on the object that should returns. As arguments to provided, you supply instances of the ParamsX class, where X is the arity of the lambda you gave to should (so, if you called should with a 3-argument lambda, you need to supply instances of the Params3 class).

Each of the ParamsX classes has a static factory method called pX (so, p2 for Params2, p3 for Params3, etc.), which, combined with Java's static imports, allows you concisely create instances of them. A separate test case will be executed and reported for each instance of the appropriate ParamsX class you supply to provided.

Putting all of the above together, we get something that looks like this:

import org.specnaz.params.junit.SpecnazParamsJUnit;
import static org.specnaz.params.Params2.p2;

public class FibonacciSpec extends SpecnazParamsJUnit {{
    describes("Fibonacci spec", it -> {
        it.should("show Fib(%1) = %2", (Integer input, Integer result) -> {
            p2(0, 0), p2(1, 1), p2(2, 1), p2(3, 2), p2(4, 3),
            p2(5, 5), p2(6, 8), p2(7, 13), p2(8, 21)

A couple of things to note about the above code:

  • It's fairly concise - at least more concise than the JUnit equivalent. The parameters are also per-test, instead of being global to the entire class.
  • It's completely type-safe (also unlike the JUnit equivalent) - if you try to change the integer 21 to a String, for example, the code will no longer compile. And, thanks to Java's type inference, the types only need to be specified once - on the parameters of the lambda passed to should.
  • We used the special placeholders %1 and %2 in the test description. These will be replaced at runtime by the values of the parameters at the appropriate index, counting from one - so, the last test above will be reported with the description should show Fib(8) = 21.
  • provided is overloaded, accepting either variadic arguments as above, or a Collection of the appropriate ParamsX instances, which is more useful when generating the parameters programmatically (by reading a file, for example).

In addition to parametrized should (and shouldThrow), you can also define parametrized describes, this way creating an entire parametrized sub-tree of tests.

Detailed documentation about this feature can be found here.

Third time's the charm

So, these are all of the new features in the 1.3 release of Specnaz. I think they really take the library to the next level in terms of making it more expressive and powerful. I encourage you to give Specnaz a shot, and, if you do, I would love to hear back from you about your experiences using it!