Kieran Hunt

Why you should use an assertion library

✦ 2023-03-01

I love assertion libraries. But I find I have to frequently explain just why they’re so great. So I wrote this post that, hopefully, summarises why I think you should use (and love) them too.

Readability and consistency

Assertions, by standardising on how to validate state, usually make it very clear what state they’re validating.

Without assertions, to validate whether a collection contains an item, you might write one of the following:

val list = listOf("foo", "bar", "baz")

assertTrue(list.any { it === "bar" })

assertEquals(1, list.filter { it === "bar" }.size)

assertEquals(listOf("bar"), list.filter { it === "bar" })

Some of those may be a little questionable, but they are all variations using JUnit’s assertion statements. They also all correctly validate the input. But the poor reader of that test needs to first mentally parse what transformations the code is performing before they can understand what the test is asserting. And the writer of the test has had to spare precious brain cycles thinking how exactly they can transform the input data a format that matches the assertion.

Compare that to a fluent assertion library. With AssertJ, this becomes:

val list = listOf("foo", "bar", "baz")

assertThat(list).contains("bar")

To someone reading this test, it is now immediately obvious what this test is validating. To someone writing this test, how to perform this validation is immediately clear, too. Less bikeshedding come review time.

Debugging

When an assertion like assertTrue(list.any { it === "bar" }) fails, it produces an failure message like:

org.opentest4j.AssertionFailedError: 
Expected :true
Actual   :false

Now sure, that tells you that the test failed, and sure you can infer that list must not contain bar, but you’re left stratching your head about what the list does contain. To find out, you’ll need to stick a breakpoint on the test line and inspect the value of list before it enters the failing assertion. But you could save precious seconds from every failed test if the assertion library just printed out what the list did contain.

Now compare that behaviour with AssertJ’s output:

java.lang.AssertionError: 
Expecting:
 <["foo", "bar", "baz"]>
to contain:
 <["qux"]>
but could not find:
 <["qux"]>

With output like that, the wily tester can get straight to fixing the bug, rather than messing around with breakpoints. A faster feedback loop means quicker validation and features out the door sooner.

In summary

I don’t really see a need for any of JUnit’s assertions. Sticking to using something like AssertJ’s assertThat will simplify your tests, make them easier to read, and keep them plain to write.