What is the "right" way to test the output of Java methods?

Question

In Python, I often have tests which look something like this:

tests = [
    (2, 4),
    (3, 9),
    (10, 100),
]
for (input, expected_output) in tests:
    assert f(input) == expected_output

What is the "right" way to write tests like this (where a set of test cases is specified, then a loop runs each of them) in Java with JUnit?

Thanks!

Preemptive response: I realize I could do something like:

assertEquals(4, f(2))
assertEquals(9, f(3))
....

But... I'm hoping there is a better way.

Answer 1

Wouldn't you just define a simple class with two fields, real result and expected result and then loop over the collection in a similar way to what your Python snippet is doing?

Answer 2

Same thing.

    int[][] tests = {
            {2, 4},
            {3, 9},
            {10, 100}
    };
    for (int[] test : tests) {
        assertEquals(test[1], f(test[0]));
    }

Certainly not as pretty as python but few things are.

You may also want to look into JUnit Theories, a future feature...

Answer 3

There are no tuples in Java, but you could use a Map or two parallel arrays to specify input/output pairs and then do a loop just like your Python example.

Answer 4

Um...

int[][] tests = new int[][]{
        {2, 4},
        {3, 9},
        {10, 100},
    };

for(int[] i : tests)
{
    assertEquals(i[1], f(i[0]);
}

Same thing, really. The only problem is Java's lack of a tuple literal, so for more complex cases, you'll have to use Object[] arrays and cast, or write a Tuple class.

Answer 5

The right way is writing separate assert statements, even if you don't like it.

It avoids unnecessary complications, and when it fails it is sometimes easier to see which value failed (no need to start up the debugger).

However, if you generate your test data automatically it is a different story.

Answer 6

Have a look at the Parameterized test runner in Junit.

http://junit.org/apidocs/org/junit/runners/Parameterized.html

It looks like it will do precisely what you are looking for.

Answer 7

Definitely not an expert on unit testing but i would prefer to have a separate method for each case that I am testing against and use some test running tools (like NUnit-GUI, for C#). That way I would exactly know which case fails, if it does. Its more work to do, but i think it eventually pays off well.