Java testing divide by zero

Question

I am trying to test the functionality of dividing by zero.

If I just do: System.out.println(5/0), I will get java.lang.ArithmeticException

I am trying to test that the exception was thrown.

Here's my unit test:

 @Test(expected = ArithmeticException.class)
    public void given_divideByZeroInput_when_divideByZero_then_verify_throwing_exception() {
        MathClass sut = new MathClass();
        sut.div(10,0);
    }

and my div is this simple:

 public float div(int x, int y) {
        return (float) x/y;
    }

However, the unit test is failing, stating:

java.lang.AssertionError: Expected exception: java.lang.ArithmeticException

What did I do wrong?

I know there are many ways to test the thrown exception, but I am trying to stick with the simple @Test(excepted = .. )

You're casting x to a float. Thus you're not doing integer division as you're doing in your first snippet, but float division. And the result is thus `Infinity`, not an exception. — JB Nizet, Sep 21 '19 at 07:51
Rather than use `@expected` try using a `try catch` in the test method to catch the expected exception. If that works I would suggest using the `ExpectedException` rule in junit 4 or its junit5 equivalent — Gavin, Sep 21 '19 at 07:52
@JBNizet oh yeah, you're right, removing (float) solved it, thanks — Jack Thomson, Sep 21 '19 at 07:53
Your code isn’t throwing an exception but the runtime is, so what are you testing here exactly? Not much point in writing unit tests for the java vm. — Joakim Danielson, Sep 21 '19 at 07:55
If you don't cast your int to float, you will get desired exception. Check https://stackoverflow.com/questions/12954193/why-does-division-by-zero-with-floating-point-or-double-precision-numbers-not — Sulthan, Sep 21 '19 at 08:00

Stephen C · Accepted Answer · 2019-09-21T10:51:15.350

JB Nizet has found the answer:

You're casting x to a float. Thus you're not doing integer division as you're doing in your first snippet, but float division. And the result is thus Infinity, not an exception.

But there is something more to explain:

    return (float) x/y;

means

convert x to a float
perform the ((float) x) / y computation using floating point arithmetic
return the value

When you remove the cast:

    return x/y;

means

perform the x/y computation using integer arithmetic
convert the value to a float (implicitly!)
return the value

So there are actually a number of (potential) conceptual errors that might have lead to the mistake that you made.

A type cast has a higher precedence than /.
An arithmetic operation involving an int and a float is performed using 32 bit floating point arithmetic. (The int operand will be widened to a float)
Widening also happens in an "assignment context", and a return statement is an assignment context.

Java testing divide by zero

1 Answers1