Property based testing in JavaScript with fast-check

Property testing is a way to test functionality by automatically generating many different inputs.

We’re going to use the fast-check library

yarn add fast-check -D

Let’s create a rather contrived example function to test

export const divide = (a: number, b: number): number => {
  if(b === 0) {
    throw new Error("Denominator cannot be 0");
  }
  return a / b;
}

As you can see, this function simply divides a by b but will fails in the denominator is 0. This was done on purpose to ensure that there is a failing test available for a specific input value (I did say it was contrived).

Now to test this we might implement the following test

import * as fc from 'fast-check';

test("Divide property test", () => {
   fc.assert(
      fc.property(fc.nat(), fc.nat(), (a, b) => {
         return divide(a, b) === a / b; 
      })
  );
});

In this instance we create the assert which wraps the property generation code, i.e. the call to property, which itself generates two natural numbers properties as input to the lambda/fat arrow function.

We then call the code that we wish to test, passing in the generated values and then returning whether the expectation was true or false.

We can use Jest or the likes in place of the return divide(a, b) === a / b, i.e.
replacing it with if we prefer and just use fast-check to generate the properties

expect(divide(a, b)).toEqual(a / b);

Now the only problem with the property based testing like this is that it will randomly generate values as opposed to inspecting the flow of your code to generate values (like tools such as Pex for .NET). Hence these values may not always include the edge cases that we might not have coded for.

For example, executing this test in Jest I find that sometimes it will pass a 0 as a denominator and other times it will not, therefore we do not always get a clear idea of all possible inputs, but it’s a great starting point for proving your functions.

In the test we generated natural numbers, we can also generate restricted ranges of values, along with integers, floating point, doubles, strings and more, see Arbitraries.