Nikola,

?

In this case, you have two things happening: random strings, and unique strings.

?

Let’s suppose your unique random string generator is called UniqueStringGenerator.

You need to inject the random generator – don’t embed it in the class itself – in this way you can control what is generated when needed (in your test).

?

For example your constructor might say: public UniqueStringGenerator ( IRandomGenerator randomGenerator).

Then you need to create the interface IRandomGenerator. There will be a method something like getRandom() or something similar.

?

In your duplicate name test, create a DuplicateTestRandomGenerator object which implements IRandomGenerator.

In DuplicateRandomTestGenerator just return the same random string from getRandom().

Inject it into the UniqueStringGenerator.

Run your test.

?

Of course, when you want to really use UniqueStringGenerator, you instantiate it with an object which implements IRandomGenerator – but which really generates random strings.

The real RandomGenerator gets tested in other tests.

?

So, two principals:

-????????? Inject things which you need to control them for testing (random number generators or clocks for example)

-????????? Separate responsibilities: the UniqueStringGenerator is only responsible for handing out a unique string (it will check against previously generated strings). The RandomGenerator is only responsible for generating a random string.

?

Hmm! More writing than I expected. I hope it’s clear.

?

John D.

?