UPDATE: For a better approach, check out MoQ Sequences Revisited.

One area where using MoQ is confusing is when mocking successive calls to the same method of an object.

For example, I was writing some tests for legacy code where I needed to fake out multiple calls to a data reader. You remember data readers, don’t you?

Here’s a snippet of the code I was testing. Ignore the map method and focus on the call to reader.Read.

while(reader.Read()) {
  yield return map(reader);
}

Notice that there are multiple calls to reader.Read. The first couple times, I wanted Read to return true. The last time, it should return false. And here’s the code I hoped to write to fake this using MoQ:

reader.Setup(r => r.Read()).Returns(true);
reader.Setup(r => r.Read()).Returns(true);
reader.Setup(r => r.Read()).Returns(false);

Unfortunately, MoQ doesn’t work that way. The last call wins and nullifies the previous two calls. Fortunately, there are many overloads of the Returns method, some of which accept functions used to return the value when the method is called.

That’s the approach I found on Matt Hamilton’s blog post (Mad Props indeed!) where he describes his clever solution to this issue involving a Queue:

var pq = new Queue<IDbDataParameter>(new[]
    { 
        mockParam1.Object, 
        mockParam2.Object 
    });
mockCommand.Expect(c => c.CreateParameter()).Returns(() => pq.Dequeue());

Each time the method is called, it will return the next value in the queue.

One cool thing I stumbled on is that the syntax can be made even cleaner and more succinct by passing in a method group. Here’s my MoQ code for the original IDataReader issue I mentioned above.

var reader = new Mock<IDataReader>();
reader.Setup(r => r.Read())
  .Returns(new Queue<bool>(new[] { true, true, false }).Dequeue);

I’m defining a Queue inline and then passing what is effectively a pointer to its Dequeue method. Notice the lack of parentheses at the end of Dequeue which is how you can tell that I’m passing the method itself and not the result of the method.

Using this apporach, MoQ will call Dequeue each time it calls r.Read() grabbing the next value from the queue. Thanks to Matt for posting his solution! This is a great technique for dealing with sequences using MoQ.

UPDATE: There’s a great discussion in the comments to this post. Fredrik Kalseth proposed an extension method to make this pattern even simpler to apply and much more understandable. Why didn’t I think of this?! Here’s the extension method he proposed (but renamed to the name that Matt proposed because I like it better).

public static class MoqExtensions
{
  public static void ReturnsInOrder<T, TResult>(this ISetup<T, TResult> setup, 
    params TResult[] results) where T : class  {
    setup.Returns(new Queue<TResult>(results).Dequeue);
  }
}

Now with this extension method, I can rewrite my above test to be even more readable.

var reader = new Mock<IDataReader>();
reader.Setup(r => r.Read()).ReturnsInOrder(true, true, false);

In the words of Borat, Very Nice!