Will actually compile into a nested class which implements IEnumerable<int> and the body of GetInts() will return an instance of that class.

使用反射，你可以看到：

Using reflector you can see:

public IEnumerable<int> GetInts()
{
    <GetInts>d__6d d__d = new <GetInts>d__6d(-2);
    d__d.<>4__this = this;
    return d__d;
}

修改 - 增加更多的信息关于 GetInts 实施： 这种实现使得偷懒的方法是根据枚举的 的MoveNext（） 方法。当产生枚举嵌套类（＆LT; GetInts＆GT; d__6d 中的例子），它有一个状态，每个状态的值连接（这是一个简单的例子，在更先进的情况下的价值进行评估时，code达到的状态）。如果我们采取一看的MoveNext（） $ C $的C ＆LT; GetInts＆GT; d__6d 我们拭目以待状态：

Edit - adding more info about GetInts implementation: The way this implementation makes it lazy is based on the Enumerator MoveNext() method. When the enumerable nested class is generated (<GetInts>d__6d in the example), it has a state and to each state a value is connected (this is a simple case, in more advanced cases the value will be evaluated when the code reach the state). If we take a look in the MoveNext() code of <GetInts>d__6d we'll see the state:

private bool MoveNext()
{
    switch (this.<>1__state)
    {
        case 0:
            this.<>1__state = -1;
            this.<>2__current = 1;
            this.<>1__state = 1;
            return true;

        case 1:
            this.<>1__state = -1;
            this.<>2__current = 2;
            this.<>1__state = 2;
            return true;

        case 2:
            this.<>1__state = -1;
            this.<>2__current = 3;
            this.<>1__state = 3;
            return true;

        case 3:
            this.<>1__state = -1;
            break;
    }
    return false;
}

当枚举被要求为当前对象它返回被连接到当前状态的对象。

When the enumerator is asked for the current object it returns the object which is connected to the current state.

为了表明code只计算在需要时，你可以看看这个例子：

In order to show that the code is evaluated only when it's required you can look at this example:

[TestFixture]
public class YieldExample
{
    private int flag = 0;
    public IEnumerable<int> GetInts()
    {
        yield return 1;
        flag = 1;
        yield return 2;
        flag = 2;
        yield return 3;
        flag = 3;
    }

    [Test]
    public void Test()
    {
        int expectedFlag = 0;
        foreach (var i in GetInts())
        {
            Assert.That(flag, Is.EqualTo(expectedFlag));
            expectedFlag++;
        }

        Assert.That(flag, Is.EqualTo(expectedFlag));
    }
}

我希望它更清楚一点。我建议看看在code与反射，并观察编译code为你改变收益code。

I hope it's a bit more clear. I recommend to take a look at the code with Reflector and observe the compiled code as you change the "yield" code.