In C++, when we see the following code:

z = x + y();

We know that the value of z is immediately computed when the control reaches past the statement z = x + y(). In fact, the act of computing the sum involves evaluating the expressions x and y() themselves. Here, y is presumably a function or a functor instance, so the call to y() will in turn trigger more evaluations. Irrespective of whether z is ever used for anything later, its value would still be computed. This is the model of eager evaluation that a lot of programming languages follow. The actual story is slightly more complex because compilers can reorder and optimize away computations but there is little control the programmer has on the process.

What if we could defer the evaluation of such expressions and any of their sub...