Suppose I have the following graph:

What will programming languages look like one hundred years from now? Where will all of those wasted cycles end up going?

Many languages have adopted some form of the “foreach” keyword, for traversing elements of a collection. The advantages are obvious: fencepost errors are impossible, and programs are easier to read. Foreach loops are not places where I expect to find bugs. But about a month ago, I found one, in a piece of code similar to the code below. The expected behavior of the program is to print out the numbers zero and one, on separate lines. Do not read past the end of the program if you want to find the bug yourself, because I explain it below.

A friend has pointed out to me that the command line and web interface in my last post do not need to interact with the main game through an iterator. He proposed that the web interface could pause the execution of the game by using exceptions. I played with the idea, and discovered that he was right. The upshot is that continuation-based web serving can be faked in any language which has exceptions. (Lexical closures are also helpful, but can also be faked using objects.) The approach given below relies on neither CPS nor monads, and so has the added advantage of being fairly idiomatic in most mainstream languages.

Though separation of concerns may be the most important design principle in software, its effective implementation is often elusive. A common problem in web design is how to link a sequence of pages together without scattering their logic all over the application. While this problem has been almost completely solved by continuation based web servers, not every language supports continuations. There is a middle ground however: coroutines. This post describes a light-weight approach to doing continuation-style web programming using Python’s coroutines.