Re: Table of functions as state machine

[Alex Sandro Queiroz e Silva <asandroq@...>]

Hallo,

Duncan Cross wrote:
> On Fri, Jul 18, 2008 at 6:26 AM, Mike Crowe <drmikecrowe@gmail.com> wrote:
>
> > 2)  I ran across something about tail calls, or tail recursion, does that
> > apply?
>
> You can use tail calls to approximate state machines, but it will mean
> the system is no longer asynchronous in the same way and will run
> until the terminating state is hit, unless you make use of coroutines.
> It also doesn't help with your meaningful-names-for-debugging problem.
> It does mean that you can use arbitrary arguments appropriate for the
> state instead of always a single data argument, however.

     "Tail calls make debugging hard" is the argument of the Python 
crowd for not implementing them. This is not true because when using 
tail calls one must pass the complete "state" of the algorithm as the 
current function will be removed from the stack. To debug, study the state.

> Tail calls mean that one of these states could return with a call to
> itself, an unlimited number of times, without overflowing the stack.

     Lua optimise all tail calls, not only tail recursion.

-alex
http://www.ventonegro.org/