He does that in Chapter 15 of ANSI Common Lisp, but I never checked it out. Thanks for the pointer. I tried working through the implementation in Structure and Interpretation of Computer Programs, but it quickly got a bit too hairy for me. I'll look at Paul Graham's version more closely, but one problem that happens with larger systems like this is the systems tend to rely heavily on recursion and this is not Perl's strength :/ Lisp and Scheme don't have this problem.

...one problem that happens with larger systems like this is the systems tend to rely heavily on recursion and this is not Perl's strength :/

You're not giving perl enought credit for what it can do. The following program executes in under 9 seconds on my (1GB) machine.

#!/usr/bin/perl -w

print deep(2_500_000);

sub deep
{
    $_[0]>0 ? deep($_[0]-1) : "done\n";
}
[download]

...and I can calculate the 36th fibonacci number using the simplistic algorithm below in about a minute. It ends up calling "fib" over 48 million times.

#!/usr/bin/perl -w

$count = 0;
my $n = ($ARGV[0] < 1) ? 1 : $ARGV[0];
print fib($n)."\ncount = $count\n";

sub fib { $count++; $_[0]<2 ? 1 : fib($_[0]-2) + fib($_[0]-1) }
[download]

---

-- All code is 100% tested and functional unless otherwise noted.

As soon as I tried your "deep" program, I immediately got a "deep recursion" error. I've a 1.6 GHz machine with 640 megs of RAM. I don't know how you could have run that. Every time Perl enters a sub (normally), it creates a new stack frame and this quickly eats memory.

On my machine, it pretty much ground the box to a halt after about 1.5 million iterations.

Cheers,
Ovid

New address of my CGI Course.