BrowserUk:

Here's what I came up with:

#!/usr/bin/env perl
use strict;
use warnings;

for my $X (3 .. 7) {
    for my $Y (2 .. 8) {
        my @orig = ((map {"X$_"} 0 .. $X),(map {"Y$_"} 0 .. $Y));
        my @wanted = ((map {"Y$_"} 0 .. $Y),(map {"X$_"} 0 .. $X));
        print "@orig\n";
        swap($X+1, \@orig, \@wanted);
        my $fl = 'Y';
        for (0 .. $#wanted) {
            if ($orig[$_] ne $wanted[$_]) {
                $fl='N'; $orig[$_]=lc($orig[$_]);
            }
        }
        print "@orig ($fl)\n\n";
    }
}

sub swap {
    my ($ofs, $list) = @_;
    my ($cnt, $src, $dst, $cmp, $tmp);

    $cnt = @$list;
    $dst = $cmp = 0;
    $tmp = $list->[$dst];
    while ($cnt--) {
        $src = ($dst+$ofs) % @$list;
        if ($src == $cmp) {
            # Loop detected, finish it and start next one
            $list->[$dst] = $tmp;
            $dst = ++$cmp;
            $tmp = $list->[$dst];
        }
        else {
            $list->[$dst] = $list->[$src];
            $dst = $src;
        }
    }
}
[download]

The two tricks are: (1) You never have to subtract from your pointers, and (2) you have to detect loops when your pointers have a smaller loop size than the buffer size.

Update: It took me a bit of time to get back to this, and I posted it before reviewing the thread. It turns out that graf already described the algorithm, as did hdb.

...roboticus

When your only tool is a hammer, all problems look like your thumb.