TL;DR

My fourth take on Conway’s Game of Life this year, amounting to more than 1% of all posts!

Well, the folks at Advent of Code did it again: squeeze the Conway’s Game of Life in a puzzle. This time we’re talking about the second part of today’s puzzle, where the grid is hexagonal and each cell has six neighbors instead of 8 as in “normal” play.

So, I thought it best to code a solution once and for all and put it inside cglib (as ConwayGameOfLife.pm, of course), so that the next time I’ll surely need it there will be an implementation waiting for me.

 1 sub conway_game_of_life {
 2    my %args = (@_ && ref($_[0])) ? %{$_[0]} : @_;
 3    my @reqs = qw< existence_condition neighbors status >;
 4    exists($args{$_}) || die "missing parameter '$_'" for @reqs;
 5    my $iterations = defined $args{iterations} ? $args{iterations} : 1;
 6    my $ash = ref $args{status} eq 'HASH' ? 1
 7       : ref $args{status} eq 'ARRAY' ? 0 : die "invalid status";
 8    my $status = $ash ? $args{status} : {map {$_ => 1} @{$args{status}}};
 9    while ($iterations > 0) {
10       ($status, my $previous, my %count_for) = (\my %next, $status);
11       for my $key (keys %$previous) {
12          $count_for{$key} = 0 unless exists $count_for{$key};
13          $count_for{$_}++ for @{$args{neighbors}->($key)};
14       }
15       while (my ($k, $c) = each %count_for) {
16          $next{$k} = 1 if $args{existence_condition}->($k, $c, $previous);
17       }
18       --$iterations;
19    }
20    $args{status} = $ash ? $status : [keys %$status];
21    return \%args;
22 }

In pure cglib spirit, I’m sacrifying a bit readability for compactness. there is the usual arguments unpacking at the beginning, as well as checks for obvious wrong inputs (lines 2 through 5).

The “current” status can be provided as a list of keys, or as a hash whose keys are currently “active” (“alive” in Game of Life parlance). In both cases, we expect it to be a reference to the data structure, placed in key status (lines 6 and 7), although we then proceed to work with a reference to a hash (line 8).

The loop iterates the required number of times, defaulting at 1 (line 5).

First of all we count the number of active neighbors for “whatever” cell. This is in theory an infinite search, but we know that this count will be non-zero only for neighbors of cells that are currently active, so this helps a lot (line 11, we only consider the previous positions).

Positions are represented through opaque keys. In the counting loop (lines 11 through 14), we call the callback function at $args{neighbors} to calculate the keys of the nodes that are neighbor to a given one. This is the first place where this becomes generic, because here we can potentially plug whatever proximity rule we want (e.g. arrange as a cube, a hypercube, a hexagonal grid, …).

After the counting phase, we do a reaping phase, where we analyze all cells with some counting and run another callback $args{existence_condition} to evaluate whether a node should be active/alive in the next round or not (line 16).

Last, we make sure to return something that is compatible with the input, setting $arg{status} to a reference to a hash or an array depending on what we received in the first place.

Last (line 21) we return the whole thing, that might be possibly fed to additional iterations, should we need to do them.

Happy Christmas everybody!