AoC 2021/6 - Lanternfishes going round and round

TL;DR

On with Advent of Code puzzle 6 from 2021: lanterfishes going round and round.

This one, i think, was designed to lure people into an ever-growing list of fishes that would eventually exhaust memory in the second part.

Being lazy, anyway, I thought that what is good for a fish at its 0 day is just as well good for all of them. Which led me to keep track of how many fishes were in each “possible” day. This led me to something like this:

sub solve ($filename) {
   my %n;
   %n{$_}++ for $filename.IO.lines.comb(/\d+/);
   ...

Using a hash was my go-to solution, but in hindsight an array would have been better. Whatever.

This would force me to do an update over all elements as days go, which I didn’t like at all. So I remembered of a trick I was told about some time ago and eventually landed on this:

sub solve ($filename) {
   my %n;
   %n{$_}++ for $filename.IO.lines.comb(/\d+/);
   my $part1;
   for 1 .. 256 -> $day {
      my $spawning = %n{$day - 1}:delete or next;
      %n{$day + 6} += %n{$day + 8} = $spawning;
      $part1 = %n.values.sum if $day == 80;
   }
   return ($part1, %n.values.sum);
}

Basically we add fishes to a growing list of days. The :delete keeps the simulation tidy because it makes sure that no more tha 9 elements are in the hash at any time, although this is overkill in this case because we would have about 256 elements or so at the end of phase 2.

Looking at this solution now, I realize that there was a potential bug in collecting the output for part 1. What if there was no fish at day 79 and next would have kicked in?!? Oh my…

The “trick” is in not decreasing the “remaining days” for each class, but to increase the day number and compare it against the spawn day, working on the difference.

Then I looked into the solutions megathread and realized that I could do much better with an array.

A popular solution seemed to be of shifting the lower position from the array (which corresponds to “day 0”) and adding a new element at the end, as well as increasing the count in the right place. This is brilliant, because it eliminates the need for the trick of the ever-increasing day.

But then I figured that the same can be obtained just as well with some modular arithmetics. Why shift and push when we can just fiddle with the indices? So this was born:

sub solve ($filename) {
   my @n = 0 xx 9;
   @n[$_]++ for $filename.IO.lines.comb(/\d+/);
   my $part1;
   for 1 .. 256 -> $day {
      @n[($day + 6) % 9] += @n[($day + 8) % 9];
      $part1 = @n.sum if $day == 80;
   }
   return ($part1, @n.sum);
}

The interesting side-effect is that only the sum is needed, and the specific day’s share of fishes remains unchanged 🤓

As an afterthought, I wonder how much this would have been the natural solution if only the puzzle was described differently…

Stay safe, folks!


Comments? Octodon, , GitHub, Reddit, or drop me a line!