Skip to content
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Go to file
Cannot retrieve contributors at this time

Snailfish numbers

Day 18 of Advent of Code 2021 is about Snailfish numbers, which are represented like a tree using square brackets and have a complex method of adding which consists of making a combined tree of two numbers and then through 'explosions' and 'splits' reduce the tree to a depth of most four. Both the explosions and the splits affect numbers on neighbouring leaves (if present) to make things complicated.

I decided to use a tree representation for my solution, but while I was doing this, I several times complementated to use a flat representation when I was struggling to implement an iterator that walks over the leaves of the tree. It took me 2:20:41 to finish the first puzzle and another ten minutes to solve the second puzzle.

Then in the afternoon, I tried to implement the algorithm using an array to store the numbers together with their depth. But again I struggled and resorted to implementing an iterator and a builder (kind of output stream) to implement the explosion and split passes. In the end, it took me even more time.

Then in the evening, I thought a bit more about it. I realized that the explosions could be done in one pass. I also realized that a split at depth four, leads to a single explosion, which combined lead to a 'distrution' of the value to surrounding values, and which can be done in a single step. splits at lower levels only lead to more nodes and never to any explosions. I also realized that with some clever moving around this can also be done in a single pass. But it would be rather awkward to implement this with arrays (requiring a lot of copying). So, I turned to a double linked list and implemented a third solution using the above optimized algorithm.

Both the first and the third solution are making memory allocations and are thus not very memory efficient. (I did not spend any time on deleting any allocated memory, as this would require more coding.) I wonder if this could be done in a smarter way taking into account the fact that Snailfish numbers are at most four levels deep and thus contain at most 31 nodes and leafs. The idea is to asume that you have a tree with fullest depth and then you number the nodes, from left to right, starting with 1. If you decide to use -1 for the representation of a node or a missing leaf, you only need an array of 31 integers to represent every possible snailfish number. (I used an array of size 32 out of convenience.) With this I implemented the fourth solution. I tested the solutions on the example input given at the end of the description for the first puzzle. The code can be found in the program day18_4sol.cpp. With the command line option -t, the program gives some detailed output of all the additions.