Today was Cellular Automaton Day at Advent of Code. You have a world that's usually represented as a grid of cells. Each cell can be in a certain state. Given a certain state, the next state is determined by simple rules like for a given cell how many of it's neighbors are the same color.

The most popular Cellular Automat is probably Conway's Game of Life where each cell can be either alive or dead in a given generation and in the next generation the state will be determined by how many of its neighbors are currently dead or alive.

I've always liked teaching with Celluar Automata and explored it deeply in the intro course I designed at Stuy. NetLogo is perfect for representing a CA (old post) and we even do things like using a CS to solve a maze. Good stuff.

Today's problem set up a CA where each cell can be one of three states - floor, occupied chair, unoccupied chair. The floor never changed but the chairs can change from occupied to unoccupied and back based on neighbors. I actually really like teh question. It's posed as passengers prefer to sit in s eats next to other empty seats which is true and the overall problem is much more realistic and therefor accessible than Conway's game of life. I think I might lead with it if I do CA again with my students.

You can find my Clojure solutions here. They were pretty straightforward. I spend too much time going down the "learn how to write a lazy sequence in Clojure" rabbit hole for part 2 before just writing it in a more traditional way.

To me there are a couple of interesting things at play when teaching CA. The first is state management and synchronization. When beginners work on a CA they frequently look at a cell, figure out it's next generation value and then replace the cell with its new value. Then, later they look at a neighboring cell and when that cell looks at its neighbors it gets the changed value not the original one. Clearly a problem.

Writing a CA makes it clear that you have to keep your state clean while you build a new state for the next generation. That's an important lesson to learn.

The other interesting teaching point is with data representation. CA is frequently taught as a 2D array exercise. This makes sense - simulations like Conway's Game of Life work perfectly in a 2D array and if you're doing graphics with your class you can also get cool animations through the generations.

That said, there are other valid representations that could be superior at times. You might have a list of living cells or a hash tabe of cells if you have more than 2 living states. This can be far superior from a space efficiency point of view if you don't have a lot of living cells compared with your world size - say you have a CA with a theoretically unlimited size.

So, there it is - thoughts on today's problem. As I said, it was pretty straightforward but I also really like the setup and will probably use it in class at some point in the future.