Create implementation plan

[jasonwebb]

This project can probably work very similarly to my 2d-reaction-diffusion-experiments project. ThreeJS for WebGL conveniences, custom shaders applied to a plane mesh oriented perpendicular to an ortho camera, ping-pong between render targets to speed things up, encode cell state data in textures, passing previous iteration's texture into current gen shader pass for coherent simulation.

Kevin Chapelier (@kchapelier) has done some great prior work that can be built in as dependencies or adapted based on personal preference:

• cellular-automata-rule-parser - takes rules encoded as strings in many common formats used by CA researchers, then spits out an object with useful parameters extracted from that string. Kind of like an "unpacker" for CA rule strings.
• cellular-automata-gpu - similar to this project, but with some differences in how it approaches the WebGL side of things. Uses more low-level WebGL calls instead of ThreeJS. Interestingly it dynamically constructs a fragment shader based on the rules parsed with the library above. It basically constructs the shader as a giant string that is pieced together by modular, conditional function calls that inject different sections of code based on the rules provided.

Piecing together a shader as a string feels kind of gnarly to me, so I wonder how feasible it would be to have a single shader that has multiple paths triggered by conditionals. There might be a performance hit, but surely that would be pretty small?

[jasonwebb]

The cellular-automata-rule-parser package is working out beautifully! All of the other fundamentals (ping-pong rendering, ortho camera, etc) are all also working nicely, using the reaction-diffusion project as a starting point.