Building a DOOM-like multiplayer shooter in pure SQL

Summary

DOOMQL: A DOOM-like multiplayer shooter in pure SQLI recently stumbled across Patrick’s excellent DOOM clone running in a browser powered by DuckDB-WASM. Ever since I’ve read that, I wanted to push his awesome idea to the logical extreme: Build a multiplayer DOOM-like shooter entirely in SQL with CedarDB doing all the heavy lifting. During a month of parental leave (i.e., a lot of sleepless nights), I tried exactly that.Here’s a sneak peek at DOOMQL: Your browser does not support the video tag.DOOMQL in actionOkay, with the flashy demo out of the way, let’s talk about details. What follows is a tour of the architecture, the SQL rendering pipeline, the game loop, and the fun metagame of cheating by issuing SQL commands against the database.Why even do this?Playing DuckDB DOOM in your browser is fun, but some things bugged me: First of all, having parts of the rendering pipeline in Javascript felt like cheating. It worked well for DuckDB-Doom where everything is contained in a single HTML page, but I wanted to see if I could do everything in SQL. DuckDB-Doom is also a little bit stuttery with just 8 frames per second and has a pretty tiny viewport. I wanted to see if I could speed that up by switching over to CedarDB. I also wanted real sprites with transparency and they should move around believably in 3D space. And most importantly, making the game multi-player should not just be possible, but easy, right? I got nerd-sniped by the perceived similarity of a database server to a traditional game server: Databases exist to synchronize shared state across clients. Thanks to transaction isolation, each player has a consistent view of the game world, no matter what the other clients are doing. Why not lean into that? I would love to lie to you and claim I did it all to push CedarDB as an awesome database system but to be honest the database nerd in me just wanted to turn all knobs up to 11 and see what breaks.Architectural overviewAt a high levelState lives in tables (map...