scattered (scatter3D)

composable library for 3D scatterplots.

[!CAUTION] This project is as Work-In-Progress as it gets.

3D scatterplots probably suck, but they're much more fun than regular scatterplots!

the basic idea is:

• input: an arrow/parquet (just with 'x', 'y', 'z' columns)
• zero dependencies (just vite, typescript for dev)
• webgpu rendering
• minimal API for displaying 3D scatterplots in canvas
• also a widget version, can consume numpy and pandas

usage

import * as sctrd from "scattered";

const c = sctrd.display("https://raw.githubusercontent.com/dvdkouril/sample-3d-scatterplot-data/main/penguins.arrow");

let appEl = document.querySelector('#app');
if (c) {
    appEl.appendChild(c);
}

goals

1. demonstrate composability principles
2. learn a bit of webgpu
   • just following https://webgpufundamentals.org/ for now

name

scatter plot -> scatter 3D, scatter3rd -> scatter3d/scattered

related and inspo:

• https://matplotlib.org/stable/gallery/mplot3d/scatter3d.html
• https://plotly.com/python/3d-scatter-plots/
• https://jupyter-scatter.dev (i don't think there's 3D option)
• https://abdenlab.org/eigen-tour/ (repo: https://github.com/abdenlab/eigen-tour)
• https://projector.tensorflow.org