pauloday / quasicrystals Goto Github PK

Have this output gifs or videos directly. For gifs rust-image has support, but for videos I'm probably going to have to use some kind of ffmpeg interface.

There are some more ways to colorize it I could investigate. Right now it uses a sawtooth wave, I could look at other waveforms. Or I could adjust the phase spacing to make the colors different as well.

This is kind of reliant/part of #2, but there's a blog article this was based on, I should add that. Also add some examples, I've got some really cool colorizations now.

Use multithreading to speed it up!

Some of the params could change over time. I'm thinking the initial batch will be scale, x/y offset and order. Order might look weird but I think that's fine. I wish there was a way to do fractions of an order. Maybe adjusting the angle from being evenly spaced over the course of the animation?

There needs to be a good commandline interface, so I can pick

• Colorization options
• Resolution
• Scale
• Order

This thing is real slow. It does a bunch of different sin/cos/asin operations for each pixel of each frame. Speed it up!
Possible optimziations:

• memoize sin/cos/asin
  This is probably just a worse version of using a lookup table
• use a lookup table for sin/cos/asin
  I only need accuracy to the thousandths on sin/cos. Specifically, I need 255 values from 0 to 1, distributed with sin/cos/asin.
• Utilize concurrency
  Each pixel and frame is independent, so it should be pretty easy to parallelize. But the code will need to be restructured a bit to use pmap or agents.

Overall I'm going to try memozing since it's trivial, then a lookup table and compare. The concurrency can wait.

For the Rust rewrite, I need to have good unit tests, and good benchmark tests.

The benchmark tests are most important, #2 needs to be done so I can script them.

This is related to #10, but the code should be split up

• main.rs: handle command line arguments and file i/o
• crystal.rs: generate the images
• color.rs: colorize the crystals

After some thought, I've come to the conclusion that it'll probably be more trouble than it's worth. The expressive power of Clojure is nice but doesn't bring much to the table here, since the program is basically just "do math -> place pixel". There is a lot of stuff that makes that process slower though.

I think I could make this a lot more efficient in Clojure, but ultimately I think my effort would be better spent looking at a lower level (i.e. faster) language. Rust is one I've heard great things about, both speed and design wise, so I've wanted to learn it for a while. This project is relatively simple, but it does have some complexity, and it's got plenty of room to really learn about Rust and the features that make it so great.

So yeah, this codebase is moving to Rust.