shady VS jsony

I have used Nim for personal projects for 6 years now and it continues to surprise me on how well versed it is for many problem domains. I am fond of it's SPA framework, karax https://github.com/karaxnim/karax for which I wrote a translation utility https://github.com/nim-lang-cn/html2karax Latest Nimv2 release candidate has improved in the ergonomics and syntax that affect compilation to js, so I was able to cleanup my webapp's code to be less verbose. On GPU programming there has been a few projects that touch GPU programming, most notably https://github.com/treeform/shady

This includes the GPU! Yep, that’s right. You can write shaders in Nim. This makes shader code much easier to write because you can debug it on the CPU and run it on the GPU. Being able to run the shader on CPU means print statements and unit tests are totally doable.

You can do a lot with Nim at compile time, check out my talk on Nim Metaprogramming not just for FizzBuzz, but real world applications: https://fosdem.org/2022/schedule/event/nim_metaprogramming/

I am working an a macro to compile Nim code into GLSL. So not only can you write Nim to C or Nim to JS, it can also (in limited way) do Nim to GLSL GPU Shaders. See here: https://github.com/treeform/shady

I am also working on a macro system similar to SWIG, where using a some macros one can write a Nim library and generate wrappers for your NIM library for many languages like C, Python, JS, Ruby. See here: https://github.com/treeform/genny

See this as one of the examples: https://github.com/treeform/shady

It makes debugging shaders much easier as you can use print statements and unit tests. You can also share code between CPU and GPU side.

Nim's default json library is terrible in performance, but there're much faster drop-in replacements like jsony[1]. I'm not sure that's the main issue for low rank, but it's definitely one of them.

1. https://github.com/treeform/jsony

One of the fastest things I have written in Nim is a JSON parsing library. Why is it fast? Well, it uses Nim’s metaprogramming to parse JSON directly into typed objects without any intermediate representations or any unnecessary memory allocations. This means I can skip parsing JSON into a dictionary representation and then converting from the dictionaries to the real typed objects.

The nim stdlib's json(due to how it works) is relatively slow. Jsony Jason are some alternatives that are more performant and speeeedy.