atomic VS euphoria

I wanted ClojureScript, without transpilation, in the browser so badly I implemented much of it in plain JS. But JS still lacks first-class protocols (although there is a T39 proposal for them) so I have to implement protocols first. And I can tell you from having implemented much of Clojure myself in plain JS that it's protocols all the way down. 80% of my library are types and their protocols.

I've watched a lot of talks, but it was Rich Hickey's which most captivated me and, ultimately, inspired big change in how I coded. After discovering Clojure I was so desiring FP (i.e. ClojureScript) in the browser without a build step and hoard of dependencies that I wrote my own library.

I fell in love with Clojure and ported most of the tools to plain ol' JavaScript. I stole reactives and FRP from Elm. I write a functional core then wrap it with an imperative shell. It's been my mainstay for 10 years now.

What I did was looked past the languages at the ideas and saw they could be applied everywhere, including JavaScript. Take a look at my example programs (via prior link) and you can see Clojure applied to the JavaScript/browser realm.

I have found that the amount of code needed to support any page/app need not be massive. I have implemented modern board games with 250-300KB of js and the core lib I use on all my projects maxes out around 250KB, unminified. The lib itself is built on the premise of build avoidance.

I do JavaScript most of the time, but I was so inspired that I ported the Clojure mindset into the web development world so I could use it even in the absence of Clojure. I know there's ClojureScript, but I wanted the goodness without transpilation.

So I don't even do Clojure per se. I do Clojure in JavaScript without ClojureScript.

I use functional core, imperative shell. It would have you write a core which is a pure simulation. Write everything in the core using immutable data and pure functions. I elaborate on that here, along with examples.

I've been working in implementing libffi into Euphoria so we can call Raylib functions directly. Previously, Euphoria did not support passing structures by value and only occasionally could we get away with "faking" it by passing int type values directly (but not float types). Raylib is the first library I've run into that makes heavy use of passing structures by value, so it's been an interesting challenge. My original proof of concept worked well with libffi built as a shared library, so now I'm working on building libffi directly into the backend of Euphoria. Then we'll be off and running with full support for Raylib!