Clojurians Log v2

sounds similar to cljs-bean?

has it become a lot faster compared to just objects though?

cljs-bean is useful, but the wrong way

ten years ago or whatever it was 1000X slower, now <10x slower or something

but that isn't the only thing we're measuring

clj->js isn't lazy and generates a lot of garbage

(meaning it just allocates like crazy on the way to producing a JS values)

I've considered doing proper immutability in squint using proxy

as a replacement for clj->js it might work nicely. I wonder if you can also accomplish "expose cljs data structure as JS object" without an explicit call to wrap it in a proxy? that might be messy and would likely cause perf issues for existing callers

I wonder if the js-interop already has done something like this

I did a bunch of research, there's no transparent way in the JS standard - doing anything in CLJS is just too magical

transparent way of doing what?

you need explicit calls to wrap in proxy

👍

exhibit A where this would be nice any JS lib way to deal w/ DOM attributes - having to always clj->js that is meh.

If have this little thing for it now: https://github.com/borkdude/reagami/blob/7e23c82c37726f9e749a28cf94677b5631df589d/src/reagami/core.cljc#L75-L86 Do you think converting in a proxy would result in any perf wins? You have to visit them all anyway so the time building a new JS obj may be just as fast as looking up things via a proxy. Need to measure this.

proxied get w/ carefully written code + caching is faster than convert then read. again that's not factoring memory issues in less trivial programs.

this is JSC, I need to verify Chrome & Firefox, though my impression is Proxy perf is good there too.

I didn't realize that Vue.js was using this as the dependency tracking mechanism since 2020 (?) - so I have a feeling this technique has been taken up elsewhere and JS engines have optimized

I remember @wilkerlucio experimented with Proxy some years ago and wrote his findings here https://blog.wsscode.com/alternative-to-clj-js/

yeah it seems most of those runs are out of date, Proxy doesn't look bad any more, especially as the size grows

the benchmarks run in your browser

I know I just re-ran them all in Safari and Chrome, Proxy wins almost always

and this is probably naive proxy, not super finely tuned like I am considering

makes sense

I always thought Proxy was cool, but 10 years to make it fast sounds about right 🙂

This is very cool!