🪅-progaming

[supple whale]

its rust black magic

[winged mantle]

use c preprocessor on java to have one codebase for multiple minecraft versions in mod

[supple whale]

https://github.com/ThaUnknown/wasm-bench/blob/main/Cargo.toml#L22-L28

i dont think my compilation flags are wrong either

[jade stone]

i think all major minifiers just remove blocks of code that are statically unreachable

[royal nymph]

ages ago I tried using a worker for discord webpack search, but creating and preparing the worker took like 2 seconds cause I had to serialise and copy over all the webpack modules so it ended up being way less efficient

[jade stone]

i love sending every discord module over websocket

[supple whale]

seems like its a copy https://github.com/ThaUnknown/wasm-bench/blob/main/pkg/sha_was.js#L16

its not a worker btw

its just raw WASM

[royal nymph]

yeah i know

[winged mantle]

why doesn't js have pthreads

[supple whale]

u cant rly not copy with WASM

[winged mantle]

[supple whale]

since u cant just randomly pass other memory as a reference

u need to set the memory

[royal nymph]

can't you use shared array buffer

I never used wasm

[supple whale]

nope

it wont help

you still need to PUT the data into the stack

imagine some1 has a file on his own computer, and goes to hash it on a 2nd computer

u cant just go "yeah use the other's PC RAM"

you need to copy it onto the first PC's ram

pretty much the way i'd explain it

thats how wasm : js works

the original data was in JS, it needs to be "added" to "appended" to WASM's memory

since the program has its own memory in its own sandbox

WASM is initialized with a specific memory buffer, and that's it's sandbox, just like how its done on a normal PC with native code

when u go to use a pointer it needs to be a a pointer to the memory you're using

it cant all of a sudden be diff memory

but yeah, wasm seems weirdly slow on newer versions of chrome

idk if JS got better or if i fucked smth up

[tired vigil]

Is this also how asm.js worked or was that used on the same thread with the same memory? (I just found out about this like yesterday)

[supple whale]

asm.js just compiled assembly instructions into JS code

so... effectively it implemented a CPU in JS

you can imagine how horrific that was for performance

[tired vigil]

I think firefox actually optimised asm.js code somehow

[supple whale]

but strictly speaking, no, you could hack it, but it would be a hellish undertaking of an ungodly level

[tired vigil]

When "use asm" was used, indicating that it should be optimised
But idk

[tired vigil]

😭

[supple whale]

yeah wtf

why did my WASM get like... x5 slower

wtf changed

it used to max out at 2M IOPS for 8b, now it maxes out at 400k

wtf

js got slower too?

.

[ivory heath]

You can. It’s purely implementation limitations. Idk what your rant about the stack is about

[supple whale]

this is going to be fun to re-test....

[supple whale]

yeah but if the original data is created in JS

and u want to hash it in WASM

then you need to copy the data into WASM memory

there's no way around that in browser

[ivory heath]

You can just share a heap ptr.

[supple whale]

sir....

the heap

is in the wasm's memory too

and the data

is outside of wasms memory

[ivory heath]

Allow js to write to WASM? where is the data coming from?

[supple whale]

english

?

i think u forgot a comma

or like 3 of them

its coming from JS

[ivory heath]

Fixed I had a stroke

[supple whale]

since shocker, wasm in browser cant interact with anything else than itself, and some atomics

so if you want to say hash a file, you need to load it via JS

and then copy into WASM

since WASM cant access any W3C apis directly for reading shit like files etc

[ivory heath]

Yeah they neutered it’s theoretical capabilities.

[supple whale]

nope

they didnt

its all 100% inline with WASM's design principles

WASM was always meant to be a safe, sandboxed way of executing native code

not "free performance for destroying JS"

[ivory heath]

WASM isn’t native it’s just a “jvm in the browser”

[supple whale]

which is why it always makes me laugh when people are delusional thinking that they can "fix the perf problems they have" with WASM

[supple whale]

yeaaaah

for the most part

it's not far off

[ivory heath]

WASM is neutered in every aspect

It’s slower than naive

[supple whale]

well yeah its bytecode

[ivory heath]

Massive overhead communicating with JS

[supple whale]

ofc its slower than native

[ivory heath]

wtf was the point

[supple whale]

WASM was always meant to be a safe, sandboxed way of executing native code

that was the point

WASM has NEVER been about perf

sandboxing it was always the #1 thing

[ivory heath]

It’s not native

[supple whale]

well

by "native code"

i mean shit like C/C++/Rust/Go

[ivory heath]

And what is the advantage of a second sandbox?

[supple whale]

native languages

or languages that are meant to compile to native code

[supple whale]

not getting hacked by a malicious website!

[ivory heath]

Porting libs that are gonna run like shit in the browser

Smart

[supple whale]

that said WASM had its share of sandbox excapes

[supple whale]

yep!

100%

thats EXACTLY the point of WASM

[ivory heath]

HORRIBLE

[supple whale]

nope

you just dont understand

because compiling a lib which will run at 95% of the perf of JS to WASM will take u a day

porting it to JS will take u 2 months

if you're a good developer, in most cases way longer than that

that's been the goal of WASM

and people HEAVILY missunderstand that

when they see the "asm" in "wasm"

[ivory heath]

But what are you even porting?
Too slow for image and video codecs.

Your proprietary api for your database I guess ?

[supple whale]

why do you care what people are porting

are you the authority of all web?

what matters is allowing developers to do that

they are plenty of highly proprietary native libraries

that are the only ones in the world that do one giga specific thing

shit, i can even think of encryption algos like that

for the longest time there was no ed25519 implementation in JS

wasm was a good lazy solution for that

because would u risk fucking up ed25519 by writing in manually in JS? or simply compile a lib to wasm in 15 minutes

[ivory heath]

WASM could have been more

[supple whale]

sure, but it wasnt meant to be more

it perfectly accomplishes what it was designed to do

without a single flaw

[ivory heath]

It could have done more easily

[supple whale]

becuase what it was DESIGNED to do, is different from what people think it is

[supple whale]

hm?

no?

legit all of squoosh uses WASM

and I use wasm myself for muxxing in browser

because shocker! you dont need to convert the codecs, browsers already support that

u just need to change the containrs

and re-wrap them using native code

like MSE

[ivory heath]

It’s a lot slower than native

[supple whale]

but noooo people need to import all of fucking ffmpeg for some reason

lol?

[supple whale]

okay and?

x1.1 speed of real time vs x10 speed of real time?

fuck cares, its real-time?

"oh no i cant watch my video at x10 speed when using WASM, what will i do"

.

[ivory heath]

Avif wasn’t and still kinda isn’t

[supple whale]

because you're likely re-encoding all of it

[ivory heath]

Speed is apart of access

[supple whale]

instead of letting browsers do the heavy lifting

[ivory heath]

In not talking about converting a video.

[supple whale]

anyways, WASM does what it was designed to do perfectly

and there's legit 0 problems with its implementations

it solved what it was set out to do, for the most part without hickups

its just for some reason, people keep wanting flying cars, which is stupid, because cars are meant to drive on roads, not fly, and a flying car will never be as good as a car and an airplane separately

that's trully the state of WASM

its why i'm writing this entire paper

[jade stone]

i love typescript i love typescript

[supple whale]

becuase people are unhinged as shit

[supple whale]

XD

CLASSIC

[supple whale]

lineOnText: boolean = false

[jade stone]

yeah

unhusk

[ivory heath]

People want more because everything is moving to browsers and browsers suck ass.

[supple whale]

see

if everything is moving to them

that means they dont suck

because otherwise every1 wouldnt be using them

[ivory heath]

Or people just cope with shitty software

[supple whale]

its trully that simple

there's not some shadow goverment who's manipulating companies into using browsers

its simple, developing for browsers is HORRENDOUSLY simple and fast

and what motivates companies more than anything?

costs!

and developing for browsers, simply has the lowest costs

because of all the development advantages it has for companies

[ivory heath]

At the cost of literally everything else for the end user.

[supple whale]

what end users get is another story

[ivory heath]

LOL

[supple whale]

because end users are not who decides what software is written

its companies

and companies only care about money

its trully that simple

and yep, end users get fucked in the process, because who would have thought, making a magic all in one wand that solves all problems isnt easy

[ivory heath]

But X software can run a literal order of magnitude faster natively. Or rely on out of date libs the browser has.

Or depending on the browser some features don’t work even between forks of chromium

[supple whale]

and that matters for some X company because?

[ivory heath]

If you target win32 and supply x86 and arm congrats your software works universally

You only have to support windows (x86/arm).

Wine.

Fucking wine

Wine is that good

At the cost of the end user

[tired vigil]

0.25s is a pretty big delay for a click event tbh

discord on WSA also reloads the whole app when you move the window to another monitor 😭

no other app does that

also WSA is goated and I hate microsoft for discontinuing it

[ivory heath]

I think people do notice.

This is the downfall of software

Enshitification

[supple whale]

wsa?

[tired vigil]

windows subsystem for android
(exact opposite of what it sounds like)

[supple whale]

oh

didnt google do their own thing for that now?

[tired vigil]

well what do users even care about
I'm sure it's not ai slop apps

[supple whale]

via WSL

or am i missremembering?

ah its not WSL

gg

[ivory heath]

I’m saying your mentality is. And yeah i remember when native apps were a thing

[tired vigil]

google does have their own little thing, they use it for the desktop Play Games or whatever

https://play.google.com/googleplaygames/

I couldn't figure out how to launch external apps tho

[ivory heath]

It’s already a problem

And the free lunch of hardware improvements is over

You have smaller transistors to look forward to and back side power delivery

[tired vigil]

nope, local actually

the performance is similar to wsa

yeah

I just found this tho, from the same person that makes those wsa builds
https://github.com/MustardChef/GPGPCToolkit

😭 all the options are either chinese adware slop or android studio 500ms latency emulator that crashes after opening two apps

oh you edited the message

cya

[tired vigil]

the development is stalled 😭

[jade stone]

since when was there a native popover api

https://developer.mozilla.org/en-US/docs/Web/API/Popover_API

[supple whale]

like a few months back

not long

[jade stone]

i will mirgate to it soon and use my own popovers

[supple whale]

I'd love to as well but Firefox exists

[jade stone]

shrimply dont

firefox is mid anyway

[supple whale]

as long as Firefox exists we can't have nice things in the world

[ivory heath]

Firefox

[tired vigil]

Firefox view transitions when

[royal nymph]

can mdn undo this ui change already man

[jade stone]

i tried looking on their github but i couldn't find anything related to it

[tired vigil]

I thought people were just overreacting but it genuinely makes it hard to read on desktop cuz of how the font wraps 😭 feels like it takes up too much space, though it would probably look good for a headline

It's more inter than inter

[royal nymph]

https://developer.mozilla.org/en-US/blog/launching-new-front-end/

FUCK NO YOU HAVENT

[jade stone]

@royal nymph https://github.com/mdn/fred/issues/586

[royal nymph]

also what the hell were they thinking with this banner

.page-layout__banner {
    display: none;
}

mdn/fres#587

mdn/fres

[jade stone]

@solid gazelle

[jade stone]

oh, i was wondering what they were talking about but i realized i blocked it with ubo

[solid gazelle]

good banner

[jade stone]

@royal nymph MDN INSANE

[supple whale]

ya'll use MDN

and not shit like devdocs?

[royal nymph]

yeah it just looks wrong

idk why

[jade stone]

yeah

[royal nymph]

it's the same font but it looks so different and is painful to read

[valid jetty]

@deep mulch do you know radix sort

use std/prelude;

fn bucket_sort<T>(T[] arr, u8 digit) {
    buckets := (0..10).map(fn() [T;]).collect();

    for x in arr {
        digit := (x / math::pow(10, digit)) % 10;
        buckets[digit].push(x);
    }

    return buckets.concat();
}

fn radix_sort<T>(T[] arr) {
    out := arr;

    for i in 0.."{}".format(T::MAX).len() {
        out = bucket_sort(out, i);
    }

    return out;
}

fn main() {
    $dbg(radix_sort([14, 23, 1242, 352, 39, 442]));
}

such a simple sorting algorithm

[deep mulch]

no

i look and then promptly forget a week later

idk how you remember any of this

[jade stone]

before

(wayback machine)

after

[valid jetty]

its the one where you have like

first bucket sort is:

numbers 1 3 1 8 9 2

and you put them into buckets of 1-10

so in the bucket of 1s, there is 1 and 1
in the bucket of 2s there is a single 2
etc

then you concatenate all the buckets (a bucket is just a dynamic array)

for numbers with multiple digits you just apply bucket sort on each digit and thats radix sort

it is surprisingly simple

[jade stone]

@royal nymph THEYRE USING JETBRAINS MONO NOW

[valid jetty]

its the first sorting algorithm ive ever made working first try

[winged mantle]

eugh

[deep mulch]

hold on

ill try kotler

[valid jetty]

@deep mulch you can just copy

its a very simple algorithm

[deep mulch]

oh rosie my rosie

[jade stone]

Fix MDN with this one simple trick

@-moz-document url-prefix("https://developer.mozilla.org/") {
:root {
    --font-family-code: monospace;
}
}

[deep mulch]

@jade stone SALAD

[deep mulch]

so hard to read

(fn() [T;])

Evil.

[supple whale]

just use devdocs like a human being

[jade stone]

@deep mulch love?

[deep mulch]

im gonna explode rosie with my mind

[supple whale]

https://devdocs.io/javascript/global_objects/array/slice

here u go

[deep mulch]

NO idea what type it returns

[jade stone]

no embed

also i prefer the UI of mdn, even after the change

[deep mulch]

wjaht does this even return fn bucket_sort<T>(T[] arr, u8 digit) {

[jade stone]

to devdocks

[valid jetty]

anonymous function that returns an empty dynamic array of T typed elements

[royal nymph]

the new one looks thicker, no?

[valid jetty]

T[]

[deep mulch]

i literally cant read this

rosie will use pseudocode

[valid jetty]

buckets is T[][], calling buckets.concat() will join together all the arrays to form T[]

[deep mulch]

guh

[valid jetty]

ok hold on

[jade stone]

it's thinner in some areas and thicker in others

[deep mulch]

https://www.geeksforgeeks.org/dsa/radix-sort/ so much easier to read

[valid jetty]

bucket_sort(arr, digit) -> sorted_arr {
    buckets = [] for i in 0..10
    
    for x in arr {
        digit = (x / digit ** 10) % 10
        buckets[digit].push(x)
    }

    return joined(buckets)
}

radix_sort(arr) {
    out = arr

    for i in max digits of int {
        out = bucket_sort(out, i)
    }

    return out
}
``` @deep mulch

[deep mulch]

love

i can read

is buckets a list of lists

[valid jetty]

yes

specifically a list of 10 lists

because there are 10 digits

[deep mulch]

oh rosie my rosie

[royal nymph]

@import url('https://fonts.cdnfonts.com/css/menlo');

.page-layout__banner {
    display: none;
}

:root {
    --font-family-code: Menlo, Consolas, Monaco, Liberation Mono, Lucida Console, monospace;
}

code {
    font-size: 0.833rem
}

it still looks wrong but better

[deep mulch]

.

make elle mascot @valid jetty

[royal nymph]

@jade stone userscript that automatically redirects mdn links to pre-redesign archive link

[valid jetty]

???

[deep mulch]

do not ???

[royal nymph]

actually can you selfhost the old version

[valid jetty]

i love elle's mascot i will NEVER make it

[royal nymph]

i despise the new In this article

[valid jetty]

potential proof it was written by ai

[deep mulch]

what is max digits of int {

[valid jetty]

format the max size of your integer type into a string and take its length

[royal nymph]

@jade stone ```css
.reference-toc a {
padding: 8px 16px;
color: var(--color-text-secondary);
}

so much better

[valid jetty]

youre doing a bucket sort for each digit

[deep mulch]

cant i just hardcode it

[valid jetty]

well you can

seems to be 10

[deep mulch]

rosie moment

[valid jetty]

but essentially you wanna do a bucket sort for each digit of the numbers

you could also probably loop through the numbers once to find the biggest one, so you can take that number's amount of digits

[deep mulch]

what is the return type of radixSort

[valid jetty]

that way youre potentially not sorting pointlessly

[valid jetty]

just an array

[deep mulch]

how

[valid jetty]

bucket_sort returns T[], radix sort will return that too

[deep mulch]

something isnt right

[valid jetty]

because its returning out which is the return value of bucket_sort

[deep mulch]

bucketSort returns an array

it would build a 2d array

h

oh

wait

i see

[valid jetty]

youre not concatting the values

youre recomputing each time

[deep mulch]

SH

i see now

[royal nymph]

mdn fioxed

[deep mulch]

shhhushh roise

roie

[valid jetty]

each iteration of bucket_sort sorts by 1 digit

lole ok

[deep mulch]

okayu ti works

[valid jetty]

i think you read out = bucket_sort(...) as out += bucket_sort(...) right

[deep mulch]

yes

[valid jetty]

silly operator overloading

[deep mulch]

fix elle syntax @valid jetty

[valid jetty]

elle syntax is perfectly understable if you try to read it for more than 1 second

[deep mulch]

nop

rust more readable

[valid jetty]

whats not readable about this

use std/prelude;

fn bucket_sort<T>(T[] arr, u8 digit) {
    buckets := (0..10).map(fn() [T;]).collect();

    for x in arr {
        digit := (x / math::pow(10, digit)) % 10;
        buckets[digit].push(x);
    }

    return buckets.concat();
}

fn radix_sort<T>(T[] arr) {
    out := arr;

    for i in 0.."{}".format(T::MAX).len() {
        out = bucket_sort(out, i);
    }

    return out;
}

fn main() {
    $dbg(radix_sort([14, 23, 1242, 352, 39, 442]));
}

[deep mulch]

no explicit returns is really hurting elle

like

[valid jetty]

how would you make it more readable (keep in mind this is a static language)

[deep mulch]

explicit returns i beg of you

[valid jetty]

i can just add return types

[deep mulch]

force them

[valid jetty]

why

[deep mulch]

how elle even compiles without them baffles me

[valid jetty]

for little functions or functions with very complex return types its very tedious

[deep mulch]

then make it optional for short functiions like fn guh() = 2

[deep mulch]

it looks weird for short functions but id rather not read a 100+ line function just to know it returns an array

[valid jetty]

its 8 lines

[deep mulch]

the pros far outweigh the cons

[deep mulch]

it requires pre-requisite knowledge of the stdlib

[valid jetty]

imo buckets.concat() is relatively easy to understand whats going on

[deep mulch]

id only use concat for joining two distint instances

flatten for one instance

[valid jetty]

this is basically a flatten

i just gave it that name idk

[deep mulch]

concat(list1, list2)

oh rosie my rosie

[valid jetty]

fn Array::concat<T>(T[][] arrays) {
    res := [T;];

    for array in arrays {
        res.extend(array);
    }

    return res;
}

[deep mulch]

;];

is there no way to remove the inner ;

whats extend

oh

[valid jetty]

yeah because there is no way to disambiguate when T is a type or a value

[deep mulch]

that cant be efficient can it?

[valid jetty]

wdym

[deep mulch]

why not allocate the array size upfront rather than extending it each time

[valid jetty]

thats a really good point

pepehba

[deep mulch]

@valid jetty meow

@valid jetty elle will be so fast

[valid jetty]

its already pretty fast

[deep mulch]

have you made fibbonacci sequence

[jade stone]

@valid jetty add c++ interop to elle

[valid jetty]

you have to remember the only reason [T;]; even exists is because the language isnt dynamic

its statically typed and still looks (for the most part) like a dynamic language

[deep mulch]

it willb e soon

interpreted, running on v8 engine

[valid jetty]

technically you can

actually that reminds me

[jade stone]

it's llvm isn't it

[valid jetty]

no

its qbe

[jade stone]

oh

[valid jetty]

who says you cant compile both to object files and call the linker yourself

elle can define c++'s mangled symbols and be given sane names

[valid jetty]

i wanted to make a node module in elle lol

[shrewd canopy]

Is that accidentialy not windows 12 move (everything is React native, the kernel will be rewritten in javascript)

[deep mulch]

say meow @valid jetty

[valid jetty]

the thing is @deep mulch that requires an extra iteration through all the values, making it O(2n) instead of O(n ammortized)

meow

[deep mulch]

yea

would htat make it more efficient with massive arrays

[valid jetty]

no

[deep mulch]

becnhmark @valid jetty

[valid jetty]

its O(n ammortized) which means the cost of reallocation and copying is fully paid off by the time you need to allocate again

[deep mulch]

idk what ammortized means

[valid jetty]

thats what it means

i just said

[deep mulch]

nop

gradually write off the initial cost of (an asset) over a period.

generalized

[valid jetty]

yeah

the cost of the reallocation

[deep mulch]

@valid jetty add reflection to elle

[valid jetty]

you reallocate once and copy everything over (n items), then the next n items are allocated at O(1) because theres no reallocation

so by the time you need to double again, you paid off the cost of reallocation and copying

each element being O(1) paid it off a little bit

[valid jetty]

its not a dynamic language

..

[deep mulch]

make it so

make elle run on JVM

50k stars on github

rosinga

[winged mantle]

rtti

[valid jetty]

you have very rudimentary RTTI

you have TypeInfo::type_of<T> and TypeInfo::type_of_val(...) which give you the string representation of the type

[jade stone]

i feel like there will be horrorcode that just switches off that

[deep mulch]

@valid jetty learn about op-amps and then teach me

op-amps so cool but so confusing

the rosie rot consumes us all

[royal nymph]

why does his pfp kinda look like a soyjak

looks like cobson

[shrewd canopy]

Or you might get an app with severe security vulnerability which is deployed to prod

[supple whale]

oh you dont need vibe coding for that

i remember for the first 3 years the utorrent electron app was built with EVERYTHING in the repo embedded in the asar

including signing keys

fucking hillarious

[shrewd canopy]

Was utorrent a native app before or it was electron initially

[supple whale]

idk tbf

never used it cuz it always was spyware

over a decade ago i used bitcomet

then started using qbit

and then ust made my own torrent clients which are specialized for the thing they do XD

[deep mulch]

@valid jetty my IR transmitter works over 10 feet now

havent actually implemented proper packets

[valid jetty]

@deep mulch im trying to make a nodejs module in elle

isnt this beautiful

[deep mulch]

no

[supple whale]

that looks more like a v8 native binding in elle

[valid jetty]

well yeah but im loading it from node

[supple whale]

yeah but it does nothing with nodejs no? XD

still

thats CBT

[valid jetty]

did you know if you compile a v8 dylib with the extension .node you can just straight up require(...) it in the nodejs repl

[supple whale]

I LOOKED at it once, and said fuck no aint no1 doing that shit

and mf is doing it

[valid jetty]

you dont need gyp or anything you just compile an .so and rename it to .node and boom you can load it

[supple whale]

huh didnt know

it also doesnt look any more fun than writing assembly

so i think i'll pass on it XD

[valid jetty]

yeah its not very fun

but if it works itll be fun

i love this sooooooooo much

im enjoying myself

[supple whale]

yeah no

its inline with what i expected

[valid jetty]

well

[supple whale]

its very close to CBT itself

ngl

[valid jetty]

if you write it in c++ its rather easy

[supple whale]

meanwhile i'm compiling nodejs for android and i have some black magic crash

#00  pc 0x000000000210ac70  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/split_config.arm64_v8a.apk!libnode.so (v8::internal::trap_handler::EnableTrapHandler(bool)+40)
#01  pc 0x0000000001514c04  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/split_config.arm64_v8a.apk!libnode.so (node::InitializeOncePerProcess(std::__ndk1::vector<std::__ndk1::basic_string<char, std::__ndk1::char_traits<char>, std::__ndk1::allocator<char>>, std::__ndk1::allocator<std::__ndk1::basic_string<char, std::__ndk1::char_traits<char>, std::__ndk1::allocator<char>>>> const&, node::ProcessFlags::Flags)+1316)
#02  pc 0x0000000001515c0c  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/split_config.arm64_v8a.apk!libnode.so (node::Start(int, char**)+240)
#03  pc 0x00000000000058ec  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/split_config.arm64_v8a.apk!libnative-lib.so (Java_net_hampoelz_capacitor_nodejs_NodeProcess_nativeStart+940) (BuildId: f04efa150d42f03fa6efa37d161e28f0f7d5dfda)
#04  pc 0x0000000000f03e0c  /data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (art_jni_trampoline+140)
#05  pc 0x00000000000f7388  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/oat/arm64/base.odex (net.hampoelz.capacitor.nodejs.NodeProcess.start+1064)
#06  pc 0x00000000000f4e7c  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/oat/arm64/base.odex (net.hampoelz.capacitor.nodejs.CapacitorNodeJS.lambda$startEngine$0+3772)
#07  pc 0x00000000000f3344  /data/app/~~PIzaPoLbRPe8r4Yr86aNXQ==/watch.miru-LO8LfSvp4JCF-2pvYiz9mw==/oat/arm64/base.odex (net.hampoelz.capacitor.nodejs.CapacitorNodeJS$$ExternalSyntheticLambda0.run+100)
#08  pc 0x00000000003754f0  /data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (java.lang.Thread.run+64)

thats what i get for modifying libs

[valid jetty]

i have a reference to look at

#include "node.h"

void Hello(const v8::FunctionCallbackInfo<v8::Value> &args) {
    printf("hello world!\n");
}

void Initialize(v8::Local<v8::Object> exports) {
    NODE_SET_METHOD(exports, "hello", Hello);
}

NODE_MODULE(test, Initialize)

but im trying to call c++ functions from a c ABI context

[supple whale]

i think android's kernel does some dogshit with trap handlers for WASM

and just randomly segfaults on some OS's when nodeJS goes to start some of its native WASM modules?

I think?

[shrewd canopy]

Why are you trying to use nodejs on android

Thats just dumb

[supple whale]

when u google this shit there are 2 results, and both are for nodejs for RiscV

[supple whale]

why not?

its like x100 faster than java

[shrewd canopy]

no thats isnt true

nodejs is slower than java

[supple whale]

we both know how untrue that is

anyways, for raw network IO java could barely do 20MB/s on my redmi note 9

for nodejs it easily did a gigabit

and i wont be re-writing my app to java, just to waste time on something that will run slower

i can simply just run nodejs and IPC to it from my webview

[valid jetty]

?????????????????

how tf is it crashing in the allocator

[supple whale]

and simply re-use my electron codebase

0 work required

[shrewd canopy]

thats just speedrunning to get OOM error

[supple whale]

sure

i have that problem a lot too

but its a LOT slower to OOM than with java

especially on e-waste such as android TV

where as much as running custom UI is enough to reboot ur TV because it runs OOM

using anything else than a webview there is instant GG

try making 1k network connections with java

[shrewd canopy]

why you are making 1k network connections in first place...

[supple whale]

because i can?

its what torrent clients do

they create connections

[fallen nebula]

me : try to solve issue with discord embeds
Ven jumpscare : https://github.com/discord/discord-api-docs/issues/7373#issuecomment-2619043847

[shrewd canopy]

thats not a good reason to not use a single request that does tons of work

[supple whale]

sure, you host pirated content on a http endpoint

so i can use a single network connection

lets see how well that ends for you

anyways its just a good example how java doesnt scale well for any actually significant amt of compute

writing good JS isnt hard, and making W3C spec reliant libraries which compate with native libs for performance also isnt a high order

it requires basic competence, sure, which is rare in the JS ecosystem, sure, but its not hard

[jade stone]

i love the new page on my site

[shrewd canopy]

It certainly does

There's no way v8 can be faster than JVM with JIT

[supple whale]

unfortunately i have not seen a single exaple of this in practice

[supple whale]

ah you see, because you're thinking of "raw js, vs full blown java with all its apis and builtins"

and what i'm saying is "js built on top of w3c and node apis, which are native code, which is orders of magnitudes faster than java"

because JS wont do the hashing, i'll use web crypto, its faster, because i wont use JS to create a raw IO socket and create the TCP protocol on top of it, i'll use node:net which runs native code

its trully that simple

and for everything that is high compute i simply use NAPI and bind to native code

so JS becomes a simple state wrapper, that doesnt do any heavy lifting

[shrewd canopy]

So same in Java

But only native thing is just syscall

[supple whale]

yeah java will simply just syscall some kernel method, and then implement the rest in java

[shrewd canopy]

Most of anything else is Java and its okay because its jitted anyway

[supple whale]

sure

still slower than C++ compiled to assembly for the exact CPU arch

which is what nodejs simply becomes

which is why im compiling it for android

[shrewd canopy]

That cannot be true lol

[supple whale]

sir, nodejs is almost entirely C++

X D

[shrewd canopy]

I know lol

But js speed cannot be same as c++

[supple whale]

then what do you think i'm doing

i never said its JS

i said its C++

as i said

i use NAPI and node's own APIs for ALL the heavy lifting

and that is all C/C++/Rust compiled to ASM

now you understand why i'm compiling nodeJS for android?

because if your code is written right, nodeJS performance is fucking crazy

dont do compression in JS, do it using node's ZLIB bindings, dont implement ur own file handling, use NAPI

JS simply becomes glue, that does very little outside of state handling

and java can't compete with that

sure java is faster than raw JS

but apps arent raw JS, they will always use native APIs, be it W3C in browser or nodejs's api

both of which run native code

\o/

[supple whale]

so this, this is why, because if i can do it right, i can get horrendously fast code, that i already have written for desktop, with no actual need to re-write any of my code to be android specific

the only real issue is the apk size, because it becomes like 80MB because i need to ship 3 dists of nodejs, for arm 32 bit, arm 64 bit and x86

still not a big issue considering many apks like tiktok are 200-300MB+

[talking no compression, on play store downloads are compressed to shit]

[jade stone]

@deep mulch you love node.js for android

https://tenor.com/view/java-duke-everywhere-computer-phone-gif-2655662629958065522

remake the gif for node

[supple whale]

its unironically really good, it baffles me noone has done this

like the performance shook me

bionic is insanely fast

even tho there are.... a few issues

maybe... a few more than a few....

a solid 2% of interactions end in some sort of sigfault

which... isnt a terrible turnout, but isnt great, 0% would be nice

its pretty much all WASM related tho for some reason

android rly doesnt like native WASM, no clue why yet

i saw some code about it in chromium itself

it has android specific patches for WASM

@ivory heath https://gitlab.cs.uni-saarland.de/bigdata/mutable/v8#mutable-specific-changes

LMAO some1 actually did it

re: 0 copy memory copying to WASM

i randomly found this

[supple whale]

https://source.chromium.org/chromium/chromium/src/+/main:v8/src/trap-handler/trap-handler.h;l=63;drc=67329df6671f8e854bff240020ea90be51ad91b8

fun

[ivory heath]

I still don’t understand why it wasn’t done.

[supple whale]

something something threading

from what i can read in the code

[ivory heath]

Something something mutex/atomics

[supple whale]

probably

not used to navigating gitlab

but i can see they are disabling some threading related shit in v8

[ivory heath]

I can’t imagine that much would have to be disabled

[supple whale]

well its like 2 flags

[ivory heath]

Or only some special checks would have to be added

[supple whale]

fuck knows what the flags actually do

they disable some v8 features via flags

but i'm not gonna dig in v8 to see what they do

way too much work

but hey, its somewhat possible!

not that i think it'd help WASM much

[deep mulch]

rosie so smart

[deep mulch]

nop

[supple whale]

TRAITOR

this is a JS/Elle household

[ivory heath]

Removing any memory copy is a good thing. There is no downside.

[supple whale]

yeah sure

but im simply saying it wouldnt help wasm's own perf

[ivory heath]

Sure

[supple whale]

since WASM itself aint the fastest, thats the problem

but yeah, it would slightly help it

a lil bit

[valid jetty]

struct V8Local { void *_ }
const V8Local::_ = V8Local { _ = nil };

external fn _ZN2v86Object3SetENS_5LocalINS_7ContextEEENS1_INS_5ValueEEES5_(V8Local *self, V8Local context, V8Local key, V8Local value) @alias(V8Local::Set);

namespace V8Isolate;
external fn _ZN2v87Isolate10GetCurrentEv() @alias(V8Isolate::GetCurrent) -> V8Isolate *;
external fn _ZN2v87Isolate17GetCurrentContextEv(V8Isolate *self) @alias(V8Isolate::GetCurrentContext) -> V8Local;

struct V8FunctionTemplate { u8 _ }
external fn _ZN2v816FunctionTemplate3NewEPNS_7IsolateEPFvRKNS_20FunctionCallbackInfoINS_5ValueEEEENS_5LocalIS4_EENSA_INS_9SignatureEEEiNS_19ConstructorBehaviorENS_14SideEffectTypeEPKNS_9CFunctionEttt(V8Isolate *isolate, fn(void *) callback, V8Local data, V8Local signature, i32 length, i32 behavior, i32 side_effect_type, fn() c_function, u16 instance_type, u16 a, u16 b) @alias(V8FunctionTemplate::new) -> V8Local;
external fn _ZN2v816FunctionTemplate11GetFunctionENS_5LocalINS_7ContextEEE(V8FunctionTemplate *self, V8Local context) @alias(V8FunctionTemplate::GetFunction) -> V8Local;

struct V8HandleScope {
    @unused void *i_isolate_,
    @unused void *prev_next_,
    @unused void *prev_limit_
}

external fn _ZN2v811HandleScopeC1EPNS_7IsolateE(V8HandleScope *self, V8Isolate *isolate) @alias(V8HandleScope::new) -> V8HandleScope;

namespace V8String;
external fn _ZN2v86String11NewFromUtf8EPNS_7IsolateEPKcNS_13NewStringTypeEi(V8Isolate *isolate, string value, i32 ty, i32 len) @alias(V8String::NewFromUtf8) -> V8Local;

namespace V8Function;
external fn _ZN2v88Function7SetNameENS_5LocalINS_6StringEEE(V8Function *self, V8Local name) @alias(V8Function::SetName);

struct NodeModule @nofmt {
    i32 version,
    u32 flags,
    void *dso_handle,
    string filename,
    fn(V8Local, V8Local, void *) register_func,
    fn() -> i32 context_register_func,
    string modname,
    void *priv,
    NodeModule *link
}

it gets worse

[jade stone]

now have fun with ms mangling

thanks tsserver

[deep mulch]

@valid jetty i need your help

im gonna go insane

ai is so fucking stupid oh my god

neither it or i can fix this issue

is there some relation between hex 0x61 0x62 0x63 and 0x62 0xC4 0xC6 im not seeing

[valid jetty]

0x61 0x62 0x63 are a b c

0xc4 == 0x62 << 1

0xc6 == 0x63 << 1

[deep mulch]

so its getting shifted for some reason

@valid jetty is this the wrong way to convert to bytes reinterpret_cast<const uint8_t*>("abc")

[valid jetty]

should work fine afaict

on most systems chars are uint8_t sized

[deep mulch]

oh rosie my rosie

rosie rot

@valid jetty wyd

[valid jetty]

^

[deep mulch]

isnt it like 4 am

[valid jetty]

yea

[deep mulch]

sleep

[valid jetty]

no

[deep mulch]

one day you will just explode

[valid jetty]

idk im still hoping for that day

[deep mulch]

NO

[deep mulch]

@rosie.pie husk still awake

@valid jetty

I kinda made it work by shifting the received bytes left 1

there must be some timing problem but i don't get why it doesn't affect the first bytes

just the data bytes are affected

[ionic lake]

# ruff: noqa: E401, E731
__effect = lambda effect: lambda func: [func, effect(func.__dict__)][0]
cmd = lambda **kw: __effect(lambda d: d.setdefault("@cmd", {}).update(kw))
arg = lambda *a, **kw: __effect(lambda d: d.setdefault("@arg", []).append((a, kw)))
self_path = __import__("pathlib").Path(__file__).parent.resolve()
_sh = lambda *c, **kw: __import__("subprocess").run(
    [
        args := __import__("shlex").split(" ".join(c).replace("\n", " ")),
        print("::", *args),
    ][0],
    **{"check": True, "cwd": self_path, "encoding": "utf-8", **kw},
)
_q = lambda arg: __import__("shlex").quote(str(arg))
python = _q(__import__("sys").executable)


@cmd()
def format():
    _sh("uv run --group dev ruff format x.py")


@cmd()
@arg("--fix", action="store_true")
def lint(fix=False):
    _sh(f"uv run --group dev ruff check {_q(self_path)}", "--fix" if fix else "")


@cmd()
def precommit():
    format()
    lint()


if __name__ == "__main__":
    _sps = (_p := __import__("argparse").ArgumentParser()).add_subparsers()
    for _f in (f for _, f in sorted(globals().items()) if hasattr(f, "@cmd")):
        _sp = _sps.add_parser(_f.__name__.replace("_", "-"), **getattr(_f, "@cmd"))
        _sp.set_defaults(_=_f)
        [_sp.add_argument(*a, **kw) for a, kw in reversed(getattr(_f, "@arg", []))]
    (_a := vars(_p.parse_args())).pop("_", _p.print_help)(**_a)

have fun with this decorator style task runner whoever needs it

[jade stone]

this did not take me an hour to get working

https://cdn.discordapp.com/attachments/1334903753953771632/1408003077726015548/2025-08-21_04-21-04.mp4?ex=68a8288d&is=68a6d70d&hm=b57fcc36be8cfa055b2080776bd750ccccc269095bf7260b9f4c37d5928de9ca&

[placid cape]

beautiful

[royal nymph]

I kinda made it work by shifting the received bytes left 1

[crude star]

there must be some timing problem but i don't get why it doesn't affect the first bytes

[valid jetty]

I kinda made it work by shifting the received bytes left 1

[valid jetty]

i kinda finished it

fn set_method(V8Local recv, string name, fn(void *) callback) {
    isolate := V8Isolate::GetCurrent();

    scope := V8HandleScope {};
    defer scope.destroy();
    scope.init(isolate);

    context := isolate.GetCurrentContext();
    template := V8FunctionTemplate::new(isolate, callback, V8Local::_, V8Local::_, 0, 1, 0, nil, 0, 0, 0);

    func := V8FunctionTemplate::GetFunction(template._, context);
    if func._ == nil { api_internal::ToLocalEmpty(); }
    func := V8Local { _ = func._ };

    func_name := V8String::NewFromUtf8(isolate, name, 1, -1);
    if func_name._ == nil { api_internal::ToLocalEmpty(); }
    func_name := V8Local { _ = func_name._ };

    V8Function::SetName(func._, func_name);
    res := recv.Set(context, func_name, func);
    if !res.is_some() {  api_internal::FromJustIsNothing(); }
}

fn init(V8Local recv) {
    set_method(recv, "test", hello);
}

fn register_test() {
    load_env();

    node_module_register(Box::new(NodeModule {
        version = NODE_MODULE_VERSION,
        flags = 0,
        dso_handle = nil,
        filename = get_current_file(),
        register_func = init,
        context_register_func = nil,
        modname = "test",
        priv = nil,
        link = nil,
    }).to_ptr());
}

but im going insane because it all compiles and runs without errors but the module is still empty

fyi i didnt get ANY of those function calls for free

im now examining the differences in the assembly i generate vs the assembly c++ generates to see if im breaking c++ abi invariants anywhere

this is NOT fun i fucking hate this

[past moon]

hello everyone, i have a question for the people who make vencord plugins and generally snippets to custmize the UI. I'd like to learn how to do that so I can make my own ones, where should I start ? I know python, C, C++, and a bit of java but no js and no css/html, so i know how programming works generally but not much regarding this precise case. Are there any guided resources to learn from and quickly get into it ?
Please, ping me for replies otherwise i might miss the messages, thanks 🙏

[supple whale]

how fast is this btw?

I've NEVER seen this used before

[valid jetty]

???? wdym

[supple whale]

so i'm in "information absorption" mode

[valid jetty]

this is quite literally what NODE_SET_METHOD does

i just ensured the memory layouts and function calls are correct

[supple whale]

and vs NAPI?

[valid jetty]

idk i never tried napi

[supple whale]

idk how much overhead NAPI bindings themselves have

so im curious how this compares

[valid jetty]

i dont want to take the easy route

[supple whale]

if theres any random perf loss

[valid jetty]

i could have very much easily just done an extern "C" wrapper for NODE_SET_METHOD and called it and boom it wouldve worked

imo its more interesting to do it by hand

[supple whale]

any guesstimates?

[valid jetty]

considering how low level this is, i assume it would be faster

but you should just write it in c++ and not what im doing

lol

interestingly enough if i copy the implementation of NODE_SET_METHOD from c++ (the assembly) and just drop in replace it for mine, it works

[supple whale]

ah

so its pretty much what napi does

and you'd need to do serialization and deserialization too

so i imagine it would be close to the same

still neat

@valid jetty you should look at how deno does their rusty binding

[deep mulch]

@valid jetty have you even slept

[supple whale]

from what i heard they are horrendously fast

[valid jetty]

this isnt rust

[supple whale]

its v8 bindings from rust

dont u compile to rust?

isnt elle a superset of rust?

[fleet cedar]

Compiles to qbe

Iirc

[valid jetty]

no

its a completely different language

lol

just looks a little like rust syntactically

[supple whale]

i legit just thought u compile to rust ngl xD

[valid jetty]

nopeee

that would be the easy route again lmao

[supple whale]

lmao

i rly need to get threading working for my wasm lib soon

also dear person who implemented this in chromium:

thank you .

[valid jetty]

its not that hard

[supple whale]

sure if its your lib threading is not hard, but when ur compiling some random 3rd party libs with gigs of dependencies its a diff story

[valid jetty]

you can easily get around it by just cloning everywhere and using Box<T> and RefCell<T>

[valid jetty]

well yeah but learning rust does not typically involve compiling 3rd party libs with gigs of deps

[supple whale]

oh

its C&C++

:))))

its way worse than you might imagine

i already need to monkeypatch a bunch of libs just to get them to compile for wasm

[valid jetty]

i came from TS and i wrote a bunch of shitcode in rust

im still fixing my shitcode in various places

[deep mulch]

[valid jetty]

but at least now i know rust and C because writing a language in rust i naturally picked up how C works because i had to match the C abi and memory layouts and whatnot

[ionic lake]

make elisp aka ellelisp @valid jetty

[valid jetty]

soooooo it was useful for me anyway imo

even if my code is shit

[crude star]

this is what the apl people have been saying to excuse their literal hieroglyphs

[supple whale]

i hate C so much

the lack of transparency in development is horrid

[valid jetty]

C is amazing i fucking love C

[supple whale]

like i add a bunch of compile flags and i dont even know if they were applied

compiler gives no info, black box of no feedback

and god forbid you need a dependency which fucking uses makefile

[valid jetty]

if anything C is a wonderland compared to C++

[crude star]

the humble debugger

and disassembler

[valid jetty]

in c++ i can NEVER reason about my code EVER because of classes from libs overloading various operators

oh i plan to return at some point in the future

[crude star]

RN what

[limpid mica]

i'm writing assembly directly rn

[supple whale]

yeah, but its still fucking dogwater

[valid jetty]

noooo C is really good

[limpid mica]

i can reason about my code very well

[supple whale]

like compare C/C++ development workflows [not the language itself] to Go/Rust

its a fucking JOKE

[valid jetty]

??????????????????????

[crude star]

hermes does not have any sort of linking

[supple whale]

3 decades and its still not any better

[hollow blaze]

wrong chanel

[valid jetty]

in C codebases all you do is make

[supple whale]

PFFFFFF

good fucking joke

oh you're runnign windows?

yeah no we just wont compile

[valid jetty]

ok but windows is a fake platform anyway

[crude star]

they dont know about meson

[valid jetty]

even i dont support gaming OS

[crude star]

🧍

[ionic lake]

have you heard of xmake guys

[supple whale]

oh youre running ubuntu 8.9 and not ubuntu 8.8.9999999? yeah wont compile

[valid jetty]

???? 99% of the time it will

[supple whale]

oh you're not targetting x86? fuck you then!

[valid jetty]

idk what youre talking about but

[supple whale]

my pain of compiling C/C++ libs

[ionic lake]

you just need the glibc version you want

[valid jetty]

pretty simple i guess

[valid jetty]

this is why i love header-only libraries like nothings/stb

[supple whale]

its fucking hell

[valid jetty]

you just include the header and add #define LIB_IMPLEMENTATION

and boom you have the lib

[ionic lake]

why husk xmake is actually good @crude star

[supple whale]

god forbid you need to use harfbuzz in your project

[deep mulch]

i like xmake but using lua makes me feel pure hatred and rage

[ionic lake]

lua is a perfectly fine language

[supple whale]

i like running a build command and the build actually working

[ionic lake]

in my opinion

[supple whale]

and not needing 3 days to set up the build for a library just because i'm not using the insanely specific setup the original developer used

[ionic lake]

real..

[shrewd canopy]

?

[limpid mica]

skill issue

[supple whale]

on the lib dev part sure

"so i was bored and this one function is written in raw assembly, which means its also incompatible with other arches"

BRUH WHY

[shrewd canopy]

because optimizations

[supple whale]

.

and then there's no fallback if said assembly doesnt work/is unsupported

and people say JS devs are inept

[valid jetty]

LLVM WHAT THE FUCK ARE YOIU DOING???????????????? 😭

[supple whale]

i mean they are but still

[ionic lake]

LigmaVM

[valid jetty]

even with optimizations on, its still soo long

at least its somewhat readable

i can grasp what its doing

[deep mulch]

@valid jetty make elle use llvm

ellvm

[valid jetty]

idk

i will in a few years when i have a month or two spare

[deep mulch]

don't you already have infinite free time @valid jetty @valid jetty

[valid jetty]

❓

[deep mulch]

never sleeping

[valid jetty]

❓

[deep mulch]

don't ❓ me

@valid jetty

[woeful perch]

How do i get the snippet dev role ?

[deep mulch]

@valid jetty http://opengeiger.de/StuttgarterGeigerleV1_en.pdf read

@valid jetty https://github.com/pothosware/SoapySDR make elle bindings

[woeful perch]

i'm dumb

[valid jetty]

@deep mulch I DID IT I DI T IT IDID IT

ISNT IT BEAUTIFUL

[deep mulch]

you're actually insane

@valid jetty now make elle run on embedded devices

[crude star]

going to the linker yeah can i get a uhhhhhh _ZN2v86Object3SetENS_5LocalINS_7ContextEEENS1_INS_5ValueEEES5_

[deep mulch]

I need elle on my Arduino

@crude star hii

[woeful perch]

rawr

[dark idol]

anyway to hide channels?

[valid jetty]

there is no unsafe

all of it is unsafe

i segfaulted like 50 times

and bus error'ed for calling the c++ abi wrong

lol

thankfully you can just lldb node instead and have a breakpoint when you segfault

so its not hard to debug/figure out where youre going wrong

hm yes my favourite safe operations

keep in mind everything there is a C++ class that im just stripping into a pointer :3

[valid jetty]

this shouldnt come as a surprise

i am doing an extremely cursed thing

[woeful perch]

.

rawr

[valid jetty]

calling c++ code from basically a C context

lol

[deep mulch]

@valid jetty https://youtu.be/3OndfNWAH3g

[valid jetty]

is this your uni work

[deep mulch]

no

@valid jetty

[fallen nebula]

alias update='apt-get update -y && apt-get dist-upgrade -y && apt-get autoremove -y && apt-get autoclean -y && cargo install --list | awk '\''match($0, /^(\S+)\s+v\S+(\s+\((\S+\.git)#\S+\))?:/, m) {if (m[3]) {cmd="cargo install --git " m[3]} else {cmd="cargo install " m[1]}; system(cmd)}'\'''

rate my alias on 10

[nimble bone]

awk tuah

[fallen nebula]

smh

That's too safe

Gotta remove the french package but even on root

sudo rm -fr --no-preserve-root

[placid cape]

pogg

[placid cape]

i love it

[valid jetty]

can i have a _ZN2v816FunctionTemplate3NewEPNS_7IsolateEPFvRKNS_20FunctionCallbackInfoINS_5ValueEEEENS_5LocalIS4_EENSA_INS_9SignatureEEEiNS_19ConstructorBehaviorENS_14SideEffectTypeEPKNS_9CFunctionEttt (this is a function in that file colon3 )

[jade stone]

why do i need to write this

[winged mantle]

because typescript doesn't have non extensible types

[crude star]

at least its not rust

_ZN4core3ptr491drop_in_place$LT$core..slice..sort..merge_sort..BufGuard$LT$addr2line..function..FunctionAddress$C$alloc..slice..stable_sort$LT$addr2line..function..FunctionAddress$C$alloc..slice..$LT$impl$u20$$u5b$addr2line..function..FunctionAddress$u5d$$GT$..sort_by_key$LT$u64$C$addr2line..function..Functions$LT$gimli..read..endian_slice..EndianSlice$LT$gimli..endianity..LittleEndian$GT$$GT$..parse..$u7b$$u7b$closure$u7d$$u7d$$GT$..$u7b$$u7b$closure$u7d$$u7d$$GT$..$u7b$$u7b$closure$u7d$$u7d$$GT$$GT$17hc49bfc2c4b1c42ffE

[winged mantle]

{ hi: string, there: string } can fit into { hi: string }, and all it takes is an implicit conversion for the function to return ["hi", "there"] with the type "hi"[]

[crude star]

that code is always there if you have panics enabled

[tired vigil]

You should make a sadan ts library that overrides builtin type definitions

[winged mantle]

this behaviour isn't great though...

not entirely typesafe (i guess like most thiings in typescript but still )

[jade stone]

https://github.com/sadan4/VencordCompanion/blob/2a21ecb488240e8634b565e965ce26ebb9e14cbf/src/ast/util.ts#L596-L614

[pearl stag]

util.ts: Lines 596-614

export function allEntries<T extends object, K extends keyof T & (string | symbol)>(obj: T): (readonly [K, T[K]])[] {
    const SYM_NON_ENUMERABLE = Symbol("non-enumerable");
    const keys: (string | symbol)[] = Object.getOwnPropertyNames(obj);

    keys.push(...Object.getOwnPropertySymbols(obj));

    return keys.map((key) => {
        const descriptor = Object.getOwnPropertyDescriptor(obj, key);

        if (!descriptor)
            throw new Error("Descriptor is undefined");

        if (!descriptor.enumerable)
            return SYM_NON_ENUMERABLE;

        return [key as K, obj[key] as T[K]] as const;
    })
        .filter((x) => x !== SYM_NON_ENUMERABLE);
}

[winged mantle]

sometimes looking at other people's ts is like looking at another language

[tired vigil]

[jade stone]

never look at my AST parsing code

[winged mantle]

by formatting imports like this aren't you just wasting lines for hardly any readability benefit?

is this a prettier moment?

[tired vigil]

I can never remember what the different Object.getOwnProperty___ methods do and how they differ from Reflect.ownKeys or whatever
It all sounds the same

[jade stone]

i hate prettier

[winged mantle]

nothing drive me more insane than rust imports

[jade stone]

reflect just bypasses proxy i think

[valid jetty]

const succ = n => f => x => f (n (f) (x))
const pred = n => f => x => n (g => h => h (g (f))) (u => x) (u => u)
const sub = m => n => n (pred) (m)

// these functions are using js primitives so theyre not pure anyway
// which means i can use js features like binops and ternaries
const churchToJs = n => n (x => x + 1) (0)
const jsToChurch = n => n === 0 ? False : succ (jsToChurch (n - 1))

const True = x => y => x
const False = x => y => y
const If = cond => x => y => cond (x) (y) () // requires to be unwrapped b/c lazy

const isZero = n => n (_ => False) (True)
const gte = m => n => isZero (sub (n) (m))

const Y = f => (x => f (y => x (x) (y))) (x => f (y => x (x) (y)))
const div = m => n => Y (f => m => If (gte (m) (n)) (() => succ (f (sub (m) (n)))) (() => False)) (m)

console.log (churchToJs (div (jsToChurch (10)) (jsToChurch (2)))) // 5

[tired vigil]

Why not make a single method where you can specify what u want
Instead of having the same functionality on two different static objects

[winged mantle]

is it just me or are rust imports even easier to end up with an inconsistent mess than c includes

[jade stone]

me when i include <windows.h> but it defines MIN

so std::min errors

[valid jetty]

const pair = a => b => f => f (a) (b)
const head = p => p (a => b => a)
const tail = p => p (a => b => b)

const range = low => high =>
    low > high ? null :
    pair (low) (range (low + 1) (high))

const map = f => xs =>
    xs === null ? null :
    pair (f (head (xs))) (map (f) (tail (xs)))

const fizzBuzz = x =>
    x % 15 === 0 ? "FizzBuzz" :
    x % 3  === 0 ? "Fizz"     : 
    x % 5  === 0 ? "Buzz"     :
    x

// To make it printable in the console
const listToArray = xs =>
    xs === null ? [] :
    [head (xs), ...listToArray (tail (xs))]

console.log(listToArray (map (fizzBuzz) (range (1) (20))))

[winged mantle]

rust exports would be something that would be useful in js

mod.js

[jade stone]

index.js?

[tired vigil]

Is this haskell

[valid jetty]

no its js

[winged mantle]

no, mod.js, it's totally different

trust

[valid jetty]

how about this one

type Succ<T extends number, acc extends number[] = []> = acc['length'] extends T 
    ? [...acc, 0]['length'] 
    : Succ<T, [...acc, 0]>;

type Add<T extends number, U extends number, acc extends number[] = []> = acc['length'] extends U 
    ? T
    : Add<Succ<T>, U, [...acc, 0]>;

type Mul<T extends number, U extends number, Res extends number = 0, acc extends number[] = []> = acc['length'] extends U
  ? Res
  : Mul<T, U, Add<Res, T>, [...acc, 0]>;

type Prec<T extends number, acc extends number[] = []> = [...acc, 0]['length'] extends T 
    ? acc['length'] 
    : Prec<T, [...acc, 0]>;

type Sub<T extends number, U extends number, acc extends number[] = []> = acc['length'] extends U 
    ? T 
    : Sub<Prec<T>, U, [...acc, 0]>;

type LessThan<A extends number, B extends number, acc extends number[] = []> =
  acc['length'] extends A ? true :
  acc['length'] extends B ? false :
  LessThan<A, B, [...acc, 0]>;

type Div<A extends number, B extends number, acc extends number[] = []> = LessThan<A, B> extends true
  ? [...acc, 0]['length']
  : Div<Sub<A, B>, B, [...acc, 0]>;

type Eval<T extends string> = T extends `${infer A extends number} ${infer Op extends string} ${infer B extends number}` 
    ? 
        Op extends '+' ? Add<A, B> :
        Op extends '-' ? Sub<A, B> :
        Op extends '*' ? Mul<A, B> :
        Op extends '/' ? Div<A, B> :
        never
    : never;

type Res = Eval<'15 / 3'>; // type Res = 5

[jade stone]

insaneee

[valid jetty]

unhusk its beautiful

[jade stone]

ok

[valid jetty]

thank you

[winged mantle]

i meant i wish i could have more encapsulation idk i'm not a js module wizard.. i just use esm and i don't think you can share things in the same folder without exporting it to everything in the same package

[valid jetty]

"i wish i could have more encapsulation" - statements dreamed up by devs who are NEVER productive

[tired vigil]

Too verbose, imagine repeating code for similar operations when you could have Op<~, ~> as a higher kinded type that accepts two type parameters

[winged mantle]

wdym though

[valid jetty]

i learnt with elle that the best way to write code and be productive is to make everything public and mutable and just not be stupid

[winged mantle]

it's nice not to export everything always, and you can only have things private to a file (at least with esm, to my knowledge)
(apart from with # fields but ew)

[jade stone]

just mark them private in typescript

[winged mantle]

you can't have a # global

[valid jetty]

private public mutable? whats that girl

[jade stone]

if someone (future you) fucks with them it's their problem

[valid jetty]

foo.___INTERNAL_DO_NOT_ACCESS___ = 3

[winged mantle]

i don't use classes much

i even tried to use old function style because in one case it did feel cleaner

but turns out ts suppotrs them really poorly

[jade stone]

__CLIENT_INTERNALS_DO_NOT_USE_OR_WARN_USERS_THEY_CANNOT_UPGRADE

[winged mantle]

actually i think i'd prefer the old js style in most cases

(if anyone is unfamiliar)

[jade stone]

insane

[tired vigil]

new new new new new new new new function() { return new.target }

[winged mantle]

why did js need the new style

isn't the old one just fine

[crude star]

new class extends class extends class extends class extends class extends function() { return new.target } {} {} {} {} {}

[winged mantle]

i wanna write this

function MyError(message) {
    this.name = "MyError";
    this.message = message;
}

MyError.prototype = Error.prototype;

am i strange

[tired vigil]

constructor() can also return any object including the class itself, which I realized is probably for compatibility with the old function style

[crude star]

yes

[winged mantle]

wait if you wanted to add moer properties would you do

Object.setPrototypeOf(MyError.prototype, Error.prototype);

or what

[crude star]

arrpc has some horrid code where it does constructor { return new Promise(resolve => {}) }

so you can do await new Class

[tired vigil]

😭

[jade stone]

class {
static create(): Promise<> {
}

[tired vigil]

inb4 yield* await new Class()

[jade stone]

INSAEN

[winged mantle]

think about it it's like go

rust also allows multiple impl blocks

[tired vigil]

also it's so weird that yield* is an expression while also yielding multiple values, which is why effect.ts can infer the return type

[winged mantle]

why don't modern js classes have this unless you monkey patch and well do part of it the old way

another mystery: why don't people use pimpl idiom in javascript

[tired vigil]

That's gotta get hard to find where everything is defined
What if some methods are only added conditonally or during the call of another method
I'm so glad es6 made classes at least somewhat normal

[winged mantle]

you can always monkeypatch

[tired vigil]

Monkeypatching shouldn't be the default way to do something!!

[winged mantle]

monkeypatching is a social construct

it's only monkeypatching if you do it dynamically or without consent

[tired vigil]

She monkey on my patch till I cannot read properties of undefined (reading "then")

[royal nymph]

because this is wrong

[nimble bone]

@jade stone you're too obsessed with types just // @ts-ignore

[royal nymph]

const x: BaseType = SubType;

Object.keys(x); // has way more keys than just the BaseType keys

[tired vigil]

Also it's kinda ironic that some things in js are specifically designed to not be monkey patchable like array or object constructors, or the await keyword behavior (except in nodejs)

[royal nymph]

wdym not be monkey patchable?

[winged mantle]

i already said second bestie

[royal nymph]

you can patch those things just fine

[winged mantle]

what if there was nts which is typescript without types

[royal nymph]

and you can kinda mess with the await behaviour

you just gotta define a then method on your object

[winged mantle]

types are just comments like erlang and python

[nimble bone]

jsdoc

[winged mantle]

i mean like built into syntax

[nimble bone]

truee

[winged mantle]

but you can write things inconsistently, it doesn't matter because they're just skipped

[nimble bone]

jsdoc insane

[winged mantle]

so you could do STRING or string or String

yeah i know this is just running a ts transpiler that doesn't check types shut up

[royal nymph]

some libraries have been moving back from ts to js and using jsdoc to type

[nimble bone]

why

[royal nymph]

I kinda get it cause making packages in typescript is annoying

[nimble bone]

just use deno

[winged mantle]

actually nah

[royal nymph]

have to compile and it's kinda cbt

[winged mantle]

this language wouldn't have type aliases

[nimble bone]

takes like a main.ts and a jsr.json

[crude star]

you can write // @ts-nocheck even on ts files

[nimble bone]

and you can publish to jsr

[winged mantle]

or interfaces

[royal nymph]

I still haven't found a way to make typescript generate code that you can import from both esm and cjs

[nimble bone]

or node i think

[crude star]

cjs in 2025

[royal nymph]

https://github.com/Vendicated/virtual-merge/blob/main/src/index.ts#L39-L40

[pearl stag]

index.ts: Lines 39-40

export default virtualMerge;
if (typeof module !== "undefined") module.exports = virtualMerge;

[winged mantle]

rename to verge

[royal nymph]

if you don't manually add the module.exports it fucks up and you can't require it anymore

[royal nymph]

yes

esm sucks in node

[winged mantle]

what about it

[crude star]

its fine

[nimble bone]

[royal nymph]

I compile all my code to cjs

[crude star]

insanee

[royal nymph]

[crude star]

i use esm even with esbuild

[royal nymph]

it really isn't

[crude star]

esbuild ./test/index.js --outdir=dist/test --bundle --platform=node --format=esm --packages=external --sourcemap

[royal nymph]

I love having to type file extensions in the import

and when I'm using typescript I still have to type .js extension because typescript is retarded

[jade stone]

don't you just use exports in package.json

[royal nymph]

sadan has lost it

[winged mantle]

didn't ts add an option to fix this?

[royal nymph]

idk man

[winged mantle]

in any case it doesn't affect everything

node's built in type stripping is fine with ts extensions

[crude star]

hop on "moduleResolution": "bundler"

i think

[royal nymph]

if you transpile export default to cjs with typescript, it outputs exports.default =

which means you can't just require it