🪅-progaming

[winged mantle]

i have a pretty vague idea of what linking is

[crude star]

dynamic linking is when you statically link at runtime

self-modifying code and all

[spark tiger]

i got demoted.

[winged mantle]

oh yea there's stuff like static where linking actually is relevant to the programmer

just add a uuid to the end of the name

[jade stone]

Oh, neat!

[crude star]

deserved

[median root]

whats the actual difference between JIT and interpreted? Every damn "explaination video" I see tells me "Its the computer reading the code" or the "computer directly executing instructions" but last I checked computers didnt just "read code". My other thought was that it could be translating source code into another language to be executed as the program runs, however is this not transpiling?

[frosty obsidian]

interpreted languages don't get compiled

jit compiles as its executed rather than beforehand

[fleet cedar]

Interpreted iterates over the list of instructions as it runs, jit generates code that runs each instruction

But there's certainly some gray zones

[frosty obsidian]

jit is a mix of interpreted and aot

[deep mulch]

wing detected

[fleet cedar]

Modern cpus have builtin jits

[median root]

yeah im asking how this is possible, since computers dont just "read code" and translating it is just transpiling no?

[fleet cedar]

The processor itself iterates over each instruction as it runs and generates micro-ops that it then runs

[crude star]

the compiler interprets the code and outputs new code

fact

[frosty obsidian]

the interpreter does the actual execution

[median root]

and im guessing I must build my own interpreter to understand this?

[fleet cedar]

It is very educational, yes

[median root]

or how this is done?

[frosty obsidian]

you can look into existing interpreters

[median root]

I dont wanna read pythons code 😭

[crude star]

you really don't

cpython is hell

[shrewd canopy]

then read Ruby's interpreter code

[limpid mica]

i love interpreters

[crude star]

ruby the language where random amounts of periods is valid syntax ```
irb(main):001> ..
irb(main):002> ...
irb(main):003* ....
(irb):2: warning: ... at EOL, should be parenthesized?
internal:kernel:168:in 'Kernel#loop': (irb):3: syntax errors found (SyntaxError)
1 | ..
2 | ...

3 | ....
| ^ unexpected '.'; expected an expression after the operator
| ^ unexpected end-of-input; expecting a message to send to the receiver

from /usr/lib/ruby/gems/3.4.0/gems/irb-1.14.3/exe/irb:9:in '<top (required)>'
from /usr/bin/irb:25:in 'Kernel#load'
from /usr/bin/irb:25:in '<main>'

irb(main):004> .....
irb(main):005> ......
irb(main):006* .......
(irb):5: warning: ... at EOL, should be parenthesized?
internal:kernel:168:in 'Kernel#loop': (irb):6: syntax errors found (SyntaxError)
4 | .....
5 | ......

6 | .......
| ^ unexpected '.'; expected an expression after the operator
| ^ unexpected end-of-input; expecting a message to send to the receiver

from /usr/lib/ruby/gems/3.4.0/gems/irb-1.14.3/exe/irb:9:in '<top (required)>'
from /usr/bin/irb:25:in 'Kernel#load'
from /usr/bin/irb:25:in '<main>'

irb(main):007>

[limpid mica]

unexpected .

[deep mulch]

.

[shrewd canopy]

these are Ranges

[frosty obsidian]

zeet loves kotlin partially being interpreted

[limpid mica]

range my ranges

[deep mulch]

@frosty obsidian

[crude star]

oh my god

irb(main):005> ..............0
=> ..............0

[limpid mica]

yeah

tho i must wonder what version of ruby you're on

[crude star]

3.4.4

python is better than ruby because they have the bigger 3.x version

go is the worst because they're still on 1.x

and said there wont be a 2.x

[limpid mica]

sorry i went

[deep mulch]

you went

[supple whale]

greaaaaaaaat

chrome introduced a regression in ffmpeg in v136

which is still in all active stable chromium versions

and it fucks seeking and loading

does this effectively

and its been like this for MONTHS

and they only fixed it in 140.0.7299.0

which is a dev build released a week ago

https://chromium.googlesource.com/chromium/src/+/4a9f8b15c5a77949d79bb972955601af1c9c0af9

mmmmmm thank you

over half a year of this shit being fucked

[ornate quiver]

compiled -> a compiled compiler translates source into machine code that is able to be directly run
interpreted -> a compiled runtime reads source code, parses it into it's own bytecode, and virtually runs it (hence, a "virtual machine")
JIT -> an addition to a an interpreter in a runtime, that transforms the parsed bytecode into real machine code, which provides a marginal performance boost as it can be directly run by the cpu, instead of being virtually run by the interpreter
AOT -> a loose combination of compiled and JIT, where portions of the bytecode is transformed into machine code ahead of time, just like compiled, but run from the context of a runtime (like in GraalVM, or .NET with AOT on)

the point is that either the cpu directly can run instructions it supports, or it can run the instructions of an interpreter, which will virtually "run" custom bytecode (aka. a VM), and accomplish the same thing

[fleet cedar]

So much to nitpick here. I'll settle for yeeting the word "marginal" onto a collision course with pluto

I wouldn't call several orders of magnitude marginal

[ornate quiver]

well I guess it really depends on the situation but yeah JIT can be really good

[runic sundial]

Yeah what

[ornate quiver]

I worded that as dumbed down as I could and it ended up being husk

[runic sundial]

JIT is the difference between a cool and good or warm and bad interpreted lang

Try running Minecraft world gen with JIT disabled and come back saying "marginal"

Prolly the best example

[ornate quiver]

it really matters for performance critical areas but not so much for other stuff
which is why android does profiling before JITing methods to not waste resources on compiling methods that won't yield a significant performance increase

[runic sundial]

Not doing JIT means you're getting the worst-case

Which makes a lot of sense for Android UIs, where the code paths need to be fast IMMEDIATELY

And the compiled caches prolly won't last as soon as the user switches to another app

[ornate quiver]

i believe it's persisted

[runic sundial]

TL;DR it matters anywhere where you're gonna be doing the same thing a lot, over and over for some period of time

[fleet cedar]

Even a very simple jit that basically compiles each bytecode into an asm call instruction gets 5-15% perf over an interpreter loop https://v8.dev/blog/sparkplug

[runic sundial]

The real power of JIT is that you have all the context in the world at runtime

You know with certainty what code path is hot or cold

[crude star]

FALSE

they grow downwards

[fleet cedar]

You're the other half then

May I offer an unhelpful compromise? I think stacks don't grow

[crude star]

who needs stacks

just have infinite registers

[runic sundial]

Don't care what direction the stack goes

In memory or otherwise

It's a stack.

When I stack things the stack grows taller.

[fleet cedar]

Fact

Direction is useful for metaphor and visualization though

[runic sundial]

They should have called it the mineshaft

If it needed to go down

Anyway who the hell visualizes the stack as going down lmao

[runic sundial]

You love memory

How do you feel knowing that Java is more memory safe than Rust?

It's true.

[valid jetty]

i visualize it as a long rope

the stack and heap are sort of touching

and the stack grows downwards towards one end of the rope while the heap grows upwards towards the other end

[winged mantle]

OK tbh i was lying in bed thinking of ideas for Unicorn so maybe I should work on that instead
(Unicorn was the name I thought of for a scripting language but i didn't think of many actual features )

[valid jetty]

this is probably a wrong way to think about it

but it works for me

(and that pivoting point in the middle changes as the stack requires more memory)

[valid jetty]

compiler or interpreter

[winged mantle]

interpreter

[valid jetty]

😭

[winged mantle]

i've wanted to work on an interpreter for a long time

seems fun

[fleet cedar]

Quite wrong indeed

Most importantly, heap is not necessarily consecutive

[valid jetty]

well yeah because of fragmentation

[fleet cedar]

And I think stack is usually closer to the end of the address space, with heap below it

But I'm not sure about that

[valid jetty]

but i was thinking more about how one grows to one end and the other grows to the other

[fleet cedar]

Especially since with threading or whatever there can be multiple stacks

[valid jetty]

with virtual memory you can't really reason about any of this in real code but yeah

[fleet cedar]

Yeah

[valid jetty]

in an ideal world my model would work i think

[fleet cedar]

In the grand scheme of things, the stack itself is just another heap allocation

[valid jetty]

fair

actually that's so real

the stack is just a bump allocator

[fleet cedar]

Kinda yeah

[ionic lake]

go back into ur stack

get to work

[fleet cedar]

Compiler has a bit more control over it than a typical bump, but conceptually yeah

[valid jetty]

hey i am NOT a plate

[winged mantle]

so good

entrypoint {
    println("hello world")

    entry := ScoreboardEntry.make({ name: "joe", score: 10000 })
    println(entry.toString())
}

type ScoreboardEntry {
    name
    score

    meth toString() {
        return f"Name: {{name}}; Score: {{score.toString()}}"
    }
}

why don't any languages use the meth keyword to declare methods

[hoary sluice]

dynamic typing 😭

[hoary sluice]

yes you are

[valid jetty]

entrypoint yet meth

lmfao

i love consistency

[crude star]

entrpt {}

[valid jetty]

main {}

[limpid mica]

start {}

[valid jetty]

why is there a meth keyword anyway

you can use the left parenthesis to differentiate properties and methods

[crude star]

because func doesn't work there

its a method not a function

[fleet cedar]

Same reason as they don't use lsd to declare fields

[valid jetty]

yeah just have no keyword

use the fact that there is a left paren after the identifier to tell you that it’s a method rather than a property

[supple whale]

https://vm.tnktok.com/ZMS7mSUGo/

ewic.mp4 (@ewic.mp4)

❤️ 26.8k 💬 308 🔁 5.8k

▶ Play video

fxTikTok · Jul 23, 2025, 6:06 AM

why is this unironically good......

[elder yarrow]

I've sent you a DM with more info!

[austere idol]

8 spaces

[winged mantle]

if i make an editor i will make it so that multiple spaces always render as one to force you to use tabs

[supple whale]

yes officer, this man right here

[lavish frigate]

reactive frameworks hate this one simple trick! see how this hacky developer avoided installing 50gb of dependencies for reactivity using web components!

[ionic lake]

okay

[winged mantle]

i have reached the point of insanity in language design

[valid jetty]

??? what do you mean name of the type

is that not the thing next to the type keyword

[winged mantle]

yeah

and the types of type is the Type type which contains name

[deep mulch]

crazy

[winged mantle]

it's genius

[valid jetty]

wait

why is Type a type

oh dynamic typing

[winged mantle]

i mean go and java have the ability to inspect types too

[deep mulch]

add reflection to elle @valid jetty

[winged mantle]

the icing on the case is this is a word document that i am typing code into

[valid jetty]

you have some reflection

[deep mulch]

@valid jetty make elle library for generating elle code

[valid jetty]

you can do this

you cant do much more

[valid jetty]

thats just macros

[deep mulch]

nop

generating code during runtime

i mean like a elle builder

[winged mantle]

RTTI

[deep mulch]

.add_const("guh", 500)

[winged mantle]

is it viable to support method chaining with automatic semicolon insertion

[valid jetty]

you can do a little more

if you take in a generic value to ElleMeta you can not only get the type's name but also its expression as a string

thats how $dbg works

thats also how type_of works btw :3

[valid jetty]

sooooo

ElleJIT

[deep mulch]

@valid jetty make Elle run on JVM

we need

[winged mantle]

asi seems generally kinda bad

but it does look clean

i thought kotlin might do it well but apparently this breaks

[deep mulch]

"""
line 1
line 2
line 3
""".trimIndent()

?

[valid jetty]

elle

"line 1
line 2
line 3"

or ```rs
"line1\nline2\nline3"

[shrewd canopy]

Both

[valid jetty]

no im saying thats what it currently is

[deep mulch]

@valid jetty rosinga

[lavish cloud]

every time I see people working on langs I want to try making my own only to end up remaking kotlin

[winged mantle]

i have managed to design a language that doesn't resemble anything i've used that much

but it is kinda cursed

also i need to impl it

and i haven't really designed that much

it doesn't even have switch statements... first things first though

oh yeah it's a scripting language

my instincts kicked in and i added a rune type

maybe that's a bit weird?

[ornate quiver]

thats becuase its almost perfect

lolll

[lavish cloud]

so true

honestly the two improvements I'd want to see:

• static extension methods
• C++ interop in KNative (somehow)
  • If we can't have that, at least the ability to use any given toolchain so I can develop switch mods with it

[frosty obsidian]

technically you can extend the companion object but that would only work if the target actually has one

[shrewd canopy]

C++ ABI is still unstable

and its different between multiple compilers (although clang tries to be same as gcc)

[deep mulch]

also collection literals and destructuring

[lavish cloud]

Exactly, if there is no companion there's nothing you can do and that needs fixing (even if it's just a synthetic companion object)

[frosty obsidian]

theres a proposal for it

[deep mulch]

@frosty obsidian will make it get merged faster

[frosty obsidian]

you did the same url twice

[deep mulch]

guhhh

https://github.com/Kotlin/KEEP/discussions/442 i dont like this KEEP @frosty obsidian

they never heard of a linter before

[frosty obsidian]

i also don't like that

i don't see it seriously solving a problem

would just be a new thing libraries force on you

[deep mulch]

@frosty obsidian hi

[royal nymph]

terrible

idk how anyone could think this is a good idea

Apple moment

[deep mulch]

yop

[royal nymph]

I love when libraries decide how I'm supposed to write my code

[ornate quiver]

YES I NEED STATIC EXTENSION METHODS
this is literally my biggest gripe with kotlin
the support for statics is shit

[shrewd canopy]

Why do you need static extension methods

[ornate quiver]

i was doing some fancy syntax stuff a while back
related to dsl for patching and I needed static extension methods but there was literally no way to accomplish what i was trying to do

actually so bad

[shrewd canopy]

How

[lavish cloud]

Same :( ended up making a TExt object for T and ended up putting them there

[ornate quiver]

funniest thing is you can't even translate a java api into kotlin without breaking backwards compatibility
because you can only make statics on singletons for some fucking reason

[jade stone]

Can c++ interop even work given kotlin doesn't have destructors

[lavish cloud]

Knative uses destructors all the time to handle refcounts

[ornate quiver]

the jvm kinda has destructors, it's the finalize method on the Object class
but it's never guaranteed to be called, as that could only happen during GC I think

not the same thing but still interesting

[runic sundial]

Phantom refs can get u destructors

[jade stone]

does that work in jvm tho

[lavish cloud]

Well for JVM you can just write JNI/JNA

[lavish cloud]

The JVM equivalent is the Cleaner

Cleaner manages a set of object references and corresponding cleaning actions.
Cleaning actions are registered to run after the cleaner is notified that the object has become phantom reachable. The cleaner uses PhantomReference and ReferenceQueue to be notified when the reachability changes.

[ornate quiver]

oh this is cool wait

this perfectly solves an issue I was having

love

[lavish cloud]

Oh wait that's a neat idea

Writing a GC impl for C++

[deep mulch]

ooo I had no idea about this

I might could use this for ffmpegkt

[lavish cloud]

I just use javacpp bindings tbh

[deep mulch]

they're horrid

since it's just 1:1

[lavish cloud]

Well yeah

Writing a proper wrapper would be a nightmare

[deep mulch]

that's what I'm doing

works good

recently I added hardware accel encoding

how have I never heard of cleaner before guhh

[winged mantle]

isn't the whole advantage of having named parameters as a proper language feature that they're optional

in javascript it would be really annoying to make it optional as you literally implement it by accepting an object

why have a lexer when you can have a grokker

[royal nymph]

aren't Closables essentially destructors

you have to explicitly close them but still

[runic sundial]

I prefer AutoCloseable over cleaners

like

neconectar

In C++, a destructor is called when that thing goes out of scope

aka; you know WHEN it will happen for that exact resource

For cleaners, the GC will run whenever it wants.

[royal nymph]

in Java/Kotlin you have to explicitly close via a statement akin to C# using

[runic sundial]

Lombok @Cleanup

or try with resources

[royal nymph]

try (var thing = new CloseableThing()) {
    thing.explode();
}

CloseableThing().use {
     it.explode()
}

the try syntax is so cursed lmao

[runic sundial]

ya

with Lombok it ends up being:

{
    @Cleanup var thing = new CloseableThing();
    thing.explode();
}

And it will do the whole try(){...} for you

Very useful when you have more than one closeable

[fleet cedar]

Well, you can try-with-resources with multiple things

But not if you want to do other things between

[royal nymph]

you can make a closeablemanager for this

lmaooo

try (var closeables = new CloseableManager()) {
    var foo = closeables.add(new Foo());
    var bar = closeables.add(new Bar());
}

[winged mantle]

idk i prefer java syntax

the try syntax is so cursed... followed by more cursed syntax

epsecially if you name the var

CloseableThing().use { thing ->
    thing.explode()
}

ig that's the unfamiliar lambda syntax

[lavish cloud]

https://light-theme-hurts.my-ey.es/2U5xZLf.mp4
Made a funky thing that works with vanilla clients to emulate a zelda hookshot

[deep mulch]

your video cant be seeked

fix

[lavish cloud]

works fine for me

it's literally the most basic conversion of ```bash
owo_video () {
FILE=$(ls -Art1 "/home/mart/Videos" | tail -n 1) # most recent file
TMP=$(mktemp --suffix .mp4) # file to write to
ffmpeg -y -i "/home/mart/Videos/$FILE" $TMP >/dev/null # convert to mp4 with ffmpeg
owo -u $TMP # upload to owo
}

[deep mulch]

probably cause image host

[hoary sluice]

does seeking mean moving to a different time in the vid

[deep mulch]

yes

[spark tiger]

it doesnt work at all for me

oh now it does

[royal nymph]

love?

[shrewd canopy]

Browser moment

Just always use VLC or default Windows media player smh

[winged mantle]

somebody i was talking to thought vlc was spyware

[tired vigil]

real gaymers use potplayer

[shrewd canopy]

Whats potplayer 🐟

[supple whale]

unfortunately the only actually functional video players in the world are mpv and kodi

everything else falls short

potplayer lacks support for a bunch of formats, vlc doesnt do color mapping correctly

and most other players dont do hdr or dv correctly

chromium patched with extra codecs like ac3 and dts is third when it comes to accuracy and featureset

but it lacks subittle format support since it only has VTT

[deep mulch]

I love mpv

[royal nymph]

i hate mpv @deep mulch

[supple whale]

mpv sucks for users

gotta write 200 lines of config before its usable

and import like 14 luas

from a usability standpoint its fucking rancid

[deep mulch]

I use without config

[supple whale]

which means HDR and DV doesnt work correctly

because mpv fucks hdr and dv out of the box

[deep mulch]

I just use jellyfin for content with HDR

[supple whale]

which

it has like 5 apps

[crude star]

evil

[supple whale]

i unironically use my own build of chromium for playing videos

because it lets me make my own video player with the shit i want

with the least cancer and fuckery with color spaces, formats, codecs and weird features

[supple whale]

but ofc chromium manages to fuck it up for me too

[deep mulch]

love

[winged mantle]

is this insane

my lexer is overengineered but i like how it has methods named peek and seek

peek 'n' seek

peekin' and seekin'

[valid jetty]

are you cheating and using a library for your lexer..

[winged mantle]

no

[valid jetty]

you dont really need a trie for a lexer, whats the point?

[winged mantle]

to make it cleaner

[valid jetty]

what..

[winged mantle]

avoid giant switch case

[valid jetty]

ok but you do know you made a giant switch case with different syntax right

[winged mantle]

yes

but i prefer it for some reason

[valid jetty]

😭

you dont even need this if you really want your lexer to be "clean"

[winged mantle]

my trie impl is cursed

[valid jetty]

for my ichigo lexer i have this

fn Lexer::next_token(Lexer *self) -> Option<Token> {
    if self.is_eof() {
        return None();
    }

    self.skip_whitespace();

    c := self.current_char();

    if _, op := c.to_operator() {
        self.advance();

        return Some(Token::new(
            TokenKind::Operator,
            TokenValue::Operator(op)
        ));
    }

    if c.is_letter() {
        return Some(Token::new(
            TokenKind::Identifier,
            TokenValue::String(self.consume_identifier())
        ));
    }

    if c.encode() == "「"　{
        return Some(Token::new(
            TokenKind::String,
            TokenValue::String(self.consume_string_literal())
        ));
    }

    if c.encode() == "※"　{
        return Some(Token::new(
            TokenKind::Comment,
            TokenValue::String(self.consume_comment())
        ));
    }

    if c.is_number() {
        return Some(Token::new(
            TokenKind::Number,
            TokenValue::Number(self.consume_number_literal())
        ));
    }

    if c.is_punctuation() {
        punct := self.current_char();
        self.advance();

        return Some(Token::new(
            TokenKind::Punctuation,
            TokenValue::String(punct.encode())
        ));
    }

    return None();
}

[valid jetty]

oh spam

whatever

but my point is, collect all punctuation into a generic "punct" token

then compare on particular strings

instead of having a seperate token for each punctuation type

[winged mantle]

i even did this sorta thing for matching binary

[valid jetty]

with a trie??

[winged mantle]

no

i just made it declarative

[deep mulch]

Koda toda crazy

[valid jetty]

idk but what youre doing seems a little overkill to me

[deep mulch]

@valid jetty

[valid jetty]

i would just have a few basic token kinds and then match on their value

[winged mantle]

https://github.com/TheKodeToad/fish8/blob/main/system/opcodes.go#L17-L52

[pearl stag]

opcodes.go: Lines 17-52

var opcodes = [...]opcode{
    {"00E0", clearScreen},
    {"00EE", returnFromCall},
    {"1---", jump},
    {"2---", call},
    {"3---", skipIfVarEqualsByte},
    {"4---", skipIfVarNotEqualsByte},
    {"5--0", skipIfVar1EqualsVar2},
    {"6---", setVarToByte},
    {"7---", incVarByByte},
    {"8--0", setVar1ToVar2},
    {"8--1", setVar1ToORVar2},
    {"8--2", setVar1ToANDVar2},
    {"8--3", setVar1ToXORVar2},
    {"8--4", incVar1ByVar2},
    {"8--5", decVar1ByVar2},
    {"8--6", shiftVarRight},
    {"8--7", setVar1ToVar2MinusVar1},
    {"8--E", shiftVarLeft},
    {"9--0", skipIfVar1NotEqualsVar2},
    {"A---", setIndexToByte},
    {"B---", jumpWithOffset},
    {"C---", setVarToRandom},
    {"D---", drawToScreen},
    {"E-9E", skipIfKeyHeld},
    {"E-A1", skipIfKeyNotHeld},
    {"F-0A", awaitKey},
    {"F-07", setVarToDelayTimer},
    {"F-15", setDelayTimerToVar},
    {"F-18", setSoundTimerToVar},
    {"F-1E", incIndexByVar},
    {"F-29", setIndexToFontFromVar},
    {"F-33", splitVarDigits},
    {"F-55", copyFromVarsUntil},
    {"F-65", copyToVarsUntil},
}

[winged mantle]

this was simpler

but i like declarative pattern matching

[valid jetty]

i mean if you say so

[winged mantle]

kinda ugly to need to group things by prefix

[valid jetty]

youre not grouping by prefix

youre grouping by class

oh you meant the opcode thing

im dumb

[deep mulch]

no you are not

[deep mulch]

@valid jetty

@valid jetty @valid jetty @valid jetty

you are world smartest person

[valid jetty]

@deep mulch minklang progress when

im gonna get nin0chat working in elle before you get your first hello world running

[deep mulch]

soon

[valid jetty]

good

[deep mulch]

@valid jetty hiiii

[valid jetty]

hiiiii

look at this beautifully formatted code

fn Token::precedence(Token self) {
    $assert(self.kind == TokenKind::Operator);
    op := self.value.as_op;

    if op == Operator::Mul
        || op == Operator::Div
        || op == Operator::Rem { return 2; }

    if op == Operator::Add
        || op == Operator::Sub { return 1; }
}

[winged mantle]

with switch case i was doing

case '=':
    if lexer.match("=") {
        // Equals Equals (comparison)
    } else {
        // Equals (assignment)
    }

i wasn't a fan of needing to group unrelated opreators

[deep mulch]

what if you could do
if op == (Operator::Mul | Operator::Div | Operator::Rem)

[winged mantle]

bitwise or

[deep mulch]

yes

[valid jetty]

..or you can group together all punctuation into a single token kind and a string value, then match on the token's string rather than giving every operator its own distinct token kind

[deep mulch]

rosie could mke some insane syntax like |||

[valid jetty]

^^^

[winged mantle]

then if you have bool = !bool it will read =! as a single token even though it's meaningless

[valid jetty]

look at specifically the last if statement

[winged mantle]

oh wait

[deep mulch]

oh rosie my rosie

[valid jetty]

what, no it won't

[winged mantle]

each char is an individual token

well that's also not that useful

but i want == to be a different token type to =

this is too much for my small brain...

[valid jetty]

i'm pretty sure there is a way to collect tokens into a single token kind while their string value represents their actual semantic meaning

(including multi char ones)

hold on lemme consult stb c lever

[deep mulch]

who

what

rosie knows something sinister

[valid jetty]

hmm

[winged mantle]

but a token for any punctuation is just a character with extra steps

[valid jetty]

well yeah

[winged mantle]

so you're back to square one

even if you do totally insane things it should still process @!|||===:=!!= as @ ! || | == = := ! !=

you need a greedy match... but not too greedy

[valid jetty]

hmmm i'll sit on this overnight i think

i'm pretty sure there's a better way

elle does the approach you didn't wanna do lmao

[valid jetty]

aka this

maybe you can have a function similar to match_keyword kinda, but instead of matching a particular keyword you match any punctuation tokens

so the function internally would match against "==", then "=" , etc, and if none match then it's not a punct token

[deep mulch]

rosinga

[valid jetty]

i'll model this in the morning

[crude star]

None()??????

stop

[valid jetty]

yes

tagged enums dont exist

[crude star]

horrible

[valid jetty]

None needs to be the same size as Some so that it can be represented the same in memory

hence it needs to be generic

hence it needs to be a function

@winged mantle ```rs
use std/prelude;

enum TokenKind {
Punctuation // we only support punctuation in our lexer :)
}

struct Token {
TokenKind kind,
string value
}

fn Token::match_punct(Token self, string punct) {
return self.kind == TokenKind::Punctuation && self.value == punct;
}

struct Lexer {
string input,
u64 position
}

fn Lexer::consume_string(Lexer *self, string value) -> Option<string> {
for i, c in value.iter().enumerate() {
if self.input[self.position + i] != c {
return None();
}
}

self.position += value.len();
return Some(value);

}

fn Lexer::consume_punct(Lexer *self) -> Option<string> {
if _, v := self.consume_string("@") { return Some(v); }
if _, v := self.consume_string("||") { return Some(v); }
if _, v := self.consume_string("|") { return Some(v); }
if _, v := self.consume_string("!=") { return Some(v); }
if _, v := self.consume_string("!") { return Some(v); }
if _, v := self.consume_string("==") { return Some(v); }
if _, v := self.consume_string(":=") { return Some(v); }
if _, v := self.consume_string("=") { return Some(v); }
return None();
}

fn Lexer::next_token(Lexer *self) -> Option<Token> {
if _, punct := self.consume_punct() {
return Some(Token { kind = TokenKind::Punctuation, value = punct });
}

return None();

}

fn Lexer::iter(Lexer self) -> Iterator<Token, SingleEnded> {
return Iterator {
env = Box::new(self).to_ptr(),
next = Lexer::next_token,
next_back = nil
};
}

fn main() {
lexer := Lexer {
input = "@!|||===:=!!=",
position = 0
};

res := lexer.iter().collect();
$dbg(res);

// now we can do things like:
token := res.remove(0).unwrap();

if token.match_punct("==") {
    // do things here
}

}

just, as long as || is before | itll be matched first

and you dont need seperate tokens for each punctuation

and you dont need to make each punctuation a single character

win-win

this is the token stream from the $dbg call

which is correct

or well, i think you actually wanted != to be a seperate punct

but i didnt add it to the list

but you can see hopefully how that would work

you just add != before normal = and it gets matched first

there

technically all of these consume functions should be called try_consume because they can potentially not return a valid punct

but again you get the idea lmao

ok good night

ok i cleaned it up a bit

REAL good night

[solemn ravine]

@woven mesa I gotta redo the cells in the clipboard manager to look like this T_T

im so lazyy

[woven mesa]

o

[winged mantle]

uh thanks but this is what i was trying to avoid because i'm weird 😭

😭 stayed up until 1:00 i'm sorry....

[winged mantle]

try not to make code look like cannibalism challenge

[winged mantle]

a go developers idea of clean code

[fleet cedar]

Won't second case always be true

[winged mantle]

how...

it's just [A-Za-z]

this could just be a regex

[0-9A-Za-z_$]

[fleet cedar]

Either second case is always true, or third case is always false, or the syntax is doing something very fishy

[winged mantle]

, is effectively the same as ||

[fleet cedar]

Then second case is always true

[winged mantle]

|| would be better but this is a go developers idea of clean code

[winged mantle]

why...

walk me through it

[fleet cedar]

All chars are either >= A, <= Z, or both

[winged mantle]

all chars are A-Z?

😨

oh wait

i wanted &&

[fleet cedar]

No all chars are U+0 through U+10FFFF

[winged mantle]

that's the operator

idk how i missed that

[fleet cedar]

10FFFF is >= A

[winged mantle]

i also have average go developer iq

the code was correct before

i will just use normal operators

less confusing

[valid jetty]

what why

hold on i came up with a better solution while i was pondering

[winged mantle]

this feels like it would be the most pleasant to add to

you can't get any simpler

[fleet cedar]

The correct implementation to this function is unicode_ident::is_xid_start/continue

[winged mantle]

ehh

[fleet cedar]

Trie seems like it'd be a lot of overhead

[winged mantle]

i mean... is anyone going to use my toy language

[fleet cedar]

No

[winged mantle]

it's an excercise in futility

parsing performance can matter sometimes

i'm just trying to learn

[valid jetty]

@winged mantle look at this

fn Lexer::try_consume_punct(Lexer *self) -> Option<string> {
    original_position := self.position;
    puncts := ["==", "!=", ":=", "||", "|", "@", "!", "="];
    best_match := None<string>();
    i := 0;

    while self.position + i < self.input.len() {
        current := self.input.slice(self.position, self.position + (i += 1));

        for p in puncts {
            if p.len() == current.len() && p == current
                && best_match.is_none_or_with(
                    fn(value, p) p.len() > value.len(), p
                )
            {
                best_match = Some(p);
            }
        }
    }

    if best_match.is_some() {
        self.position += best_match.unwrap().len();
    } else {
        self.position = original_position;
    }

    return best_match;
}
``` pattern matching on the puncts :]

this might be a little overkill

[winged mantle]

that's was the idea i started with but i felt a trie was a better structure

[valid jetty]

hm

[winged mantle]

let me write uncle bob clean enterprise grade code

[valid jetty]

but you cant have a string value if you do it like that aaaa

needing a token kind for every single punct feels so..

idk

[winged mantle]

i'm sorry...

isn't that what you do in elle though

you have a separate token for Comma and Equal

maybe i should copy you and use consume instead of eat too

eatDigits

[valid jetty]

yeah i do but elle is my first attempt at writing a language

i didnt use good practices i just wrote whatever i first thought of

[winged mantle]

why not make a punk token type

[valid jetty]

punk

[winged mantle]

what the hell is a punct meant to be

punk is cooler

lol

feels like when you see a bsod at an airport

the go compiler falling apart because my code is just too good and vomiting out its internals

but hey it compiles really quickly

the compiler crashes within less than half a second

ok in all seriousness how 😭

i should just riir 🚀

rust would unironically be more suited for this but i wanted to write in a gc language first

write in java like the interpreter guide i was reading did

or just work out what the hell is going on

is this a common go issue

or did i just trigger an edgecase

ah i'm using go dnf package

i will try official distribution

[valid jetty]

@winged mantle what do you expect this to print

use std/prelude;

fn main() {
    xs := [1, 2, 3];
    it := xs.iter().map(fn(x) x * 10);
    xs.push(4);

    $dbg(it.collect());
}

[winged mantle]

10, 20, 30, 40

[valid jetty]

yes :3

its so silly

[fleet cedar]

I'd say error, but if not, yeah that

[winged mantle]

no idea what happens in rust

[valid jetty]

why would it error

[valid jetty]

rust errors

because borrowing

[winged mantle]

ah

/run

import java.util.*;
import java.util.stream.*;

public class Main {
    public static void main(String[] args) {
        // java 8 based just in case
        List<Integer> myInts = new ArrayList<>(Arrays.asList(1, 2, 3));
        Stream<Integer> stream = myInts.stream().map(x -> x * 10);
        myInts.add(4);
        System.out.println(stream.collect(Collectors.toList()));
    }
}

[rugged berry]

Here is your java(15.0.2) output @winged mantle

[10, 20, 30, 40]

[winged mantle]

it does the same in java @valid jetty

[valid jetty]

yeah i know this is expected behavior

[winged mantle]

yeah it'd be weird if it didn't work like this

but theoretically you could make it throw

store crc with list

i should learn c# and do linq everywhere

i don't even know how it works

probably like

List<String> myStrings = new List<String>();
List<String> aaa = SELECT item FROM myStrings WHERE position("aa", item) < 3;

wait it's literally pretty similar 😭

[valid jetty]

that is not how you do it in c#

[winged mantle]

can you get brew for linux

i just wanna do brew install go

[valid jetty]

var xs = new List<int> { 1, 2, 3 };
var it = xs.Select(x => x * 10);
xs.Add(4);
Console.WriteLine(string.Join(" ", it.ToArray()));

this is how i would do it

Select is like .map in any sane language

[winged mantle]

wdym it is https://learn.microsoft.com/en-us/dotnet/csharp/linq/

[valid jetty]

well yeah it is but thats not the "sane" way to write it

i fear nobody actually writes sql-style queries

[winged mantle]

you would've thought this would exist for a reason...

this code compiles

this code crashes the compiler

how??

fixed 14 hours ago

https://github.com/golang/go/issues/63285

probably

i can also remove the equals

[winged mantle]

lol

[placid cape]

built-in SQL 👍👍

[lucid trail]

ahh macros, but its basically the same as rosie's method, so i have to place the 2 char ones first

[valid jetty]

i feel like my new method is actually more efficient than this

because you’re not doing as many comparisons

you instantly rule out some puncts based on their first characters

[lucid trail]

the pattern matching?

[valid jetty]

yes

[lucid trail]

yeah

my way of checking is really horrible, actually

[spark tiger]

linq so cool but in our codebase i was always told not to use it due to performance issues 🤷‍♂️

but yeah it looks really cool to do sql queries that way

[valid jetty]

i mean the way elle does it is also technically like that but not as nice

it checks for the = character then advances and checks if the next is also an =, if it is then the punct is == otherwise it’s =

so basically manual pattern matching

[tired vigil]

Why is discord's i18n key hash function so bad
It uses 36 bits (6 characters) from a 64 bit hash, but their base64 implementation is wrong so the last 4 bits are duplicate, effectively lowering the entropy to 32 bits
So like with their intl maps already having over 20k items, there's almost a 5% chance of a collision between all the keys

I tried bruteforcing it but there are just so many valid hashes, even with 37 characters (uppercase + numbers + underscore, I didn't even try lowercase letters)

[dense sand]

how does one fix this vulnerability?

i tried to override it but it doesnt work

[deep mulch]

impossible

[nimble bone]

@deep mulch is this a @cinder egret reference

[deep mulch]

yes

[valid jetty]

nin0 do you unironically listen to taylor swift

[nimble bone]

i'll let my lfm speak for itself ykyk

[valid jetty]

what the fuck

[nimble bone]

why am i in programming i thought this was regulars

[ionic lake]

what package manager? pnpm?

[dense sand]

yes

[nimble bone]

you're a hater you need more ykyk

[ionic lake]

Do this:

  "pnpm": {
    "overrides": {
      "esbuild": "^0.25.0"
    }
  }

[dense sand]

okay thaks

[ionic lake]

you're welcome champ

[dense sand]

this is radiohead reference

[nimble bone]

true

[crude star]

whenever there's a collision they just change the string slightly

(I have no idea)

[tired vigil]

That would be actually funny ngl 😭
Ik some of the keys have been misspelt for a long time

I might do a pr cuz it's kinda irking me that the code is clearly wrong
Hopefully it doesn't break any vencord patches

[jade stone]

@nimble bone i'm working on vencord compnaion again

[deep mulch]

@crude star hii

[crude star]

zt

[deep mulch]

rini

[jade stone]

@deep mulch my code was so bad, adding basic caching (which hits 99% of the time) makes it read 60x less data from disk every time it reloads

[deep mulch]

insane

[winged mantle]

you should pr a fix so that vencord needs to fix a load of patches

[jade stone]

That wouldn't really change anything.

All we would need to do is change the one hash function.

[winged mantle]

technically it would look like an apocalypse with vencord reporter though and the scare would be funny

[jade stone]

yeah

[jade stone]

horrid types

export type AssertedType<
    T extends Function,
    E = any,
> = T extends (a: any) => a is infer R ? R extends E ? R : never : never;

export type CBAssertion<U = undefined, N = never> = <
    F extends (n: Node) => n is Node,
    R extends Node = AssertedType<F, Node>,
>(
    node: Node | N,
    func: F extends (n: Node) => n is R ? F : never
) => R | U;

[deep mulch]

[chilly viper]

hi

please send me code for minecraft 2

[tired vigil]

TIL type assertions can be inferred

[winged mantle]

writes typescript

complains about rich type system

[jade stone]

not complaining

just saying they're a bit cursed

[jade stone]

you can do things like this which i loev

[winged mantle]

discrimination

[jade stone]

sadly it's still missing autocompletion

https://github.com/microsoft/TypeScript/issues/52898

[jade stone]

https://en.wikipedia.org/wiki/Tagged_union

i love discriminated unions

i wish more languages had support for them

@deep mulch add discriminated unions to kotlin

[tired vigil]

Doesn't the type T also accept class objects since they're subtypes of the Function type

[jade stone]

i mean if your class somehow has an assertion signature, then sure

otherwise user error

[deep mulch]

idk what that means

[limpid mica]

unions that know what element is currently active

[deep mulch]

@limpid mica hiii

[limpid mica]

hi zt

nice to poke you as always

[tired vigil]

This is just Extract<U, { [D]: K }> why do you need a whole new type utility for this

[royal nymph]

to discriminate

[jade stone]

no autocomplete

[jade stone]

you also get errors on invalid variants

[tired vigil]

😭

[nimble bone]

saladinanity

[jade stone]

javascript should get if let

[winged mantle]

HORROR

[jade stone]

declare function foo(): string | undefined;
if (let bar = foo()) {
  console.log(bar);
}

[supple whale]

declaring in ifs

perish

[winged mantle]

i thought you meant like rust patter natmching

if (let string(str) = input) {

}

[jade stone]

wouldn't that mean a whole scoped enum system for js

that would be insane

[winged mantle]

just to check type

[jade stone]

i want it because i would rather not type out this longish type

wait

i'm stupid

[winged mantle]

yea

[jade stone]

but i still think if let would be cool

[winged mantle]

do if walrus

[jade stone]

i dont really care for walrus

[winged mantle]

like go

[jade stone]

(i also hate go)

[winged mantle]

if myThing, err := doThing(); err != nil {

}

[jade stone]

insaneeee

[winged mantle]

:= is the best feature of go

I will not call it by the official name

[crude star]

I would take this ```js
if (let bar = foo(); bar != null) {
}

[supple whale]

its horrendously simply to implement in js tho

[jade stone]

yeah, same thing

[supple whale]

i considered refacoring all my code to use it

[winged mantle]

there's a library to implement the (CENSORED) operator in js

😭

[crude star]

let being a condition is kinda weird

makes even less sense it being !== undefined

C++ has that syntax though

(also go)

[jade stone]

yeah, i use it often in c++

[winged mantle]

ever met the type of person to use thing !== null && thing !== undefined

[jade stone]

insane

[crude star]

no

!= null is exactly that

[jade stone]

@winged mantle let's comprimise with != void 0

[winged mantle]

isn't there a node util

void 0 is horror

[crude star]

i always write void 0

[supple whale]

async function err <T>(Promise<T>: v): T {
   try {
      return [null, await v]
    } catch (e) {
      return [e, null]
    }
}

const [err, res] = await err(fetch(url))

or you could patch promise directly

[crude star]

bundler brainrot

[winged mantle]

do void new function(){}

[jade stone]

why not reverse

res, err

[supple whale]

because thats how go, and old nodejs callback style is

[crude star]

return ((function() {})(), undefined)

[winged mantle]

huh

go always uses the last parameter for error

[jade stone]

yeah

same with node iirc

[supple whale]

Promise.prototype[Symbol.iterator] = async function () {
     try {
      return [null, await this]
    } catch (e) {
      return [e, null]
    }
}

but idk how to do typedefs for this

and then simply

const [err, res] = await fetch(url)

[winged mantle]

i think node had it as first so you didn't accidentally miss it out

[deep mulch]

@crude star @crude star

[jade stone]

oh

node has the err first

[winged mantle]

go deffo has it last though

[supple whale]

wait this is wrong

[royal nymph]

😭

[supple whale]

i remember i did this somehow

i dont remember how

XD

but drunk rn

[crude star]

that cant work

[jade stone]

would this fuck with async iterators

[crude star]

after you await its no longer a promise

[supple whale]

yeah it patched then

[winged mantle]

does js support implicit conversion operators

[jade stone]

valueOf is all i think

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Object/valueOf

[supple whale]

im retarded, this is wrong

you need to patch then

but no

async iterators are symbol.asynciterator

or asynciterable

i dont remember

im fucked up

[jade stone]

ah

[supple whale]

XD

[crude star]

const [{ value, reason }] = await Promise.allSettled([fetch(url)]);

[winged mantle]

why doesn't this work in js

#include <node/console.h>
class CoolBool {
    constructor(value) { this.value = value; } 
    operator bool () { return this.value; }
}

console::log(CoolBool{true});

[supple whale]

at that point this is easier

but the point is to define it as a lang feature

[supple whale]

uuuh

i think you can define how an object converts to asymbol

but idk if that works

i dont rly do rust

[jade stone]

what

[supple whale]

const object1 = {
  [Symbol.toPrimitive](hint) {
    if (hint === "number") {
      return 42;
    }
    return null;
  },
};

[jade stone]

that's c++

[supple whale]

damn

im a c98 mfer

godd damn wasm

[royal nymph]

no?

callbacks yes

but go nuh uh

[supple whale]

very likely im wrong

[crude star]

!!{ [Symbol.toPrimitive]() { return false } } == !!{}

[royal nymph]

not a real go lover..

[supple whale]

my go experience is limite

[crude star]

doesnt work

[crude star]

yeah

you cant change how an object behaves as bool

[deep mulch]

insane

[royal nymph]

well yeah cause all objects are truthy

!! new Boolean(false) // true

[crude star]

except window.all

wait what was the

*document.all

[royal nymph]

insane

[supple whale]

there are always exceptions XD

WE LOVE JS

[jade stone]

is that instances of HTMLAllCollection or just document.all

[crude star]

document.all is the only HTMLAllCollection

@royal nymph they should let Proxy create exotic objects like this

[supple whale]

problem is, its a language feature really

not an object call

!obj doesnt read a property so u cant trap it

its kinda like ===

[jade stone]

wonder if that would prevent a bunch of optimization

[limpid mica]

https://tc39.es/ecma262/#sec-IsHTMLDDA-internal-slot

[royal nymph]

only all

https://stackoverflow.com/a/10394873

[crude star]

lily faster than me

[supple whale]

its probably some hardcoded backwards compatibility shit

where its meant to not exist

[royal nymph]

yes

the thing I linked answers it

if (document.all) {
  // code that uses `document.all`, for ancient browsers
} else if (document.getElementById) {
  // code that uses `document.getElementById`, for “modern” browsers
}

it's meant to make code like this use the getElementById but also still support websites that only use .all

[jade stone]

vchars “

[elder yarrow]

​​ “ ​ LEFT DOUBLE QUOTATION MARK

[jade stone]

insane

[supple whale]

mmmm, im happy that i understand js enough to be able to guess it

[limpid mica]

old code used to check if document.all exists to detect for ancient IE

new browsers implement document.all for legacy compat reasons but don't want to be detected as ancient IE

[supple whale]

chromium: oh no our backwards compatibility to 2002 browsers

also chormium: yeah lets just break all video playback and make constructors randomly segfault on 32 bit systems

[crude star]

i think your usecases are a bit abnormal

[supple whale]

new ImageData() is abnormal?

[limpid mica]

i assume 32-bit systems don't exist when writing software these days smh

[supple whale]

ALL android TVs are 32 bit sir

well, actually there are like 3 exceptions from what i know

out of roughly 3k

ooooh 4 now not 3

china released a new atv box

[royal nymph]

it's about Websites not browsers 🤓

they want ancient websites to still work in latest chromium

[supple whale]

nvidia shield, ugoos am6b+, and z9x pro

[supple whale]

and new websites to not

.

[royal nymph]

if the website is new it means you will fix it 😉

[supple whale]

i think its been 3 years since i reported that filters dont work on offscreen canvases when they were first introduced

still no dice

[royal nymph]

what the hell are you making

filters in offscreen canvas

[supple whale]

still need to piss performance away by computing colormaps in raw js, instead of using gpu accelerated svg filters

[crude star]

ur the kind of user every developer fears existing

[royal nymph]

hop on wasm or smth

[supple whale]

const canvas = document.createElement('canvas')
document.body.append(canvas)
const ctrl = canvas.transferToOffscreen()
worker.sendCanvas(crtl) // do giga heavy stuff

[supple whale]

i already do, its too slow

from what i calculated we'd need 9GHz single cpus for wasm to be as fast as native 2.5GHz cpus

for this specific task

aka text rasterization

[supple whale]

correcting canvas rendered shit, to match the colorspace of a video thats playing under it

and the cheapest way by a LAAAAAAAAAAAAAAAND slide is to do it using SVG filters, as they support color matrixes

so you can pre-compute all possible combinations of color space conversions

map them to a color matrix

and simply apply that color matrix to an svg filter, and boom, you've got fully GPU accelerated color matrix conversion at 0 overhead for the thread the canvas is running in

be it the main thread or the worker thread

well, that is if chromium got off their ass and fixed it not working in worker threads

here @royal nymph enjoy https://github.com/ThaUnknown/jassub/blob/main/src/jassub.js#L3-L25

computed via https://github.com/ThaUnknown/colorspace

didnt do bt.2020 because... the math for it is black magic and i didnt understand it!

this._ctx.filter = `url("data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'><filter id='f'><feColorMatrix type='matrix' values='${colorMatrixConversionMap[subtitleColorSpace][videoColorSpace]} 0 0 0 0 0 1 0'/></filter></svg>#f")`

[crude star]

self super react is crazy

[supple whale]

its a vencord plugin

so its always on

XD

[royal nymph]

wtf is it reacting to

[supple whale]

cursed svg filters

[royal nymph]

bro has a plugin to selfbot react to jsslop

[granite geyser]

this._ctx.filter = url("data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'><filter id='f'><feColorMatrix type='matrix' values='${colorMatrixConversionMap[subtitleColorSpace][videoColorSpace]} 0 0 0 0 0 1 0'/></filter></svg>#f")

aw

[supple whale]

sir its your plugin

its not auto

[royal nymph]

oh you mean the always super react

[supple whale]

but yeah that filter is very not what it was made for

but chromium snuck url() as one of the options

and the flood gates opened

cuz the expected use case is

ctx.filter = "blur(4px)";

XD

[crude star]

does anyone actually notice their subtitle is in the wrong colorspace

[supple whale]

oh yes

very much yes

[crude star]

insane

[supple whale]

nope

not insane

its GIGA visible

[supple whale]

without color space correction

and this is with:

[royal nymph]

just write better code

[supple whale]

????

sir do you understand how color spaces work?

you need to correct color spaces

like

what if the video is simply a different color space

you can apply subtitles to many videos, and they can be encoded differently with different color spaces

so you need to read the video color space, and the subtitle color space

and correct

[solemn ravine]

https://tenor.com/view/sky-striker-raye-master-duel-yugioh-1girl-blonde-gif-15970066289156319915

[leaden crater]

https://tenor.com/view/sky-striker-raye-master-duel-yugioh-1girl-blonde-gif-15970066289156319915

[jade stone]

programming chat

[deep mulch]

they are insane

[valid jetty]

https://tenor.com/view/sky-striker-raye-master-duel-yugioh-1girl-blonde-gif-15970066289156319915

[nimble bone]

you remind the teacher of homework

[viscid grove]

any vim motion enjoyers?
I'm making a calendar program for a school project. Should j be next month or previous month
next seems the most vim-y, but it's on the left so seems more previous to me

[jade stone]

just use h and l

[viscid grove]

h and l will be for year

[jade stone]

oh

then j is next

k is previous

[viscid grove]

that feels so wrong to me

[jade stone]

nop

[viscid grove]

ok it doesn't anymore

idk why it was feeling wrong

[viscid grove]

should h and l be month and jk for year?

[jade stone]

depends on ui

[viscid grove]

h and l seem more "extreme" to me since they're further on the outside

[jade stone]

this depends on ui

[viscid grove]

i'm making the ui rn

[jade stone]

oh

tui

[viscid grove]

yeah

[jade stone]

idk man

just pick one

not important

[viscid grove]

this is the current ui that's barely functional that i just used for testing

i don't think i even have time to make the ui good

it's due tomorrow

well thanks

[jade stone]

@deep mulch @nimble bone rate horrorcode

parseClassDeclaration(clazz: ClassDeclaration): RawExportMap {
    const ret: RawExportMap = {};

    for (const member of clazz.members) {
        if (isMethodDeclaration(member)) {
            if (!member.body)
                continue;
            ret[member.name.getText()] = [member.name];
        } else if (isConstructorDeclaration(member)) {
            // the ConstructoKeyword
            ret[WebpackAstParser.SYM_CJS_DEFAULT] = [member.getChildAt(0)];
        } else if (isPropertyDeclaration(member)) {
            ret[member.name.getText()] = [member.name];
        } else if (isAccessorDeclaration(member)) {
            if (!member.body)
                continue;
            ret[member.name.getText()] = [member.name];
        } else {
            outputChannel.warn("Unhandled class member type. This should be handled");
        }
    }

    // name ?? ClassKeyword
    ret[WebpackAstParser.SYM_CJS_DEFAULT] ??= [clazz.name ?? clazz.getChildAt(0)];

    return ret;
}

[nimble bone]

please never touch a keyboard again

[jade stone]

horrorcode pt. 2

tryParseClassDeclaration(node: Identifier): RawExportMap | undefined {
    const varInfo = this.getVarInfoFromUse(node);

    if (!varInfo) {
        return;
    }
    // classes should only have one decl.
    // if someone proves me wrong on this (with an example in discord's code), ill support it
    if (varInfo.declarations.length !== 1) {
        return;
    }

    const [decl] = varInfo.declarations;

    if (!isClassDeclaration(decl.parent)) {
        return;
    }
    return this.parseClassDeclaration(decl.parent);
}

[hoary sluice]

klass > clazz

[woven mesa]

@hoary sluice

[hoary sluice]

why me

is this a switch 2

[shrewd canopy]

Thats a first switch pretty sure

Switch 2 doesn't have colored joycons afaik

[woven mesa]

swift on nintendo switch

https://cdn.discordapp.com/attachments/1026504914131759104/1391123104167891074/togif.gif

[valid jetty]

Ｃlass when

ｃｌａｓｓ

[hoary sluice]

you can attach colored ones

[hoary sluice]

allow spaces in elle variable names

[shrewd canopy]

Ah

[winged mantle]

https://github.com/moonlight-mod/lunast knockoff ?

[dense sand]

@nimble bone you use kotler compose for android right

Well im trying to decide if i should use flutter or kotlin

[nimble bone]

use that yeah

[crude star]

@supple whale

ykyk

[dense sand]

Use what

[lavish frigate]

That is a switch 1 you cannot attach switch 1 joy cons to a switch 2 and switch 2 has only had one set of joy cons released yet (the grey ones with coloured accents)

[supple whale]

not calibration, correction

still the wrong thing XD

[lavish frigate]

cool

[shrewd canopy]

Its probably not aware of php 8/8.4 lol

[jade stone]

Mine has been around for longer 😭

Also there’s like no code in that repo

[winged mantle]

it seems to delegate parsing to another lib

fair

[jade stone]

Yeah, I'm doing all the parsing myself

Right now I'm working on Getting a UI for all the flux events so you can jump to where any one event is dispatched or handled.

[nimble bone]

salad vibe coded her parser

[winged mantle]

i admire it when people work on projects for a long time

[jade stone]

I think this is the longest project I have going 😭

Only a year

[winged mantle]

after writing go ts gives me migraine

i wanna rewrite everything in go

[jade stone]

(Unless you count my nix dotfiles)

[jade stone]

I could never switch to go

[winged mantle]

i was really starting to like js but my js code looks like slop now

[jade stone]

Module system is too confusing and I don't like the forced casing for variable names.

[winged mantle]

i think i probably just overengineered things

[nimble bone]

i love js

[shrewd canopy]

Forced casing is a good idea

[winged mantle]

and stuff like this isn't inherently a js problem but after writing a bunch of go i hate looking at this

[nimble bone]

just don't make a schema you know you know

[winged mantle]

true

maybe i should use less !== undefined throughout the code

it's kinda ugly

[jade stone]

Sometimes I wish typescript strayed farther away from js

With things like actual overloading (probably impossible)

[winged mantle]

overloading isn't that good

[jade stone]

And runtime generics

[jade stone]

I like it a lot for some things

[winged mantle]

have you never had that ambigious overload error in java and c++

[jade stone]

I have in kotlin

Not Java or c++

[winged mantle]

anyway i'm kinda looking at my code and thinking how did i get here

tbf this is over a year

[jade stone]

Real

I think the horrid tests I added a while back and the (somewhat sparse) comments are the only thing keeping me sane rn

[winged mantle]

i think maybe a good goal is to reduce lines of code

because the loc is probably an indication i made things more complex than they need to be

maybe i should just revert the last 100 commits it will probablybe much better

[jade stone]

https://github.com/sadan4/VencordCompanion/blob/main/src%2Fast%2Fwebpack%2FWebpackAstParser.ts

[winged mantle]

one thing good about my project is the largest file is 250 lines

[jade stone]

Yeah

I can't really do that with that file

[winged mantle]

just make it less oop

[jade stone]

I'm not partial to OOP or more functions, but I think it works well for this file with the only downside being it's massive.

[lavish frigate]

oop has been a curse on the programming community

[winged mantle]

i totally didn't just rewatch https://www.youtube.com/watch?v=QM1iUe6IofM for the 17th time

[jade stone]

least half-baked youtube ui

[crude star]

how did you end up urlencoding your slashes

[jade stone]

gh mobile urlencodes the filename

[crude star]

horror

[winged mantle]

i made an extension point system where plugins declare a set of "contributions" which modify the behaviour of other plugins

would it be better just to use functions

is there a good guide on making a nice plugin architecture with js/ts

[crude star]

are rust developers okay https://bsky.app/profile/orhun.dev/post/3lutxbgu23s2n

Orhun Parmaksız (@orhun.dev)

Tachyonfx brings web-level animations to the command line.
Then Ratzilla puts it back in browsers where it belongs 🤷‍♂️

Witness this beautiful chaos: junkdog.github.io/exabind

🦀 Powered by Rust & @ratatui.rs ecosystem
⭐ GitHub: github.com/junkdog/exab...

#rustlang #ratatui #animations #terminal #web

▶ Play video

Bluesky · Jul 26, 2025, 6:38 AM

what is this

[winged mantle]

yeah honestly my code got worse 😭

best bet to revert before my rework and backport stuff

[deep mulch]

kode tode

[deep mulch]

no