🪅-progaming

[dense sand]

sonarqube is screaming at me

[placid cape]

xdddd

[dense sand]

i wonder if paper dev team have any plans on reworkign this

since they hard forked

[spark tiger]

are recursive structs considered a bad practice? not sure what would be the better approach for a file tree
(ignore the lifetimes, i have nfi what these are, just used 'a so rust-analyzer stfu

[dense sand]

i mean how else do you want to do this

[fleet cedar]

Nah they're fine, though in this instance a Vec<String> would do the job better

[spark tiger]

dunno

[spark tiger]

for parent or wdym?

[dense sand]

i really want to rewrite my jvm parser from ground up

[royal nymph]

the naming is odd. what exactly is Entry.folder? the folder the Entry is in?

[dense sand]

fuck it im doing it lol

[spark tiger]

yep. this is just a poc (actually it doews't even have any impl) yet

[fleet cedar]

folder: Vec<String> has exactly the same set of values

[dense sand]

is there anyone knowledged in avr libc? i know it has malloc/free fucntions, but do they work as i expect them to do? or do they behave differently?

[ivory heath]

The only issue with recursive structs is memory fragmentation and non sequential reads but that probably isn’t a problem here.

[royal nymph]

weird design

[spark tiger]

oh true. also i guess i can use [string] here insteaf of vec<string> for the sake of uh smaller struct?

[royal nymph]

inconsistent

[fleet cedar]

[String] is unsized though so it's kinda inconvenient

[spark tiger]

why. i don't really get what's so wrong about it. or do you mean just the naming or

[spark tiger]

not sure what any of that means tbh lol

[royal nymph]

why is there both Folder::parent and Entry::folder

this makes no sense

why are you differentiating between folders and files like this instead of using the same struct

[unkempt shuttle]

rust???

hello

[royal nymph]

also you should name it Directory instead of Folder

[spark tiger]

i thought someone was going to tell me so

i'm a proud windows user

ig i can just do it like uhhh so?

struct Entry<'a> {
    name: String,
    bytes: &'a [u8],
    path: Vec<String>,
}

[royal nymph]

windows calls it directory too https://learn.microsoft.com/en-us/windows/win32/api/winbase/nf-winbase-getcurrentdirectory

[unkempt shuttle]

what is bytes in this case

[royal nymph]

either it should be path: String or path_elements: Vec<String>

storing a Vec as path is unintuitive

[unkempt shuttle]

(you could also use a PathBuf)

[spark tiger]

oh it was supposed to be the raw file data in bytes but then i remembered that i should NOT make each file struct has megabytes of the data

i forgor

[unkempt shuttle]

so now it's..?

why can't you just store the entries as names and paths and then load them as needed

i have no idea what's going on to be fair but that's what i would do

[spark tiger]

i want the entry struct to have the begin address and size. i'm building a winrar-like app but for a proprietary file format used for storing game assets

[ivory heath]

It’s not applicable here as far as I can tell but when you have a struct that contains a list of structs (generally) each sub struct will be its own memory allocation and will fragment along with other allocations the memory allocated to the process artificially boosting the memory usage.
If you want to read all the data sequentially instead of storing just a list for each property you must do 2 pointer dereferences and due to not being linear the compiler cannot make any SIMD optimizations.

Here neither apply. Just logically it’s weird

[unkempt shuttle]

"the begin address and size"?

i'm assuming you mean the beginning address and the filesize?

[spark tiger]

yeah. i can send the explanation from the prompt i asked chatgpt (i was so lost i had to ask chatgpt how to reverse engineer the file format)

I’m reverse engineering a proprietary
binary file format used for storing game assets. Well, there are actually three different files:

1. bigfile.bfn — used for storing file names and file tree. The structure is the following: first there are 4 bytes (u32) that represent the size of the folder’s name. Then there is the string - the folder’s name - that has that many amount characters as the u32 we just got. After that we have another 4 bytes (u32) that represent the amount of files in the folder, let’s call it files_len. After that there are files_len amount of the following structure: 4 bytes (u32) that amount of bytes that represent a utf-8 file name. Then there’re another 4 bytes (u32) — the amount of subfolders in the folder we’re in. And the cycle repeats
2. bigfile.bfdata — the biggest size-wise file out of these, which as could be guessed contains the binary data for the files from the previous file
3. bigfile.bfdb — the least researched file which, it seems, should link the file names with the binary data they contain. The issue is that I haven’t managed to reverse engineering it yet. I decided to pick one file from bigfile.bfn and found the address and size of the binary data it has in bigfile.bfdata. The way I could confirm that this file name exactly links to this data is by using binwalk. It shows which files are contained in the file’s binary by searching for signatures and magic bytes. That way I could find all the PNGs and I decided to take the last one, thinking that the order of files in bigfile.bfdata and bigfile.bfn. I was correct. That way I had the matching file name and the data it represents. Searching for the address of the file’s name from bigfile.bfn didn’t get any results in bigfile.bfdb. But searching for little-endian byte representation of the address of the file data was successful and just 8 bytes before it I found LE-bytes that represent the size of the file data.

[unkempt shuttle]

this already exists in the stdlib, looks to achieve what you have here already? you can directly call readDir and then handle it that way: https://doc.rust-lang.org/beta/std/fs/struct.DirEntry.html

[unkempt shuttle]

chatgpt?????

[spark tiger]

yes

i was like at the final step of finishing reversing this file format but i couldn't find the way to link file 1 and file 3

[fleet cedar]

Things with big fancy trees like compilers often use fancier things than boxes, but for casual stuff it's fine

[spark tiger]

also i wonder what would be better:

1. store the beginner address of the file data and the filesize in the entry struct
2. store the beginner address of the file data and the final address in the entry struct
3. don't store these in the entry struct, instead read it from file 2 when extracting the files from the 'archive'

[unkempt shuttle]

so you're reading out of the file into your own structs and not from an file system

[spark tiger]

yes, these files/entries do not actually exist on the uh machine

[unkempt shuttle]

okay

hmm

[spark tiger]

i have an idea - what if i just look at the implementations of already existing archive managers made in rust

[scarlet tendon]

(async () => {
    // Get the token using Vencord
    const tokenData = Vencord.Webpack.findByProps("token");
    const token = tokenData?.token;

    if (!token) {
        console.error("Token not found.");
        return;
    }

    console.log("Using token:", token);

    // Set headers 
    const headers = {
        "Authorization": `Bearer ${token}`,
        "Content-Type": "application/json",
    };

    // Make the GET request to the Discord API
    const response = await fetch("https://discord.com/api/v9/users/@me/relationships", { headers });

    // Handle the response
    if (response.ok) {
        const friends = await response.json();
        
        // Filter real friends (type 1 = real friends)
        const realFriends = friends.filter(f => f.type === 1);

        console.log(`You' have ${realFriends.length} friends!`);
    } else {
        // Log full response for debugging
        const responseText = await response.text();
        console.error(`Failed to fetch friends: ${response.status} - ${responseText}`);
    }
})();

i wrote this for #📜-js-snippets but i cant quite get it to send my friends i have

I have a valid token and i have a python equaivlent that fully works

[spark tiger]

i don’t have much experience with this but i don’t really think you’re supposed to manually write the headers, extract the token and send the request that way. i’m pretty sure there are helper methods inside either vencord or discord that do all that stuff for you

[scarlet tendon]

Hmm wdym ? do you know of these helper methods?

[spark tiger]

from reading a few snippets i assume it's this

ofc replace the body with the needed endpoint

[spark tiger]

that seems to work

[scarlet tendon]

Got it

(async () => {

    // Use Vencord to fetch the relationships (friends list)
    const response = await Vencord.Webpack.findByProps("getRelationships").getRelationships(); // Using Vencord's helper function

    // Check if the response is an object and contains the friends list
    if (response && typeof response === 'object') {
        // Count the number of real friends (type 1)
        const realFriendsCount = Object.values(response).filter(f => f === 1).length;
        console.log(`You have ${realFriendsCount} friends!`);
    }

})();

[dense sand]

macro

[fallen nebula]

Here's a valid reason to be scared

[dense sand]

[balmy lintel]

Horror

doing raw http requests should probably be a last resort

[scarlet tendon]

Y

[void leaf]

Damn you cant nest pseudo-elements

[dense sand]

[daring kite]

hes just chatting sh*t tbh, this aint raw, its using fetch lmfao i dont think he knows code

[keen spoke]

Nope, he's right, thats a raw request. He should use discord clients api to get the relationships

[daring kite]

erh depends if hes doing this for bot or user token

but yk better

[dense sand]

isnt this bs?

or i just cant read

calculus people assemble

[keen spoke]

Only users can have relationship

[jade stone]

Wonder if you can do it with nested selectors

[void leaf]

https://tryitands.ee/

/joke

[jade stone]

In bed

[spark tiger]

tfw rust’s from_ne_bytes is just transmute

[austere anchor]

well I mean, what else are they supposed to do

yk

[spark tiger]

i’m pretty sure it can be done without unsafe code like so:

fn read_u32_be(bytes: &[u8]) -> u32 {
    ((bytes[0] as u32) << 24) |
    ((bytes[1] as u32) << 16) |
    ((bytes[2] as u32) << 08) |
    ((bytes[3] as u32) << 00)
}

[austere anchor]

how to handle ne to current endian

[spark tiger]

?

[austere anchor]

from ne bytes

[spark tiger]

if cfg!(target_endian = "big")
read_be_u32()
else
read_le_u32()

[austere anchor]

blehh too much of a hassle

safety ensured anyways because it's enforced by type system

[dense sand]

no <endian.h>?

wow

bad language

(the header file is totally not gnu specialty)

[austere anchor]

there are no header files in rust generally

[spark tiger]

i mean that cfg! part is literally what from_ne_bytes does ?

oh wait no

[dense sand]

poo language

[spark tiger]

or yes

[austere anchor]

I love not having to copy paste all of my function signatures to header files for it to be able to be used elsewhere

[austere anchor]

I actually have no idea what the code for from_ne_bytes is

[dense sand]

header file is the first thing you do bruh

then you implement it inside C file

its not that hard

also, you can just directly define it inside .h

[austere anchor]

ok but like, what if I don't need to do that

[austere anchor]

also idk what funcs I'm going to write exactly before I start writing it

I'm like

"oh yeah I should separate this out"

then I have to copy paste to headers

annoying

[dense sand]

just use extern bro

[austere anchor]

mid

[fleet cedar]

Wouldn't surprise me if part of the reason for using transmute rather than bit stuff is that it makes it easier to generate the impl for all types with one macro

[spark tiger]

https://github.com/rust-lang/rust/blob/master/library/portable-simd/crates/core_simd/src/to_bytes.rs#L94

[pearl stag]

to_bytes.rs: Line 94

fn from_ne_bytes(bytes: Self::Bytes) -> Self {

to_bytes.rs: Lines 94-100

fn from_ne_bytes(bytes: Self::Bytes) -> Self {
    // Safety: transmuting between vectors is safe
    unsafe {
        #![allow(clippy::useless_transmute)]
        core::mem::transmute(bytes)
    }
}

[austere anchor]

yeah that's true as well

[spark tiger]

i have a crazy idea

what if i try to implement my parser in zig instead of rust

zig seems surpassingly good for doing these kinds of things at first glance

[buoyant sparrow]

Hi sorry what do you think? Would you like to try this, like we do with the filters for voices but to change the sound of the headset, for example, we would have a surround button, a bass boost button, a rock button... etc. all with a menu on the headset icon, so nothing to do with it and it would be a full plugin because it is not integrated as standard.

[dense sand]

i think no

Im thinking miku miku oo ii oo

[formal belfry]

@woven mesa wondering how i should document modifiers on ddbkit docs in a way that isnt like a wall of text and unreadable

I tried tables but uhh..

yeah

[woven mesa]

you mean unlike the wall of text i wrote

[formal belfry]

idk

[woven mesa]

no the joke is i wrote a wall of text

[formal belfry]

I am thinking i might just hint at modifiers' existence via example codeblocks which i am already doing and i think it might work fine

im not sure if people will try to rely on xcode autocomplete or not because some might ignore that lol

I would heavily prefer to just not document them because of how self explanatory most of them are but i don't know..

[woven mesa]

idk either

its difficult

[formal belfry]

I wish ddbkit server wasn't so dead so i could ask for input

[spark tiger]

zig experience:

create a variable
code won't compile because of error: local variable is never mutated
make the variable a const
code won't compile because of error: unable to evaluate comptime expression

like ??????

oh i think i get why that happens

why is zig so damn hard

like it builds fine

i run it

and i basically get "it failed, idfk why"

[fleet cedar]

Why is "not mutated" a hard error

Smells like golang's unused variable errors

[spark tiger]

(it's a thing in zig too)

[runic sundial]

Show code

Zig is like piss easy compared to Rust

Hardly ever gets in my way

[runic sundial]

Oh, you want to use _ = functionWithUselessReturn();

You must use local variables.

[spark tiger]

it's p bad as i know nothing about zig, i just read the features page and sample page, but i was expecting the compiler to help me

const std = @import("std");

pub fn main() !void {
    const file = try std.fs.cwd().openFile("C:\\Program Files (x86)\\Steam\\steamapps\\common\\KarmaZoo\\resources\\cookedData\\bigfile.bfn", .{ .mode = .read_only });
    const reader = file.reader();
    try read(reader);
}

pub fn read(reader: std.fs.File.Reader) !void {
    const folder_len = try reader.readInt(u32, .little);
    var list = try std.ArrayList(u8).initCapacity(std.heap.page_allocator, folder_len);
    defer list.deinit();
    try reader.readAllArrayList(&list, folder_len);
    try std.io.getStdOut().writer().print("{s}\n", .{list.items});
}

ig it's something with the list

[runic sundial]

Do you have IntelliJ?

[spark tiger]

i'm using zed but like i do have other jetbrains' ides

[runic sundial]

Or RustRover better if on Windows

[runic sundial]

Yeah.

Get that, and the ZigBrains plugin

(I am biased as my bestest friend ever made it)

This, plus zls

zls needs to be version matched with zig

I usually just use latest upstream for both

[spark tiger]

done

i think i'm using tagged version of zig as i didn't figure out how to use nightly

[runic sundial]

You'll prolly need to for zls to work, I think the plugin can help with that

I can get on computer in a bit, can u vc or nah?

[spark tiger]

i can just listen if it's fine with you

[runic sundial]

Perfectly valid

[spark tiger]

i can't guarantee that i'll actually start using zig, but if you don't mind it, i'm down to learn something new

[runic sundial]

Alr I got it setup on my Windows install

can

demonstrate

[spark tiger]

let me quickly connect my headphones

[formal belfry]

i looked in quickCSS for something to hide the nitro emojis on reactions, but i couldn't find anything besides just making the nitro bits less flashy. the problem is that anything below what's hidden doesn't want to render properly, does someone have any ideas?

or do i go to theme development for this kinda stuff sorry idk

[fleet cedar]

Cssing lazy-rendered lists usually doesn't work well

[formal belfry]

would i need to do javascript then?

[fleet cedar]

https://github.com/Kyuuhachi/VencordPlugins/tree/main/Anammox does that, among other things

[formal belfry]

thank youuu

[supple whale]

<script>
const volume = persisted('volume', 0)
$: exponentialVolume = $volume ** 3
</script>

<video bind:volume={exponentialVolume} />

did i ever mention how much i fucking love svelte

just for shit like this

[valid jetty]

wtf is that syntax

[olive mesa]

would anyone know or be able to help me make it so i can get notofications on my phone from discord dms while still being online and ac tive on my pc? like no idle delay etc

[supple whale]

svelte!

its peak

compiler level reactivity is peak

TLDR updating volume in any way, automagically updates the exponential volume

so i can do $volume = 0.05

and it becomes 0.000125

in the video

and persisted just syncs it with storage, so localStorage or idb

so i can have peristent variables

[valid jetty]

let volume = use_state(|| 0.0);
let exponential_volume = (*volume).powi(3);

let onloadedmetadata = Callback::from(move |e: Event| {
    if let Some(video) = e.target_dyn_into::<HtmlMediaElement>() {
        video.set_volume(exponential_volume);
    }
});

html! {
    <video onloadedmetadata={onloadedmetadata} />
}
``` xd

[supple whale]

jesus

[supple whale]

case in point:

15% of a 5800x3d

on a single poor animation

this is why you should ONLY ever transition and animate transform and opacity

[runic sundial]

TO ALL WEB DEVS

Stop doing this

Cheers

[balmy lintel]

horror

[spark tiger]

wakatime has a vdp???

[cursive beacon]

tryna run my multi-purpose discord bot

[dense sand]

ok

[balmy lintel]

hop on the dev portal and enable all intents

[supple whale]

you mean navigation button?

the shit on the side of ur mouse?

[runic sundial]

Browsers have a back button.

[supple whale]

so navigation

[runic sundial]

Like the "go back a page"

Yeah

Fancy sites break that shit

Allll the time

[supple whale]

well yeah

thats what SPAs are

[supple whale]

i'm about to loose my shit

declare global {
  interface MediaSession {
    setActionHandler: (action: MediaSessionAction | 'enterpictureinpicture', handler: MediaSessionActionHandler | null) => void
  }
}

BRUH

like why isnt this possible in typescript

you cant override already defined types

D:

[nimble bone]

declare global {
  interface MediaSession {
    setActionHandler: any;
  }
}

[ionic lake]

what about & {}

[supple whale]

yeah waht about it

[supple whale]

1:1 same issue, you cant override

you can define new, but not override

lib.dom.d.ts takes priority over anything you'll ever do

[nimble bone]

// @ts-ignore

[supple whale]

dogshit language

[hoary sluice]

agree

[supple whale]

like how the fuck do you force types

that are incorrect

[placid cape]

easily: you don't

[supple whale]

well typescript disagrees

clearly

[dense sand]

if you want to force types which are incorrect then youre in a wrong language

[supple whale]

BRUH READ

i want to correct types which are incorrect by default

[valid jetty]

if you define interface X and interface X already exists you can define more types on X or redefine existing ones

[supple whale]

you cant redefine existing ones

[dense sand]

make a patch with bun

[supple whale]

you can only append types to an existing interface

[dense sand]

trolleylag

[valid jetty]

oh fun

[supple whale]

legit what do

[dense sand]

make a patch with bun

[supple whale]

had to do a forced asertion

setActionHandler: async (...args) => navigator.mediaSession.setActionHandler(...args as [action: MediaSessionAction, handler: MediaSessionActionHandler | null]),

but i dont like it

[dense sand]

i fell in love with makefile

its so simple its complex

[supple whale]

idk how one can look at this and go "yeah this is fine" https://github.com/ThaUnknown/jassub/blob/main/Makefile

[dense sand]

its the type of code that you write once and then never touch it because it can break the whole build process

[supple whale]

yeah

i wanted to add pthreads to that build config

gave up after like 3 days of work

[spark tiger]

justfile looks interesting though

[dense sand]

crazy

can we add sql transactions to C dynamic memory allocations

[supple whale]

@dense sand did ya see the animation perf i posted above btw?

[dense sand]

yes

[supple whale]

gut

https://cdn.discordapp.com/emojis/470215099559444492.webp?size=48&name=whip~2

[dense sand]

this is crazy though

the perf difference

[supple whale]

yeah

its going from worst to best case

[winged mantle]

does that sound like a good idea to you

[supple whale]

yes it does, because the default navigator types are often missing shit

esp if you're working with new chrome only apis

[winged mantle]

it would break type safety

[supple whale]

break what now

type safety

in typescript

bro what

[winged mantle]

i mean yeah

[supple whale]

you clearly dont understand what i want to do

but that's fine

navigator.mediaSession.setActionHandler('enterpictureinpicture', console.log)

is valid code, but the default ts types are invalid

and erorr on it

[winged mantle]

in terms of language design i don't think they should add the ability to modify interfaces which are not yours

you can just use ts-ignore

and i'd argue using browser specific apis is bad

[supple whale]

sure, for C or Rust it makes sense

but for a language where u can do Function.prototype = null

its not good design

[winged mantle]

typescript is meant to make things safer...

[supple whale]

no, its meant to add typing to js

those are not the same things

[supple whale]

just as related to the conversation

[supple whale]

i have 0 idea how a screenshot of typescript marketing keypoints, relates to the fact that the default types for the language it provides are invalid an cant be fixed in any way

without forking the types and maintaining them yourself i guess

[hoary sluice]

What is TypeScript? A pile of leaves that makes a shitpile invisible and even easier to step in

[winged mantle]

i assure you this image of david tennant has a profound relation to this topic

[dense sand]

has anyone here worked with java xml validation and custom schemas?

im getting this weird error

[winged mantle]

may i ask why you are using xml instead of json

[dense sand]

cause xml fits this better

[fleet cedar]

Xml validation is the third circle of purgatory

[dense sand]

would you use json for designing uis?

[winged mantle]

maybe

[fleet cedar]

I would use whatever the ui library I use does

[dense sand]

it fucking declared there, whats your problem 😭

[winged mantle]

java class name in attribute makes me worry about thiis api design

[dense sand]

i can always make it lowercase

thats no problem

id just like it to work

[winged mantle]

no i mean the class="org.jaibf.plugin.TestInventory"

[dense sand]

oh thats ok

[winged mantle]

means it's using reflection....

[dense sand]

yes

im aware of that

[winged mantle]

and i fear what other things it's doing

[dense sand]

i love refelction

[winged mantle]

it just brings back memories of eclipse plugin dev

[dense sand]

horror

[fleet cedar]

Nasty

Go wash your codebase with soap

[winged mantle]

what's the point of duplication

[dense sand]

lmao

i just want it to work though 😭

xml namespaces work please

[fleet cedar]

Futile

I have never in my whole career seen a xml namespace do anything that could be remotely described as "work"

[winged mantle]

if i were to do this, i'd use a library that can just pass the ast of the xml and do everything else myself

[dense sand]

wtf did i just found

https://www.youtube.com/watch?v=vAnPGtmkM1s

this is chatgpt answer rewritten into wall of text, rendered as a video published to youtuhbe

[balmy lintel]

ai slop

[valid jetty]

@hoary sluice this is what i think kotlin feels like

[ornate quiver]

what are you doing....

[dense sand]

Horror java mvc

[ornate quiver]

with reflection is crazy

[dense sand]

How do you expect me to do DI

[ornate quiver]

gu

[dense sand]

?

[hoary sluice]

lmao wtf

[placid cape]

what the fuck is this

[jade stone]

does anyone know any good local dns server

i dont want to use pihole to just redirect a few queries

@deep mulch zootie

[deep mulch]

i think theres one in Go thats really good

[jade stone]

ive heard of unbound but the config for that looks scary

but i might go witht hat because it has nix support

[deep mulch]

i will give you shock therapy so you will hate nix

[jade stone]

NEVER

you will never get me to hate nix

[deep mulch]

guhh

[winged mantle]

sorry to interrupt your.. whatever this is

i am stuck trying to think of a good system for persistent state in discord commands

[deep mulch]

database

[jade stone]

use a discord message as the state

just update the message whenever the state changes

use discord search feature to find the message

[winged mantle]

i am trying to have buttons which remember some context about the command

[deep mulch]

i have no hands

[winged mantle]

i don't want to put the data in the buttons themselves

i have a custom TTLMap implementation

the data only needs to last a short time and not between restarts so db is pointless i think

db is being used for other things

i just wasn't sure about a good way to structure things

wasn't sure whether i should do

run(context) {
    context.set_state({ current_page: 0 });
    context.respond({ components: [/* ... */] });
},
component_event(context, id) {
    const state = context.get_state();
}

or

run(context) {
    context.respond({ components: [/* ... */], data: { current_page: 0 } });
},
component_event(context, id, data) {
}

context is basically something which implement stuff relating to the context as well as a respond function

currently there is a SlashContext and a PrefixContext but i guess i'd need a ComponentContext as component responses work differently...

so it wouldn't actually be the same object

so maybe the former is a little weird

wait

better idea

just design the api naturally and hopefully the implementation will not be insane

run(context) {
    context.respond({
        content: "Press the button",
        components: [{
            text: "Click me",
            callback: click_context => {
                click_context.ack();
                context.respond("You pressed me");
            }  
        }]
    })
}

(do you even need to acknowledge the interaction if you are editing the message anyway)

(also i forgot async+await lol)

i had an unsane idea for generating button customIDs

use URLs

moderation/ban?callback-id=0

[placid cape]

Store everything in button id

You can use Base1114111

https://github.com/CRBT-Team/Purplet/tree/main/packages/serialize

[winged mantle]

ehh no

that's cursed and sounds like a bad idea because you might forgot that you're storing the data in a button and accidentally leak something

[spark tiger]

guh why did it suddenly stop working

[fallen nebula]

What do you rate my solution (in Java)
https://onlinegdb.com/G97XgM_Tq

[keen spoke]

Since #🗳-plugin-requests are closed, can I PR a new plugin?

[fleet cedar]

PRing is always fine (but be prepared to be ignored for months)

[keen spoke]

thats okay

[fleet cedar]

Plugin requests is fir begging without contributing

[keen spoke]

Well usually you first request, then triage then PR

according to the contrib guidelines my plugin should be fine, basically discord selects the last sent msg and if you sent a msg for which the server has not sent a response that it was posted yet, discord selects the last sent msg when you use the arrow key, extremly annoying. I want it to select the last sent msg, regardless of the server having replied yet. If it has not at the time of sending the PATCH, it should queue the patch, wait for the server to reply and then send the edit

[spark tiger]

seems like the extension's 0.2.2 broke it with the symlinks

[placid cape]

working on fix

oh wait you're the one that made the issue

[spark tiger]

yeah lol

[placid cape]

thanks :D

i hate windows

[spark tiger]

i think this might be related to https://github.com/zed-industries/zed/issues/20559

or maybe not

[placid cape]

idk but there's a bunch of issues with LSP

https://github.com/zed-industries/zed/issues/25358

[spark tiger]

i've only faced with this issue

it's so annoying

and for some reason the zed team doesn't really seem to work on a fix despite it being a really major issue

[placid cape]

they're not working on windows

[spark tiger]

sad

[placid cape]

only community works on it

that's also why they dont provide builds

[spark tiger]

there've been like 3 or 5 PRs attempting to fix that issue but all of them got closed because "it's not a proper fix"

like i'd rather have some fix than most extensions being broken

[placid cape]

yeah

[spark tiger]

and the worst thing is that i'm not really sure like what to blame

like is it wasmtime's issue or wasi's issue or zed's issue

[placid cape]

well it makes sense they don't want bad code in their codebase but it's still better than nothing and you can just add comment in the issue

[spark tiger]

i'm not so sure how exactly extensions work in zed, but it seems like it basically tries to execute the commands with wasm paths (prefixed with /) outside wasm runtime or whatever

because wasmtime can open /* paths just fine

[placid cape]

https://github.com/lilnasy/zed-windows-builds here you can obtain builds that contains https://github.com/zed-industries/zed/pull/22600 patch

💀 amazing idea, let's modify std behaviour

[spark tiger]

that silly bug is stopping me from starting to use wakatime :(

[placid cape]

:(

well you can build zed with the patch

[spark tiger]

yeah i can, it's just that i want to have a comfy way of managing it - currently by using scoop

i don't really want to like manually add it to path, to apps list, etc.

[placid cape]

well yea understand

that's way easier on linux

like i have my own packages repository for pkgs that are not on void repos

so everything is managed by xbps

[spark tiger]

windows is funni

and i'm too uh lazy to get used to linux

[placid cape]

its not that hard

the hardest thing is to do it

i fully switched to linux like 2 years ago

[spark tiger]

i tried to move to linux (or at least get used to it) a year ago

i installed mint and it seemed good for the first two weeks

but then i just kinda lost any point in using it

so i was just booting in windows

[frosty obsidian]

they'd rather focus on an ai that people with fake jobs use

[astral current]

then ai to do the prompts

[placid cape]

that's true lol

[hoary sluice]

mint has the wrong phisosophy

they try to make it easier to migrate from windows by making it look similas to windows and super bloated

u need to install arch manually, without install scripts, get a tiling window manager and either delete windows or promise urself not to use it at all for a month or two

[winged mantle]

what the hell

my code worked

scary...

i literally only made one mistake and first time it worked

(just forgot to register event handler)

[hoary sluice]

average day using rust

[winged mantle]

it's interesting how code made by copying and pasting other code works best usually

[winged mantle]

I wrote a Discord bot and when it was in production there were virtually no bugs and i was terrified

[hoary sluice]

did u write it in rust

[winged mantle]

yes

[hoary sluice]

crazy how i was able to guess that just from the fact that everything worked right away

[winged mantle]

in this case earlier i was using typescript

[hoary sluice]

oh then ur lying

[winged mantle]

i was using rust for one thing but not another thing

[winged mantle]

how does v8 "know" not to garbage collect i

does it just keep all variables loaded

or do that when there is an eval call

(as i believe that's the only function which can see local variables without needing to pass them)

[royal nymph]

using direct eval forces it to make a bunch of assumptions

[winged mantle]

ah

makes sense

thanks

now i can stop being paranoid about a few bytes extra memory usage

[hoary sluice]

@valid jetty why is stdlib qsort only barely faster than random geeksforgeeks merge sort impl

with 1 million random numbers between 0 and 1 million

[austere anchor]

bc its supposed to be standard

works on all data

unspecialized

[valid jetty]

yea ^^^^

[hoary sluice]

does it not optimize for ints if it sees ints

[valid jetty]

specialized ones probably have better pivots for the qsort

it will always be an int bestie

[hoary sluice]

so unspecialized as in it doesnt specialize for specific ranges of ints?

[valid jetty]

also your input set is not integers lol

[hoary sluice]

wdym

[valid jetty]

you said between 0 and 1

that implies a float

oh

[hoary sluice]

u cant read i think

[valid jetty]

i love mobile and outside

[hoary sluice]

lmao

[hoary sluice]

why tho

can u not define a sorting function for structs

[valid jetty]

you can???

that’s the point of the cmp thing in qsort it takes a comparator function

but that function still compares ints lol

[hoary sluice]

but like theres radix sort which only works on large ints, i thought thats the kind of specialization u were talking abt

[hoary sluice]

but like theres radix sort which only works on large ints, i thought thats the kind of specialization u were talking abt

i love having bad internet and sending msgs twice

[austere anchor]

yes

it takes in a void*

it cannot know it's an int that it's comparing

the only thing it does is call the cmp function on 2 elements every turn

[valid jetty]

^^^^

i made my own in elle xd

fn Array::sort<T>(T[] self) {
    if self.is_empty() || self.len() <= 1 {
        return;
    }

    __internal_qsort(self.elements, 0, self.len() - 1);
}

local fn __internal_qsort<T>(T *array, i32 low, i32 high) {
    if low >= high {
        return;
    }

    T pivot = array[high];
    i32 i = low;

    for i32 j = low; j < high; j += 1 {
        if array[j] < pivot {
            T temp = array[i];
            array[i] = array[j];
            array[j] = temp;
            i += 1;
        }
    }

    T temp = array[i];
    array[i] = array[high];
    array[high] = temp;

    __internal_qsort(array, low, i - 1);
    __internal_qsort(array, i + 1, high);
}

[winged mantle]

looks so good i need to learn elle....

[hoary sluice]

@valid jetty is claude 3.7 neferious

[spark tiger]

i decided to try claude today but i didn't know it's literally fucking 10 free msgs a day or something

and unlike on chatgpt there isn't like free model that you can use without any limits

and the subscription is even more expensive than chatgpt plus

$30 a month

i guess i'll just keep using 4o(-mini)

[hoary sluice]

im using chatgpt plus from a girl on snap rn

might buy claude again

[hoary sluice]

yea cause they dont have 10 trillion dollars to burn from vc money

[spark tiger]

also what kinda annoys me a bit is the lack of any markdown in claude

i'm so used to it in chatgpt

it's kinda hard to read text in claude

[native spruce]

@woven mesa you use uikit for ipod right?

I was wondering if you use a uikit navigation controller as well

trying to use any sort of collection view causes my swiftui navigationview to act strange

holy shit nevermind

.ignoresSafeArea(.all)

why does that work even

okay

[native spruce]

why the fuck does a scrollview act so weird in swiftui holy shit

[valid jetty]

i finally added allocator switching to elle xd

[woven mesa]

yea

sowwy my mom had my phone

she was playing pokemon go

[native spruce]

its okay

[royal nymph]

you will make a claude wrapper that adds markdown support

it's easy to do

just need to tell the ai in the prompt that it can use markdown and then well render it

[spark tiger]

sadly this won't fix me being broke as hell

[royal nymph]

does it not have a free tier

[spark tiger]

either i did something wrong or the free tier is only like 10 messages a day or every 6 hours

[royal nymph]

ohhh

[worldly sigil]

you could try duckduckgo's frontend https://duck.ai

[spark tiger]

amazing

[placid cape]

use it inside zed

In zed you have 10$/month for free

[spark tiger]

i mostly use chatgpt for some coding stuff at school from my phone

i'm just often very bored there so i just read some code and ask chatgpt to explain some stuff for me

wtf does that mean

[still jolt]

it means that the function is defined here ​​​​​​​​​​​​​

(you probably have an error or warn on a different line and it's misinterpreting extra info about it as something it should display)

[spark tiger]

yeah looks like that was it

yippie it compiled

[spark tiger]

is there a limit or something

that looks so ugly, but i hope it's fine

[winged mantle]

why is type not narrowing here...

emoji.name should become keyof icons, no?

(trying to make type checking stricter because i keep missing bugs )

[woven mesa]

@spark tiger hiii

[spark tiger]

hiii hru

[woven mesa]

eepy

hbu

[spark tiger]

me too

[austere anchor]

that would be implicit conversion

there's no guarantee that emoji.name is a key of icons as far as typescript is concerned

[winged mantle]

which is meant to happen

https://github.com/microsoft/TypeScript/issues/10485

i guess it only narrows the value on the right side

not the left

[austere anchor]

whar why doesn't that work

that's confusing

does it not check at runtime ??

[woven mesa]

@austere anchor gonna work on more ipod

https://tenor.com/view/yummy-opioips-playboi-carti-gif-5866923219055296347

[winged mantle]

if typescript was to narrow the type, the type of s would be keyof typeof a, when there are actually more keys which are just not required by the type

[austere anchor]

:3

[austere anchor]

what does the last part of your sentence mean

"there are actually more keys which are just not required by the type"

[valid jetty]

@placid cape you can finally switch allocators lol

(its logged after #free because i wanted to show that its just a stub for the gc and arena as you dont free like that)

you can also do --nogc to use arenas instead of a gc as the default allocator

[placid cape]

noice

i haven't worked on blom for few days bcs i don't have any time 😭

[valid jetty]

i dont either really but i had just a little bit of time so i did it

[winged mantle]

took several attempts lol

[woven mesa]

Warning once only: UITableView was told to layout its visible cells and other contents without being in the view hierarchy (the table view or one of its superviews has not been added to a window). This may cause bugs by forcing views inside the table view to load and perform layout without accurate information (e.g. table view bounds, trait collection, layout margins, safe area insets, etc), and will also cause unnecessary performance overhead due to extra layout passes. Make a symbolic breakpoint at UITableViewAlertForLayoutOutsideViewHierarchy to catch this in the debugger and see what caused this to occur, so you can avoid this action altogether if possible, or defer it until the table view has been added to a window. Table view: <FLEXTableView: 0x115d26a00; baseClass = UITableView; frame = (0 0; 390 844); clipsToBounds = YES; gestureRecognizers = <NSArray: 0x302513960>; backgroundColor = <UIDynamicSystemColor: 0x3030ae280; name = tableBackgroundColor>; layer = <CALayer: 0x302b64a00>; contentOffset: {0, 0}; contentSize: {390, 0}; adjustedContentInset: {0, 0, 0, 0}; dataSource: <FLEXHierarchyTableViewController: 0x11ec4b700>>

ok thanks bro

[austere anchor]

huh, then next i shall ask why const c: typeof a = b is even valid given that b is clearly not typeof a unless it somehow does some restriction thing such that you can only access the keys that a has

also audio explanation fire

[winged mantle]

i thought it would be easier than typing essay

but i paused

[austere anchor]

nvm reheard and you did say this

[winged mantle]

because if you have a type like { a: string } it does allow other keys

[austere anchor]

but why

QHAR

why is access like that even allowed

surely you should be forced to have typeof b to access b's keys

(i am coming from rust's typesystem for context)

[winged mantle]

ah

a very different system

[valid jetty]

it does not

[austere anchor]

indeed

no easy implicit casts unlike *script

[winged mantle]

you can do const thing: { a: string } = { a: "a", aijfkwajfjkwf: "aweiawejkewa" }

[winged mantle]

.unwrap() ❤️

[valid jetty]

only if you do { a: string } & Record<PropertyKey, any>

[austere anchor]

panics

vexedtomato

[austere anchor]

if let Some(x) = y

[winged mantle]

WHAT

my life is a lie

[austere anchor]

ok good

the type system is not completely ass

[valid jetty]

if you do this it lets you

[austere anchor]

the fact that typescript even has any is husk

it should be marked uhh

unsafe

[valid jetty]

and you still get autocomplete so its better than just Record<PropertyKey, any>

[valid jetty]

do you love that elle has any

[winged mantle]

oh wait

of course

[austere anchor]

hate

[valid jetty]

trolley~1

[winged mantle]

but that's specific to object literals

[woven mesa]

any

[winged mantle]

const a = {a:"eawewa", yormum:"afwjkwafkwajf"}
const b: { a: string } = a;

works

(i think)

[austere anchor]

var body: some View

[winged mantle]

(99.99% sure)

[valid jetty]

yeah i guess

[woven mesa]

what does this mean

[winged mantle]

my core point was

because if you have a type like { a: string } it does allow other keys

i forgot this doesn't work if you use the object notation

[valid jetty]

@austere anchor do you love

use std/prelude;
use std/allocators/arena;
use std/allocators/heap;

fn do_stuff<T>(T *allocator) {
    #set_allocator(allocator);
    $dbg(#env.allocator.kind);
    ptr := #alloc(i32, 5);
    #free(ptr);
    $dbg(ptr);
}

fn main() {
    $println("By default:");
    $dbg(#env.allocator.kind);
    $println("-".repeat(40));
    do_stuff(HeapAllocator::new());
    $println("-".repeat(40));
    do_stuff(ArenaAllocator::new());
    $println("-".repeat(40));
    $println("Default allocator:");
    do_stuff(#env.default_allocator);
}

[winged mantle]

but if you implement an interface in a class for e.g. yo ucan also add loads of _properties to keep internal state like i do

[valid jetty]

you can finally have temporary buffers and stuff

[winged mantle]

[austere anchor]

means apple is mad about you not recording this explicitly in NSPrivacyTrackingDomain

thumbsup~4

[woven mesa]

recording what tho

[austere anchor]

allocator pretty nice

[valid jetty]

before it was just GC lol

[austere anchor]

uhh whatever <private> is after u've managed to unhide it

[woven mesa]

fr

[valid jetty]

if you wanted to change it before you had to change the allocator at the time of compiling the compiler

skulley

[austere anchor]

@valid jetty does elle support no_std for embedded

[valid jetty]

yes

ellec foo.le --noalloc --nosm --nofmt --nostd

disables all

actually you dont need --nogc

if --noalloc is there it wont make gc anyway

although the issue is that it compiles to qbe lol

so youre not gonna be running it on embedded anyway

[austere anchor]

does qbe need an os or sumn?

[valid jetty]

no just the targets

i guess rv64

[austere anchor]

is arm64 not for embedded?

[valid jetty]

what i meant is like

youre not gonna be able to compile for sh3eb-elf

or whatever

[woven mesa]

30 warns

so good

[austere anchor]

eh most hobbiest stuff is just like arm64 or armv7 anyways

[valid jetty]

qbe doesnt have armv7 either lol

[austere anchor]

that should probably be added

[valid jetty]

probably hard

[austere anchor]

oh it has zero suport for 32-bit

nevermind then

[valid jetty]

so many

[valid jetty]

lmao yeah

[austere anchor]

@valid jetty elle on LLVM

troll~4

[valid jetty]

maybe in the distant future

it shouldnt be that hard tbh

just time consuming

consuming time i dont have

[woven mesa]

@austere anchor tiny spikes yippee

[austere anchor]

troller

[valid jetty]

i wonder if i can make more zero cost abstractions like this

[austere anchor]

why doesn't java abstract System.out.println to just println

vexedtomato

[valid jetty]

lmao

meanwhile i decided io::println was too long and made aliases so you can do $println

trolley~1

among others

[austere anchor]

https://stackoverflow.com/a/29337429 oh this is so cursed

java went the other way and decided nothing is global

[valid jetty]

im about to chat spam

so apologies

but is this a good explanation

is there anything else i should have

♡ Allocators

Elle has a moderately complicated allocator system. Here's how it works:

• By default:
  • garbage collection
• Using the --nogc flag at compilation:
  • arena-based allocation

Changing the allocator:

#set_allocator(MyAllocator::new());

Resetting to the default allocator:

#set_allocator(#env.default_allocator);

Make sure you don't forget to free any memory leftover when switching allocator! #set_allocator does not call the free_self method on the previous allocator when switching allocator, to allow for programs designed like this:

fn main() {
    arena := ArenaAllocator::new();
    #set_allocator(arena);
    x := [1, 2, 3]; // x is allocated through the ArenaAllocator
    #set_allocator(#env.default_allocator);
    $println(x); // allocates via default allocator
    #set_allocator(arena);
    #env.allocator.free_self(); // frees the arenas
    #set_allocator(#env.default_allocator); // go back to default allocator
}

What should an allocator have defined on it?

• Allocators should have the following methods defined on them:
  • MyAllocator::new() (preferrably allocating the allocator structure itself via mem::malloc)
  • MyAllocator::alloc(MyAllocator *self, i32 size) -> void * (size in bytes to allocate, should return void *)
  • MyAllocator::realloc(MyAllocator *self, void *ptr, i32 new_size) -> void * (new_size in bytes. should return void *)
  • MyAllocator::free(MyAllocator *self, void *ptr) (may be omitted if permitted by the allocation model, will become noop)
  • MyAllocator::free_self(MyAllocator *self) (destructor for the allocator itself, NOT objects created by it)

Disabling allocation altogether:

• You can pass the --noalloc flag during compilation. Keep in mind that, while this will remove allocators, you won't be able to use any of the standard library, as all of it depends on these allocators. This goes well with the --nostd flag.

its all a trick which is funny imo

the arbitrary allocator is a struct that has fields for the underlying allocator and function pointers to its methods

#set_allocator gets the type of that expression passed in, ensures its a struct or struct *, then sets the fields of the arbitrary allocator to the function pointers of the allocator you specify, or if it cant find a method it sets it to a noop method that returns nil

silly

probably better ways to do this

[fleet cedar]

#set_allocator(MyAllocator::new()); should return the previous allocator so that you can nest them hierarchically

[valid jetty]

that

is a bit cursed but i like it

the issue is that you lose access to the information about the allocator at compile time when you do #set_allocator

as it works with pure void *

so if it returned prev it would need to return a bunch of void * together or something idk

#env.allocator.kind is set to a string literal but that makes it runtime based

[runic sundial]

Just use, Zig?

Part of the standard library over here

[valid jetty]

the point is that im making my own language lol

[runic sundial]

LLVM front end?

[valid jetty]

nope qbe

[runic sundial]

Wharg

[valid jetty]

https://c9x.me/compile

[runic sundial]

What backends does qbe even have

[valid jetty]

@austere anchor good idea to bring $scoped back? idk how i feel about it

[runic sundial]

Where SPIR-V

[valid jetty]

so you can do

fn main() {
    x := 39;
    $scoped(ArenaAllocator::new(), fn(i32 *x) {
        arr := [*x, *x * 2, *x * 3];
        $println(arr);
    }, &x);
}
``` and this time it *will* call `ArbitraryAllocator.free_self` before setting to the previous allocator

[runic sundial]

Also kinda strange that free_self doesn't free associated allocations

[valid jetty]

it does

it frees the entire structure which includes allocations created by it

[austere anchor]

sure why not

[runic sundial]

Your spec said otherwise

[valid jetty]

that thing above was meant to more like, free_self shouldnt be used like free in c stdlib

as in, if you alloc something, dont use free_self to free it because that frees the entire structure

[runic sundial]

What's wrong with a model doing that?

It's easier in some contexts

Create allocator, do a bunch of dumb stuff, free the whole allocator

[valid jetty]

nothing is wrong i just said that free_self specifically frees the whole allocator including all of its allocations instead of just an object created by the allocator

its important to make that disctinction

because you might use it thinking it frees an object

idk

[valid jetty]

yea i know

this

fn main() {
    arena := ArenaAllocator::new();
    #set_allocator(arena);
    x := [1, 2, 3]; // x is allocated through the ArenaAllocator
    #set_allocator(#env.default_allocator);
    $println(x); // allocates via default allocator
    #set_allocator(arena);
    #env.allocator.free_self(); // frees the arenas
    #set_allocator(#env.default_allocator); // go back to default allocator
}

maybe i should be more specific

[runic sundial]

Yeah

Also get yourself the

deferred

Keyword

ASAP

[valid jetty]

better?

* `MyAllocator::free(MyAllocator *self, void *ptr)` (frees a specific object passed by pointer, may be omitted if permitted by the allocation model, will become `noop`)
* `MyAllocator::free_self(MyAllocator *self)` (destructor for the allocator itself including all of its allocations, **NOT** objects created by it)

[valid jetty]

elle already has defer

[runic sundial]

Epic

[valid jetty]

https://github.com/acquitelol/elle?tab=readme-ov-file#-defer-statements

use std/io;

fn main() {
    let size = 10;
    i64 *numbers = mem::malloc(size * #size(i64));
    defer mem::free(numbers);

    for let i = 0; i < size - 1; i += 1 {
        numbers[i] = i * 2;
        let res = numbers[i];
        io::printf("numbers[{}] = {}", i, res);
    }

    if numbers[2] + 1 * 5 == 10 {
        // Calls `free` here
        return 1;
    }

    // Calls `free` here
}

[runic sundial]

It's very useful for memory management

[valid jetty]

or in the case of custom allocators actually 1 sec

[runic sundial]

Also errdefer

[valid jetty]

use std/prelude;
use std/allocators/heap;

fn main() {
    #set_allocator(HeapAllocator::new());
    size := 10;
    numbers := #alloc(i64, size);
    defer #free(numbers);

    for i in 0..size - 1 {
        numbers[i] = i * 2;
        res := numbers[i];
        io::printf("numbers[{}] = {}", i, res);
    }

    if numbers[2] + 1 * 5 == 10 {
        // Calls `free` here
        return 1;
    }

    // Calls `free` here
}
``` """*modern*""" elle

[runic sundial]

Also allocators

How do you handle it between thread scopes?

[valid jetty]

threading is not exactly something im working on at the moment lol

[runic sundial]

No like looking into the future

Would this need a whole ass rework for it?

[valid jetty]

i have no idea xd

[runic sundial]

Trolley

Global state moment

Guess async corotuines would also be out of the question then

[valid jetty]

making the current allocator be a mutex sounds kinda painful

[runic sundial]

Zig just treats em as a struct

Shrimple as

[valid jetty]

probably because theres no inline asm yet so coroutines cant really be done anyway unless you like link with custom asm objects when compiling (which tbf could be part of the runtime, at one point i was doing something like that with asm for finding the top of the stack for the GC i made)

then i found a better solution (or well simpler and more portable) so platform specific asm wasnt needed

unfortunately thats not possible for coroutines because you need to save registers which is very platform specific

[winged mantle]

javascript sexism???

[native spruce]

@woven mesa how do u do your uikit list with swiftui

[woven mesa]

r u having issues

[native spruce]

I just wanna use a uicollectionview instead of a lazyvgrid and a scroll view

but i find uiviewrepresentable odd

[woven mesa]

appswitcher example

[native spruce]

app switcher?

[woven mesa]

icon *

[native spruce]

I will need to experiment this is good code to reference from ty

[woven mesa]

no problem oomf

sorry its not the simplest and also even buildable code sample

[native spruce]

question does this update like if a @State changed

[woven mesa]

im just eepy

uhh its not configured for it

but it wouldnt be difficult to implement

actually @native spruce it would work i think bc it calls reloaddata on the table

[native spruce]

oki

[valid jetty]

did i yap too much

fat unit test
https://github.com/acquitelol/elle/blob/rewrite/examples/tests/allocators.le

[winged mantle]

it'd be so nice if you could just get typescript to return unknown instead of any when indexing on unknown key

currently it seems you have to use as keyof whatever to get it to work but if circumstances change this could stop being safe

[valid jetty]

i cant find it but there is a package that turns a lot of stdlib types to return unknown instead of any

[winged mantle]

how would that work with lsp though

[valid jetty]

wdym

its just a package that edits a bunch of types to return unknown instead of any

it would just work as if it was returning unknown to begin with

[winged mantle]

ah i do not believe you said stdlib types before editing it

the issue here is indexing objects

this is any instead of unknown

I tried turning on noExplicitAny again but it is such a pain

even if you put unknown it has an error

[native spruce]

@woven mesa this took forever

but is it beautiful

I use a uicollectionview because of speeds

list and lazyvgrid sadly don't meet up to my standards :c

[native spruce]

swiftui is pissing me off I think

[formal belfry]

https://cdn.discordapp.com/emojis/1326737881062051954.webp?size=48

[native spruce]

ngl I may just choose uikit for this view because holy shit lol

combining swiftui and uikit stuff has never been so difficult

[formal belfry]

I think SwiftUI is a good option sometimes and then I use it and then I regret it immediately because one really valuable modifier is missing or something

[native spruce]

swiftui is good for simple views with not much complexity

but when you want something that’s performant (like using a uitableview or collectionview) there will be problems

For example pushing a swiftui view from a uiviewrepresentable is horrible

[native spruce]

I wanted to use this api but iOS 16 only

there goes my hopes and dreams

[frosty obsidian]

swiftui just keeps sounding like compose but if a quarter of its apis were deprecated

[native spruce]

is there a uikit version of compose

[frosty obsidian]

closest would be the view system

[native spruce]

but this is not related to deprecated apis

[frosty obsidian]

but it uses layout xml files

if you hate yourself you can do it all programmatically

[native spruce]

well uikit is like that but it also has an option to not use xml layouts

I do it programmatically

(imo the proper way of doing it)

[frosty obsidian]

its just extremely verbose in android land

most devs will just do the basic layout in xml and make it functional in code

[native spruce]

programmatic uikit https://github.com/khcrysalis/Feather/blob/main/iOS/Views/Sources/SourcesViewController.swift

lovely isnt it

[frosty obsidian]

telegram devs were a little insane though and only did it programmatically

(it is technically faster since no xml parsing is needed)

[native spruce]

telegram devs so good

[frosty obsidian]

its only really beneficial for very low end devices

typically very old ones

as telegram supports some ancient android versions

[frosty obsidian]

nvm its min sdk 23

my apps actually support older versions

[icy needle]

@valid jetty look what you made me do

https://cdn.discordapp.com/attachments/1320430127334821949/1343999037162127463/GkpXDOpWMCYAs4m.png?ex=67bff8e4&is=67bea764&hm=b0711296fbbc1a33be57bcb1bee1883b2632575d2e07fb48bbb1ecd846c1ccf9&

[valid jetty]

beautiful

[woven mesa]

disagree

[native spruce]

what do you do then

uihostingcontroller?

navigationDestination?

[woven mesa]

UIHostingConfiguration works perfectly

[native spruce]

I said pushing a swiftui view not UIHostingConfiguration

[woven mesa]

o

mb

[native spruce]

like onto the navigationstack

[woven mesa]

is navigation link breaking

[native spruce]

how would I use a navigationlink inside of a uicollectionview (representable) without a uihostingconfiguration

[woven mesa]

introspect and manually push a new view to the stack

wrap your new swiftui view with uihostingconfiguration and push that

[native spruce]

wouldn't that just use a uihostingcontroller

reminder I said without a uihostingconfiguration

[woven mesa]

mbmb

[icy needle]

this is how it looks in the code

i spent several days trying to get method access to work properly like an idiot

then realized pratt parsing just makes it work out with no effort anyway

[valid jetty]

why are curly brackets unary operations

[icy needle]

i thought it made sense

if you have one item in brackets, it fills up the only slot

if you have 2, you use a comma which is a binary op

for more, you chain commas

tho i'll probably have to restructıre them into a list

[icy needle]

and if you have 0 items, lexer inserts a control token

[placid cape]

do you want it to be compiled?

[valid jetty]

interesting

[valid jetty]

i didnt bother looking up popular parsing methods and using them lol i made my own from scratch

which means that the field access thing is unpredictable (its the wrong way around)

[icy needle]

yep

[icy needle]

i was gonna do that but a friend told me to look into pratt parsing a bit

so i did, and i'm glad that i did

[icy needle]

💀

[valid jetty]

although i dont do namespace accessing like that lol

1 sec

lo

i honestly dont know what else to do now

lmao

maybe its finally time to make some enums

or like refactor the 4.5k line compiler into multiple files

[icy needle]

yeah that's kinda weird imo too

basically for me they work like regular math operators

so left to right and respecting priority

if an op has a later order, it swallows the earlier order ones before it in the tree

that's basically how pratt parsing works

[valid jetty]

thats how i wish i did it tbh but oh well

i do that for ranges tho lol

.. is an arithmetic operator that takes in lhs and rhs and calls a function that creates a range out of them

[placid cape]

into which backend?

and is it open source?

[icy needle]

half of me wants to use llvm but i gotta be more reasonable

so probably qbe

[placid cape]

yeah qbe is nice for start

[icy needle]

idk how usable tilde is

[placid cape]

i need to finish my parser refactor

[valid jetty]

i dont personally like sea of nodes IR

also theres like no documentation on it yet

qbe doesnt have lots but it as at least a litt;e biot

[placid cape]

what are you talking about?

[valid jetty]

https://yasserarg.com/tb

[placid cape]

interesting

[placid cape]

this is smart

[valid jetty]

if youre planning to make your compiler more complicated or eventually represent a higher level language i dont recommend qbe

its too simple for that

i started it off with it and now i regret it because the compiler is too complicated to represent some programs in qbe and i need to do a lot of heavy lifting in the front end

[placid cape]

any examples?

[valid jetty]

generics, deferring, custom allocators are all compilation gimmicks, qbe has no way to do any of that stuff

dynamic dispatch too

i want to have dynamic dispatch because its very useful but qbe just makes it very hard as its so simple

not that llvm has that stuff either but its type system is a lot stricter which makes it easier in the end

[placid cape]

understand

[valid jetty]

a simple program like this

use std/prelude;
use std/allocators/heap;

fn main() {
    #set_allocator(HeapAllocator::new());
    x := #alloc(i32);
    #set_allocator(#env.default_allocator);
}

compiles to all of this

oh

lol

the assembly is even worse

and keep in mind this is with a runtime so most functions that are used dont even have their bodies

im thinking

would it be a good idea to get rid of the /std prefix?

so you can do use prelude; and use allocators/heap; directly

[placid cape]

nah

[valid jetty]

i can do that really easily

[placid cape]

keep it

[valid jetty]

hmm ok

[brave moss]

it's look like rust but something tell me this is not rust

[placid cape]

this is not that bad

[valid jetty]

i was thinking because some modules can be quite lengthy to type out like use std/collections/hashmap;

or use std/runtime/allocators/arbitrary;

[valid jetty]

nop its a thing im making

[placid cape]

i would keep it as it is

[valid jetty]

maybe thats for the better tbh

[brave moss]

what is the language ?

[valid jetty]

elle

[placid cape]

elle

and i have blom

[valid jetty]

blom!!

bloom

[placid cape]

fun @native printf(format: string, ...) -> i32;

fun @infix lol(left: i32, right: i32) -> i32 {
  return left + right;
}

fun @infix lol(left: string, right: i32) -> i32 {
  return 99 + right;
}

fun @infix lol(left: i32, right: string) -> i32 {
  return left + 99;
}

fun @public main() {
  printf("%d\n", 5 lol 3 lol 9 lol "a");
  return 0;
}

[valid jetty]

broom

[placid cape]

infix

you can put the annotations anywhere

so you can also do @public fun main() now

[valid jetty]

that sounds really hard to make i would struggle with that

not the annotations

but overloaded infix functions

[placid cape]

you mean infix

yeah its horrible

parsing infix functions is the biggest mistake i ever done

i need to figure out how to do it properly

[valid jetty]

i literally got rid of the incrementing operator (aka postfix and prefix) from elle because it caused so much parsing and compiling slop lol

[placid cape]

i want to keep the syntax as simple as possible

[valid jetty]

maybe one day ill do the go thing and make it a statement and ONLY let you do x++

yeah thats what im doing mostly

[placid cape]

so i really dont want things like ?. etc like in js

[valid jetty]

i have a small set of things and im extending existing ones

[placid cape]

ill also rework the field access to be just a math operator, like @icy needle

[valid jetty]

im adding a lot of things to this

maybe ill add a #reset_allocator(); which does #set_allocator(#env.default_allocator) for you

[placid cape]

compile time functions 👍

[valid jetty]

ya

imo better than stuff in C where it looks just like a regular function

the # thing is inspired from rust with the ! for macros

[placid cape]

i just have @ like in zig

but ill probably change it bcs now i have problems to figure out what is an annotation and what's a compile time function

since each statement can potentially have annotations

[valid jetty]

yeah thats why i made them different lol

i just have a thing called sigils

[placid cape]

👍

i also dont want to end up like ruby where you dont know what are you parsing lol

[valid jetty]

lol so easy

fn parse_reset_allocator(&mut self) -> AstNode {
    let location = self.current_token().location.clone();
    self.advance();

    self.expect_tokens(vec![TokenKind::LeftParenthesis]);
    self.advance();

    self.expect_tokens(vec![TokenKind::RightParenthesis]);
    self.advance();

    AstNode::SetAllocator {
        value: Box::new(AstNode::FieldAccess {
            left: Box::new(AstNode::Environment {
                value: None,
                location: location.clone(),
            }),
            right: Box::new(AstNode::Literal {
                kind: TokenKind::Identifier,
                value: ValueKind::String("default_allocator".into()),
                location: location.clone(),
            }),
            value: None,
            location: location.clone(),
        }),
        location,
    }
}

[placid cape]

implement tail call recursion optimalization

https://lists.sr.ht/~mpu/qbe/<CALzWbOzJY9pUEgt9gdMF8bBbpca7EJz+fVksKddPLtr-zp3zRA%40mail.gmail.com>

[valid jetty]

eventually

ok it works

that was fast

[icy needle]

yeah that's what i figured

or i could do the vlang meme and transpile to c

[valid jetty]

true

[austere anchor]

coding for Elle in the morning goes crazy

gm

[valid jetty]

gm

[placid cape]

gm

[valid jetty]

so easy to add new features

https://github.com/acquitelol/elle/commit/44d39423d4d82b623fe23cde86889a09a769e0d1

literally 32 additions for a whole new builtin

[placid cape]

the hardest thing is to start and make a proper structure

[austere anchor]

@valid jetty support for aligned allocs?

[valid jetty]

allocs on stack are aligned to 8 bytes

alocs on heap are aligned to ptr width which is usually also 8 bytes

but no way to set custom alignment yet

qbe supports it, there is alloc4, alloc8, alloc16 for stack

and i could just allow for custom alignments in the allocators thats not particularly hard