development

[neon leaf]

https://github.com/0x7d8/draw-together/blob/main/src/main.rs#L117

[quartz kindle]

drawing together

[neon leaf]

https://github.com/0x7d8/draw-together/blob/main/src/main.rs#L146

https://github.com/0x7d8/draw-together/blob/main/src/main.rs#L160

[quartz kindle]

ah i see

isnt it a bit overkill?

like

if you have 100 people drawing

you have 300 threads lel

[neon leaf]

apparently this is how its properly done

🤷

not like there is another way to really do it

[prime cliff]

Performance

[quartz kindle]

async?

lmao

[neon leaf]

ehh

sadly not how it works with tokio

[quartz kindle]

well idk rust so im just guessing

but i know too many threads = bad

[neon leaf]

I mean

its also in their docs https://docs.rs/tokio/latest/tokio/task/fn.spawn.html#examples

[prime cliff]

Just get a better server

[quartz kindle]

come https://draw-together.rjansen.dev/

[neon leaf]

ah @quartz kindle

not all spawn calls go to different threads

Spawning a task enables the task to execute concurrently to other tasks. The spawned task may execute on the current thread, or it may be sent to a different thread to be executed. The specifics depend on the current Runtime configuration.

[quartz kindle]

ah

then it should be fine

you can then do the same thing for the pixel array

create a task that receives the pixel messages and updates the snapshot

[wheat mesa]

You could have a different thread process each chunk of pixels, something like that

That way you are never concurrently writing on the same section of the array

But you still benefit from concurrency on a per-line basis

[neon leaf]

welp ill look into it tomorrow 💤

[lyric mountain]

ah, it's like pixelcanvas

when I did something similar I simply stored the operations, instead of the canvas itself

then simply re-generated the canvas on the clientside

this way u dont need to worry about concurrency or flooding your server with threads

[celest monolith]

hello mate

[frosty gale]

out of all the messages above he decided to say hi to tim specifically

[quartz kindle]

lel

[pearl trail]

better luck next time

[queen needle]

Fittingly with this talk, how would you implement websockets for something like honk.me?

[solemn latch]

what is that? 👀

rip, it died before I even knew what it was

[queen needle]

I couldn't link it, but it was a chat app where you saw the typing live

and I've always thought what would be the best way to implement that websocket wise, because with something like a collaborative google doc there is still delay in the typing

[solemn latch]

chat apps are typically one of the first things people do with websockets, tons of guides on how to do it.

[queen needle]

Well I meant specifically that instant typing aspect, if you visit the website you can see an example of it

Because the first thought is, send every keystroke, but I guarantee there's a better way

[solemn latch]

Are you talking about the server side or client side?

Server side it just receives the start typing, and end typing and broadcasts it to whatever clients it needs to.

Client side it just has a timer that goes off after x time once the user stops typing and sends the event or if they send a message.

(or well thats how discord handles it)

[queen needle]

But i meant on there it shows what you are typing

[pearl trail]

no its not about start and end typing. if you see on the website, every character a user typed, itll be visible to the other user

[solemn latch]

Maybe its just a stream then? 👀

I wouldnt expect it to be handled with websockets, the overhead on that would be pretty large 👀

[warped summit]

Rate

[solemn latch]

Maybe they would buffer requests and make it appear as though you're seeing the live typing.
Ie, a user starts typing, buffer the first word, send it. type it the way the user typed it, buffer the next segment

[solemn latch]

fancy 👀

[warped summit]

☺️ Sml

I need to give it a aesthetic look 🥲

Ik this project is gonna blow up

[queen needle]

that's what i thought

like a debounce effect, that displays different on the client if that makes sense

[solemn latch]

It could be per character, but it wouldnt use websockets

[solemn latch]

My mind just exploded.
TCP is handled by the os, not the browser

[solemn latch]

Looking at other protocols than websockets lead me so far down the rabbit hole.

I somehow ended up on graphql

[lament rock]

That is a very way out there destination on the journey

The mention of graphql reminded me of this cool mc project I saw a long time ago. https://github.com/WorldQL/mammoth

[radiant kraken]

smallest rust project

[neon leaf]

@quartz kindle got it down to 7bytes per msg for now, instead of 9
x (u16) + y (u16)
turned into
xy (u24)

action (u8) + height (u8)
turned into
actionHeight (u3 + u5)

height is multiplied by 4 to look better, aka 4-128 (same as before)

I was also thinking of maybe doing x (u13) and y (u11) since usually screens are wider, not taller

that would be 8192x2048 instead of 4096x4096

[frosty gale]

cargo is basically the new node_modules meme

its the same thing

you import one package that package imports 50 different packages each of which has different code/binaries depending on the OS youre using which further adds to the package size

[radiant kraken]

yup

at least its blazingly fast in the end right 🚀

[frosty gale]

one thing i appreciate about C++ is that packages are usually standalone unless its some big feature which realistically cant be standalone

i just drop it into my dependencies folder and it works

package systems spoil library developers a bit so they dont feel as restricted importing packages for small things

[radiant kraken]

except i fucking hate LNK2001 and LNK2019 so much

what caused this 😭

i've already used TOPGG_EXPORT in every class declaration which should expand to __declspec(dllexport / dllimport)

it literally says dllimport here

and ```cmake
option(BUILD_SHARED_LIBS "Build shared libraries" ON)

if(WIN32)
target_compile_definitions(topgg PRIVATE $<$BOOL:${BUILD_SHARED_LIBS}:TOPGG_BUILDING_DLL:DPP_STATIC TOPGG_STATIC>)
endif()
c
#ifdef _WIN32
#if defined(DPP_STATIC) && !defined(TOPGG_STATIC)
#define TOPGG_STATIC
#elif defined(TOPGG_STATIC) && !defined(DPP_STATIC)
#define DPP_STATIC
#endif
#endif

#if defined(_WIN32) && !defined(TOPGG_STATIC)
#ifdef TOPGG_BUILDING_DLL
#include <dpp/win32_safe_warnings.h>
#define TOPGG_EXPORT __declspec(dllexport)
#else
#define TOPGG_EXPORT __declspec(dllimport)
#endif
#else
#define TOPGG_EXPORT
#endif

@wheat mesa

😭

[quartz kindle]

yeah thats cool, just keep in mind that pushing a canvas to those resolutions is gonna blow up the browser lmao

canvas performance drops exponentially with canvas size

[neon leaf]

ah well I implemented 90% of it now, only issue is loading existing state

since its a whopping 67.1mb

[quartz kindle]

xDddd

[neon leaf]

takes ~13sec to download for example

and 1min to load

...

[quartz kindle]

lmao

[neon leaf]

    fetch('/history_2.raw')
        .then((res) => res.arrayBuffer())
        .then((buf) => new Uint8Array(buf))
        .then((arr) => {
            let offset = 0

            while (offset < arr.length) {
                const [ type, height, color ] = fromFormatInternal(arr.slice(offset, offset + 4))
                const x = (offset / 4) % 4096
                const y = Math.floor((offset / 4) / 4096)

                draw(x, y, color, type, height)

                offset += 4
                messages++
                bytes += 4
            }
        })```

[quartz kindle]

i wouldnt make a canvas bigger than 2k res

[neon leaf]

I mean I think itll still take ages with 2048

how could I "compress" this data

[quartz kindle]

the loading can be improved by ordanizing the data into a format that can be applied directly to the canvas imagedata object

for the size of the data itself you cant really do much except for some sort of compression like repeating pixels of the same color

[neon leaf]

hmmm

actually I forgot something

displayed resolution is 16384x16384

height is multiplied by 4

[quartz kindle]

lmfao

[neon leaf]

well maybe ill decrease then

to 1024x1024 or something

4.1MB

[quartz kindle]

thats literally a 1gb image

lmfao

[lament rock]

16k is insane even in game engines

8k is typically seen as a max

and thats insane and seen as for atlases

[quartz kindle]

16384 x 16384 x 4
= 1073741824

that should be the size of the canvas in memory

[neon leaf]

ok it loads a lot faster now

1024x1024 with real res 4096x4096

~7s

https://3000.webhook.pet/

impossible writing on a trackpad

still need to implement ws buffering

and make client drawings show immediately

[radiant kraken]

oh nvm it was apparently caused by this 😭

[neon leaf]

and need to properly scale the canvas

[radiant kraken]

gotta love when my http request never returns ❤️

@green kestrel does the bot needs to be online for dpp::cluster::request to work?

because when i tried m_cluster.request without the bot starting the function just never returns

[delicate zephyr]

@harsh nova

[green kestrel]

in 10.0.x, you just need to have instantiated a cluster

in 10.1.x you need to have called dpp::cluster::start() to run the reactor loop

[radiant kraken]

i see

[green kestrel]

10.1 uses a unified event loop for everything

which is spun up in bot.start()

without that, it cant make any requests

10.0 just launches requester threads

[radiant kraken]

i think i'll do bot.start(st_return);

okay it worked now!

wth 😭

oh the body just stops there??

@green kestrel do you have a limit on http response bodies? because i've been getting invalid string: missing closing quote errors

[radiant kraken]

here body stops at 672 bytes

what makes it stop at 672 bytes?? is it encoding-related? i'm not sure

what's wrong with the W character????

[vague flare]

Phoenixthewilld

Is my roblox name

[green kestrel]

no there's no limit

[radiant kraken]

then how

did it just stop

[green kestrel]

but wtaf is this

three bar equals fif?

idk,.works on my machine

use Linux like a normal bot dev

10.1 hasn't been as tested on windows as elsewhere, could just be iffy there

real production bots aren't using windows with dpp

[radiant kraken]

i presume an ASCII rendering of an UTF-8 emoji #development message

[radiant kraken]

what????

howwww

@green kestrel i rewrote my code to just this

  dpp::http_headers headers{};

  headers.insert(std::pair("Authorization", "token"));
  headers.insert(std::pair("Connection", "close"));
  headers.insert(std::pair("Content-Type", "application/json"));
  headers.insert(std::pair("User-Agent", "topgg (https://github.com/top-gg-community/cpp-sdk) D++"));

  bot.request("https://top.gg/api/bots/264811613708746752", dpp::http_method::m_get, [](const auto& response) {
    std::cout << "===BODY START===\n" << response.body << "\n===BODY END===\nSIZE: " << response.body.size() << std::endl;
    g_sem.release();
  }, "", "application/json", headers);

  g_sem.acquire();

and now it gives 2,041 bytes??? what?

out of curiousity i reran the executable again (without recompiling) and i got the entire body (3,746 bytes)

?????

i feel like this is from dpp's end

i'll file an issue for this in the DPP repo if you want

i reran the executable again, and i got 3,746 bytes, and then 3,410 bytes

it's very much random

[neon leaf]

@quartz kindle everything implemented https://draw-together.rjansen.dev/

[quartz kindle]

nice

[neon leaf]

last issue is something you may see when refreshing

[quartz kindle]

i see

the pixels are cut off?

[neon leaf]

pixels are displayed as size 4, and stored as size 1

so 4 pixels can be on one spot even with smallest size

[radiant kraken]

https://github.com/brainboxdotcc/DPP/issues/1382

[neon leaf]

i think ill store both history and pixel data

pixel data so I can remove unused history entries

history so I can accurately recreate pixel data

[green kestrel]

try taking connection close header off

idk why you'd send that, but that's at odds with http 1.1 and handled by dpp

you should only need token and content type

are you able to duplicate this without the top.gg api

if instead of top.gg API you query for example this URL... one sec

https://freetestdata.com/wp-content/uploads/2024/01/sample2zip.rar

[neon leaf]

@quartz kindle final solution I swear https://github.com/0x7d8/draw-together/commit/717c4766279a5ff6c01c955efca20f540b6b17ff

[radiant kraken]

ohh taking connection close header off actually fixes it 😭

i thought connection: close was mandatory for single HTTP requests done with HTTP v1.1

because i tried it once before and it worked

kinda weird that undoing this fixed it

[green kestrel]

it is, but dpp is adding one

so you are sending two

[radiant kraken]

😭

you should've wrote it in the documentation

[green kestrel]

it might be nice to make a pr for that to stop people making this mistake

but there seem to be other issues with long running bots on windows, im looking into that

[radiant kraken]

cuz i thought dpp::cluster::request is very low level

[green kestrel]

tbh didnt expect people to be adding random headers that affect how dpp handles its connection lol

that is confusing the server, i bet

[radiant kraken]

yeah that's weird

[green kestrel]

its not the lowest level, this is: https://dpp.dev/classdpp_1_1http__request.html

bot.request/bot.co_request create one of these and insert it into the queue

below that is THIS, which you really cant directly use
https://dpp.dev/classdpp_1_1https__client.html

[radiant kraken]

if (!header.has("connection")) header.insert(std::pair{"connection", "close"});?

[green kestrel]

hm, the opposite, you shouldnt be able to override it

something like

if (header.has("connection")) { ... drop that header like a mic ... }

[radiant kraken]

lmao nice

throw an exception?

or just delete it

[green kestrel]

deleting it is the least breaking change

isnt going to crash anyones bot

[radiant kraken]

aight nice

[quartz kindle]

shouldnt headers be deduplicated?

[radiant kraken]

yea

[green kestrel]

no

thats the reason its a vector of pairs and not a map

nothing in the http spec says headers should be deduplicated, and many headers can happen multiple times

like cookie headers

admittedly, connection: close is a weird one

[radiant kraken]

interesting

i've yet to see code in the DPP repo that adds a Connection: close header though

[green kestrel]

From RFC2616 section 4.2:

    Multiple message-header fields with the same field-name MAY be present in a message if and only if the entire field-value for that header field is defined as a comma-separated list [i.e., #(values)]. It MUST be possible to combine the multiple header fields into one "field-name: field-value" pair, without changing the semantics of the message, by appending each subsequent field-value to the first, each separated by a comma. The order in which header fields with the same field-name are received is therefore significant to the interpretation of the combined field value, and thus a proxy MUST NOT change the order of these field values when a message is forwarded.

so

the server may reinterpret:

Connection: close
Connection: close

->

Connection: close, close

[radiant kraken]

wow i thought the problem was fixed and now bot.start(dpp::start_type::st_return); just crashes 😭

[green kestrel]

https://stackoverflow.com/questions/11477108/is-this-http-header-wrong

this SO post says the outcome of two connection headers is non-deterministic and depends on the server side

[radiant kraken]

how could bot.start() crash tho?

[green kestrel]

no idea

like i said

windows doesnt have much testing, wont get much testing

[radiant kraken]

dpp::cluster bot{discord_token}; discord_token here has been checked previously if it's a null pointer

[green kestrel]

nobody uses it in prod 😄

[radiant kraken]

😭

i guess i am one of those 0.1%

[green kestrel]

hmmm this isnt a constructor

is it?

its an initialisation

[small tangle]

this RFC is apparently superseded by RFC9110 https://www.rfc-editor.org/rfc/rfc9110#section-7.6.1 and also there is says that the connection header is a list of connection options

[radiant kraken]

same thing happens if i used ()

[green kestrel]

i cant help you without a stack trace

i might as well just guess and say "it was space aliens"

[radiant kraken]

how do i get a stack trace

there's no stack trace

[green kestrel]

use debugger

on a debug build of dpp

[radiant kraken]

OH WAIT

I REALIZED I CALLED bot.start() TWICE

😭

[green kestrel]

#825411275458740304 message

[small tangle]

nulllll

[green kestrel]

how exactly are you able to call it twice if its st_wait

[radiant kraken]

its st_return

[green kestrel]

lol

why

you got a loop after it?

[radiant kraken]

no

[green kestrel]

it wont work then

the minute the cluster goes out of scope it'll be torn down, wont wait for requests to finish

[radiant kraken]

i was ctrl-zing my test.cpp file (cuz i previously need to write it as an example for the github issue)

i wrote another bot.start() line and commented the older one

[green kestrel]

you might want to do this

https://dpp.dev/webhooks.html

before returning:

    /* Wait here for the webhook to complete, but we could do anything we need here */
    while (bot.active_requests() > 0) {
        std::this_thread::sleep_for(std::chrono::seconds(1));
    }

this may or may not do what you want, but its better to use bot.start(dpp::st_wait)

[radiant kraken]

i wasnt doing it intentionally

[green kestrel]

ah ha

ok

[radiant kraken]

[green kestrel]

should it let you bot.start twice?

[radiant kraken]

no

theres no checks for it

just crash

[green kestrel]

    /* Wait here for the webhook to complete, but we could do anything we need here */
    while (bot.active_requests() > 0) {
       std::this_thread::sleep_for(std::chrono::seconds(1));
    }

gonna add this in a pr

its such a fundamentally weird thing, i never even considered it lol

[vocal knot]

while loops infuriates me

[green kestrel]

while (midnight_wolf_infuriated) { infuriate(midnight_wolf); }

[radiant kraken]

yippeeee

finally got to know that my C++ SDK is working well by testing it with top.gg's small handful of still-working endpoints

god why does get_bots take like 2-3 seconds to finish

[green kestrel]

https://github.com/brainboxdotcc/DPP/pull/1384
https://github.com/brainboxdotcc/DPP/pull/1383

first one will make sure nobody else calls bot.start twice again

[radiant kraken]

oh damn

you're very committed

[green kestrel]

second one is nothing to do with the http request headers, but i noticed it while looking

[radiant kraken]

lmao

[green kestrel]

...git pun?

[radiant kraken]

yeah pun intended

[green kestrel]

you mean i should be committed

[radiant kraken]

@harsh nova are you proud of me

[harsh nova]

huh?

[radiant kraken]

of my dad joke 😭

[harsh nova]

OHHHH

hahahha yes

I love dad jokes

[green kestrel]

whyd the blind man fall down the well

because he couldnt see that well

[harsh nova]

hahah good one! Might perhaps steal that one

[proven lantern]

is there a rate limit on these at mentions? it keeps going back and forth between being a valid and invalid at mention

here's the same message

[limpid onyx]

It's just a cache issue, whether your client has the user in cache or not

[green kestrel]

@radiant kraken https://github.com/brainboxdotcc/DPP/pull/1385

[proven lantern]

shouldn't the discord desktop client fetch user info if there is a cache miss? and if there are too many misses it would get rate limited because it does too many API calls right?

[limpid onyx]

It's probably to avoid spamming the api, yes. There's a vencord plugin to fetch the user if you hover it

[radiant kraken]

nicee, good job!

[lament rock]

Typically, the api sends an object map of members that were mentioned, so you dont need to fetch them

[neon leaf]

I think the host header would be useful to keep changable

sometimes you need to send stuff to an ip with a host header because multiple sites are on it

[green kestrel]

no

it's set by the hostname you provide when you make the request

setting it a 2nd time in the header would be pointless

[neon leaf]

yeah I know, thats not always what you want though

[green kestrel]

yes it is

the interface isn't that low level

if your host doesn't resolve, that's not something that's a dpp problem

[amber rose]

Does anyone use JDA wrapper?

[lyric mountain]

I do

[amber rose]

Nice

How is the experience?

[lyric mountain]

normal, ig

idk what to compare against, only other I've used was d.js when I barely knew how to code

but being java it's a lot more pleasant to me

[violet flicker]

where can i get webhook authorization token

[pearl trail]

i wonder why the data somehow is not the same from what being logged and inserted to db

[pearl trail]

maxspoilerlength:200 doesnt work, but maxspoiler:200 works? fixed with a very bad way

[lyric mountain]

in your bot edit page, webhook section

but only after it's approved

[violet flicker]

Yeah thanks!

hey i need help with vote registration
i have set the webhook to discord webhook on top.gg

but when i vote but didn't registers it niether sending webhook
but when i manual testing !testwebhook and !testvoteapi they works fine

should i send code here?

[lyric mountain]

yes

[quartz kindle]

and he was never seen again

[frosty gale]

do not send the code he will steal it and make millions

[frail cypress]

hey yall, i've been relying on the query parameters when voteing, but now it looks like there is no query data passed to the webhooks? any changes or something i can work with?

[humble gyro]

hey, what query value do you add in there? 🤔

[frail cypress]

the guildId that the user voted from

it was working before but ig something changed and now an empty query is passed

[humble gyro]

Sorry about that -- that's most likely an issue

Could you send me your vote URL?

[frail cypress]

can you escalate that to the development team and get it fixed?

[humble gyro]

That's me

[frail cypress]

https://top.gg/bot/1079456745971261591/vote?guildId=1119014578316128426

it should look like this

[humble gyro]

thanks!

[frail cypress]

please let me know as soon as its fixed (dm's fine)

[humble gyro]

will do!

[quartz kindle]

ehmahgerd its verudo-san

[humble gyro]

hello tim

[quartz kindle]

herro!!!

[neon leaf]

perfectly cursed, as all things should be

[quartz kindle]

nice

i love binary formats

[pulsar marten]

how to post shard count using python sdk ?

i am confused

post_guild_count() ?

[quartz kindle]

according to the docs you can use topgg.DBLClient(bot=client, token=topggtoken, autopost=True, post_shard_count=True)

[proven lantern]

I'm not fetching in this case, it's the discord desktop app. But i'm thinking about moving away from those at mentions and just using their global_name instead.

[lament rock]

I'm just saying. The desktop client should have the user info so long as the members are part of the guild since the API does return that info

[proven lantern]

that's a good point. i was able to see those two users in the Member List so that data should have already been loaded, but it still couldn't show the at mention. maybe at mentions need more data for some reason

it's doing it right now

and now it's showing up again

[lament rock]

More than likely just a client bug

Err actually idk how mentions are handled when under embeds

But when they're in the content, the mentions is definitely having the user data sent

[proven lantern]

the only difference i know between embed and content at mentions is the embed at mentions dont ping people

[solemn latch]

pings include user data iirc

when you send a webhook message you can disable pinging at all, so it can happen in message content too.

[frosty gale]

yeah embeds have had this bug for years now i think and discord still hasnt fixed it

it looks like a caching issue, if you go on the users profiles on your client they should show up normally

the only reliable way for mentions to show up in embeds is to add it to the message content

[primal fog]

Hello quick question what do you think is more suited monetize option for discord bots

A) One time payment
B) Subscription

I feel like users are more inclined on A but B can be more profitable long term

[viral wedge]

I imagine it depends on a few factors:

1. Frequency of meaningful updates(This more applies to game bots, I suppose)
2. How much time did/do you spend on development?
3. How much does it cost you to maintain?
   4.Value of the features(Does it do something original or in a better way that users will care about, that they can't find in a free bot?)

[radiant kraken]

*posting shard count will soon be deprecated btw

i think veld is deprecating anything shard-related in Top.gg

[amber rose]

What build tool do you use?

Maven or Gradle?

[lyric mountain]

Gradle 100%

I only ever use maven for publishing libs

[amber rose]

Good stuff 🥶👌

[lyric mountain]

I just hate xml

[amber rose]

Fr

[proven lantern]

subscriptions are blocked in some countries like India because they have consumer protection laws that don't allow subscriptions. so if you do subscriptions make sure to support purchasing subscriptions via consumables too

[prime cliff]

Doing good progress

[humble gyro]

so based

[frosty gale]

i just realised docker is like electron but for servers

uses lots of storage and memory for each instance

[neon leaf]

no not really

cgroups are very efficient

~1-2mb ram overhead

[humble gyro]

you can use micro VMs for minimum resource consumption

[neon leaf]

^ lxc

[humble gyro]

adding indexes is like magic man

50% less db usage

[acoustic bough]

would want to know how much money you'd save by completely removing mongo from the stack

[humble gyro]

several thousands

unironically

[coral trellis]

mongo stinky fr

[acoustic bough]

I used to defend mongodb so much when I didn't know sql

now I see how much it sucks

[humble gyro]

zzz

[coral trellis]

Mongo is good for super small projects

[humble gyro]

mongo is decent for writing lots

[acoustic bough]

even at a scale of just 4,000 discord servers it's just painful

[humble gyro]

but there's a lot of better alternatives

[coral trellis]

I really wonder how you scale with it

[humble gyro]

mongo has a specific usecase

[acoustic bough]

you just dont

[humble gyro]

all of github audit logs are on mongodb

you just write a LOT to it

and read sometimes

[acoustic bough]

a timeseries database is prob still better, dunno

[humble gyro]

not all data is timeseries tho

like audit logs

[acoustic bough]

arent auditlogs kinda?

[humble gyro]

wdym

timeseries are good at aggregating

[acoustic bough]

audit logs are events that happen sequentially over time, and are usually sorted after time, so it would be practical to use timescale db or smth

[humble gyro]

yeah, but one big issue is that time-series dbs aren't really built for consistency

they are built for performance ingesting tons of data

which is great for many events, but not if you want to guarantee your data being sent

[acoustic bough]

fair I guess

[delicate zephyr]

it also depends on the time-series db

mongo has a time-series option

[lunar junco]

Hey why i cannot sumbit discord bot on topgg?

When i click sumbit it just reloads the button

😃

[earnest phoenix]

wrong channel --> #support

[bitter granite]

Bro did u use a magnet

[wraith ember]

How fix the application did not respond

[deft wolf]

Respond to the interaction

[pearl trail]

@oak cliff morning

[oak cliff]

[frail cypress]

@humble gyro any updates on the query issue?

[humble gyro]

I've slowly been implementing it

has it not worked for you yet?

[frail cypress]

ohh yeah

just now it started working

thank you

[humble gyro]

good to hear

@severe pike for u too

[severe pike]

amazing thanks, seems to be working now 🙂

[radiant kraken]

me when this works

but this doesn't

am i missing something?

[acoustic bough]

doesn't seem like it

spelling also looks fine

[radiant kraken]

YEAH 😭

the top.gg api must hate people who use node.js frfr

[neon leaf]

0 is not truthy

it may be omitted (for some stupid js reason)

[radiant kraken]

oh 😭

what the fuck

[bitter granite]

Topgg mods hates me

[acoustic bough]

if user_agent.includes('node'): error()

[radiant kraken]

then how do i JSON.stringify() it without truthy?

[neon leaf]

actually try with the same number as in py

i think its on topggs side

not urs

[acoustic bough]

possible that topgg thinks it doesn't exist since 0 is falsy

might just me like !server_count || !shards

[radiant kraken]

maybe

[acoustic bough]

which would error if both are undefined, or even just 0

[sharp geyser]

I hate that I see js, py, bash, c++ and c all in one screenshot

[radiant kraken]

batch

not bash

[sharp geyser]

[radiant kraken]

L

[sharp geyser]

close enough

[neon leaf]

even worse

[sharp geyser]

batch is just bash

lets be real

[radiant kraken]

batch is good

[neon leaf]

dont disrespect bash like that

[radiant kraken]

BATCH IS GOOD

[acoustic bough]

batch is ass

[sharp geyser]

batch is the unwanted cousin of bash

[radiant kraken]

batch is literally my first ever language

[neon leaf]

batch is painful

[neon leaf]

same

doesnt stop it from being bad

[acoustic bough]

sadly same...

[sharp geyser]

i'd rather learn VBScript than batch

[neon leaf]

nah

[radiant kraken]

VBScript is also one of my first languages

[neon leaf]

gotta disagree on that one

[radiant kraken]

shit start ik

[sharp geyser]

I started out with python

then moved to js

[radiant kraken]

same

[sharp geyser]

then to C#, to rust, to c++ then finally back to C#

💀

[neon leaf]

batch -> js -> ts -> rust

[sharp geyser]

To truly outline my journey

[acoustic bough]

batch -> Lua -> js -> TS -> go

[neon leaf]

some other stuff too but thats for main projects I did for fun

not for money

[sharp geyser]

py -> js/ts -> c# -> go -> elixir -> rust -> c++ -> c#

[radiant kraken]

batch -> vbs -> js (vanilla) -> py -> js (node.js) -> typescript -> c# -> c++ -> c -> rust (briefly, then gave up) -> go -> v -> rust

that's mine

[neon leaf]

i did nodejs before any frontend js

[sharp geyser]

I now only use c#

Theres no need for another language

C# has it all

[neon leaf]

I think im going rust + ts for life

[sharp geyser]

I will use rust if I need to use low level programming on like embedded software

but C# is the goat

[neon leaf]

ye all of ther c# I saw was pretty decent

rust is cool because of macros

1000 lines less for the same performance 🔥

especially useful for parsing

#[derive(Deserialize, Clone)]
struct Reroutes {
    name: Option<String>,
    reroutes: HashMap<String, (String, u16)>,
}

fn main() {
    let file = File::open("reroutes.json").unwrap();
    let data: Arc<Reroutes> = Arc::new(serde_json::from_reader(file).unwrap());

    let listener = TcpListener::bind(("0.0.0.0", PORT)).unwrap();
    println!("Server started on 0.0.0.0:{} (v{})", PORT, VERSION);
    println!("{} reroutes active", data.reroutes.len());
// ...```

[pearl trail]

😎

[small tangle]

is it just me or does using unwrap feel dirty

not like there are many other options

[wheat mesa]

unwrap is meant for when you know it's safe to unwrap

Otherwise you should be handling it

[sharp geyser]

I unwrap everything

Cause idc if its safe or not

[indigo holly]

Any free chatbot to add on my bot?

[solemn latch]

Honestly, there is so many cheap ones right now you're better off paying the few dollars a month itll likely cost you.

[scenic kelp]

he's enlightened

also C# has the coolest pattern matching syntax i've seen in any language yet

being able to do like age is >= 18 and < 25 is so cool

[sharp geyser]

Yup

Short hand

Love it

[prime cliff]

Isnt that just standard in any other language hu?

[humble gyro]

no

c# 🔛 🔝

[prime cliff]

JS dosen't have >==/>= or something?

[humble gyro]

they mean the pattern matching

if (user is { Age: >18 and <25 }) {
   ...
}

idk if that is valid syntax

but its pretty sweet

[prime cliff]

Hu but that's just the same as user.Age >=

I've never actually seen anyone do it like that in C# before ik you can do object is User user but not like that

[prime cliff]

Btw don't bind on 0.0.0.0 by default that opens up your program to the internet use 127.0.0.1 (localhost) instead xD

[neon leaf]

well

the difference is this is expected to be public

[prime cliff]

Ok good

[acoustic bough]

i always bind to 0.0.0.0 because it runs in a container anyway, who cares

[neon leaf]

that too

[acoustic bough]

• I host at home wo/ portforwarding, who wants to access it?? my mum gonna hack me?

[pearl trail]

best practice™️

[humble gyro]

it's not, its a value between 19 and 24

[prime cliff]

I understand that but the fact that you can do is that way is too weird even for a C# nerd xD

[humble gyro]

fair lol

it's a .net 8 thing i think

so fairly new

its cool because you can also do stuff like

user is { Username: "veld", Verified: true }

[queen needle]

So it's like querying?

[humble gyro]

it is but only for in-memory data

[prime cliff]

Not querying more of a if statement structure if (user.Username == "veld" && user.Verified == true)

[humble gyro]

it compiles to effectively this in js:

"username" in user && user.username === "veld" && "verified" in user && user.verified === true

[scenic kelp]

nah it's been in C# since like C#9 or so

they've changed the syntax a bit in the successive versions but

[humble gyro]

can they do that with async please

[scenic kelp]

like you used to not be able to do x is not null you had to do !(x is object)

[humble gyro]

oh fuck i remember that

soon C# is gonna be sql#

[scenic kelp]

x is {}

[humble gyro]

var approvedBots is select from bots where approved is true;

[scenic kelp]

can't you already do that with linq queries

[humble gyro]

only one keyword here is invalid

[scenic kelp]

ig not with full pattern matching

[humble gyro]

assigns != is

[scenic kelp]

have you ever seen someone use linq query syntax

like i'm very conscious it's a thing but i much prefer using lambda syntax and im pretty sure every other human on earth is the same way

c# needs two features and i will marry it and those are discriminated unions and blocks in switch statements a la java

[humble gyro]

with unions you mean like AnyOf<MyType, MyOtherType, Etc>?

[scenic kelp]

yeah like F# unions or rust enums

type Option<'a> =
| Some of 'a
| None

then we'll have

ParseInt
TryParseInt
TryParseIntOption

can't wait

[humble gyro]

oh hell yea

please give me this

[scenic kelp]

there's a proposal for it that also features like

A | B union types

which is more than i wanted originally but pLPEASE

c# and typescript are actually just going to be merged atp

[wheat mesa]

Types are cool but I don’t want it to get to a point where I have to be a rocket scientist to understand what type to pass into something

Typescript has fallen victim to that in some places

[delicate zephyr]

you can just use overloads for this most of the time

people just forget they exist

[scenic kelp]

yeah typescript types can be pretty disgusting but i don't think union types are terrible as a whole

and it should be possible to refactor the really evil types a bit

[humble gyro]

this is what ive been saying for a year now

please merge it tho

imagine the possibilities

[scenic kelp]

they're actually gonna pull a crazy twist and merge F# with TS

fable is already a thing that compiles F# to JS

[elfin fractal]

https://github.com/supermacro/neverthrow

[scenic kelp]

F# supremacy

[scenic kelp]

not C# though

there's not really an easy way to do unions in C# without a preprocessor

[humble gyro]

type Result<TData, TError> = { ok: true, data: TData } | { ok: false, error: TError } is all u need

[nocturne dagger]

I am trying to fetch members who have a specific role

const role1 = await guild.roles.fetch(relevantRoleIds[0]);

console.log(role1?.members);

But it seems like no one is cached, is there a better way to go about doing this?

I have tried just fetching members and filtering it

guild.members.fetch()

but this ends up timing out, the server has about 25k members

[delicate zephyr]

Have you got the members intent?

I'm assuming discord.js which relys on cache for fetching members roles

[prime cliff]

Yea you need members intent to fetch the entire list

[nocturne dagger]

Yea figured it out thanks! Member intent was enabled but not in constructor.

[unkempt aspen]

🤔

[sharp geyser]

That’s an analytics endpoint iirc

[deft wolf]

Yea, it's been down for as long as I can remember and has no effect on whether the site works or not

[unkempt aspen]

I was going to send my bot but it didn't work and I found out it was calling this api

[acoustic bough]

since it's just analytics it doesn't matter

[deft wolf]

You probably missed some required field like prefix

[acoustic bough]

iirc it's an api.top.gg/graphql post that's relevant

[unkempt aspen]

I pressed submit registration but there was no notification message so I don't know if it was sent or not. But now I know that the bot is checking it.

[acoustic bough]

does smb have experience with discord bots in bun (performance wise compared to nodejs)

[neon leaf]

slower

[acoustic bough]

bleh rly

idk how to even properly benchmark it

in terms of like mps/ips

or average response time to messages/interactions (ignoring network ping/latency)

[lament rock]

There are some libs which support both node and bun. I'd say just setup a bot and try to run it in the same code base and yeah

[lament rock]

But really if you're trying to look to bun for more performance, I think you're looking in the wrong places.

I'll give an example from my own bot: A lot of libs depend on the ws package from the npm registry for gateway connection and personal experience with a bot in over 10k servers has shown that just gateway message processing can really eat cpu time. At the time, it was all on one cluster. However, Tim provided some great code for a stupid fast WebSocket impl and that gave a significant performance boost. This was before interactions were a thing. Just processing messages was baaaad. It all depends on your use case.

Really, the first place you should start is your dependency tree. Vet the code you install and run. Replace deps with more lightweight ones if you can and maybe also look into taking a different approach into how you build your bot. An example is to move away from libs that try to do it all like Discord.js and instead look into more modular libraries that can get you great mileage without having a lot of cpu and memory overhead from just lib internals. The raw Discord API really is not that bad. I'd plug the libs I maintain but not trying to get bonked

When I made the switch away from Discord.js, I was able to manage what data I really needed to cache in memory and the performance shifted from being the fault of libs to my own code and that helped me realize what patterns I was doing was bad for performance

[earnest phoenix]

there is some wrapper made for bun, don't remember the name though

[acoustic bough]

yeah I know that I could use the discord api without djs, (or at least without the wrapper and just use djs/ws and djs/rest) but that would require me to rewrite the entire bot which is not my goal. I am pretty selective with the packages I install. From some projects bun is "just" faster, and would like to know if this also is the case if it comes to discord bots.

I currently run 12 clusters with 12,000 guilds and use 1.2-1.5gb ram (though I rn have about 350k messages in cache which I have to kill, that would fix most of the memory usage)

I just would like to get "free" extra performace by "just" switching runtimes

[earnest phoenix]

well you can tias

[acoustic bough]

I still would work with djs - at least the forseeable future. if I would decide to drop djs, I probably also gonna drop js as a whole

[neon leaf]

getting closer

real    0m0.169s
user    0m0.168s
sys     0m0.001s
root@remotedev:~/projects/0x7d8/bad-lang-2# time node testing/fib.js 

real    0m0.080s
user    0m0.074s
sys     0m0.008s
root@remotedev:~/projects/0x7d8/bad-lang-2# time python3 testing/fib.py 

real    0m2.858s
user    0m2.813s
sys     0m0.005s
root@remotedev:~/projects/0x7d8/bad-lang-2# ```

[neon leaf]

lets go

real    0m0.065s
user    0m0.061s
sys     0m0.004s
root@remotedev:~/projects/0x7d8/bad-lang-2# time node testing/fib.js 

real    0m0.086s
user    0m0.088s
sys     0m0.000s
root@remotedev:~/projects/0x7d8/bad-lang-2# time python3 testing/fib.py 

real    0m2.765s
user    0m2.756s
sys     0m0.009s
root@remotedev:~/projects/0x7d8/bad-lang-2# ```

interpreted language faster than js

@quartz kindle time to switch

[quartz kindle]

the real surprise here is how it python so slow

[neon leaf]

well

do you want to see the code

    let a = 0
    let b = 1
    let temp = 0

    if (#eq(n, 0)) {
        return a
    }

    let i = 2
    loop {
        if (#gt(i, n)) {
            break
        }

        temp = a
        a = b
        b += temp

        i++
    }

    return b
}

fn fib_until(until) {
    let results = []

    let i = 1
    loop {
        if (#gt(i, until)) {
            break
        }

        array#push(results, fib(i))

        i++
    }

    return results
}

let il = 0
let it = 100000
loop {
    if (#gt(il, it)) {
        break
    }

    fib_until(55)

    il++
}```

for python its ```py
def fib(n):
a = 0
b = 1
temp = 0

if n == 0:
    return a

for i in range(2, n + 1):
    temp = a
    a = b
    b = temp + b

return b

def fib_until(until):
results = []

for i in range(1, until + 1):
    results.append(fib(i))

return results

Run the benchmark loop

il = 0
it = 100000
while il <= it:
fib_until(55)
il += 1```

[quartz kindle]

how about using an internal timer instead?

[neon leaf]

sure wait

bruh my lang only has unix time

ill need to add ms precision time rq

67
root@remotedev:~/projects/0x7d8/bad-lang-2# node testing/fib.js 
66.869545
root@remotedev:~/projects/0x7d8/bad-lang-2# python3 testing/fib.py 
2892.989013671875
root@remotedev:~/projects/0x7d8/bad-lang-2# ```

[quartz kindle]

nice

[quartz kindle]

can you also show the js code?

[neon leaf]

const start = performance.now();

function fib(n) {
    let a = 0;
    let b = 1;
    let temp = 0;

    if (n === 0) {
            return a;
    }

    for (let i = 2; i <= n; i++) {
            temp = a;
            a = b;
            b = temp + b;
    }

    return b;
}

function fibUntil(until) {
    const results = [];

    for (let i = 1; i <= until; i++) {
            results.push(fib(i));
    }

    return results;
}

// Run the benchmark loop
let il = 0;
const it = 100000;
while (il <= it) {
    fibUntil(55);
    il++;
}

console.log(performance.now() - start);```

[quartz kindle]

looks good

i can think of a few things to change but it wont show any meaningful difference

[half scroll]

want to share other ways of handling mem/compute for python

import timeit
import numpy as np
from functools import lru_cache

# Define Fibonacci Implementations
def fib_recursive(n):
    if n <= 1:
        return n
    return fib_recursive(n - 1) + fib_recursive(n - 2)

def fib_iterative(n):
    a, b = 0, 1
    for _ in range(n):
        a, b = b, a + b
    return a

@lru_cache(None)
def fib_memoized(n):
    if n <= 1:
        return n
    return fib_memoized(n - 1) + fib_memoized(n - 2)

def fib_numpy(n):
    F = np.matrix([[1, 1], [1, 0]], dtype=object)
    return (F**(n-1))[0, 0] if n > 0 else 0

# Set test value for n
n = 55

# Benchmark each implementation
results = {
    "Iterative": timeit.timeit(lambda: fib_iterative(n), number=10000),
    "Memoized": timeit.timeit(lambda: fib_memoized(n), number=10000),
    "NumPy": timeit.timeit(lambda: fib_numpy(n), number=10000)
}

# Recursive is too slow for n = 55 at high repetitions, so we run it separately for fewer iterations and n = 30
results["Recursive"] = timeit.timeit(lambda: fib_recursive(30), number=100)  # Limited to 100 runs

print(results)

results would reflect what you saw as realtime in sec

$ python3 performance.py
{'Iterative': 0.008456800016574562, 'Memoized': 0.00037169991992414, 'NumPy': 0.09832059999462217, 'Recursive': 6.185182299930602}

[quartz kindle]

whats the difference between recursive and memoized?

[half scroll]

they both use recursion but the memorized add caching (avoiding/reducing redundant computations)

[quartz kindle]

does lru_cache do that automatically? whats the implementation for that decorator?

[half scroll]

you could set it to say
@lru_cache(maxsize=128, typed=False)
where max size = how many function calls to cache
typed if different arg types should be cached seperate

default is maxsize=128 anyway

[quartz kindle]

so basically it just takes the function arguments and caches the return value for those arguments

[half scroll]

exactly 🙂

[quartz kindle]

well thats kinda cheating

lmao

[half scroll]

😉 just offering different performance tests

[slender wagon]

i have a function that expands into multiple interactions. I need to gather that info from multiple menu's and handlers. How do you guys usually approach these as in what do you use to store the info through the different interactions

[quartz kindle]

you can store state directly inside them via uniqueid

alternatively you can have some global object/cache/db

[slender wagon]

i see, im gonna experiment a lil bit and come back if i somehow get stuck

thanks

[half scroll]

how is it cheating though?

[neon leaf]

not actually doing the calculations yknow

[half scroll]

and if you had a very repeated computation that would benefit from cache you wouldnt take it?

[neon leaf]

I would, but this is purely for testing performance

[sharp geyser]

@scenic kelp how’s go

[half scroll]

that's kind of my point though and offering different performance tests - to show why one functionality differs so much from another

[slender wagon]

@quartz kindle

.setCustomId(`start_debate_modal_${uniqueId}`)

looking good?

i will have to modify my listeners tho

thats fine i guess

[half scroll]

get what you're saying - you were after an unbiased comparison of langs, 😅 this is where I wanted to show...its kind of up to the person to know what abilities they really have

[lyric mountain]

we were comparing langs?

[neon leaf]

not really

[neon leaf]

just this

im assuming python parses the code multiple times, not quite sure why its so much slower though

[scenic kelp]

i'm not even writing it i was just in vscode for smth

i will say the go team had some very strange design decisions

why the fuck did they feel the need to make x + y automatically format to x+y kiss my ass go team

[sharp geyser]

Huh

Bruh

[wheat mesa]

It could be JIT warmup, you have to account for that

[neon leaf]

python doesnt jit

theres a seperate python runtime that does jit

[wheat mesa]

It does have some JIT

Pretty sure at least

Looks like python 3.13 has it, idk if it's used by default

[neon leaf]

pyroot@remotedev:~# python3 --version
Python 3.12.3

[wheat mesa]

Theree's probably some sort of runtime optimization causing it to be slow

I remember java used to have a problem with VM "warm up" times, I can't remember exactly why (I think it still does have this)

[lyric mountain]

it's just jit

java starts with compiled bytecode, then optimize it further as it goes

[wheat mesa]

I thought that even pre-jit java had warmup issues

[lyric mountain]

probably old issue

[versed coral]

What all is possible with the top.gg API ? I'm building a command to reward users who vote but other than that I'm not sure what I could do to utilize it more

[deft wolf]

I would use webhooks if I were you for this purpose

[versed coral]

I am using web hooks for the voting but was just trying to see other ideas

[deft wolf]

Later it's up to you what you want to do with this fact. Some bots give access to commands that are unavailable without voting, others give some currency in the bot or items and the biggest bots also make vote streaks thanks to which people have a reason to vote every day to get better rewards

[versed coral]

Vote streaks is a W. Low-key didn't think of that

Thank you

[lament rock]

Trust me, I understand that rewriting a bot is a very daunting task but also https://amanda.moe/to/stats

You can see where I ditched discord.js

Some of the data is missing and it stopped after 2023 because I got lazy to readd performance stats, but

Currently hovering at 150MB of ram usage

[white gyro]

Telling someone to rewrite their whole bot isn't great advice. Discord.js is the most well maintained library in js and they'd either use something less maintained that would get features slower (or potentially be abandoned) or essentially write their own lib.

It also shows a misunderstanding of the real issue. You can just customize the cache in discord.js - to reduce memory usage, you should disable the intents that your bot isn't using and limit/sweep the cache.

https://discordjs.guide/miscellaneous/cache-customization.html#limiting-caches

Other libs simply cache less or not at all (and you trade less ram usage for slower performance as you need to be fetch everything, and you may hit rate limits).

[lament rock]

I think you might not have a full understanding of the situation. Even with customization of cache, discord.js operates under a large amount of assumptions and some parts of the library simply cannot be fully bypassed. I was in the exact same situation and discord.js led me towards having a large amount of technical debt. I relied too heavily on its class system and sometimes, you can't even initialize the structures just to do rest actions without a good amount of data. Even if you literally limit the amount of data to what you'd theoretically need and use the built in Rest to make your own requests, you are basically not even using the lib at that point and are opting for internals. Just ditch the memory overhead of even importing the library. The imported string size of Discord.js already blows past what my bot's memory usage sits at during peak hours. Discord.js' approach to utilizing OOP which is what JS is known for shoots itself in the foot as you can't fully realize what you could achieve.

Your usage of the phrase "slower for performance" just because you "need to fetch everything" is such a bad take as the API actually gives you a LOT of data you'd need depending on what your use case is. You completely miss the point of trying to guide users towards making better choices just because they'd prefer to just swap runtimes and hope that magically solves all their problems. You're enabling bad choices. I would think that they would have already customized their caches which is why they'd be looking towards another runtime to fix their problems. Switching runtimes isn't just such a simple decision to make. 1.2-1.5GB of memory usage is not good by any stretch.

I also outlined in an earlier message that discord.js makes use of libs which are actually detrimental to performance. @discordjs/ws depends on ws and I have done my own tests. I've seen the before and after switching to a more lightweight ws implementation, switching to an entirely different gateway lib entirely. The CPU performance gains are nothing to scoff at.

"you may hit rate limits"
Libraries implement guards to prevent this. The only limits you'd have to worry about are the 50 requests per second globally. Even then, just cache what you actually need either in memory or in a database/object storage like Redis.

And to drive home the fact that I was hell bent on trying to maintain some semblance to Discord.js even when I was trying to migrate away from it:
https://github.com/AmandaDiscord/NEKO I wrote this before I fully gave up on Discord.js (v12) . Only later was cache customization a thing, but a global presence cache for users is still yet to be seen.
https://github.com/AmandaDiscord/ThunderStorm When I started maintaining some raw API packages, I wrote this instead of rewriting my bot entirely just so that I could avoid a total rewrite which was a waste of my time, a huge mental strain which I realized the hard way, and I ultimately stopped working on it

[white gyro]

You switched in v12 before both customizing the caching and intents were a thing so how would you know that it doesn't fix their problem? Regardless, telling someone to rewrite their entire project rather than simply use the cache controls in the library first, or if they still use message commands to switch to an entirely interactions-based bot and disable the messages intent, is a pretty wild premature "optimization" (in reality it'll probably just be a huge time sink more than anything, since all the other libraries do is less/no cache).

[prime cliff]

Both of you are right and discord.js is known to have a some memory performance issues especially at high number of servers but not really feasible to rewrite a bot if it's only in a few servers.

[lament rock]

The bot is in over 10k

[prime cliff]

How much memory you using?

[lament rock]

I would think that they would have already customized their caches which is why they'd be looking towards another runtime to fix their problems

[lament rock]

Currently sitting at 150MB which is mostly occupied by caching images in memory

[white gyro]

Why would you assume something they never mentioned

[lament rock]

Why else would they want to make such a big decision to jump to a totally different runtime

You'd think they'd exhaust their options

[prime cliff]

Wait only 150mb that's actually impressive

[lament rock]

https://amanda.moe/to/stats

Old data, but the amount of servers never mattered

My bot is rather stateless

[prime cliff]

Yea figures with that amount of memory but a lot better still

[white gyro]

The fact that they wanted to switch runtimes suggests that they're a beginner who doesn’t understand what’s causing the issue, as that isn't very relevant as the memory usage is due to how much data they’re caching. You should still get confirmation that they tried it before assuming.

You just seem like a disgruntled user of discord.js v12 but most of the issues you had were fixed with the new caching controls in my opinion. If you still disagree, you could do a more updated benchmark on your memory usage with unused intents removed and a limited cache before jumping to "ditch discord.js".

[lament rock]

I highly doubt that doing the tests would be worth my time as the imported string size blowing my peak memory usage is still an issue

[latent marten]

mates, maybe there is a react enjoyer here. im buildung a nextjs app with routes and somehow while prod building i get the following issue

TypeError: (intermediate value).default is not a function```

how ever, the error comes from chunks and the following lines say `at r.metadata.icon`on files like .next\server\chunks\856.js
i haven't set any metadata at all except for title within the <head> in root. i can't find any relations while searching for keywords, it doesn't matter if /upload, /users or settings, i couldn't find a pattern yet and (lol) copilot isnt that helpful. also it's my first project. my research told me i need to handle 2 files for each route: layout and page and nextjs would ask for metadata anyways. but i wonder why dev build works without errors or db issues. i can't even tell what this error means, lol
would appreciate some help :)

[acoustic bough]

I'm not just trying to switch runtimes because it's the holy grail that will fix ally issues - it won't. My biggest issue isn't even memory, even with 1.5gb ram usage (I have a fix for that by limiting cache sizes to 0 and using keep over limit to only store what I need). My point rather is CPU utilization/CPU usage. things like parsing hundreds to thousands of gateway message takes time and resources, which another runtime may be better at. Bun is generally perceived to be faster, and I only really wanted to know if this would also be the case with djs (as bun may be slower than nodejs under certain conditions).

it's just not feasible to rewrite the entire bot with another library. If I was about to do this, I would drop JavaScript entirely for Golang or C++

[lament rock]

As things stand with discord.js, I don't believe switching runtimes would have any meaningful impact like you think it would. At best case, you should have your gateway clustered across multiple machines and only ingest what messages you actually need at your workers. And ideally replace the gateway backing. the ws package is pretty bad for CPU performance at scale as I mentioned. I don't know how feasible it is to do this in @discordjs/ws, but Tim helped me write a slimmed version of a WebSocket and also an etf reader/writer

[white gyro]

The difference of bun is really negligible at best it does have faster package install but you can just use pnpm. But you can simply switch to an interactions-only bot then you will only process the interactions you will receive and not the "hundreds of thousands of gateway messages"

[acoustic bough]

I use pnpm - I need the gateway (at least until http events release)

[lament rock]

What they mean is that If you process commands through messages, you shouldn't as MESSAGE_CREATE events can easily saturate the CPU for processing time. If you have it so that you use / commands for commands, you save up so much processing. You can still use the gateway for other events

And as a word of advice, dont use the gateway to receive interactions

[delicate zephyr]

the main issue with d.js is it's built to be a tool for developers to learn basically

they've built it in a way to allow not just experience devs to use it but everyone

after about 8-9k servers depending on your usage, can use way more memory than any other lib even with cache customizations

at the end of the day it falls on your experience and your willingness to learn honestly

[acoustic bough]

my current solution

should kick out at least 300k messages out of my cache

[quartz kindle]

why maxSize 500

[acoustic bough]

I fear that djs breaks if I dare to call
reply on a message that's not cached anymore

wouldn't at all surprise me

[quartz kindle]

it wont break because of that

messages is something you can safely set to 0

[acoustic bough]

will test it anyway before pushing to prod

[quartz kindle]

as long as you're not replying to a message from a delete event

[acoustic bough]

oh yeah no lol

even with 500 it's way better than my current 34k per cluster lmao

[quartz kindle]

isnt the default 200 tho?

the limit is per-manager, not global

[acoustic bough]

no idea

[quartz kindle]

each channel has its own manager

[sharp geyser]

ima manage u

[acoustic bough]

oh wait is message manager per channel?

[quartz kindle]

yes

[acoustic bough]

oh

ok

makes sense

[harsh nova]

Making my own "library" for my webhook bot and can strongly recommend doing that if you can. It's fun

[sharp geyser]

They made it per channel?

[quartz kindle]

it's always been per channel

[sharp geyser]

I never bothered looking into it

but thats cool

[quartz kindle]

you set the limits for a single MessageManager object, and every single Channel has its own instance of MessageManager

[acoustic bough]

there is also guild message and dm message manager but I think I can ignore it if I set a message manager

[quartz kindle]

they all extend MessageManager

[acoustic bough]

but ok will def set a lower max then

but the default 200 explains why a cluster is always at the 33k messages for me

[pearl trail]

thanks for the idea

[harsh nova]

Discord API docs are sometimes a bit strange but with enough trial and error it works

Nonetheless a really fun project

[pearl trail]

haha yeah, better than having no project

[sharp geyser]

The new docs I feel are terrible

They used to be a lot easier for newbies to follow but now its jumbled with a bunch more jargon than needed.

[harsh nova]

Wish everyone had the quality of docs that Microsoft does

[sharp geyser]

I hate ms docs as well tbh

😭

Sure they are good

[harsh nova]

Oh really? Always been good for me

[sharp geyser]

but I swear to learn one concept you have to follow 3 hyperlinks

only to be sent back to the beginning

[harsh nova]

Kinda the opposite of discord. They tell you that you need to send something to some certain route but don't link what exactly you need to send so you have to dig for it

[sharp geyser]

yup

[neon leaf]

shits cursed but finally scopes work properly https://github.com/0x7d8/bad-lang-2/commit/37a44bddd759dc67ca5f7e8ddc427824ca352d61

[frosty gale]

Learn More buttons are even better

they send you to some random article filled with a dump of irrelevant information that you need to scan through to find what you need

[sharp geyser]

Yup

[latent marten]

fixed by downgrading to node 14 and react 18
next/metadata doesnt work work with newest versions

[earnest phoenix]

isn't node 14 EOL since a few years?

it's like 20 something by now

[latent marten]

nextjs, my fault :>

[earnest phoenix]

ah yeah v15 had some changes i'm not a fan of

[wheat mesa]

Aaaaaaand support scam

@zinc fable hi

[zinc fable]

tyty

[latent marten]

tbf, this is my very first react project and wasnt able to build prod at all. it was weird since dev worked without any issues. after a whole day i figured i have to use standardexports

(still not working, but new errors at least :>)

[surreal sage]

i love seeing

nameservers propogate to my ns

but then they went back to the registrar ns

[sharp geyser]

So something I still have not been able to wrap my head around is keeping state that the user is logged in on the frontend.

What I mean by that is, auth is handled on the backend in my aspnet api using microsoft's Identity and OpenIdConnect libraries, the frontend makes a request to an endpoint that calls the signin function to these libraries that redirects you to the oidc auth page of my auth server then back to the frontend on successful auth. The problem is that a cookie is saved on successful auth with all the information about the user, but its a signed cookie and encrypted so I can't use it on the frontend I would have to request the user info from the backend.

The problem I see with that is stale data, the data i'd save to state from the userinfo endpoint could become stale in the event they update something about their user, my idea was to call a mutation whenever something is updated (by mutation I mean a function that updates the state locally that fetches the user info endpoint to update this "stale" state).

I also think I may be overthinking this and there is a better solution to keep track of this state on the frontend, because all I really need to do is be able to keep track of the user info so I can use it when querying for things from the backend.

[prime cliff]

You realize you can just implement controller routes to return user info that you can use with javascript fetch too xD also what front-end are you using?

[sharp geyser]

Yes I know, thats what I was referring to when saying "hitting a userinfo endpoint"

I am using react

I plan on using a state library like zustand or something

[prime cliff]

There should be state stuff in react that you could look for it's a very old framework

[sharp geyser]

Yea but its not as good as an external lib

zustand gives me more features to work with that'd be more beneficial

[prime cliff]

https://react.dev/reference/react/useState

[sharp geyser]

yes I know of it

but thats component specific and its harder to saturate that amongst numerous components

so a state library is better since it makes it easier to access state anywhere

[prime cliff]

This one seems to be more global https://react.dev/reference/react/useContext

[sharp geyser]

Right but this still doesnt solve what I perceive is the root problem

Keeping the userinfo up to date on the frontend.

[prime cliff]

How important is it to keep the userinfo?
You could just set a background timer every 15s to update or just refetch the userinfo when doing crucial tasks.

Either that or you switch your front-end to a server side rendered app

[sharp geyser]

I guess its not super important

The only real thing I would need is their user ID which I know will never change, everything else is just info like their name, email and roles they have

I can fetch it again in cases I need to check if they have permission to do something in case they did get a new role added to em

I know their name and email will almost never change in my use case

[prime cliff]

Well you would need to check permissions server-side anyway so that's not a good point xD

[sharp geyser]

Well yea

honestly I dont think I will ever need to refetch user info

Thinking more about it

None of it is important to the frontend except their ID

Should I keep track of their access token in state as well or nah?

[prime cliff]

That's already in the cookie?

[sharp geyser]

hm fair

[prime cliff]

Or jwt whatever way you do it

[sharp geyser]

I was only asking because the frontend would need a way to authenticate it but I guess that wont matter since the Identity library does that already iirc

[prime cliff]

Yea oauth stuff should just redirect the front-end to auth on the backend

[sharp geyser]

👍

[prime cliff]

Best case is you shouldn't really trust the front-end for user/security stuff because it can be manipulated or browser console changed

[sharp geyser]

Yeah

Well thank you builderb

[prime cliff]

Np i'm not the best with web either but i'm doing a server-side blazor app

[sharp geyser]

Yea once it comes to front end I’m a fish out of water

[prime cliff]

You could take some ideas from this and implement 2FA stuff 😄

[sharp geyser]

Luckily my auth server handles all that

I just need to make api calls and it does the rest

[prime cliff]

Ah you doing a self hosted auth like keycloak/authentik or just standard oauth stuff with google/discord

[sharp geyser]

I’m using a self hosted auth server

It’s called zitadel

It handles different forms of auth between multiple providers or can be used as a local auth provider with email and password

Or a combination of both

It’s rather neat

[neon leaf]

@quartz kindle I think its now possible to write a discord bot in my language

fn handleconn(connection) {
    let data = tcp#readstr(connection)
    let head = array#get(string#split(data, "\n"), 0)
    let parts = string#split(head, " ")

    array#pop(parts)

    let method = array#get(parts, 0)
    let path = ""

    let i = 1
    loop {
        let part = array#get(parts, i)

        if not (part) {
            break
        }

        path = string#concat(path, part)
        i++
    }

    io#println(method)
    io#println(path)

    tcp#write(connection, "HTTP/1.0 200 OK\n")
    tcp#write(connection, "sus: ok\n")
    tcp#write(connection, "\r\n")
    tcp#write(connection, "Hi sir, from bad-lang\n")
}

loop {
    let connection = tcp#getconn(listener)

    thread#launch(handleconn, connection)
}```

tcp streams work now

[neon leaf]

some perf fixes later

[quartz kindle]

nice

why is the slowest 16 seconds lmao

[neon leaf]

not sure

but I need to do a lot more optimizing

[quartz kindle]

xD

[neon leaf]

variable lookups are still very expensive

[quartz kindle]

you're using rust right?

[neon leaf]

yes

[quartz kindle]

nice

[neon leaf]

mainly this is the issue

it runs on each lookup instead of each scope

[solemn latch]

Wow, it only took forever.
https://nextjs.org/blog/next-15-2#nodejs-middleware-experimental

15.2 was a really good update

the error page updates are ugly af though 👀

[queen needle]

what?!?! I love the new error pages

https://discord.mx/v/AAEK81pnfz.mp4

this is very funny however

[neon flicker]

Hello, I had asked the same question in 2024 and now I would like to ask it again in 2025, do you think even skilled programmers will lose their jobs due to the rise of AI technology?

[lyric mountain]

no

not even unskilled programmers will

[neon flicker]

Well, my opinion at this point is that it possibly can replace people who've chosen computer science just because they thought they would make much money for free and are not willing ti stay updated against any updates in the field/industry

[lyric mountain]

not really, AI has a major flaw when it comes to anything: it cannot create

besides, non-programmers rarely can explain exactly what they want, AI would have a hard time figuring out

[neon flicker]

You got such a good point, you still need to know how to program in order to provide the AI service decent prompts

[lyric mountain]

yep, there's even a comic for that

[neon flicker]

Aw lol

[wheat mesa]

AI can replace people the day that clients know how to describe exactly what they want

Aka never

Not to mention that AI inbreeding is going to fuck over the quality of LLMs soon enough

[solemn latch]

lmao

[wheat mesa]

Companies that are replacing juniors with AI will find themselves in deep shit 5 years down the road when they aren't creating senior developers that actually do a good chunk of work

[neon flicker]

The breeding of trying to produce them faster and trying to make them magically "better" besides?

[frosty gale]

for (let i = 0; i < messages.length; i++) {
  add_to_training_set(predict(messages[i]));
}

[wheat mesa]

LLMs are just predicting the next token in a sequence of tokens. Their quality is based almost entirely on the data they are trained on. Without good training data, you get poor output.

[lyric mountain]

basically they are chewing pre-chewed content

[wheat mesa]

When you have a bunch of these "AI developers" making shit and training the newer LLMs, it starts to hurt their quality because the AI developers make terrible quality code which gets fed to the training data of the newer LLMs

You reduce your "genetic diversity" so to speak, just like inbreeding irl

[lyric mountain]

AI is training on AI generated content

[frosty gale]

i want to see some more research and tests in this area, how much is the quality of LLMs impacted by generating & adding predictions back into the training set, also across different models

[queen needle]

so do you think in some time we'll see ai's having to be completely retrained on new data like they were once trained? to get rid of this

[frosty gale]

technically if you raise the temperature and adjust the token prediction algorithm you can come up with completely new sentences not seen in the training set

[wheat mesa]

I think that more people are going to have to work on knowing what code is good and what code is bad to feed to LLMs for them to produce better quality responses... aka software engineers...

[neon flicker]

Besides that, it sounds way too utopic to have AI handling everything

[queen needle]

so like approve and decline but for code feeding llms lmao

[lyric mountain]

yes, and you get the cordless phone effect

[wheat mesa]

Not to mention that these models are fundamentally flawed; they don't "think". Even the reasoning models aren't "thinking", there is just some clever tricks used to make it appear that way

[neon flicker]

Yes well, it is that most people don't know what a LLM is

[lyric mountain]

where they start making stuff up leading to incorrect or straight up unrelated answers

[frosty gale]

im not sure your use of terminology had the intended effect 😭

[lyric mountain]

huh

the hell google lmao

actually where tf did google take that from

[frosty gale]

maybe from the so called cordless phone effect you mentioned

[lyric mountain]

I meant the kid game where you gather everyone in a row and one side whisper a phrase to the next, propagating till the last player

[wheat mesa]

telephone game lmaoooo

[lyric mountain]

usually the message will be completely unrelated to the original

[frosty gale]

i think they usually call it chinese whispers

just "telephone" too apparently

[lyric mountain]

here we call it "telefone sem fio"

[humble gyro]

gamer as fuuuuuck

[solemn latch]

the middleware update might be useful to you too.

[deft wolf]

We call it "deaf telephone"

[limpid onyx]

finally node.js runtime in middleware 🙏

[solemn latch]

Helps with nextauth

[queen needle]

I'm confused on what that adds to next.js middleware

[acoustic bough]

you being able to use nextjs middleware on nodejs

and not just when using vercel.com

[slender wagon]

question, if i want to quickly check if a channel id matches another id. Would i rather use a json file to match it or a query on my postgres?

the question would be which one is faster

[acoustic bough]

would use some KV for smth like that

[frosty gale]

is there a date limit to how far you can read messages in a channel in a bot?

as in be able to get the messages for an entire channel

planning to export a channels messages which i use for notes offline

[acoustic bough]

you can get everything

100 messages at a time (per request)

[wheat mesa]

huh?

[lament rock]

What do you mean by this

[lyric mountain]

the rule is that ids within a specific domain cannot overlap, but they aren't unique globally

so 2 channels cant have the same id, nor 2 guilds, but those channels can share the same id as the guilds

[slender wagon]

i am sorry for the confusion

i have a forum in my server and this forum has posts. Now i have a bot that creates posts and it only grants access to certain people.
My question would be which way would be the best and fastest way to store an id and compare it with the channel id. That way i can delete messages of users that aren't allowed to post in that forum post

I'm doing all of this because discord doesn't give you a way to limit permissions on forum posts

[wise burrow]

Whatever method you have for storing the user IDs, your bot would have to run the sender's ID through the list, and if it's not on the list then it would just autodelete. You can't directly compare User IDs to Channel IDs.

[slender wagon]

https://tenor.com/view/arthur-morgan-arthur-death-arthur-morgan-dying-arthur-crawling-gif-985191194630978600

[lament rock]

I'd say we'd need to know how you determine if users are allowed to post. Is it a role? Some form of acknowledgement with your bot?

If it's something like a ticket system or something where it's your bot's mechanisms that decides who can use it, I'd have two tables in pg. One that describes necessary info on the forum post channel and then another table to describe who has access to it.

If you only need to store user ID and channel ID, you can have just 1 table and then create an index in that table off of channel IDs. No pkey.
For instance, the table could just have two columns which is the user ID and channel ID. Then you
SELECT user_id FROM posts WHERE channel_id = $1 and since the channel ID is the indexed column, it'll be stupid fast to return all of the users who have access to channel_id.

If you do need more info wherein you'd need 2 tables, your query might look more like
SELECT post_access.user_id FROM posts INNER JOIN post_access ON posts.channel_id = post_access.channel_id WHERE channel_id = $1
In this case, you should index channel_id on post_access and pkey channel_id on posts or however else you wanna pkey it

[sharp geyser]

  Assembly 'Mercatus.Database' with identity 'Mercatus.Database, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null' uses 'Microsoft.EntityFrameworkCore, Version=9.0.2.0, Culture=neutral, PublicKeyToken=adb9793829ddae60' which has a higher version than referenced assembly 'Microsoft.EntityFrameworkCore' with identity 'Microsoft.EntityFrameworkCore, Version=9.0.0.0, Culture=neutral, PublicKeyToken=adb9793829ddae60'

What the fuck does this mean. No where am I using EFCore 9.0.0.0

Unless its a dependancy issue where one of my deps is using it

Its what I thought it was

well thats fucking annoying since the latest package only uses 9.0.1

and all my other EFCore packages use 9.0.2

Downgrading everything to get it to work is too much of a hassle

[sharp geyser]

Fixed it, luckily only had to downgrade Design and Tools and it did all the other deps itself

[prime cliff]

Portainer vs my project, it actually looks pretty nice now

[lucid vessel]

I sent my bot for verification... but it didn't appear in the logs, and a few days have already passed... what should I do?

[ancient shale]

You've to be patience until bot gets verified

[lucid vessel]

I know, what made me curious is that last time there were logs, but this time there aren't.

[ancient shale]

sorry, am new to entire platform i just gone through docs of topgg about review and other stuff's.

[languid dragon]

Anyone know of any (working) embedded-able embed visualizers? preferably for TS/NPM
ping if reply

[sharp saddle]

I mean, I read several docs, and I didn't understand anything

[weary drum]

who can help me add bots to my discord server

[prime cliff]

bruh

[warm halo]

a

[quartz kindle]

b

[neon leaf]

c

[deft wolf]

d

[vocal knot]

e

[solemn latch]

f

[zinc fable]

z

[solemn latch]

-warn @zinc fable

[lucid vessel]

Is there any place that can guide me on integrating votes with my bot's system? (Such as rewards, for example.)

[solemn latch]

You'll want to use webhooks.
https://docs.top.gg/docs/Resources/webhooks

[lucid vessel]

Exactly what I was looking for, thank you!

[latent marten]

broskis. i am working at my very first react project and for days i wasnt able to build. i've tried literally everything and even bought 20$ cursor subscription. i've head problems with nextjs's metadata and always problems with the favicon.ico
claude 3.5 and 3.7 looped into a fix hole, until i realized bruh, i have 2 favicons.ico, renamed the wrong one and vuala
dunno if i wanna cry or hug the world lol

[frosty gale]

welcome to AI

ai is very clutch but you cant trust it to do anything significant for you without handholding

[frosty gale]

is the intel cpu hetzner has that bad to the point where the amd variant costs more with less ram lol

[quartz kindle]

its a xeon gold, but idk model

the amds are epyc 7002

[frosty gale]

im a bit torn between the x86 and arm variant, i heard the arm machines are much faster on hetzner

but some of the projects i want to have on there have libraries that may not work well with arm64

i'll try compile them for arm and see how that goes because ill get much more out of the server if i go with arm

[quartz kindle]

run some benchmarks

im on the epyc one rn

the CPX11

[frosty gale]

it looks good but im quite a bit pissed about half the ram

[quartz kindle]

afaik the other one used to be the same ram

did they increase it recently?

i remember 4gb ram being 7+ bucks not 4

[frosty gale]

this is all i see atm

or i can go balls to the walls with the more expensive AMD plan

yeah not a good start trying to compile one of my main rust projects for arm

probably because its trying to link my X86 libraries

[latent marten]

I mean, by given errors, no other react dev was able to fix it so these were my only options

[frosty gale]

@quartz kindle i think the threads i was looking at were when hetzner had ancient intel cpus

it looks like they are much better now

the amd one looks worse in this comparison(?)