development

[quartz kindle]

so async is not for cpu tasks, async is for network/disk tasks, stuff that is not cpu related and requires waiting for something to happen

the only correct way to make a cpu intensive task async, is to send it to another process/thread

99% of an async task is spent "waiting", therefore it has 0 cpu usage for 99% of the time

0.1ms to make some network call, then 50-100ms for the server to respond with some data

so first of all, what db do you use?

[surreal sage]

mongo 😓

[quartz kindle]

so you wanna make 15 mongo calls, one for each user? or can you make 1 call that returns all 15 users?

[surreal sage]

one call

.find without a filter returns all documents

[quartz kindle]

alright, so

the mongo call is already async, and its only 1 task, so no need for Promise.al

[surreal sage]

Promise.all(users.map((user) => user.asyncMethod()))

[quartz kindle]

ah, what does the user method do?

[surreal sage]

stuff 🤫

more db calls

[surreal sage]

will be caching this data for like a minute

[quartz kindle]

lmao ok

it doesnt sound like a cpu-intensive task then

or is anything in there that is atually cpu-intensive?

[surreal sage]

not that i can think of

it mainly gets a bunch of documents from a different collection with a filter
then grabs all the values of a specific key, and adds it together

[quartz kindle]

the you should be fine

[hidden gorge]

yes

[quartz kindle]

its gonna be waiting for 99% of the time

[hidden gorge]

i managed to make it so i can write and read to the json and it be updated during runtime

[quartz kindle]

you can have a million promises running at the same time, if all of them are waiting, then you have 0% cpu usage

[surreal sage]

im trying to get the last midnight with timezone offset in hours

for some reason the returned date is also on an offset of the host system's tz

first date is utc+2
second date is utc+12

my host system is utc+2
(currently 19:27, or 7pm)

[surreal sage]

i might just also be stupid

export default function getLastMidnight(timezoneOffset: number = 0): Date {
  const now = new Date();

  const timezoneOffsetMs = timezoneOffset * 60 * 60_000;

  const localTime = new Date(now.getTime() + timezoneOffsetMs);
  localTime.setHours(0, 0, 0, 0);

  const lastMidnight = new Date(localTime.getTime() - timezoneOffsetMs);

  return lastMidnight;
}

[past field]

@quartz kindle got a sec to take a look at another game?

I have a "confessions" game .. basically everyone has 3 minutes to submit a confession and the bot will display the submissions anonymously, players have to @ mention who they think wrote it

but sometimes the modal gives and error, only a hand full of times

here's the game https://pastes.dev/j4D252LWbN

and this is the error

[ConfessionsGame] Modal submission error: DiscordAPIError: Interaction has already been acknowledged.
    at RequestHandler.execute (C:\Users\Maurice\Desktop\Click War Bot\node_modules\discord.js\src\rest\RequestHandler.js:350:13)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async RequestHandler.push (C:\Users\Maurice\Desktop\Click War Bot\node_modules\discord.js\src\rest\RequestHandler.js:51:14)
    at async ModalSubmitInteraction.reply (C:\Users\Maurice\Desktop\Click War Bot\node_modules\discord.js\src\structures\interfaces\InteractionResponses.js:103:5)
    at async InteractionCollector.<anonymous> (C:\Users\Maurice\Desktop\Click War Bot\commands\confessionsgame.js:181:29) {
  method: 'post',
  path: '/interactions/1260283218536566785/aW50ZXJhY3Rpb246MTI2MDI4MzIxODUzNjU2Njc4NTpzYzFmcTUxc2p6UXA2ZnZsb2JYVHRyY2ZzYVRWTno4MVQ0TDBISXNaSHNDVkUxaTBIcWhKU2tUUUpsN0xubFlWZmN5b0RYc0pxbGZPNXcybnJXZVU0d05MRFNZNlVmWEMyUWRjZENZczZDMENxZVlpSDJLN05XSHpTa3ZQVUxMNA/callback',
  code: 40060,
  httpStatus: 400,
  requestData: { json: { type: 4, data: [Object] }, files: [] }
}

[quartz kindle]

so you want to get the Date object for the last midnight for a given timezone?

[surreal sage]

yes

[quartz kindle]

hmm i dont see anything wrong on a first glance, i'd need to look deeper

[frosty gale]

nodejs does though to an extent correct?

libuv after all

otherwise io operations would kill it

[frosty gale]

how long are they going to make the interaction tokens goddamn

thats enough to thwart any brute force for more than a couple of lifetimes of the universe

[quartz kindle]

const timezoneOffsetMs = timezoneOffset * 60 * 60_000 + now.getTimezoneOffset() * 60_000;

you have to correct for the host's own timezone (using Date with timezones is dumb and unintuitive)

[quartz kindle]

yeah, it has a thread pool for managing I/O ports and streams

[quartz kindle]

ye like what the heck, what even is an interaction token?

[hidden gorge]

now i see why my function wasnt working..

[surreal sage]

what is this

[hidden gorge]

thats the part i found

[quartz kindle]

ah yes, the famous await() function

[hidden gorge]

i forgot axios.get

[quartz kindle]

:^)

[hidden gorge]

to be honest i wrote that at 2 am

[quartz kindle]

it happens

[hidden gorge]

and now its failing this is greay

grewt*

great*

yes that makes sense...

[surreal sage]

@quartz kindle

does this look correct

i can't tell

[quartz kindle]

cant you use a lib like Luxon? thats a major pain in the ass to do in js

we still dont have the new Temporal api

[surreal sage]

magic..

[quartz kindle]

xD

[surreal sage]

okayyyy sooo

my mongo model has a startedAt and endedAt

I have 2 date objects, dayStart and dayEnd

I want to get all documents where at least one of the date objects (startedAt and endedAt) are between dayStart and dayEnd

my poopy brain is too dum dum to know what operations i have to do

wait i mightve done it

this no?

uhhh no

(there are documents for every hour on the 9th)

https://tenor.com/view/rescue-me-from-myself-save-me-i-need-help-help-me-i-need-saving-gif-14711644

[surreal sage]

actually I go tit

[past field]

only if you don’t mind!

it still functions like it’s supposed to, but when this error pops up, players get an error in the modal and it causes the bot to record the submission multiple times, so it’ll display during the guessing round multiple times

[surreal sage]

@quartz kindle So I have my array of database users

I want an array with it's items being

{
  "day": "promise result here",
  "week": "different promise result here"
}```
im not sure how to do this with promise.all though

how can i users.map correctly so that i can promise.all the result

is this right?

[quartz kindle]

yeah looks fine

but if getDayStats and getWeekStats are methods that belong to the user already, wouldnt it make more sense to call them directly? and store the result inside the user itself instead of an external object?

[surreal sage]

the methods depend on other data in the db that has to be fetched

different model

[quartz kindle]

should still be doable

for example

is this how User is defined? ```js
class HydratedUser {
async getDayStats() { ... }
async getWeekStats() { ... }
}

you could do this

class HydratedUser {
  async getUserStats() {
    if(!this.userStats) {
      this.userStats = {
        day: await this.getDayStats(),
        week: await this.getWeekStats()
      }
    }
    return this.userStats;
  }
  async getDayStats() { ... }
  async getWeekStats() { ... }
}

const result = Promise.all(users.map(user => user.getUserStats()))

[quartz kindle]

can you check if you have any duplicated logs? for example this one:

console.log(`${LOG_PREFIX} Confess button clicked by ${i.user.username}`);

[quartz kindle]

i dont know how but the only explanation i can come up with is that the await i.awaitModalSubmit is catching the wrong modal, or the modal is somehow duplicated

you can try generating a unique id for each modal

const unique = Date.now().toString(36) + Math.random().toString(36).slice(2);

.setCustomId(`confessionModal-${unique}`)

const filter = interaction => interaction.customId === `confessionModal-${unique}` && interaction.user.id === i.user.id;

[past field]

ok i’ll try this

[warm carbon]

what is the value for this? ( channel permissions, discord.py )

[real rose]

neutral. So it just takes the parent role's permission

[warm carbon]

oh ok
thx

[sharp geyser]

Whatever your highest role right?

Like if you had one role giving you the perm, but another taking it away and it was much higher, do you still have that permission

[real rose]

yep, the higher role will always take priority

[sharp geyser]

Figured, but I know discord perms are weird

💀

[real rose]

which is why we have a Muted role

above most other roles

with everything set to ❌

[warm carbon]

how do i set it to neutral tho ( discord.py )

[real rose]

because it will overwrite all other perms

[sharp geyser]

?

[real rose]

you dont touch the permission at all

neutral is default

[sharp geyser]

You dont touch them at all it is the default

[real rose]

so you say

[sharp geyser]

well

unless you touch them

then you. can just set it back

💀

[real rose]

then they are touched

[sharp geyser]

https://tenor.com/xYnK.gif

tickle tickle

[frosty gale]

just received the most disgusting leet code question for my tech interview at samsung

dont know if im just stupid or was it hard

[sharp geyser]

what was it

[frosty gale]

given an array of numbers of any length, and a sum value which is the largest number in the array, return whether any combination of numbers can add up to the sum value, in C

the largest number part was very simple but i couldnt figure out a good enough solution for the sum part

[sharp geyser]

you lost me at sum

[frosty gale]

absolute bs question

[sharp geyser]

So basically

just keep adding numbers until you reach that largest number?

That sounds like a very expensive algo

[frosty gale]

basically yes except the combination can be any number and any amount of numbers in that array

[sharp geyser]

How did you even solve it

[frosty gale]

i settled for a recursive solution but another one i thought was compute all possible combinations and see if any of them add up to the sum

the solution itself is simple but working it out is very tricky

a lot to think about

[sharp geyser]

Indeed

[frosty gale]

the recursive one i wrote is basically a step better from the all possible solutions

but not by much

[sharp geyser]

My first thought was just a loop of some sort adding up numbers until you reach the one you want

but uhm probably not the best

[frosty gale]

thats what i thought but it cant be that simple because not all numbers can be used to make a proper sum

you have to basically start from the end of the array and work your way up to the beginning trying different combinations along the way using recursion

[sharp geyser]

hm yea

[quartz kindle]

didnt we do something similar on here before?

with someone working on an ELO system

and matching teams by ELO

[sharp geyser]

yea

Ben

[quartz kindle]

i remember doing a brute force loop by testing every possible comboination

but i dont remember what the final solution ended up being

[frosty gale]

this is sort of what the solution for this was

bool isSubsetSum(int arr[], int arr_length, int sum) {
    if (sum == 0)
        return true;
    if (arr_length == 0)
        return false;

    if (arr[arr_length - 1] > sum)
        return isSubsetSum(arr, arr_length - 1, sum);

    return isSubsetSum(arr, arr_length - 1, sum) || isSubsetSum(arr, arr_length - 1, sum - arr[arr_length - 1]);
}

[quartz kindle]

is the array ordered?

[frosty gale]

any order

otherwise the solution can be a lot simpler

[sharp geyser]

That looks interesting to say the least

[quartz kindle]

ah so sum is already the highest value in the array

no need to find what the highest value is first

[frosty gale]

yeah but you need to know it beforehand here?

otherwise it has no idea which combinations to flat out throw away

[quartz kindle]

wait what exactly is n and sum lol im lost

[frosty gale]

yeah badly named ill change the names

but sum is the value that the combinations need to add up to

so the highest number in the array worked out beforehand

[quartz kindle]

so array is always guaranteed to contain at least one item that is equal to sum

[frosty gale]

there

[quartz kindle]

ah n is array length

of course C doesnt have a good way to measure arrays

[frosty gale]

so basically {1,5,7,4,9} = true because 5 + 4 = 9

[sharp geyser]

why would it

Who needed to know the length of an array in 70s

[quartz kindle]

i mean, if the highest value is contained inside the array, why is sum a separate variable?

[sharp geyser]

youd have to grab it each time?

its a recursive function grabbing it inside the function would be bad no?

[frosty gale]

so you can subtract it each time a combination is tried in each recursion

but in the final solution i had another parameter called max_value which made sure to ignore the biggest value which is the sum from being counted

[quartz kindle]

it just feels weird having a function that requires two parameters when the second parameter exists inside the first one anyway

[frosty gale]

theres not even booleans by the way

so i had to use chars with values 1 or 0

[quartz kindle]

thats the og boolean

[frosty gale]

you can include a header which has them but i couldnt

anyways ive had enough of this question

[quartz kindle]

ever heard about the therac-25?

[frosty gale]

stupid stupid stupid question

[quartz kindle]

i would honestly not be able to solve it, i would start overthinking and enter an infinite loop in my head

[frosty gale]

i think this type of thing youd have to come by beforehand by practicing leet code

in reality this type of problem wouldnt be an issue in the workplace since youd be able to look it up or use an existing solution

[quartz kindle]

or even just have a real use case for it to guide you

sometimes you cant solve something if you look at it purely from a mathematical/algorithmical way

but you can if you relate it to some real use case scenario, it gets much easier to visualize it

[frosty gale]

recruiters when you cant solve an unrealistic leetcode question: 😧😮😦😯😑🫤

[wheat mesa]

booleans weren’t standard in C until C99

And even then they were just a macro lol

[quartz kindle]

technically they still are, no?

[wheat mesa]

Maybe? Don’t know, I don’t use C very often

Yeah I think it’s a typedef

[sharp geyser]

0 and 1

[wheat mesa]

So basically just a define macro

[sharp geyser]

I dont think i've used the bool thing yet

I have only used ints as bools

[wheat mesa]

It’s in stdbool.h

[quartz kindle]

The header stdbool.h in the C Standard Library for the C programming language contains four macros for a Boolean data type. This header was introduced in C99.

The macros as defined in the ISO C standard are :

bool which expands to _Bool
true which expands to 1
false which expands to 0
__bool_true_false_are_defined which expands to 1

lmao

[sharp geyser]

wtf

[frosty gale]

yeah theyve added them in now but most compilers stick to older more standardised versions

other than that you can either use stdbool, define your own macros or just dont use them

gotta love emails trying to detect whether youve opened them or not

i set up a CSP policy in my email client that blocks image urls from loading unless i explicitly allow them to

[past field]

there were duplicated logs

getting rid of them

[quartz kindle]

i mean, its not about getting rid of them, its about if they are appearing more than once per user/modal, then something is wrong

[past field]

ah well yes there were a few that were duplicated

and i created the unique id’s

looks like it working smoothly so far, but then again i’m testing it by myself 😂

[quartz kindle]

nice

[quartz kindle]

nice try scammer

[past field]

finna dm u right now. can't pass up 50k in 72 hours

[sharp geyser]

Guys

That doesnt look like a kelly

[past field]

sure it does

i trust him

[quartz kindle]

its an r kelly

[sharp geyser]

How should I handle fetching information from routes that require you to be logged in to access? Like my service doesn't require you to be logged in to use it, but if you login you get extra features. Problem is, I am now realizing my api design is bad because I am having to make extra routes just for those who aren't logged in to fetch the same information.

My first thought was a public api key used by anyone whos not logged in, but at the same time that can be bad if people wanted to just make arbitrary requests. Also not to mention I have to find a way to protect other routes that DO need you to be logged in if I went this route.

[quartz kindle]

the standard way of doing it is to have an optional api key

no key = free access
key = special registered access

[sharp geyser]

hm

Im not sure how i'd make a middleware for that

[sharp geyser]

the way it works right now is if you access a protected route and you aren't logged in you can't access any of it. I am not sure how I will be able to pick and choose which routes are protected or not

especially if I have nested routes

like

/creators/settings -> logged in to access
/creators/:id -> doesnt need to be logged in to access

Also, if I do it that way, wont it mess up if they are logged in?

[scenic kelp]

i always thought this was really stupid then i thought about it and it actually makes a lot of sense

since technically a boolean is quite abstract, you can't really represent a single bit value in a CPU

[sharp geyser]

Like if they are logged in, and I still make a call to /creators/:id then it will deny it, because I am checking if they dont have one. I mean I guess I can check if they do or dont at the same time and allow both if the case is true but thats repetitive checks and I dont necessarily like that

[quartz kindle]

the concept of a REST api is to be stateless, independent of login status

[sharp geyser]

[quartz kindle]

by login you mean session cookies right?

if your endpoints are acessed by browser pages, ie a dashboard, then the endpoints should indeed check for session cookie

[sharp geyser]

yes by cookie

thats how my routes are protected rn

[quartz kindle]

but if the endpoints are accessible via arbitrary http request, ie node, etc, then there is no browser, therefore no session and no cookie

of course you can technically send cookies from a node http request, but thats pretty much a wonky design

i would do it like this:
public rest api accessible with or without key (no sessions)
private dashboard endpoints (with sessions) that make backend calls to the public rest api if needed

[sharp geyser]

well as it stands now I was just going to make it send over via an auth header

like a Bearer auth header

[quartz kindle]

sure

but the point is to have stuff like dashboard endpoints separate from the public api

because dashboard endpoints are not supposed to be called from outside a browser

[sharp geyser]

what exactly do you mean by public?

[quartz kindle]

public api = stuff you tell people they can make requests to from anywhere they want

private api = stuff that only your own website can access, only when the user visits via a properly logged in browser

[hidden gorge]

i actually did something like this

[quartz kindle]

example, the discord api and the discord app api

the discord api we all use is the public api

[sharp geyser]

well I odnt have a public api

[quartz kindle]

but the requests the actual discord app makes are different from trhe public api we use

[sharp geyser]

but if they know the routes they can make requests to it

idk how to make a truly private api

💀

[quartz kindle]

a private api is mostly just session cookie

optionally you can check for referral url and user agent, but those are not really secure

if you dont have a public api, then there is no such thing as api keys or auth keys

everything should be done with server-side cookies

even if the user is not logged in, you generate a session cookie when they visit your page

and use that

[hidden gorge]

keys

[quartz kindle]

basically you need 2 cookies

browser session and account session

when user visits your website, you generate a session cookie, which you will use to validate that your endpoints are being called from inside the browser

then if the user logs in, you generate an additional cookie for the account

the browser session cookie should be deleted when the browser is closed

the login cookie should be persistent (if you want to remember logins)

[hidden gorge]

like connect.sid?

[quartz kindle]

yeah i guess

[hidden gorge]

i only really know about connect.sid bc of express session

[sharp geyser]

I mean sure ig

[eternal osprey]

hey guys, i am kinda new to environmental varialbles, i only used them in a mathemtical context like:

e1, e2, e3 _| P union X -> e2 etc etc

how exactly do they work? When i deploy a webapp are those variables not able to be tracked down using developer console etc?

[lament rock]

Env vars arent global like in js you can do global.setTimeout or just setTimeout. They're usually accessed through a container like process.env or if you're loading from a file, js requires the dotenv npm module. You can access them through a repl assuming you're using the default eval. You'd have to specify a special customEval to disallow accessing things like process.env and even then wont work 100%

Env vars are usually containers for secrets like a config. Github runners make use of env vars for stuff like your gh access token and such

[eternal osprey]

owh damn i see

how are people that create websites hiding their api keys? Using dotenv..?

[slender wagon]

That can be done for backend which is not accessible by your users

Nothing is secure in frontend

[lament rock]

Yeah hiding stuff backend is super easy

With like Discord for example, you verify interactions with your client ID and a client secret

You dont need to send that info to the frontend anywhere

I just have everything in a js file so I can omit some config and build it later in the file like client ID can be inferred by token

Buffer.from(token.split(".")[0], "base64").toString("utf8")

https://github.com/AmandaDiscord/Amanda/blob/restart/packages/runtime-website/src/paths/discord.ts

[crystal wigeon]

I’ve been wondering

If people even use queues

Like there’s aws sqs but there no alternative outside right? The closest one is rabbitmq and celery

Would people even use something like sqs outside of Amazon

[lament rock]

Something like consumer queues are pretty good even in Discord bot spaces. If you have a microservice which just handles Discord gateway connectivity, it can send events to a queue for a worker to process the event and cache relevant data then that cache worker sends the event as a whole or partial to another worker which depends on the cache to be there

[eternal osprey]

oh shit i know what to do

i can just generate a public and private key pair

and use the public keys on frontend

[crystal wigeon]

But like rabbitmq already offers this mechanism why isn’t anyone building it

[lament rock]

Wdym

[crystal wigeon]

If I offered a solution that takes care of routing etc and you just have to push the message and consume it. Will people use it

Rabbitmq offers queues, dead letters multi queue

Routing etc

Like sqs

This isn’t specific to discord bots tho

[lament rock]

Likely not and building a system for that can be very expensive as the messages have to exist somewhere like in memory or on disk which depending on payload size can be extreme. It wouldnt be absurd for someone to pipe images through the queue for a worker to process either

images can be massive

[crystal wigeon]

I would just offer an sdk with the wrapper

Everything else is on you

So I don’t have to worry about expense

[lament rock]

libamqp for js already exists

[crystal wigeon]

Yeah but you have to like figure out the routing key binding etc

Or do I not make sense hmm

What if I offered a serverless solution for rabbitmq cause rn cloud hosted rabbitmq seem expensive

I couldn’t find anything else other than this cloudamp

https://www.cloudamqp.com/plans.html

[crystal wigeon]

Mmmm

Instead of you making api calls you just call the sdk method

Superset of libamqp

ofc i could make it open source

That’s what celery is doing

Queue Management

With different brokers

[lament rock]

You are not making any sense

[crystal wigeon]

Mmm

[lament rock]

A single amqp instance can handle multiple queues at the same time. While there are message consumers, the queue stays active. Of course you'd need a "routing key binding"

If you wanted a default queue, maybe, but that isnt worth making a whole lib for

[green kestrel]

i really love coroutines... makes my code so much cleaner

neutrino swear_check(event.from->creator, config::get("neutrino_user"), config::get("neutrino_password"));
swear_filter_t swear_filter = co_await swear_check.co_contains_bad_word(guild_name);
if (!swear_filter.clean) {
    bot.log(dpp::ll_warning, "Potty-mouth guild name " + guild_name + " censored for id " + event.command.usr.id.str());
    guild_name = swear_filter.censored_content;
}
auto exists = co_await db::co_query("SELECT id FROM guilds WHERE name = ?", {guild_name});
if (!exists.empty()) {
    embed.set_description(tr("GUILD_EXISTS", event));
} else {
    co_await db::co_query("UPDATE guilds SET name = ? WHERE id = ?", {g[0].at("guild_id")});
    embed.set_description(tr("RENAMED_GUILD_TO", event, dpp::utility::markdown_escape(g[0].at("name"))));
}

[crystal wigeon]

Yeah so you’ll have to configure those right. Like for each consumer etc

Hmm

[green kestrel]

this would be a nasty mix of a callback and several blocking calls, without

[lament rock]

Async is cool

[crystal wigeon]

What if there’s a UI like sqs? With rabbitmq?

[green kestrel]

i just wish they'd called the keyword await not co_await

[crystal wigeon]

would that be worth building

[lament rock]

What would you even show in this supposed UI

[crystal wigeon]

Same stuff what sqs shows

[green kestrel]

an interface, for users

[crystal wigeon]

[eternal osprey]

i somehow see async like pressing down the clutch in your manual car.

[lament rock]

I have no idea what sqs is

[crystal wigeon]

and you handle the queue

Aws queue basically

[eternal osprey]

detaching that bitch from the cpu's regular clock timers 🗣️

[green kestrel]

in C++ its like putting a bookmark in the book, and going to read a different chapter for a bit, or even putting down that book and picking up another

then later on, coming back to the bokmark (co_return)

[lament rock]

You'd have to consume a message without actually consuming it to show things related to the queue

[crystal wigeon]

Welp ig what I’m trying to solve doesn’t make sense

[green kestrel]

C++ coroutine async has no built in runtime, you have to make that bit yourself, it just provides the keywords

[lament rock]

Though tbf I havent touched libamqp in a minute

[lament rock]

Based

[crystal wigeon]

Does anyone else have any ideas

For a start up

Saas

Just brain storming ideas

[lament rock]

Solve a problem devs have

Or create a problem

[crystal wigeon]

Can’t think of any atm. What do you think

[lament rock]

I think I need someone to work on https://github.com/AmandaDiscord/Volcano

[crystal wigeon]

Lavalink

[lament rock]

alternatively, make a js lib that transcodes audio formats to opus and can apply audio effects without using ffmpeg

Must make use of native bindings to convert formats to pcm then apply any filters then encode to opus

[crystal wigeon]

So basically to stream music?

[lament rock]

Yeah. My main thing is that Volcano has to make use of ffmpeg which sucks for performance and I have to do weird hacks for ffmpeg to work properly and even then sometimes the ffmpeg process dies for no reason

LavaLink is a cool piece of software, but since it runs on the Java virtual machine, the memory usage is higher than I'd like

[green kestrel]

that would be sooooo slow

loops in js be like....
1..... 2............

access to AVX and other matrix math functions etc helps native transcoding go so fast. in js, you dont have access to any of that stuff

it would be hundreds of times slower

i mean parts of libopus are legit raw asm

[lament rock]

Luckily, you can do everything in a native addon and then transfer the result to js

[green kestrel]

but then youre not using js

youre using a native plugin

😄

[lament rock]

Id rather have the heavy audio processing be in native where it makes sense and then write my app in js

because js is fun and easier to work with imo

[green kestrel]

indeed, it would be crazy talk to do audio processing in an interpreted lang

unless you made a domain specific language for the task, and wrote it in that

there are such languages, you can find them in synthesizers and high end music hardware

[lament rock]

Im very aware js is not meant for audio processing but god I wanna die when I write rust and I am not at all confident in my abilities with any clang

[green kestrel]

actually

a DSL for audio processing accessible from js would be really cool

so you can do effects and stuff without needing to write low level C

in the same way you wouldnt do raw files as dbs and youd use sql, would be cool if audio had "this"

[lament rock]

There was some library that provided bindings for libavconf but it was so absurdly complex for me to get started that I dropped it entirely

Like I think it expected me to know the audio formats, but I can't really reliably probe them

The lib I used for probing wanted me to read to eof in some cases which sucks

especially when there sometimes isnt an eof with streams from icecast

I dont think it supported streams either

streams is another big thing the supposedly library would have to support

If it provided a callback where it pushed chunks, that wouldn't be the end of the world as I can just pipe them through a stream.Passthrough if I needed a stream.Readable/Writable somewhere

[quartz kindle]

env is like:
my code uses env, so by itself it does not contain any keys
i can post my code anywhere, github, etc, it does not contain any sensitive info
i clone my code into X machine
i add env vars inside the X machine
when my code runs, it will use the sensitive info i defined inside that machine
i can copy paste my code into a different machine and it will use the env vars of the different machine, no code change required

[sharp geyser]

What would be the best way to chunk upload videos?

[quartz kindle]

actually it wouldnt be that bad, provided your js runs in v8

v8's compiled code is not as fast as cpp but its not an order of magnitude slower either

[plain bough]

dawg this bot's literally called "nazis" what 💀 #mod-logs message

[quartz kindle]

The current performance level of V8-produced assembly is a little bit below that of g++ -O0 — and an order of magnitude higher than any other “interpreted” language
(an article from 2021)

[frosty gale]

unacceptable performance

v9 engine better be faster

this is what we get when we decide to use languages like javascript instead of real men (and women) compiled languages

[quartz kindle]

lmao

[lyric mountain]

cut the bytes into chunks, prefix with metadata (part number, upload id, etc) and send

then at the server u reassemble it

[eternal osprey]

<div id="output"></div>
 document.getElementById('output').textContent = 55+"\n55";

why does this show as "55 55"?
like i know in reality i should be using innerHTML and use <br> but i am trying to maintain the html syntax.

So, "<p>test</p><br><p>hey</p>" ->
<p>test</p>

<p>hey</p>

[sharp geyser]

[lyric mountain]

because \n cannot become number

[sharp geyser]

How do you cut these into chunks

[eternal osprey]

that ain't the prolem

[sharp geyser]

arent mp4 files notoriously hard to work with because they lose meaning when cut into chunks?

[lyric mountain]

uh...bytes.split()

[eternal osprey]

i am more like, how do i maintain html syntax while still accounting for \n:

"<p>test</p><br><p>hey</p>" ->
<p>test</p>

<p>hey</p>

[lyric mountain]

you dont need the files to have meaning

[quartz kindle]

another interesting article to read:

// #2
for (let i = 0; i < input.length; i++) {
    const x = input[i];
    output[i] = x * x + 2 * x + 1;
}

[lyric mountain]

I dont get what u mean

[frosty gale]

ive seen many platforms stream mp4 files in chunks so i doubt it

as long as the browser or whatever youre using gets the initial header i dont think it really matters

but im not sure how the mp4 format works under the hood would need to have a look

[sharp geyser]

I see

[frosty gale]

but i would assume its just a header at the beginning and the data following it (potentially compressed)

[sharp geyser]

Also naother question, what all would I need to be able to stream my videos myself rather than relying on something like cloudflare streams or bunny.net streams

I eventually want to reach that point

[eternal osprey]

i have a simple text area.
I input text -> it returns me html syntax.
but it should account for newlines as well:

assume my text area has the following input -> <p>test</p><br><p>test</p>
then the output should be:
<p>test</p>

<p>test</p>

but if i use inerHTML it is simply showing
test

test

and textContent is doing
<p>test</p> <p>test</p>

[frosty gale]

cdns are quite simple to design with modern langs like nodejs

[sharp geyser]

I think thats what is holding me back

[lyric mountain]

well, that's because your text area isn't returning html, it's returning markdown

[frosty gale]

or you can self host one which is probably more optimal in terms of performance ig

[sharp geyser]

I dont have a CDN myself

[frosty gale]

theres things like open source aws s3 clones

probably for streams or media theres something like that too

[sharp geyser]

My issue is running a CDN myself would be expensive

[lyric mountain]

you cant simply grab the inner content of it and expect to be valid html

[sharp geyser]

probably more than just outsourcing

[sharp geyser]

Maybe

but it'd likely be expensive asf to run effectively

I mean at large scale something like cloudflare streams or bunny.net streams is going to be 300$ a month at best

[lyric mountain]

you can convert from markdown to html relatively easily, but be wary of html injection

[eternal osprey]

this is what it returns me:
<p>dsfsdf</p><p><br></p><p>sdfsdf</p>
i assume that <p><br></p> is an actual newline

it seems like valid html tho

[lyric mountain]

p is paragraph

[eternal osprey]

i know

but the construct <p><br></p> represents a newline

[lyric mountain]

<br> is the newline

the surrounding p is just a side effect of markdown

[eternal osprey]

i see

[frosty gale]

its not too bad if youre not going to have too many videos and have limits on the size of the videos

but if you want large scale then yeah

youtube isnt cheap to run google is losing money supporting it

[eternal osprey]

how would i then take the markdown from the textArea, and parse it in the resulting div?

[lyric mountain]

but well, if you want to put that inside something, dont write to innerHTML

use append

[eternal osprey]

oowh

[sharp geyser]

Yea

[eternal osprey]

append? how so?

[lyric mountain]

...typing

all nodes have append method

use getElementById to get the enclosing node

then append to it

[frosty gale]

you can take advantage of techniques like "archiving" lesser viewed videos by compressing them highly and then decompressing once theyre needed and keeping them in some kind of cache

you can do that more aggressively on lower scale

[eternal osprey]

but if i append data to it, doesn't it simply lose it's html syntax?

so <p>test</p> -> test.

[lyric mountain]

that'd make no sense

you're not appending text

[frosty gale]

and also once theyre uploaded do a lot of lossy compression and smart codec adjustments to lower the file sizes

[sharp geyser]

my main issue is delivering the videos via HLS

[lyric mountain]

innerHTML shouldn't ever be used for security reasons, unless you're the only one adding content to it

[sharp geyser]

It'd honestly probably be better to just use one of the bigger companies

I mean if they let me use my own storage i'd 100% be down to use em to just deliver the videos

[lyric mountain]

https://www.geeksforgeeks.org/how-to-append-html-code-to-a-div-using-javascript/

it's appendChild

[eternal osprey]

the problem with this though is that my html tags get lost

[lyric mountain]

?

do you want to print the html tags together with the text?

[eternal osprey]

yeah!

[lyric mountain]

then append

it'll print as text

[eternal osprey]

i am appending and it's still only showing the actual text

[lyric mountain]

using append()?

[eternal osprey]

appendChild.

[lyric mountain]

...use append if you just want text literal

[eternal osprey]

yo i finna kill myself

i am using append

but it's still displaying text only

[lyric mountain]

https://stackoverflow.com/questions/6763148/how-to-show-literal-html-script-in-a-web-page

[sharp geyser]

Would a vps be fine for a cdn server?

[neon leaf]

why not use cloudflare r2

[frosty gale]

yes

you can also have one on the same server as the app youre hosting

[sharp geyser]

oh cool

[sharp geyser]

Im streaming videos

[eternal osprey]

for (let i = 0; i < paragraphs.length; i++) {
               console.log(paragraphs[i].textContent);
                outputContainer.append("&lt; p &gt;" + paragraphs[i].textContent + "&lt; p &gt;");
                if (i < paragraphs.length - 1) {
                    outputContainer.append(document.createElement('br'));
                }
            }```
I am trying to follow what you told me, but for some reaosn it deadass prints the whole line includes the created lt and gt tags

[lyric mountain]

did you use the <pre> and <code> tags?

[eternal osprey]

oowh no

[lyric mountain]

pre is for preprocessing the tags

it only works if there's also a code tag in it

[eternal osprey]

i see.
I should put it in this part here right?

 outputContainer.append("< p >" + paragraphs[i].textContent + "< p >");```?

[lyric mountain]

yes

do note tho, it'll appear like this due to code tag

[eternal osprey]

it aint working brev

this is the whole code:

function parse() {
            let cleanOutput = document.getElementById('outputHtml');
            cleanOutput.textContent = ""
            let inputText = quill.root.innerHTML.trim()
            console.log(inputText)
            let tempDiv = document.createElement('div');
            tempDiv.innerHTML = inputText;
            let outputContainer = document.getElementById('outputHtml');
            let paragraphs = tempDiv.getElementsByTagName('p');
            
            for (let i = 0; i < paragraphs.length; i++) {
               console.log(paragraphs[i].textContent);
                outputContainer.append("<pre><code>&lt;p&gt;" + paragraphs[i].textContent + "&lt;p&gt;</code></pre>");
                if (i < paragraphs.length - 1) {
                    outputContainer.append(document.createElement('br'));
                }
            }
        }```

[lyric mountain]

const > let btw

did u try writing the html text on the browser and then checking how it inserted into the tree?

[eternal osprey]

yeah it is indeed working

but for some reason i think that the append is fucking is up here

maybe it only works with textContent =...

no that's stupid

[lyric mountain]

why did the pre and code tags appear there?

[eternal osprey]

outputContainer.append("<pre><code><p>" + paragraphs[i].textContent + "<p></code></pre>");

because it is reading and working with the tags like an actual string

not html code

[lyric mountain]

yes, but you see, WHY did they appear and not become html

if your whole issue is that paragraphs[i].textContent was being parsed as html

[eternal osprey]

that's strange indeed yeah

[lyric mountain]

what does your console.log show?

the one above the append

[eternal osprey]

i quickly used typeof, it is reading it as string though (the paragraphs[i].textContent)

the 2nd one is the textContent

[lyric mountain]

so your issue isn't that append is printing parsed html, you're parsing it beforehand

textContent wont print html, it'll pring text content

if you want the tags as they're written, use innerHTML

but you use that on the parent node

so if u want the paragraphs, get the innerHTML of the enclosing div

[eternal osprey]

innerHTML?

hmm

so you are telling me that instead of:

[lyric mountain]

it's a read-write property

[eternal osprey]

  outputContainer.append("<pre><code>&lt;p&gt;" + paragraphs[i].textContent + "&lt;p&gt;</code></pre>");``` we should use ```js
  outputContainer.append("<pre><code>&lt;p&gt;" + paragraphs[i].innerHTML
 + "&lt;p&gt;</code></pre>");```

[lyric mountain]

that's still get u the text, because you're getting the innerHTML

as in, "what's inside"

you wont get the <p> tags

also your < are wrong

[eternal osprey]

i see, but ain't i adding the tags later on with that pre code shit/

[lyric mountain]

those are supposed to be replacements to < and >

[eternal osprey]

i know

[lyric mountain]

what u wrote there is <<the content>>

[eternal osprey]

no right?

&lt = <
&gt = >
so<p> = <p>

[lyric mountain]

ah right, didn't see the p there

ur missing a / after the content btw

[eternal osprey]

i still don't get why it ain't working.

 function parse() {
            let cleanOutput = document.getElementById('outputHtml');
            cleanOutput.textContent = ""
            let inputText = quill.root.innerHTML.trim()
            console.log(inputText)
            let tempDiv = document.createElement('div');
            tempDiv.innerHTML = inputText;
            let outputContainer = document.getElementById('outputHtml');
            let paragraphs = tempDiv.getElementsByTagName('p');
            
            for (let i = 0; i < paragraphs.length; i++) {
                outputContainer.append("<pre><code>&lt;p&gt;" + paragraphs[i].textContent + "&lt;p&gt;</code></pre>");
                if (i < paragraphs.length - 1) {
                    outputContainer.append(document.createElement('br'));
                }
            }
        }```this seems to be taking the innerHTML already

[eternal osprey]

ah tahnks!!

[lyric mountain]

try just outputContainer.append(paragraphs[i].innerHTML);

[lyric mountain]

yes, because your code is iterating the <p>s

what's inside a paragraph? the text

if you want to include the paragraph itself, you need to get the parent's innerHTML

imagine <p> being a cup

if you look inside the cup, you'll se what's inside it, not the cup

I mean, you'll see the cup in real life, but imagine the cup is invisible from inside

[eternal osprey]

i see

[lyric mountain]

no you dont

[eternal osprey]

tf you mean i don't

[lyric mountain]

lmao, that was a joke, ur not supposed to see the cup

[eternal osprey]

hahaha okay

[lyric mountain]

anyway, remove the whole loop

print tempDiv.innerHTML

[eternal osprey]

ayoooo

[lyric mountain]

or if u want to keep the loop, hardcode <p>

idk why ur doing this whole part instead of using inputText

[eternal osprey]

that's what i done did

wait so it can be done easier?

[lyric mountain]

I mean, if you append inputText directly it'll be the same as doing tempDiv.innerHTML

oh, I see, ur doing this to break newlines in case it was written as a single line

in this case u could just inputText.split(/(?<=\/p\>).*?(?=\<p)/g)

it'll split between the p's, without removing the >< chars

[eternal osprey]

ahh regex i hate it

i think i'll just leave it like this

it works well, finally!

Thank you!!

do you know by any chance how i can transform multiple <br> tags to a single <br> tag?

<p>t3st</p>

<p>test how Is everyone doing!!!</p>

<p><strong>test</strong></p>
<p>sfsdfsdfsd</p>




<p><a href="google.com" rel="noopener noreferrer" target="_blank"><strong>this is a link!<br></strong></a></p>````

[lyric mountain]

regex

[eternal osprey]

https://tenor.com/view/sad-emoji-gif-25749073

[lyric mountain]

/(<br>\W+)+/g

[sharp geyser]

Use your imagination

It works trust

[eternal osprey]

let inputText = quill.root.innerHTML.trim()
            
            inputText = inputText.replace(/(<br>\W*)+/g, "<br>");```
this returns me

it basically fucked my shit

[lyric mountain]

change <br> to whatever you want to coalesce

[eternal osprey]

i deadass hate frontend so much

[green kestrel]

mmm quilljs

quill is nice

[lyric mountain]

if your repetition is <p><br>p&gt;</p>, then use /(<p><br>p&gt;<\/p>\W+?)+/g

might need to do some escapes there

[eternal osprey]

what the fuck

[lyric mountain]

it'll replace all but one with nothing

[eternal osprey]

  inputText =  inputText.replace(/(<p><br><\/p>\W*)+/g, "<p><br></p>");```
you see this right

[Log] <p>test</p><p><br></p><p><br></p><p>test</p> (html.html, line 46)
[Log] <p>test</p><p><br></p>p><br></p><p>test</p> (html.html, line 48)```

the 2nd ooe is the new output after the replace

[lyric mountain]

it's not the same as what I wrote

[real rose]

chat

[eternal osprey]

but the repeated sequence as you can see is <p><br></p>

[sharp geyser]

no

[lyric mountain]

\W+?

not \W*

[eternal osprey]

bomboclat

[sharp geyser]

no

[eternal osprey]

man i am so done with frontend waddafuck

(this is backend tho, you're just incapable chittykat)

[lyric mountain]

first, test on regexr not on code, it'll be easier if u can see what's being affected

[green kestrel]

hmmmm

[lyric mountain]

second, send the input here

[green kestrel]

is this a bug in discord client?

[lyric mountain]

?

[eternal osprey]

input:
<p>test</p><p><br></p><p><br></p><p>test</p>

[green kestrel]

this mod got a profile here just for this server that has different pfp, image and everything, click their pfp and it shows it

but right click and 'show profile' it shows the global one

should it always show the local one, if the click to view profile was from here?

[lyric mountain]

just stale cache

it's showing the same for me

[green kestrel]

stoopid discord

[eternal osprey]

i love you kuu

did you know that

[lyric mountain]

[sharp geyser]

https://tenor.com/bBVwh.gif

[plain bough]

I love blacklist

[quartz kindle]

anyone using vercel/nextjs? how is support for native node addons?

i got someone trying to use my lib in vercel and getting errors

i never used vercel nor nextjs, no idea what to do about this

[solemn latch]

This feels like a user error, not your problem. I've never used vercel so I dont really know either.

[sharp geyser]

https://stackoverflow.com/a/66997150

[quartz kindle]

ye i just linked them some node loader

[sharp geyser]

One possible solution

if its vercel and a module not found error

typical culprit is git

oddly enough

[quartz kindle]

that doesnt really seem related

[sharp geyser]

vercel deploys via git

[quartz kindle]

the dude's issue is about loading native addons

[sharp geyser]

I just know vercel uses git

[quartz kindle]

theres nothing about native addons on that issue

anyway i just linked them this

https://github.com/eisberg-labs/nextjs-node-loader

[sharp geyser]

I mean its a module not found error

thats as vague as it can be

[quartz kindle]

ye but there is a major difference between a native addon and everything else

[sharp geyser]

Do you know if they were using vercel pkg?

It was a known issue in pkg that .node files could not be found

[quartz kindle]

no idea

but

[sharp geyser]

Indeed, but you never know with people

💀

[quartz kindle]

indeed

they can always are have stupid

[sharp geyser]

I am too tired

I just had a stroke reading that

[quartz kindle]

:^)

[past field]

sigh

[solemn latch]

👀 for what you showed before. are you using canvas?

[quartz kindle]

from what i remember yeah he's using canvas

[past field]

yes

so so close

[solemn latch]

Add a default offset, its easier to place things in the right spot.

IE whatever you need to put on top, offset it so the point is in the center.

It makes it much easier to place things in a grid

[past field]

well

the numbers generating are canvas

the board is a png

so right now

with

const coordinates = Array.from({ length: 25 }, (_, index) => {
        const rowIndex = Math.floor(index / 5);
        const colIndex = index % 5;
        const x = 150 + (225 * colIndex) + (11 * colIndex) + (225/2 - 128/2);
        const y = 700 + (215 * rowIndex) + (11 * rowIndex) + (225/2 - 128/2);
        return [x, y];
    });

looks like

i need N G and O moved over just a little
bit

[solemn latch]

The last image you sent had the numbers more centered.
What did you change? 👀

the 11 was something else?

[past field]

i made the numbers smaller

[solemn latch]

Try making it larger ^-^

Assuming you made the png, look at what your offsets actually are and use those numbers.

[past field]

1 sec

const coordinates = Array.from({ length: 25 }, (_, index) => {
        const rowIndex = Math.floor(index / 5);
        const colIndex = index % 5;
        const x = 200 + (225 * colIndex) + (11 * colIndex) + (225/2 - 128/2);
        const y = 700 + (225 * rowIndex) + (11 * rowIndex) + (225/2 - 128/2);
        return [x, y];
    });

is what generated the first image i sent

[past field]

how do i get those numbers again?

[solemn latch]

What software did you use to make the png?

[past field]

i had a graphic designer make it

[solemn latch]

ah, then its going to be easier to guesstimate

[past field]

number font was 100px here

[past field]

80px here

[solemn latch]

I'd leave the 150 where its at, and mess with the 11, try setting it to higher, like 15.
It should hopefully send it too far by a bit, then just lower it until its centered.

do const x = 150 + (225 * colIndex) + (15 * colIndex) + (225/2 - 128/2); first and see what happens.

[past field]

@solemn latch this is at 25

for y, should i go higher or lower than 11?

[solemn latch]

Yeah

[past field]

changed y from 11 to 5

@solemn latch

and the font is 80px

[sharp geyser]

Ima be real

why the white in the background

it looks out of place

[past field]

figured the numbers sitting nicely in there would be a nice vintage touch lol

[sharp geyser]

I mean maybe, but it looks too small to fit hte numbers properly

You will never be able to center it perfectly

[past field]

this is honestly good enough for now lol

now i need to get the daubs placed properly

[solemn latch]

const x = 150 + (225 * colIndex) + (11 * colIndex) + (225/2 - 128/2);

The first number(150 is the offset from the left or top
The second number(225) should be the exact width/height of the box.
The third number(11) should be the space between the boxes.
The last bit I'm not sure what its doing.

In a perfect world, your code would look something like this.

const coordinates = Array.from({ length: 25 }, (_, index) => {
        const rowIndex = Math.floor(index / 5);
        const colIndex = index % 5;
        
        const base = (225 * colIndex) + (25 * colIndex) + (225/2 - 128/2);
        const x = 150 + base;
        const y = 700 + base;
        return [x, y];
    });

Then once you know what the numbers actually are; you can offset by half the width, and height of the numbers.
That way the numbers are as centered as possible.

But guestimating is good enough that no user would notice.

Man, I hate canvas 😄

[solemn latch]

That should be easier ^-^

[sharp geyser]

Yea

Im seeing too much math

im outta here

[past field]

is it the 128 that places the daubs?

[solemn latch]

I dont think so. I'm not really sure 👀

I feel like (225/2 - 128/2) is just for scaling things. Ideally it shouldnt be used, but if its working theres not really an issue.

[lyric mountain]

that's for centering

half target width - half source width

[past field]

for example, the daub on the free space

marked.forEach(({ row, col }) => {
        const [x, y] = coordinates[row * 5 + col];
        console.log('Dabbing at:', x, y);  // Debugging coordinates
        ctx.drawImage(dabImage, x, y); // Draw the dab image without resizing
    });

[lyric mountain]

honestly the numbers still aren't centered

[past field]

yeahh i know, i'm guesstimating lol

[solemn latch]

Centering the text? the text has a dynamic width. that code is not accounting for it.

[lyric mountain]

why aren't u just getting string width/height?

[lyric mountain]

was supposed to

last time they came here we told them to get actual string dimensions

[solemn latch]

Ah

[past field]

idk, i'm new to building bingo and using canvas - i'm not very knowledgeable of this math lol

i tried to in paint and i struggled a bit

[solemn latch]

I think for now just ignore that part, guesstimate for now.

[solemn latch]

its not a monospace font, so you'll have to do it though code

[lyric mountain]

fixing the text would unironically fix the chip position too

all that's needed is a generic placeAtSlot function

[past field]

but the daub is way off from the text no?

[lyric mountain]

a generic function would place it centered on the slot regardless what you passed to it

[solemn latch]

The first time you go through this its so rough 👀

[past field]

it is lol i should have waited before trying to build this

[past field]

hmm let me look at this

[lyric mountain]

you'd need to write it yourself, it's not a built-in function

https://www.w3schools.com/jsref/canvas_measuretext.asp

canvas.measureText(text).width for string width

const metrics = canvas.measureText(text)
const height = metrics.actualBoundingBoxAscent + metrics.actualBoundingBoxDescent;
``` for height

all that's left is fitting those values in the formula u already have

[quartz kindle]

also, for 99% of the fonts out there, height = font size in px

[past field]

const slotSize = 225; // Define the slot size
    const offsetX = 150; // X-axis offset for the first slot
    const offsetY = 600; // Y-axis offset for the first slot
    const spacing = 15; // Spacing between slots

[real rose]

YO we got a bingo update

[crystal wigeon]

do devs even use rabbitmq?

is there a demand for serverlress rabbitmq?

[lyric mountain]

btw it's worth to note, "bingo" is supposed to be aligned with the slots

it's not just a title

and the text isn't aligned because you're still not getting text width correctly

[crystal wigeon]

bump

[lyric mountain]

bumping 15 minutes later wont really work

that's a very niche question I doubt you'll find answer here

[past field]

😭

[past field]

yeahh i’m gonna ask him to change this

[quartz kindle]

there sure is demand, but good luck competing with google cloud pub sub, microsoft azure bus, amazon sqs, etc...

especially because demand is mostly heared towards high scale

small projects dont need message queues, only very large projects do

and very large projects require very large infrastructure

[crystal wigeon]

yeah i dont think its worth building since there's alreayd so many, but there arent any serverless ones tho. even for redis. upstash is the only platform that has serverless redis

[surreal sage]

i hate when isps use "better streaming quality" as argument. i hate corporate marketing

[quartz kindle]

same

watch hd videos with our incredible 1000KB plan*
*quality may change depending on network conditions

[neon leaf]

what are good postgresql providers to use with my cloudflare worker

cfs sqlite is just horrendously slow

[frosty gale]

what is rabbitmq even for?

from what you said i assume its to be able to queue long tasks and distribute them to servers for processing?

and have the reply be fed back where its supposed to be

i swear redis can do this too

[quartz kindle]

indeed it can

although proper message broker software has more tools for more advanced use cases

[frosty gale]

a good use case for something like this is probably chatgpt or ai image generation services where generations take ages and you might have multiple servers that can process these generations

and it makes distributing everything easier

[quartz kindle]

indeed

yoi have a large queue and hundreds of consumers each taking one task

also rabbitmq is not the only such software, there are tons of others

rabbitmq is one of the slowest tbh xd

[crystal wigeon]

Tim

Do you have any ideas let’s build a product

just brainstorming ideas for now

[quartz kindle]

i hate product design lmao

i like the whole engineering part, but when it comes to actually convert/transform my engineering into a usable/commercial product, i absolutely hate it

even my libs, once i figure out the engineering part of it, packaging it into an actual js library for github/npm is so damn annoying to do

you need to write tests, types, docs, readmes, presentation, marketing, visuals, etc...

even organizing the code into reusable functions and classes is extremely annoying, you have to guard it against all possible types of misuse by idiots

[sharp geyser]

so basically your average js dev

[quartz kindle]

yup, im your average js dev

[surreal sage]

are there any websites

that allow you to edit images with css

[lament rock]

No. They meant the average js dev is an idiot who misuses things

[quartz kindle]

also me (look at how i misused discord.js)

:^)

[lament rock]

I mean. Sure you fuck around with the internals of node but god damn is that smart

[quartz kindle]

wdym edit images with css

[lament rock]

I also fucked around with discord.js and made a light addon but like lol

[quartz kindle]

tbh djs deserves to be misused, abused, mistreated and fucked around with

:^)

[lament rock]

So true

type safety be damned we're fucking with the raw api

which is also type safe…

[frosty gale]

i looked if theres any zig discord libraries and theyre all dead lol

[sharp geyser]

makes sense

[sharp geyser]

make your own

what even is the appeal behind zig

it just looks like a knockoff rust

[proven lantern]

zig builds much faster

i just spent a day debugging a zig library bug

https://github.com/softprops/zig-lambda-runtime/blob/main/src/lambda.zig#L255

the library isn't handling chunked responses. I had to modify line 255

        if (req.response.transfer_encoding == .chunked) {
            // For chunked encoding, we need to read and parse chunks manually
            var buffer: [4096]u8 = undefined;
            while (true) {
                const bytes_read = try req.read(&buffer);
                if (bytes_read == 0) break;
                try payload.appendSlice(buffer[0..bytes_read]);
            }
        } else if (req.response.content_length) |length| {
            const content_length = @as(usize, @intCast(length));
            log.debug("Content length: {}", .{content_length});

            try payload.resize(content_length);
            // make a copy of the response data that we own
            const data = try payload.toOwnedSlice();
            errdefer self.allocator.free(data);
            _ = try req.readAll(data);
        } else {
            log.warn("No content length and not chunked encoding", .{});
            return null;
        }

the lambda test UI sends events in without any transfer_encoding. when you actually invoke a lambda transfer_encoding=chunked

once i replace rust with zig my build times will be ~2m30s faster

[sharp geyser]

Build times != good language

[proven lantern]

i'm also thinking about building the binary on my local machine instead of the pipeline build machine. the language has really good tooling and lets you cross compile

[sharp geyser]

Language can build fast but if its terrible performance it doesnt matter

[proven lantern]

i've never seen anything about zig having bad performance. all else equal

Rust cant event compile without zig in some cases like with AWS lambdas

[sharp geyser]

I never said it wasnt performant

I was simply saying having fast compile times does not mean its a good language

[proven lantern]

so all else equal it does make it a better language

[sharp geyser]

I mean if all zig does is increase build times by a bit theres no point in using it imo

Just use rust

[proven lantern]

have any other langaues created a full toolchain like zig?

zig builds faster than rust

by a lot

[sharp geyser]

Most of the build times with rust comes from installing all the features along with the crate

Most of which you dont need

People really just throw the deps into cargo.toml and wonder why they take so long to build

When they dont even use all the features of the crate

[proven lantern]

will tokio ever be built into the language

[sharp geyser]

Likely not

I mean maybe one day

I doubt anytime soon though

Tokio is so vast itd take a while to natively support it

[proven lantern]

these are all the deps in my rust code.

[dependencies]
aws-config = "1.5.1"
aws-sdk-lambda = "1.32.0"
anyhow = "1.0.86"
ed25519-dalek = "2.1.1"
hex = "0.4.3"
http = "1.1.0"
lambda_http = { version = "0.12.0" }
serde_json = { version = "1.0.117" }
tokio = { version = "1", features = ["full"] }
tracing = { version = "0.1", features = ["log"] }
tracing-subscriber = { version = "0.3", default-features = false, features = ["ansi", "fmt"] }
once_cell = "1.19.0"
serde = "1.0.203"

so much junk

ed25519 is built into zig

https://ziglang.org/documentation/0.13.0/std/#std.crypto.sign.Ed25519

i dont need anything like tokio

i'll compare the code and performance after i finish the rewrite

[summer torrent]

https://thehackernews.com/2024/07/twilios-authy-app-breach-exposes.html

[frosty gale]

oh great

the fines and consequences for businesses suffering these data breaches through sheer incompetence when it comes to security is basically none

so some companies have little incentive to be serious about security

[quartz kindle]

i still use the deprecated desktop version

because

1. it still works
2. icba having to check my phone every time i need some stupid code

2fa code generation algorithm is always the same, so the codes should in theory work forever, regardless of app changes

[frosty gale]

how big is chromium lmao

i wanted to make one small change to the chromium browser for my specific dev use case but im still on the cloning stage

[quartz kindle]

dude trying to build chromium lmao

everyone knows thou shal not build chromium

it takes literal days

[frosty gale]

i really hope not

what do i do if i cant build chromium

chromium might as well not be open source

since it takes so long to build

i think i can place chromium in a dev drive so windows defender doesnt slow it down too

can i possibly also exclude certain components of chromium from building

i dont need every single feature chromium has to offer in this case

[quartz kindle]

On an 80-core machine with -j 72 you should be able to build Chromium in maybe 40 minutes. On a 4 core laptop it probably takes six to seven hours

It took an ARM Launchpad builder 100 h to build it once

How did we get here? Well, C++ and its stupid O(n^2) compilation complexity. As an application grows, the number of header files grows because, as any sane and far-thinking programmer would do, we split the complexity up over multiple header files, factor it into modules, and try to create encapsulation with getter/setters. However, to actually have the C++ compiler do inlining at compile time (LTO be damned), we have to put the definitions of inline functions into header files, which greatly increases their size and processing time. Moreover, because the C++ compiler needs to see full class definitions to, e.g., know the size of an object and its inheritance relationships, we have to put the main meat of every class definition into a header file! Don't even get me started on templates. Oh, and at the end of the day, the linker has to clean up the whole mess, discarding the vast majority of the compiler's output due to so many duplicated functions. And this blowup can be huge. A debug build of V8, which is just a small subsystem of Chrome, will generate about 1.4GB of .o files which link to a 75MB .so file and 1.2MB startup executable--that's a 18x blowup.

@offtopic
a paying client for my api is asking for help getting it to work on his website
show me his client side code (absolute mess)
i tell him to add a console.log to check what is actually being requested from my api (with a copy-paste example)
he says he doesnt understand much and asks me if i can do it instead
proceeds to send me a 497mb rar file containing the website's code

@_@

[pearl trail]

trojan incoming

[quartz kindle]

the rar file contains 490mb of audio files

lmfao

[proven lantern]

want to test this binary i wrote in zig?

[sharp geyser]

I mean hes paying

[quartz kindle]

jesus christ what is this

code is full of global variables and weird ass code flow

and everything is in italian

but the mp3 files are in portuguese

and the code gives a bunch of 404's because of that

this is literally a guy trying to make a product to sell while having no idea how to make the product itself

but you know, making money is more important than your product actually working :^)

[sharp geyser]

Bro couldnt decide what language to use

[quartz kindle]

some files are in spanish

there is a name translator function with everything hardcoded

there are so many switch functions

dude doesnt know how to use arrays and objects

and uses global vars inside functions

@_@

[proven lantern]

but does he use null?

that's the main question

[quartz kindle]

hardcoded coordinates for countries

if you're not from one of those countries, it doesnt work for you

if you are from one of those countries, you were born exactly in the middle of it, or wherever that coordinate points to

:^)

man, reading this code is as fun as reading luca's code

ah yes, my age can only be a multiple of 10

[proven lantern]

need to make sure age >= 20 right after checking age < 20

[quartz kindle]

yup

[proven lantern]

that makes it extra safe

[quartz kindle]

you never know

might get a bit flip between lines

:^)

uses both axios and fetch in the same script

[sharp geyser]

💀

tim

this is 100% chatgpt generated

[frosty gale]

im sorry to tell you but even chatgpt doesnt generate this bad of code

[sharp geyser]

are you sure?

I've had chatgpt generate virtually the same code, but it got worse each time

"This doesn't exist"

"Oh im sorry, let me regenerate that for you" uses the same fucking thing that doesnt exist but in a different way

[quartz kindle]

lmfao

[sharp geyser]

💀

yea

typical ai

Cant really fault the AI though

its what its trained on

[quartz kindle]

:^)

one of my most fun moments with chatgpt was working with math formulas

i give it certain parameters, and then see it use them all wrong

for example x = ..., y = ..., z = ... cartesian coordinates in AU

and then chatgpt randomly decides to use those values in other units

[proven lantern]

i'm using claude ai. it's supposed to be better at algorithm stuff

[sharp geyser]

well ofc

[quartz kindle]

or i give it input in degrees and chatgpt uses it as radians

[sharp geyser]

how does it know what you want if you tell it directly what you want

it has to guess obviously

[quartz kindle]

i which i had saved a conversation i had with it a few days ago

it was so funny

[sharp geyser]

well

it keeps track of all of them

I still have some from 6 months ago

funny stuff

[quartz kindle]

remember the sponge bob meme? the one with patrick's driver licence

[sharp geyser]

kinda

https://tenor.com/qZVuWGhWujQ.gif

[quartz kindle]

this one

that was legit me and chatgpt

[sharp geyser]

💀

Spongebob was a lowkey lit show

[quartz kindle]

it was something like this:
chatgpt: speed of light in AU per day equals WRONGVALUE
me: what is the speed of light in kilometers per day?
chatgpt: X
what is an AU in kilometers?
chatgpt: Y
how many AU do you go in one day at the speed of light?
chatgpt: Z
then how much AU per day is the speed of light?
chatgpt: WRONGVALUE

lmfao

[sharp geyser]

💀

[proven lantern]

zig has null and undefined for some reason. weird.

[sharp geyser]

likely were js devs at one point

[scenic kelp]

zig probably forces you to initialize a variable at the declaration site

if you want to leave the memory uninitialized you have to do so explicitly

[sharp geyser]

most likely

[proven lantern]

the comptime keyword is pretty tricky

i'm trying to read a JSON file and create a std.StringHashMap([]const u8) at compile time, but it cant use an allocator at comptime, but all the functions require an allocator.

tricky language

[summer torrent]

I will switch to Bitwarden. It is password manager but has TOTP too

• available everywhere

[lament rock]

having null or undefined is useful. the type void is also undefined

undefined is self explanatory. If you have to initialize something without providing a value, leaving it up to the compiler/runtime based off type isn't fun

[proven lantern]

it makes more sense in a language like zig that has more control over memory

it shouldn't be in JS though

[lament rock]

I strongly disagree

[proven lantern]

isn't undefine/null a legacy issue

they added null in for Java people at some point

[lament rock]

In JSON, undefined isnt serialized but null is

Most places communicate in JSON nowadays

[proven lantern]

it never made sense to send the thing you aren't sending

like send null instead of just not sending it

just omit it

[neon leaf]

Oh no Ben is arguing over null/undefined again

[lament rock]

Well then should the value be undefined or null when someone serializes it

[proven lantern]

it does make sense in Zig

but not JS

[lament rock]

the values can have different meanings depending on how the API is built

Like user.avatar is nullable

But also undefined doesnt exist in some languages

[proven lantern]

most languages dont have two values like that

that's why i was suprised that zig did it

[lament rock]

yeah but some languages error when you try to access something that doesnt exist

undefined and null are both pretty necessary imo

[proven lantern]

what other languages beside JS and Zig have undefined and null?

[lament rock]

Idk. Not like I care either since I stick to what I'm good at or need to know

Perhaps theres jank for situations like that in other languages

[neon leaf]

imagine not having null and undefined in your language
-# Only you can see this • Dismiss message

[proven lantern]

imagine not having null and undefined in your language
-# Only you can see this • Dismiss message

Rust just has Some and None right?

[quartz kindle]

perl has both undefined and null

[proven lantern]

undef and what?

null?

really

[quartz kindle]

ye

[proven lantern]

idk about that