dev-general

[brazen ether]

hilarious readme

[hushed prairie]

Some say "this README reads like they belong in a hospital."
And you're goddamn right.

nice

[tribal plinth]

aight this caught me off guard

[agile galleon]

That was entertaining lmao, thank you for sharing

[pastel imp]

what algo did they use lol

someone should implement this in paper

lol

[ocean quartz]

You mean the one they did? https://github.com/bsommerfeld/pathetic-bukkit

[prisma wave]

i think he means in the server impl?

[potent nest]

technical players wouldn't survive that

[prisma wave]

we should use lean to prove their equivalence

or disprove it

[potent nest]

wow you just came up with topics for 3 PhDs

[prisma wave]

yay!

[agile galleon]

Paper'd get another hate video from cubic lmao

"Paper'd" what the fuck did I do here is that correct English?

[potent nest]

everything is correct English

[agile galleon]

ghoti (pronounced fish)

[pastel imp]

not necessarily phd

probs any thesis

[potent nest]

yeah just far more then

[unkempt cloak]

is USpigot actually what it says it is?

I heard many mixed things about it and idk

[nocturne apex]

am using it

[dapper pawn]

@visual bluff dms

[brazen ether]

A*

[agile galleon]

On the one hand it's quite shady as it's entirely closed source (even though as per the GPL license it would have to be open) and suspected to be straight up stealing patches from other forks. Additionally if you need custom patches on new versions or something isn't exactly working then support can be slow and you can't just hire somebody to do it as they don't have the source to make the necessary changes. On the other hand I have heard very good things about the performance increase servers with many players are experiencing. Thus, imho it's a double edged sword.

[cold prawn]

i would try looking in the network tab, and finding the way they get the data, and unless they use SSR or sum, they probably just fetch the list with a page param, so you can just make a for loop and scrape them on ur browser, doubt they rate limit it much for a logged in user

[abstract osprey]

and suspected to be straight up stealing patches from other forks
who what when where why

[potent nest]

yeah why would someone who benefits from the community that they actively hurt do something unethical

[nocturne apex]

yea Ive been told that too and they also shown me what was just grabbed

I mean I do get both sides

but its tru Machine could have communicated it better w them

the worst part is that as it is close sourced, it contains private patches from servers even Hoplite patches and people can just steal it. Ive already seen dumped classes of the fork

and all Hoplite stuff was there

they should have made it open sourced at this point

[pastel imp]

It's like said, double edged sword

[agile galleon]

I can't give you an exact source sadly so take my words with caution. I just tried summarizing the different things I've heard lately

Ah nice that this message went through now... I sent it 4 hours ago lol

[pastel imp]

any ideas which packets TAB sends or are sent in general for scoreboards?

is it Scoreboard_objective?

listening for SCOREBOARD_OBJECTIVE and DISPLAY_SCOREBOARD, none of them being triggered for some reason

guess they using UPDATE_SCORE

[hushed prairie]

https://github.com/NEZNAMY/TAB/blob/master/bukkit/v1_8_R1/src/main/java/me/neznamy/tab/platforms/bukkit/v1_8_R1/NMSPacketScoreboard.java

idk which version you wanted but here's 1.8

you can check the other modules for other versions

[pastel imp]

Yeah figured it out, I will never support 1.8 but ty though

[hushed prairie]

smart choice

[nocturne apex]

I have got a simple handling for scoreboard on modern versions

if you want I can show u

[pastel imp]

sure, using PE but that doesn't hurt.

[nocturne apex]

ye there's not that much tbh

https://paste.learnspigot.com/pamiqideya.kotlin
https://paste.learnspigot.com/enibuganoq.kotlin
https://paste.learnspigot.com/ebupahenup.kotlin

here u go

it is just a regular scoreboard without the score shown

[pastel imp]

ty. also regarding your showcase, would be cool if it supported kotlinx.serialization

also why you using learnspigot's paste 💀

[nocturne apex]

am lazy to use gist

xD

[nocturne apex]

Yeeeh uh I was thinking of that tho none of my systems use it

[cold prawn]

isint 1.8 finally basically obsolete

or do servers still cling onto it

[wind patio]

I would guess there are still some mainly because of the PvP mechanics

[hushed prairie]

I wish

well, it is

but that doesn't mean people will stop using it

[wind patio]

I'm not sure if it's possible to replicate the 1:1 1.8 PvP mechanics in newer versions

[hushed prairie]

it absolutely is, and has been done

it's server sided, no reason for it not to work

[brazen ether]

some people will swear by their lives that it can be done, others will swear that it always feels different

[hushed prairie]

but people are people and say that it's not the same

[hushed prairie]

yeah

[brazen ether]

hypixel definitely doesn't feel like 1.8 on newer versions, idk about anything else

[wind patio]

I mean I would guess so too, but those servers exist for (basically) that sole reason

[brazen ether]

id imagine their emulation is just not very good

[brazen ether]

dumb players, devs who don't wanna upgrade

[half harness]

it's not server sided??

[hushed prairie]

damages are applied by the server, not the player

[wind patio]

that's true, yeah

I mean, even if the replicated mechanics don't feel 1:1, at some point you just have to adapt and move on lol, it's not 2008 anymore

[half harness]

there are much more to it than just damage

but that being said

I think there are several plugins that very closely replicate it

[hushed prairie]

the player just sends the click, server detects it and applies damages based on attributes and items used

[brazen ether]

damage is not what people refer to when taking about 1.8 combat

[hushed prairie]

what is then?

[wind patio]

knockback, durability, heal ratios, etc

[brazen ether]

okay now I'm much less trusting of your experience 😭

[half harness]

movement, blocking, and I think the OldCombatMechanics had several more stuff that I'm overlooking

[brazen ether]

knockback and movement

[hushed prairie]

and how is that not server sided as well?

[brazen ether]

blocking yeah

[half harness]

movement & blocking are client controlled (shield workaround is... meh)

[hushed prairie]

how will the client know how much kb to take without the server telling it?

[brazen ether]

if only the client predicted movement or something

I wonder what that would be called

[half harness]

wdym

[brazen ether]

sarcasm

[hushed prairie]

predicted and based on what? items and their attributes/enchants ? that can be altered with packets to match the player's version

[wind patio]

there are bunch of other "hidden" mechanics in 1.8 PvP like what happens when player you hit is sprinting at you, jumping, and running away, W tapping, etc., that are not always considered when making a 1.8 combat replica

[half harness]

wait what does the mc client predict movement? I thought it just its own movement and reported it to the server

[brazen ether]

actually I might be misremembering

I know client does some sort of movement prediction

but I don't think it actually applies to knockback

[half harness]

oh idk I haven't really looked deep into it
I heard that in like the new version it also report the controls to the server or smth

[hushed prairie]

I thought so too since when your connection lags, you can see players floating on the air without falling or moving

[brazen ether]

there's lots of strange things as well with how clicks (not attacks) impact how you take knockback

[half harness]

speaking of
there is indeed https://www.spigotmc.org/resources/LegacyKB.86080/ which is pretty neat for pvp servers
idk if it replicates 1.8 kb though, especially since that plugin itself is supported on 1.8

[brazen ether]

for example if you click ~20-40 cps on an entity while getting attacked, you take no knock back

[wind patio]

yeah, its not that easy to make a 1.8 pvp clone lol

[brazen ether]

like

this is not a thing in new versions

[half harness]

oh also this isn't directly pvp related (but it comes up in speedbridging) but didn't they change shifting as well on newer versions

the animation

[hushed prairie]

they did yes

there's a smooth animation instead of being instant

[brazen ether]

https://youtube.com/shorts/90fwCIFWvAg?si=0S7v7zg0_4l-wVYt

[half harness]

oh yeah, but tbh as someone who doesn't have a double clicking mouse, I remember this being a bit annoying lol

[brazen ether]

notice this guy literally not taking knockback at the start

while literally flying in the air

its bullshit like this that makes 1.8 hard to replicate

[hushed prairie]

that's because it's Techno in disguise

[brazen ether]

the change in animation also contributes to it feeling differently

and even if everything was perfectly replicated, people who don't believe that will have placebo

[hushed prairie]

that is true

[wind patio]

I believe its somewhat both, client can be authoritative if it doesn't exceed abnormal movement thresholds, so minecraft is using client authoritive + server authorative + server authorative with rewind

[brazen ether]

I don't play the game but ngl I wish we still got a combat update

like

[hushed prairie]

they were working on one

[brazen ether]

look at at hytale combat

[brazen ether]

several

and it got cancelled

[wind patio]

I believe the server auth rewind happens when you get a messages like "... moved wrongly" or whatever

[hushed prairie]

we had experimental snapshots around 1.14

yeah

[brazen ether]

maybe they're just scared to have another divide but man

[hushed prairie]

I'm still so sad it hasn't been looked into again

[hushed prairie]

I had heard mostly good comments on it tho

but yeah

[wind patio]

should be a toggle (gamerule?) tbh

[hushed prairie]

that too

[brazen ether]

ideally much more configurable than that but yeah..

I'm acting as if it's gonna happen anyway

[cold prawn]

cant u emulate it good enough on 1.9+ tho?

[cold prawn]

i disagree, i think its in a decent spot from a casual perspective

ive played mods that basically do hytale combat

and it just makes it less fun as its "easier"

[pastel imp]

did I just get back here to witness a war?

[brazen ether]

i don't know about that

i don't think that's a feature of the combat

just problems elsewhere

[quiet depot]

@oblique heath have you self hosted supabase before

[oblique heath]

never supabase, only pocketbase

[quiet depot]

damn

I think I got it 99% working now anyway

[oblique heath]

i have heard that selfhosting supabase is much more painful

[quiet depot]

it is very painful

like was trying to get it working with npm

just wasn't happening

then I got it working with npm

but there was no auth

so I was tryna get auth working

just wasn't happening

gave up on that and just shoved it all into a cloudflare tunnel

[oblique heath]

lmao

[quiet depot]

now I need to get cloudflare access auth working

it's not sending me an otp email even tho my email is on the allowed list

[oblique heath]

pocketbase also has ways to send emails but i have never made use of them so far

are you using an external email service or selfhosting everything

[quiet depot]

just self hosting supabase now

cloudflare isn't sending my email

[oblique heath]

oh i see hmm

[quiet depot]

like with cloudflare tunnels u can have an auth page infront of it

and it's showing

just doesn't send me an email when I click send code

[oblique heath]

i've heard people talking a lot about cloudflare tunnels but have never really looked into it, i took a quick look at the options available for that now and it's very overwhelming

good luck piggy maybe you just missed configuring the email in another one of those dozens of pages

[pastel imp]

This kids, is why you use appwrite

Way better UX for self hosting

And better overall

[oblique heath]

yeah single binary superiority

[agile galleon]

Isn't Pocketbase almost cross compatible with Supabase APIs?

I honestly don't think that self hosting Supabase is a grand idea haha

[quiet depot]

is alright

got it working

we chillin

[oblique heath]

no

supabase has a lot of features pocketbase does not afaik

though the basic crud stuff is probably similar

[cold prawn]

u can get auth on cloudlfare tunnels?

is it on their free tier too?

[agile galleon]

ye

Zero Trust

It's a bit annoying since you almost always need to verify with an email sent to your inbox

[quaint blaze]

dont they have an option to login with an auth provider. I've seen one that allowed you to login with GitHub

[quiet depot]

Yeah they do

loads of auth options

[agile galleon]

Ah nice didn't know that

[lavish notch]

cc: @agile galleon

Just encase yous are interested to know - I managed to hack an export by making a chrome extension to automatically scrape the data. Took 10-15 minutes due to attempting to avoid cloudflare bot detection

Ended up with a total of 94,130 lines of JSON data - now just gotta move it all to a suitable DB.

[cold prawn]

sounds good

i would do mongo as u dont have to mess with schemas

[lavish notch]

I honestly never really used Mongo and was going to go MariaDB since I don't plan on doing anything crazy with this data other than an automatic-ish (ish is because I'll be manually adding new purchases) discord verification system and basic management overview web dashboard.

[unkempt cloak]

https://tenor.com/view/mysql-gigachad-giga-chad-my-sql-avarage-gif-4178532470036852832

[cold prawn]

fair, but for loose stuff and for things that dont need hella verbose structure, i prefer it over sql anyday.

[agile galleon]

Awesome!

[lavish notch]

I've only just noticed that when I exported, I only got 15,688 out of the total 15,726 reported publicly. I don't suppose if anyone knows if SpigotMC has any sort of hidden purchases system?

I've ran my export twice and two different ways/versions of code which both produced the same output. And it's not like I'm missing a single page which holds 50 as adding 50 to the export brings the total of 15738 and I've more than likely captured the last page as the export count is a number that isn't an interval of 50.

[distant sun]

If you multiply the number of licenses per page by the number of pages what do you get?

All licenses should be there ig

[lavish notch]

15700 (50*314) which is odd... that must be total downloads isn't just purchases.

Gets even more interesting, just noticed SpigotMC is listing more than 1 purchase for the same user.

[distant sun]

Same date wth

Probably a refund / error during payment?

[lavish notch]

Who knows but this 'issue' doesn't help answer wth 'Total Downloads' tracks.

I original thought it was Total Purchases but now I don't know.

[distant sun]

Technically downloads should not be tied to purchases

[lavish notch]

I fully agree but SpigotMC is dumb.

Total Downloads for free plugins is Total Downloads but for paid, it's total purchases afaik.

but this export proves otherwise I hope

[distant sun]

Ig take what you have and check later for missing purchases manually

[lavish notch]

Hmm... random thought. Have you got ShopGUI+ on your account?

[distant sun]

Yes

[lavish notch]

Balls - I assume you bought it and weren't given for free?

[distant sun]

Ye

[lavish notch]

DAmn.

[distant sun]

I had a server a few years ago

[lavish notch]

I was thinking maybe I could try adding you, then you download and then I remove you to see if the Total Downloads number is effected.

[distant sun]

Doubtful the value updates instantly

[lavish notch]

Even then, I don't think BRC has revoked many, if any, licenses.

I guess I'll just hope and pray my export is successful/complete.

[crude cloud]

given this number ^ it sounds like your scrape is fine, it's just spigot being stupid

[nocturne apex]

from time to time there was a bug of being able to purchase 1 resource twice

so ig this is the case

2015 💀

[pastel imp]

this is why we don't use spigotmc kids

[nocturne apex]

well spigot was not updated since then xDD

[lavish notch]

Just thought I'd throw it out there - interesting theory from BRC, Total Downloads could possibly not be accounting for chargebacks and thus revoked licenses in that regard. 😄

[distant sun]

Most likely

[agile galleon]

Tsoding is peak

[smoky oasis]

who is that

[smoky oasis]

what are those

how to download paper?

[smoky oasis]

isnt spigot the best of the best?

[pastel imp]

I hope you are sincerly joking 🙂

[pastel imp]

databases

[smoky oasis]

no?? isnt it
1st: Spigot
2nd: Purpur
3rd: Paper

[pastel imp]

invert that list and you are right

Spigot hasn't been the best nor relevant since like 1.10 or so

For instance, paper and forks of paper, hold over 85% of the market

paper alone holds 68% across all versions. If you only look at latest versions, paper is dominating by a land slide with over 95% adoption

the only servers bumping up spigot numbers are servers that don't know any better, or old servers like 1.8, etc.

[smoky oasis]

leaf better than paper

[pastel imp]

scroll up and read about the scam it is

[smoky oasis]

what does it have

[pastel imp]

we already had that talk here (or in #minecraft), leaf is just another scam, stealing old and/or rejected and/or experimental patches from paper and other forks.

The only difference between leaf and USpigot is the fact that Leaf is free.

How misleading and scummy both are, are almost equal

[smoky oasis]

so they steal code from others

[pastel imp]

among others yes

[smoky oasis]

isnt that illegal- since the other proyects have licenses.

[pastel imp]

it is.

but there is no one enforcing the license

so, nothing rlly happens

well.. sometimes it does happen

USpigot already got in trouble cause of this (without giving more details)

so, there are SOME people enforcing their licenses

but still...

[crude cloud]

leaf is less of a scam and more of jus[ Yatopia all over again, it mixes every patch ever with no regard, broken untested code that breaks your server in unrecoverable ways

call it "bleeding edge" instead of "broken untested" and you're cool and better now

[potent nest]

I'll call myself bleeding edge now too

[brazen ether]

a better programmer than anyone in here will ever be

probably

[pastel imp]

Am I stupid for not knowing him?

At least not by name

[brazen ether]

no he's just a guy who streams his different programming sessions

and really humorous

notice most of these vids are lik 1h+

he's just a fairly competent guy all around

he wrote his own coroutines in C, amongst many many other things

[prisma wave]

he is but i think he ragebaits too hard sometimes

on twitter at least

[nocturne apex]

hey chat

don't let this channel die

[half harness]

it'll never be as dead as #developer-chat

-# only the OGs remember...

[nocturne apex]

what Abt

builders chat

[half harness]

😳

I forgot those existed as well lol

[nocturne apex]

there was a message like once a year

mostly it was my message

[nocturne apex]

a private channel too

[nocturne apex]

@halcyon leaf u use mixins or how do u handle the IDs of the keys while still being able to merge those items without any issues?

just curious, bcs ive been designing an anti dupe system before and have seen quite a bit of people use mixins for this

[lavish notch]

Never have I ever seen a more valid response from a plugin developer. 🤣

[desert horizon]

Well, some people are just weird and may be unable to read

[oblique star]

dubs to bencodez

he goes through a lot of idiots (including me), and provides a neat plugin

[brazen ether]

learning cpp in 2 days 💀

Zig had spoiled me

[potent nest]

C++ is impossible to learn

[brazen ether]

cpp feels like a clusterfuck in comparison

I need to learn it for physics internship

[potent nest]

You learn some subset of it for project A, and then you re-learn a whole other subset of it for project B

[brazen ether]

specifically rn bcs I wanna do a project and it being in cpp would be the best way to show that I am Okay at cpp

[brazen ether]

I'm literally just reading the entire learncpp book

im mostly done

it's just. a lot of bullshit

I need to write relatively fast code with it while not being just C because I want to actually learn it sort of properly

so can't bs it

[potent nest]

Just make everything constexpr

[brazen ether]

oh my god

I hate that shit

all of it

templating

Zig completely mogs cpp

in a way it's sad because all the fucking physics shit is written in like

fortran, cpp, python

i don't like ANY of those languages

[prisma wave]

Have you heard of RUST

[brazen ether]

I hope I get to use newer languages for this kind of stuff eventually yeah

though tbh not Rust 😭

Rust is a language for programmers

it's too complicated

[prisma wave]

as opposed to...?

[brazen ether]

scientists

[crude cloud]

the only language scientists use is python

and julia

and maybe R

and that's pretty much all there is

[brazen ether]

no they learn cpp really badly and do that too

it's quite incredible

[prisma wave]

wow

theyve gone too far

[brazen ether]

maybe a bit of Fortran also

but actually you're right

Rust is seeing use in this kind of stuff

so maybe

[prisma wave]

kill all rust users

[brazen ether]

😭

[crude cloud]

so true

[brazen ether]

better than cpp tho

[prisma wave]

yeah and world war 1 was better than world war 2

[brazen ether]

i would try to convince future employers to let me use Zig for part of the code but it would just make it unmaintainable to everyone else

[potent nest]

wait until you see world war 3

[brazen ether]

also a funny thing I've noticed is that in academia they call programs "codes"

like you say "I'm writing a simulation code"

it's kinda interesting

[crude cloud]

what if i'm blind

[pastel imp]

You'll feel it

[prisma wave]

rust 2

[agile galleon]

You forgot Matlab

[prisma wave]

matlab more like… fatlab

[oblique heath]

brister "1-indexed" matlab

[brazen ether]

cpp:

constexpr const int* const getXPtr() const { return &m_x; }

zig:

fn getValue(this: *const Testing) *const i32 { return &this.value; }

1. no constexpr needed because zig can determine comptime-ability by itself
2. explicit this making it clear that it requires a const pointer to itself (struct is called Testing), so it can't be modified; much more readable than a const
3. *const i32 is much more clear to me "pointer to const i32" than const int* but this is personal preference

zig mogs

[potent nest]

figuring out is one thing, but can you enforce that it is constant at compile time?

[brazen ether]

what

constexpr doesn't guarantee comptime execution either

so in either case it depends on use-site

zig is just able to do more at comptime with less hassle

[potent nest]

it does (unless you use a constexpr function in a non-constexpr context, but that makes sense)

[brazen ether]

yeah that's what I mean

that's the same thing as comptime

[crude cloud]

more characters = bad

[prisma wave]

elara =

getValue = value

best language

Those who know

[brazen ether]

I won't fall for this bait

[quiet depot]

IEvent

feels like it's 2014 again

I'm gonna go out on a limb and say this entire api is gonna be rewritten before public release

[hushed prairie]

PAPI Hytale?

[quiet depot]

ye

might aswell

[hushed prairie]

fair enough

[quiet depot]

the api's are so similar and papi has barely any hooks into actual minecraft

[hushed prairie]

I'm gonna have to finish my plugin faster

[quiet depot]

I reckon the luckperms expansion would work by simply changing the bukkit import to a hytale import

[nocturne apex]

jesus christ, havent seen that for years

[quiet depot]

yeah I mean

prefixing interfaces with I in java has never really been the go

but like it was more common over a decade ago

which I reckon there's a good chance is when this code was written lol

anyway cube already added hytale support to the ecloud so now I just gotta make papi work

and then change it all again when hytale most likely changes their api

hopefully they just hire all the talent from paper, sponge etc so we can have the most beautiful api ever created

or open source it

[nocturne apex]

on the other hand whys it actually bad?

[quiet depot]

it's not necessarily bad

do you mean the naming convention or the api

[nocturne apex]

ah ye in API it is quite strange imo

ye am not really using it

but was just curious whys it not a convention actually

[quiet depot]

I mean I don't really think it's bad per say

Like I can't think of any legitimate reasons why prefixing an interface with "I" would cause issues

it's just not common practice because people realised it's unnecessary

it's a really old practice, that's why I mentioned it

perhaps a remnant from the days pre modern IDE's that can tell you if something is an interface

and as for the api, I mean it's not really bad, it's just heavily spigot inspired, they've got a lot of play room to make a great api here so hopefully they do that

[nocturne apex]

ye we need to keep in mind that this is a progress of something that they have re-acquired 2mo ago

and they have mentioned that they reverted all riot changes

[wind patio]

I don't think it's necessarily an old practice, it came from microsoft, starting from COM, later in C++ before actual "interface" keyword became a thing and then got formalized for .NET and C#

but yes, its not common to prefix them in Java

also dont forget EMyEnum

[quiet depot]

ah yes

I think i used to do that

[wind patio]

personally I don't hate it, imo its more helpful to prefix it than not

[abstract osprey]

I think most devs are hoping so as well, the api is in such a bad state right now

[prisma wave]

nah if it's a good enough name it doesnt matter

it just makes autocomplete etc harder

[potent nest]

I use an I when I mean Implementation 😎

[crude cloud]

C++ doesn't have an interface keyword

I agree with BM that it doesn't really matter, or rather it shouldn't

I have used Interface suffix on certain types but to more so mean "this is a layer to" or "this connects with"

but that's entirely unrelated

[prisma wave]

yes

[wind patio]

yeah I didn't mean to imply it got it at some point

[crude cloud]

does it support &§x&§a&§b&§c&§d&§e&§f ?!?!?!?!?!

[wind patio]

when minimessage support?

when component support?

[quiet depot]

papi has component support now

sorta

it's in a dev build

[wind patio]

https://tenor.com/view/absolute-cinema-raccoon-absolute-cinema-gif-17862327649353748812

[quiet depot]

probs never gonna have minimessage support tho

seems out of scope?

[wind patio]

yeah I guess im just yapping to yap

https://media.discordapp.net/attachments/668996051583369228/1141424239774269451/cat-chomp-fireball.gif

[desert horizon]

Then you haven't checked JDA I suppose.

[woeful flame]

this looks funny
https://github.com/spygamingog/SpyCore/blob/main/src/main/java/com/spygamingog/spycore/managers/PacketManager.java

#showcase message

[wind patio]

what does that have to do with packets

[crude cloud]

the name

[pastel imp]

Lmao

Is that AI generated?

[fresh glade]

spy?? the name is very relaxing.

[woeful flame]

yeah that looks AI generated 😭

[pastel imp]

Anyone here has a good knowledge on llm models? Looking to find the cheapest yet still reliable model that would fit the following scenario: Gets a link to a mediawiki page, scrapes it and converts it into a specific json format for my use case. From what I researched, instruct models are pretty good for these.

[nocturne apex]

ask AI

xD

[prisma wave]

surely even a simple self hosted model would be fine for something this simple

[pastel imp]

lol

[pastel imp]

I am currently not self hosting anything sooo, not really worth it self hosting solely this

if I were to self host anything else then I guess yeah, might be worth it but even then

[oblique heath]

if it is just text reformatting then like bm said you can probably get away with a very weak model and get good results

you can start at like qwen3:1.7b and work your way up (idk if anyone even offers qwen3:1.7b but if they do it will be dirt cheap)

i have actually been considering creating a free faucet for qwen3:1.7b api requests so if you express enough interest i can set that up soon

[wind patio]

why do you need LLM for all of this, you can quite easily do the first part yourself, LLM can process the rest if you want to convert it to specific format

there are quite a few public solutions for wiki scrapers

https://github.com/openzim/mwoffliner
https://github.com/itkach/mwscrape
https://github.com/mediawiki-client-tools/mediawiki-dump-generator
https://www.mediawiki.org/wiki/Web_scraping_access

or at least that's how I would do it

[pastel imp]

interesting, I was leaning towards an LLM due to the layout/content of each page being very different from each other. Hence an LLM would "understand" (not rlly) what is what and how to split it. But I will 100% look into this!

I was also today years old when I found out you can just use ?action=raw at the end of a mediawiki page and it will give you the raw content lol (with mediawiki formatting)

[cinder flare]

MediaWiki truly is a great platform

[pastel imp]

is that with or without sarcasm?

[wind patio]

knowing Star I'd say without

[pastel imp]

eh ig

[cinder flare]

it is quite impressive that it powers virtually every wiki in the entire world and is completely free and open source!

and was written like 20 years ago in PHP lmao

[pastel imp]

yeah tbf that is indeed impressive

[wind patio]

@cinder flare did you see Laravel plugin now being integrated (I guess pre-installed) in phpstorm?

iirc it was paid before

[cinder flare]

yeah pretty nice tbh, honestly surprised it wasn't before, that was kind of an annoying barrier to entry

you did get it free with the student thing, but you had to do it separately, so it's nice that it's just included no extra steps now

[wind patio]

yeah I didnt have it before

its so nice now

didnt know it resolves inertia (frontend components?) stuff too

[cinder flare]

yeah it's kinda fucked up that Eloquent has like no intellisense without it lol, definitely one of my biggest gripes

[radiant schooner]

Hey I'm a bit confused where to properly asked to get access to a ecloud extension that's currently under Legacy. Could anyone tell me if this is the right spot or where I have to go?

[pallid gale]

Here works, let me know your username and link to each expansion with their ecloud.placeholderapi.com link

[radiant schooner]

It's SlyOtters and it's only one expansion: https://ecloud.placeholderapi.com/expansions/healthbar/

[pallid gale]

done

[fresh glade]

LLMs are horrible at understanding anything.

[pastel imp]

well, yes but this is also just patterns, but anyways, I most likely won't use LLMs after I found the raw stuff

[brazen ether]

they're not that horrible

you know you CAN have a nuanced opinion on AI right 😭

they're good enough at understanding things to be a great tool for learning difficult concepts via having it provide resources and answer questions that are hard to find answers for

[wary kestrel]

if it works ig ¯_(ツ)_/¯

[halcyon forge]

yet another vibe coder

gotta love em

[halcyon forge]

thing is, the class has no packets stuff at all

[halcyon forge]

right you're a vibe coder too

[woeful flame]

[nocturne apex]

well its meant to have NMS xD

however he does not even have NMS imported so

[halcyon forge]

yeah lol

[nocturne apex]

xDDD

[crude cloud]

why

[plain eagle]

[brazen ether]

writing proper async safe code feels nice

I think every serious developer should implement some basic concurrency safe structures by themselves just to get a better feel for what they're actually working with

[potent nest]

Nah it‘s more fun if you know there is a concurrency bug

[brazen ether]

still new to cpp but implemented parallelization for my simulation and it was nice to understand all of it, manually working with mutexes and condition variables

[brazen ether]

No.

[potent nest]

Fun fact: the complexity of the C++ memory model is partly because of Java

[brazen ether]

I haven't learned much about cpp memory model

everything I've dealt with so far had been really obvious

i also ended up not applying to the HPC internship I was looking at so I doubt I'll ever actually need to be crazy good at that kind of stuff

hopefully will get into their laser program tho...

[potent nest]

How would you write concurrency safe code without learning about the memory model

[brazen ether]

I mean just knowing how mutexes work and memory visibility in general is doing me fine

im not doing anything crazy complicated in cpp

[crude cloud]

absolutely yes

[crude cloud]

yeah that's what the memory model is about

[brazen ether]

I meant that I personally don't need to know the nuance of every little detail, probably

sorry should have been more clear

[quiet depot]

https://github.com/ichoco-milk/PlaceholderAPI/blob/0f35362a0c3d700d9fd7a6e115b9ebf6d92e3aa8/src/main/java/me/clip/placeholderapi/replacer/CharsReplacer.java#L73 might be the first time I've ever seen someone use named blocks

[potent nest]

I do this all the time

Well whenever I can’t be bothered to otherwise refactor my code to avoid labels

[crude cloud]

true and same also we have named blocks already they are called methods

[oblique heath]

damn they really added ruby to java

[hushed prairie]

I didn't even know this was a thing in Java 🥲

[potent nest]

How did you break from your ifs then before??

[crude cloud]

oh i just reenter the event loop from that point so it never returns

[hushed prairie]

I called the main function so it would continue and skip to after the if with a switch case from a string in the main's arguments

[forest pecan]

Named blocks are cursed

Only scenario I can think of is C code error handling when cleaning up stuff but like even then bruh

[nocturne apex]

how so?

[forest pecan]

Makes code hard to read arguably, also it can usually be structured better

[nocturne apex]

ah ye well usually

but it is useful for iterations

i haven't been using it in java, but i do sometimes use it in Kt

[forest pecan]

Yeah but in my like 7-8 years of Java I’ve never had to use it once

You shouldn’t really use it tbh

[nocturne apex]

I've used it like twice so xD

[prisma wave]

they are evil and awful

we have nice things now like "functions"

[forest pecan]

Don’t mention that word!

It’s a slur

[crude cloud]

I'll show you a slur if you show me yours

[forest pecan]

Another day, another paper-dev question that pisses me off

🙂

[potent nest]

If you don’t use labels, is your code even optimized???

[nocturne apex]

huh

[halcyon forge]

if you don't use a keyboard, are you even coding???

[brazen ether]

agent + voice AI now and some people might say you don't even need a keyboard

[potent nest]

In which case you aren’t coding

[plain eagle]

s3 download bucket down?

[wind patio]

might as well let the AI please their wife too at this point

[plain eagle]

Wrong server

[lavish notch]

Does anyone know how to best emulate SpigotMC's Purchase Resource transactions in Paypal Sandbox? For the life of me, I can't seem to get Webhook events to fire off of purchases, or even basic transactions, and I'm unsure what I'm doing wrong.

Like the account linked to the webhook notifications has received emails for payments completed but no webhook events got fired.

For reference, the webhook testing tool shows that my webhook listener works fine so I'm assuming it's something on Paypal's end that either I'm doing wrong or isn't working.

[wind patio]

Step 1 - don't use PayPal

most dogshit seller unfriendly payment provider

[nocturne apex]

u cannot use anything else on spigotmc

[pastel imp]

Step 2 - Don't use spigot

Most dogshit unfriendly platform provider

[wind patio]

I thought he was trying to emulate it, so he isnt using Spigot?

[lavish notch]

I am using SpigotMC, I'm attempting to make some sort of automatic verification system where by Paypal hits a laravel webhook receiver to input new purchase data into a database.

A small part of me thinks Webhooks aren't triggered for IPN transactions but even then, that doesn't explain why in-app transfers aren't sending webhooks.

[half harness]

out of curiosity, how does Hytale handle text colors in its API? (ex minecraft with legacy/components)

if anyone here has worked with it yet

since there doesn't appear to be any api documentation yet

[quaint blaze]

here is an example of formatting the chat with color - https://hytalemodding.dev/en/docs/guides/plugin/chat-formatting

[half harness]

lol TinyMessage
sounds awfully close to MiniMessage 🥲

[quaint blaze]

yeah lol. a library someone made inspired from that

[pastel imp]

wonder why they didn't go with minimessage

[half harness]

ig since it's bound to Adventure, which has minecraft-specific stuff like bossbars

it wouldn't really be viable to have the same library across two different games

[halcyon forge]

pfft voice AI? neuralink connected to AI agents is the NEW coding

[quiet depot]

been using these docs for hytale https://hytale-docs.pages.dev/

[plain eagle]

yeah

[eager fern]

Lmk if you figure it out we need it here also lmao

[eager fern]

Hot

[lavish notch]

Not my proudest moment but as a last resort, I quickly ChatGPT'ed a query and it came back saying Webhooks aren't fired for legacy (cant remember the proper name) transactions that utilise the IPN.

I also found a thread on Spigot where people were requesting a IPN forwarding feature from SpigotMC since Paypal only allows 1 IPN url. You can set the IPN url on your Paypal account however if one is set in the transaction call, like Spigot has, only the one in the transaction call is given the notification.

I believe this leaves me with the only option of polling Paypal API for recent transactions.

Still doesn't explain why my standard/modern transaction test weren't firing webhook events though.

[eager fern]

but we need spigot username

[lavish notch]

What do you mean?

[eager fern]

to sync spigot purchases

[lavish notch]

Just use the ID, no?

Paypal Transaction has the user ID.

SpigotMC Forums has an API to get username from ID.

My data for example: https://api.spigotmc.org/simple/0.2/index.php?action=getAuthor&id=310075

[eager fern]

rn we got a weird way but want funny to automate it

cuz when they aint online yk its weird

[lavish notch]

What I was planning on doing is bulk fetching purchases via a Chrome Extension (basically web scrape the buyers list) just to initially populate a purchase database I control (and thus can query however the fuck I want and not be limited by spigot, cloudflare, etc) and then use some sort of Paypal Integration to keep the database updated.

What's the weird way yous have got going currently?

[eager fern]

so from what I understand they export and import it manually

not automated

[lavish notch]

Interesting. I mean Paypal does have an API, so it is fully possible.

Also just found this which appears to be the same issue I have... https://stackoverflow.com/questions/71561151/does-paypal-rest-webhook-supports-instant-payments-notification

tl;dr rest webhook events not getting sent for any kind of payments

[lavish notch]

I don't suppose you want to hop in a VC and see if I'm being stupid or something with this Paypal Webhook Sandbox rubbish?

[eager fern]

busy atm

[static zealot]

Is GitHub down for anybody else? Can't create, see or close PRs

[potent nest]

https://www.githubstatus.com/

[desert horizon]

Would apreciate some help.

I want to convert my formatter expansion from its current structure, which is the normal src/main/... path, to a multi-module one, so that I can make a hytale version and MC version while using the same core formatting stuff.

But I'm not very experienced with Gradle (Kotlin DSL), so I'm not sure what necessary config steps I would need to take.

Current build.gradle.kts is this one:
https://codeberg.org/Andre601/Formatter-Expansion/src/branch/master/build.gradle.kts

[pastel imp]

Matt knows these things xd

[cobalt marlin]

you just want one build.gradle.kts on the root project and one on each module

depend on the stuff that you need to depend on on the module build files

[static zealot]

You'll need:

1. Split project into 3 components: core, minecraft, hytale
2. Have 1 main build.gradle.kts file and a settings.gradle.kts file. In the settings file make sure to declare the projects. For example include("core")
3. Have a build.gradle.kts file in each module. You can then declare subprojects as dependencies using implementation(project(":core"))

[cobalt marlin]

and apply the plugins on the root one

you also need to include the modules on the settings.gradle.kts file

[crude cloud]

include rabbitmq

https://tenor.com/view/кролик-gif-7585471000370564066

[desert horizon]

Okay... I've made some main changes now to the expansion.
Still not sure what the individual build.gradle.kts files should look like, so apreciate some help there (Also apreciate any feedback on current setup, as I'm sure it can be improved). The current progress is found in this PR:
Andre601/Formatter-Expansion#15

[halcyon forge]

highkey nice

[quiet depot]

ye they're the most thorough I could find at the time

might be better ones now but when I was looking a couple weeks ago that was the best one

[crude cloud]

https://media.discordapp.net/attachments/1432185066343235614/1470706021617106995/abudlq.jpg?ex=698c4537&is=698af3b7&hm=d8aee1aa920a75ad01fb546b4f89de4de8399b5ffa8f5fbd427f9618cc8f6b8f&

[quiet depot]

@compact perch explain the meme please https://cdn.discordapp.com/attachments/1432185066343235614/1470706021617106995/abudlq.jpg?ex=698c4537&is=698af3b7&hm=d8aee1aa920a75ad01fb546b4f89de4de8399b5ffa8f5fbd427f9618cc8f6b8f

@compact perch explain the meme please https://cdn.discordapp.com/attachments/1432185066343235614/1470706021617106995/abudlq.jpg?ex=698c4537&is=698af3b7&hm=d8aee1aa920a75ad01fb546b4f89de4de8399b5ffa8f5fbd427f9618cc8f6b8f

useless clanker

[pallid gale]

@compact perch ?

[compact perch]

ah, the classic '?'. you gonna elaborate, or is that the whole question, funnycube? because i'm gonna need more than an existential crisis to help you out.

[crude cloud]

classic '?'

[quiet depot]

hardly a classic at all if you ask me

[pallid gale]

@compact perch explain the meme please https://cdn.discordapp.com/attachments/1432185066343235614/1470706021617106995/abudlq.jpg?ex=698c4537&is=698af3b7&hm=d8aee1aa920a75ad01fb546b4f89de4de8399b5ffa8f5fbd427f9618cc8f6b8f

[compact perch]

#dev-general for meme explanations now? wild. i'd ask if you tried googling it, but i don't want to overload your system. next you'll be asking us to explain why your nullpointerexception isn't a feature.

[quiet depot]

@compact perch we did try googling it

literally racist

[pallid gale]

feelsbadman

u can probs still use the context one

[quiet depot]

nah

I don't wanna

fuck em

[pallid gale]

woah

[quiet depot]

barry is dead to me

bringing back gary rn

[crude cloud]

OMG GARU

[pallid gale]

need to bring back the papi bot

[crude cloud]

Gary

[quiet depot]

gary will usurp barry

[crude cloud]

wasn't there also Larry and Terry what happened to those bots

sobbing

[quiet depot]

there's literally every arry

[crude cloud]

maybe Barry needs a Mary

[quiet depot]

2 sec

ye

https://helpch.at/bots/Marry.png

https://helpch.at/bots/Tarry.png

[pallid gale]

mt

[crude cloud]

yeah but Xarry doesn't mean anything to me

[quiet depot]

o

[nocturne apex]

me likey Larry

xD

[pallid gale]

NFTarry is where it's at

[desert horizon]

The B in Barry stands for B[censored]

Somehow gives me an idea for an april fools joke on this server: Anti-HelpChat

[desert horizon]

Also, still would apreciate some feedback on this, as I want to publish a hytale version of the formatter extension

[wind patio]

code reviews aint free bro

[pastel imp]

fr

pay me 20 nugget box at mcdo

and we have a deal

[desert horizon]

your payment will be the satisfaction that you could share your knowledge in a review to improve a project.

Tho, for real, I'm not familiar with gradle kts, so anything to improve it and making it work as a multi-module project would really help.

[wind patio]

I mean, build.gradle and build.gradle.kts are pretty much the same thing, just different lang

wdym by "making it work as a multi-module project"?

as in, what is not working?

[prisma wave]

kts has some more idiomatic usage patterns i think

[wind patio]

yeah well conceptually and I guess functionally it's the same thing

but they do have some differences

[crude cloud]

ohhh do we get to do the meme
rust is the same as python, they are conceptually different but are functionally the same thing

[wind patio]

https://tenor.com/view/buh-b-u-h-cat-silly-кот-gif-2242355352608793340

[potent nest]

my turing machine is just like rust

[prisma wave]

i mean unironically sort of

aside from lifetime semantics and stuff it would be fairly* easy to make an automatic translator between the two

[desert horizon]

I have no idea how to configure it to shade in core module and build the different jars

[quiet depot]

@desert horizon u need to have a subprojects or allprojects block in ur root build.gradle to apply plugins to the subprojects

in there u can also have repos/deps that each module uses

any common plugin configurati0ons

[abstract osprey]

buildSrc my beloved

[quiet depot]

https://github.com/PlaceholderAPI/PlaceholderAPI3/tree/master here's an old example

what even is buildSrc

I've seen that a bunch

[abstract osprey]

it's a nicer way of doing the subprojects bs

https://github.com/GeyserMC/Geyser/tree/master/build-logic
example ^

and then in your actual module build.gradle(.kts) you'd be doing id("geyser.publish-conventions") or geyser.publish-conventions in your plugins block

it's way nicer than the subprojects stuff imo

as for the actual docs, https://docs.gradle.org/current/userguide/sharing_build_logic_between_subprojects.html

[ocean quartz]

buildSrc is great but I love build-logic

[abstract osprey]

naming preferences

[ocean quartz]

They are different

[abstract osprey]

fucking what

are they actually

oh you're right

icic

literally everything i've been doing should be build-logic then

not buildSrc

[ocean quartz]

Yeah lol

[abstract osprey]

if it works it works

[ocean quartz]

But yeah buildSrc can make build slower because changing anything triggers a full rebuild of everything, while build-logic being a separate project only re-builds the parts that are needed
Plus since it is a completely separate project from the root it is very powerful and you can basically write full plugins inside of it (not just convention plugins but custom tasks, extensions etc)

i know you can already do custom tasks and extensions with just a normal build but not shared, and a bit more limited

[desert horizon]

Thanks.

Tho, still a lot confused on this entire thing.

Right now I have this:

• Root build.gradle.kts: https://paste.helpch.at/esuvupawak.kts
• Core build.gradle.kts: https://paste.helpch.at/yujubesusa.kts
• Hytale build.gradle.kts: https://paste.helpch.at/atofozoyih.kts
• MC build.gradle.kts: https://paste.helpch.at/yipomaziqo.kts

And right now, executing the shadowJar task creates Formatter-unspecified-<version>.jar files.

which is kinda weird

[crude cloud]

you should leave the archiveBaseName as just Formatter

the version is a separate component from the final archive name and is set to the project's version by default

[desert horizon]

That fixes the unspecified thing

Now I just need to get the Hytale version tested to see if it works.

[abstract osprey]

Oh shit, that's game changing

buildSrc rebuilding every 3 nanoseconds gets on my nerves

[brazen ether]

strong typedefs are a big missing issue in java

say you have:
map: Map<UUID, YourData>
and you are in a project where you have multiple things which have uuid's. it is absolutely not immediately obvious what the UUID here actually is supposed to me. so you might make a choice like:
controllerUUID2Data: ...
but this is obviously annoying. the thing is that the UUID is really only meant to be something like ControllerUUID. thus the ideal solution is for a strong typedef to be created, which would solve this issue

you cannot do this in java without wrappers

[potent nest]

How would you solve that without wrappers

Like there are languages that have aliases, but that isn’t a solution to that really

[brazen ether]

strong typedefs

i just said

[brazen ether]

kotlin value classes

it's just a compiler enforceable thing

having them be separate types at compile time has no performance overhead

it's not an alias

because alias implies it's the same type, which this is specifically not (at compile time)

[potent nest]

How‘s the Java interop story if there is no overhead?

[brazen ether]

it just compiles to the underlying type

so it's a compiletime only thing

oh

icwym

it generates wrapper classes for java interop

[cinder flare]

kotlin stays winning, truly

every time i have to put five lombok annotations on a java class, i cry a little inside

[brazen ether]

i'm not 100% sure because this hasn't been my usecase though

it does generate the class but doesn't use it in the kotlin code, so you don't get the perf benefits from java

[potent nest]

Seems like the wrapper is even used by Kotlin directly sometimes

So not zero overhead

[brazen ether]

source?

[potent nest]

Another half-baked Kotlin feature

[brazen ether]

it's probably only used when boxing primitives

in generics

which it doesn't do when not needed btw

so you can have strongly typed indices

[potent nest]

https://kotlinlang.org/docs/inline-classes.html#representation

[brazen ether]

which is incredibly useful

[brazen ether]

generics

[cinder flare]

isn't Java literally adding exactly the same thing with Valhalla?

[brazen ether]

that's 0 overhead for classes

[cinder flare]

they just have the ability to actually add stuff to the runtime lol

[brazen ether]

no it's not actually

java value classes don't inherently lead to the same thing

it would only be such if it inlined the reference itself, which isn't how it would work i don't think?

[potent nest]

No, Java value classes are a more general concept

[potent nest]

But the example you showed before was using generics too?

[brazen ether]

overhead because Int is a primitive

that's the only reason it's boxed in the kotlin example

let me just check this right now

DAMN

you're right 😭

it's still useful for methods and fields in general but yeah damn

i didn't know that it actually still used the wrapper even for non primitives in generics

[potent nest]

I didn’t take a closer look but it has to if stuff like is is supposed to work at run time

[brazen ether]

Yeah

i used it only for a data structure in code before to have different types for indices

which is where it really shown i guess because no generics and only primitives

damn

):

kotlin takes an L here

though tbh it's just because of how JVM works

[cinder flare]

i mean there's only so much they can do yeah

[brazen ether]

yeah i mean anything that stores type information like jvm does can't get around this

still better than nothing but it isn't much better than a java wrapper in usecases like this then

[cinder flare]

idk, i've grown so used to stuff like this, the performance overhead of these little wrapper classes is like basically nothing in most scenarios

even if you save five microseconds and 16 bytes of memory, your backend will still be waiting like 500ms for the database I/O yk

[desert horizon]

I don't really get what you talk about tbh. Could you elaborate?

[prisma wave]

The UUID as a type doesn’t provide any context about the situations it’s used in

Languages like rust, kotlin, and Haskell (and I suppose Java with the value classes) solve this by letting you define wrapper types that have no runtime overhead but have to be explicitly boxed / unboxed

eg you could write value class PlayerID(UUID) and then have Map<PlayerID, UserData> and it becomes a lot more obvious what the uuid represents

It also means you can’t accidentally pass arguments in the wrong order (eg if you had a source uuid and target uuid), because they’re different types and so you get a type error

[nocturne apex]

we are using the value classes for this at my job

but yeh in Kt

[nocturne apex]

oh yeye exactly

[cerulean ibex]

we have this in java

[brazen ether]

no we don't

not without wrappers and a runtime performance cost

[cerulean ibex]

implemented with user code even

[brazen ether]

though as sir ywell pointed out it's not possible without that in any language like java

while it's possible in something like zig because no types at runtime

[cerulean ibex]

and no boxing since you're just using the real type

https://checkerframework.org/manual/#subtyping-checker

[brazen ether]

okay you can do anything with external programs

[cerulean ibex]

doesnt that mean it's better?

[brazen ether]

erm

how

you can do anything in any language like that

well not anything

[cerulean ibex]

because it doesnt require the authors to maintain it?

[cerulean ibex]

wym?

most languages dont support type annotations like that

[brazen ether]

I think they should

oh sorry I misunderstood

[brazen ether]

I don't see how this is a benefit

someone else has to maintain it

then the language has no extra ability to add features that work with that added on feature

it's also an extra program that users have to become familiar aside from the programming language itself

[cerulean ibex]

more competition, more options, generally faster development

[brazen ether]

though that's not that big a deal

[brazen ether]

that only makes sense here because its an analyzer with no language integration

[cerulean ibex]

yeah maybe the implementation could be better if it was more integrated into the language

[brazen ether]

making it inherently more limited

[brazen ether]

I was talking about improving the language / problems with it

so talking about external programs is missing the point

[cerulean ibex]

but checkerfrramework is still great

is it really that big of an issue if its solved though?

like its not that important who solves it

[brazen ether]

.

and it's not enforced

so it's definitely overall much worse than if it was builtin

and you cannot add specific language features which take advantage

[cerulean ibex]

users have to become familiar with it regardless of who implements it

if its a language feature you still have to learn it

[brazen ether]

no?

because they can simply not use it

if it's not a language feature

your library authors probably don't use it

so you won't get the benefit of them doing such

like it doesn't make Java better

because it's not like analyzers are some Java specific thing

hm actually that makes me think about the difference between Jai then

though I guess that mostly lies in ease of use and default integration

I guess I think that it's a feature that should be builtij because it would cause people to use it much more

[brazen ether]

also I don't fully buy this argument

3rd party software tends to not be very good

just on average

and has no guarantees to keep up with changes

but I haven't used the framework you're talking about so that doesn't refute your specific stance

[potent nest]

if you set it up correctly, your code does not compile (i.e., the compiler produces an error on type mismatches)

you can also provide annotations for libraries separately, so you can benefit there too

the benefit of the checker framework is that you can add exactly the checkers you want, so you're flexible without being annoyed by something you don't care about

type annotations are indeed a java feature and not necessarily equally available in other languages, so pluggable type systems might not be as easy in such case

[cerulean ibex]

maybe im not understanding what ur saying but you dont have to use it if its a language feature of checker framework?

[cerulean ibex]

yeah ig

these sorts of things usually only matter for domain-specific types though so i dont see it as a huge issue anyway

and the jdk is annotated with cf so you get that too

[cerulean ibex]

yeah i can agree

[potent nest]

well it's annotated rather partially

[cerulean ibex]

but it really depends

[cerulean ibex]

oh

it seemed quite good in my use

my only issues were jetbrains annotation fucking everything up since they target too many things

[potent nest]

regarding nullness it's probably fine

most of the other checkers aren't as well-maintained though

[nocturne apex]

boys have u seen the difference in j25 vs j26?

Ive seen some reports from owners who are already running on j26 rc and they are saying that its abt 10-15% faster

Java is kind of balling last few years

[potent nest]

these numbers sound rather odd

in that case you'd clearly be able to tell where that difference comes from

[oblique star]

pepe_hmm

[nocturne apex]

imo its pure computing perf

not like 10-15% on MC server itself

[potent nest]

I mean in specific microbenchmarks, fair

but in any normal program such change would be crazy

unless they e.g. tried the valhalla build rather than normal java 26

[nocturne apex]

ik that they have made some cool changes tho

and it was known that in j21 there was like 20-30% computing difference with j24

at least from what Ive been reading

[crude cloud]

where are you pulling these numbers out of 💀

[nocturne apex]

https://www.youtube.com/watch?v=renTMvh51iM&t=158s

[pastel imp]

well that's 21 vs 25 not 25 vs 26 or 21 vs 24 xD

[nocturne apex]

.

okay it was 15-20%, but still

such difference is huge

even in benchmarks

[crude cloud]

that is highly specific to the architecture of the application in question

[nocturne apex]

yeye ofc

[crude cloud]

not all of java ever lol

and a very specific test too

[nocturne apex]

Ive been watching multiple of such vids and they were all mentioning such improvement

there are no actual benchmarks w j26 yet as its still not officially released

there are just RCs

[pastel imp]

tbf jvm is already fast, I'd assume the problem of mc servers is not java itself lol

[nocturne apex]

yea it really is

[pastel imp]

how so

[nocturne apex]

idk why people still think its super slow

[pastel imp]

oh okay

you agreed lmao

confusing

[nocturne apex]

sorry xD

[pastel imp]

but yeah, doubt problem is java itself

mc servers are just... old.

the code probably has a ton of tech debt

[nocturne apex]

Ive been also checking some like leaderboards of language performance - like computing performance again

[pastel imp]

lmao

[nocturne apex]

and java was usually like 4th?

regarding most used languages

[pastel imp]

yeah

[nocturne apex]

which is crazy

[pastel imp]

compiled languages be like

[nocturne apex]

Ive been also reading that there's really no reason of doing MLs in Py, etc. its just that Py has all the libs that Java does not

but while someone uses it seriously, then hes mostly running it on JVMs

[pastel imp]

my old high school teacher was a huge fan of python, so I always teased him with this xD

[nocturne apex]

yea xD

[pastel imp]

yeah I really wish we had more llm stuff in java/kotlin

I'd love to mess around with it but there is barely any libs

and the libs that exist are a headache to work with lol

[nocturne apex]

Koog?

am using Koog for LLM

not for fine-tuning obv

but for like AI Agents

its really cool

and really complex xD

but for ML I havent found much

[pastel imp]

first time I am hearing about it lol

but well that's not what I meant tbh

I meant for training and running custom models

not a wrapper for other models

[nocturne apex]

ah yea

Ive tried using LoRA for this

[pastel imp]

what happened to training models via cpus.... good old times

[nocturne apex]

but yeh it was Py

[pastel imp]

reminds me of https://fasttext.cc/

text classification lib by facebook

running via cpus

but it got discontinued I think

[nocturne apex]

AI took it over

ye its archived for almost 2 yrs

[pastel imp]

yeah...

it's pretty sad

[nocturne apex]

whats becoming from AI is sad lol

AI is a cool thing and pretty useful in LOADS of cases, but many people are just using it wrong

and its being overused imo

everyones trying to implement it absolutely everywhere

for no reason

[still portal]

openclawd on the mac mini 😍 💀

[fresh glade]

If I'm doing some type of docs, should I use Starlight or Vitepress. I was looking for one that lets me tweak it a lot.

no answer? 🥺

[cinder flare]

it's mostly about what platform you prefer/are used to. If you're doing Vue, VitePress is gonna be awesome for you (I prefer it). If you like Astro components more, Starlight is based on Astro, so that'll probably do you better. As far as I know, you can basically completely customize both of them pretty easily, but I've only really had expeirence with VitePress.

[pastel imp]

https://youtu.be/UuQQVY9jxSs @nocturne apex

[nocturne apex]

👀

what is that BS?!

[pastel imp]

Xd

I am just the messenger

[potent nest]

I mean it's not exactly wrong

[brazen ether]

im gonna read up on the new witness thing

supposed to be something like Rust traits

[potent nest]

it's still a very rough design

[prisma wave]

what is this

[potent nest]

operator overloading for java

at least that's kind of the goal

[brazen ether]

but only for numeric types

[nocturne apex]

huh

[prisma wave]

wow

that sounds pointless

link?

[brazen ether]

it was stated in a talk

witnesses are basically a way for external code to allow other external code to fit an interface

literally rust traits but a little worse

operator overloading they stated will be "if you want it at all, you need it ALL" meaning you will need to implement every numeric operation, as a way to have it only be used for numeric types

idk if they're still going to do that

I'll find the video later

[brazen ether]

tbh I don't really like this

I just want them to add operator overloading

[prisma wave]

when you say "numeric types" does that mean a type that behaves like a number or specifically int/float/etc

[brazen ether]

I work with a lot of math and scientific stuff where it would be incredibly helpful

yes

oh

the former

[prisma wave]

wank!

[crude cloud]

the idea is that you can define your own "numeric" types, "integral", "decimal", "orderable" (for < <= > >=) etc, you provide a "witness" for it which determines exactly how + (and other operators) behave, it's pretty much just a method that takes lhs rhs and you return the new value, but it's not directly a method on the type, just done through type classes

but yeah it's a very rough design and more so just playing around with what an implementation of that would look like

rather than actually setting anything in stone

tho once value classes do land, it will eventually be nice to have those too, if i have my own Float128, Vec3i etc types

it does not only apply to custom operators, but you could also provide an identity value for your type

[nocturne apex]

yea lol

[prisma wave]

yeah i mean this sounds nice but the fact that it doesnt generalise very well is kind of annoying

but i guess i should see it before judging

[prisma wave]

atp just add proper type classes you know

[ocean quartz]

Do you have the link? I love watching that stuff

If anyone else enjoys watching talks on language design stuff, this one is a really fun one
https://youtu.be/P0mo8Tzrapo

[crude cloud]

well, this is step one, and they are testing the ground

just like they did with pattern matching, first it was very limited in scope and application, then we got record patterns, then pattern matching over switches, etc

[crude cloud]

https://youtu.be/Gz7Or9C0TpM?si=hiL3Ecst7LHsUBV7

[brazen ether]

55:30 for what I mentioned about operator overloading

[ocean quartz]

Tyty

[cinder flare]

that does sound really nice for a lot of bigdecimal stuff

it gets real annoying nesting Math.addDoublesPrecisely(lhs, rhs) a million times lol

[forest pecan]

structured bindings in Java when

When can we get for ((key, value) : map.entrySet())

instead of this whole Map.Entry<Key, Value> entry 😭

[cinder flare]

mmm Kotlin

[forest pecan]

i was originally thinking C++, but i realied kotlin has them too

[potent nest]

That’s more or less the plan with carrier classes

[cinder flare]

i think like most languages have destructuring lol

javascript, python, kotlin, C#, C++

[potent nest]

And some languages have actually powerful pattern matching, like Haskell or Scala (the homework Kotlin unsuccessfully copied)

[prisma wave]

var?

[prisma wave]

truth nuke

[prisma wave]

ok it seems i was being overly harsh on this

kinda funny that this is literally just type classes

also intrigued how they plan to do the collection literal witnesses without adding higher kinded polymorphism

[crude cloud]

well, what he talked about in that keynote was largely daydreaming of what one day could be, there are existing works and greater concerns in java's language space currently, but it would allow for some very funny things

I trust Brian knows what he's doing and saying, after all language design is his daily job and it goes far beyond "hehe syntax"

I would argue, while the syntax of a language gives the first impressions, it does not live among the most important aspects of language design

By far

[ocean quartz]

Pretty sure he is talking about destructuring, to not need to do entry.getKey() / entry.getValue(), sure var would make the line shorter but wouldn't change the issue

[nocturne apex]

yeh var does not really destructure it

I thought java already has something for that

[prisma wave]

ah yes

no i dont think so

[nocturne apex]

Ooof

[prisma wave]

the best in the short term is just a record pattern

[nocturne apex]

I got used to Kt too much

how would record pattern solve this?

[prisma wave]

switch (entry) {
  case Entry(var key, var val) -> ...
}

but this relies on Entry being a record, which it is not

[nocturne apex]

ah