help-development

[wet breach]

they could have just done a profile dump as well

[pine forge]

i wanna get remote going though

[wet breach]

why?

[pine forge]

because im always using my remote server so others can join and it can stay running

theres my configuration

im not gonna start up a local server every time i want to debug

[wet breach]

don't need to

a profile dump can be ran remotely

[pine forge]

whats that

profile dump

okay its liostening now

i see in the console

still timeout connection its my tunnel then

[wet breach]

Here is a tool that should help you in whatever it is you are wanting

[pine forge]

never works

[wet breach]

https://visualvm.github.io/

[pine forge]

whats aprofile dump

[wet breach]

it dumps whatever the server was doing into a file

you can play them back sort of

or just look to see what was or was not executing

[pine forge]

so how do i use this

[wet breach]

https://visualvm.github.io/documentation.html

could just click the link at top for the documentation

[pine forge]

okay got remote debug working

so it does not look like intellij can map the codepoints to my local source code

i cant see what code its executing

[wet breach]

well we have no idea what you are looking at

[pine forge]

not all of it atleast

i set a breakpoint at dispatchCommand

[wet breach]

well some of the code steps should go through your plugin if you are executing a command

which won't be in the server source

since well its your plugin

[thorn isle]

he's trying to get the server to run an arbitrary command and capture the command output by passing it a spoofed command sender, which is supposed to log anything it receives

so he needs to see what's happening inside the nms/cratbukkit parts of the command execution, specifically whether either of them unwraps the commandsender somehow

[wet breach]

I would imagine it would need to?

[thorn isle]

not for bukkit commands i don't think

vanilla commands get the unwrapped nms command sender

either way you'll probably want paper userdev's mojang-mapped sources, and ideally those for whatever build your server is on, so all the line numbers and other things align

[wet breach]

yeah that would be handy

especially if you are new at doing this lol

[thorn isle]

the remote server doesn't exactly send your debugger the classes you're stepping through, just the class name and line numbers, so shit goes whack when you're looking at some mangled decompiled class in your ide

[pine forge]

i can just look at the source code myself

https://hub.spigotmc.org/stash/projects/SPIGOT/repos/craftbukkit/browse/src/main/java/org/bukkit/craftbukkit/CraftServer.java

[pine forge]

how do i get these

[thorn isle]

https://docs.papermc.io/paper/dev/userdev/

basically instead of a specific .jar artifact , you depend on paperweight.paperDevBundle("1.21.11-R0.1-SNAPSHOT"), which adds a bunch of jar artifacts, one of them being the mojang-mapped server and another its sources

the downside is that it requires gradle which i would like fucking use apache ant over, but it is what it is

[pine forge]

i use gradle so it sfine

[thorn isle]

or if you're building against/using spigot, you do the whatever buildtools nonsense, but i don't remember how that works

[pine forge]

do they not have 1.12.2?

[thorn isle]

i doubt it

[pine forge]

okay

whatever

[slender elbow]

mojang mappings did not exist back then

[pine forge]

should be the version of my minecraft server anyways

right

[thorn isle]

glhf basically

[slender elbow]

if you are compiling against and running paper, there is api for a faux feedback CommandSender, but that also did not exist in 1.12 lol

[thorn isle]

oh neat, that's new to me

what's it called

i might have to do this again later and i'd rather not nms into command blocks a second time

[slender elbow]

Bukkit/Server#createCommandSender i believe

[thorn isle]

neat

[pine forge]

https://hub.spigotmc.org/stash/projects/SPIGOT/repos/craftbukkit/browse/src/main/java/org/bukkit/craftbukkit/CraftServer.java#902
https://hub.spigotmc.org/stash/projects/SPIGOT/repos/bukkit/browse/src/main/java/org/bukkit/command/SimpleCommandMap.java#133
Dont really get it, it seems to only run plugincommands

guess i need to debug

ah i dont think paper dev thing works with my 1.12.2 setup

io.papermc.paperweight.PaperweightException: Exception executing remapMinecraft

org.gradle.api.tasks.TaskExecutionException: Execution failed for task ':1.12.2-spigot:paperweightUserdevSetup'.
at org.gradle.api.internal.tasks.execution.ExecuteActionsTaskExecuter.lambda$executeIfValid$1(ExecuteActionsTaskExecuter.java:135)
at org.gradle.internal.Try$Failure.ifSuccessfulOrElse(Try.java:288)
at org.gradle.api.internal.tasks.execution.ExecuteActionsTaskExecuter.executeIfValid(ExecuteActionsTaskExecuter.java:133)
at org.gradle.api.internal.tasks.execution.ExecuteActionsTaskExecuter.execute(ExecuteActionsTaskExecuter.java:121)
at org.gradle.api.internal.tasks.execution.ProblemsTaskPathTrackingTaskExecuter.execute(ProblemsTaskPathTrackingTaskExecuter.java:41)
at org.gradle.api.internal.tasks.execution.FinalizePropertiesTaskExecuter.execute(FinalizePropertiesTaskExecuter.java:46)
at org.gradle.api.internal.tasks.execution.ResolveTaskExecutionModeExecuter.execute(ResolveTaskExecutionModeExecuter.java:51)
at org.gradle.api.internal.tasks.execution.SkipTaskWithNoActionsExecuter.execute(SkipTaskWithNoActionsExecuter.java:57)

i am doing this though
paperweight.paperDevBundle("1.21.10-R0.1-SNAPSHOT")

[lilac dagger]

Remapping goes brr

Spigot doesn't have runtime remapping

But also paper shouldn't have runtime remapping on 1.12

[vagrant stratus]

Gotta love using older versions of things 😎

[pine forge]

isnt there a similar thing for just craftbukkit stuff

[young knoll]

Crazy how we went from no mappings in 1.12 to no obfuscation in 26.1

[dreamy glade]

Do most players prefer to play the game with or without the "Force Unicode Font" set to true?

[slender elbow]

i doubt most players even know that setting exists

[chrome beacon]

^

[thorn isle]

i do see a decent number of players using fonts from their resourcepacks, but i don't remember seeing many using the unicode font

i'd say it's about as popular as using pirate speak as your language

[slender elbow]

lmao

[young knoll]

I know it exists!

I don't really know what forcing unicode font does though

[slender elbow]

wanna find out?

[dreamy glade]

It forces the game using specific font that support a lot of characters

[young knoll]

Does it just disable the default font and make everything use unifont

[chrome beacon]

yeah

[slender elbow]

one could say that it forces unicode font, even

[young knoll]

i'll unicode your font

[slender elbow]

go ahead

[lilac dagger]

I use that setting, i like the small font more

[ancient plank]

meow

[lilac dagger]

I can't wait to have just one nms module going forward

Hopefully it works for future versions too

[chrome beacon]

I mean breaking changes will still happen

[lilac dagger]

That's sad

[dreamy glade]

I like both, but I'm trying to make a resource pack with icons and stuff, but the icon is displayed very inconsistently between the two options. I have to find out which one is the majority and optimize the resource pack for them.

[young knoll]

MD on his way to continue relocating Craftbukkit per version

[lilac dagger]

It'd be better if he will leave it up to the devs to update and check

Ewch new drop i'll add it to a stable version and then if something is untested the players will know

[young knoll]

Mojang does like refactoring things

[lilac dagger]

True but think of all the saved nms modules

[thorn isle]

ever since paper rolled out mojang mapped nms plugins and server, maintaining nms plugins has been a breeze

most of the time they work without building, build without changes, or need only some minor refactorings

i find that the bukkit api breaks more often than nms now, with sound names and potion effects and shit being changed

[distant cosmos]

when should I depend on netty for my plugins?

[lilac dagger]

When you need to sniff or modify packets

[pine forge]

okay i will stick with console logging, this does not seem to work

looking at the bukkit code, consolecommandsender is converted to getServer().createCommandSourceStack with no reference to the sender methods

and my blockcommandsender cant be cast to craftblockcommandsender

id have to use reflection or something

theres also ProxiedCommandSender which is cast to ProxiedNativeCommandSender though which i cant create

[lilac dagger]

Not sure if it's supposed to use that way

There's player precommand event

I'm not sure what you wanna do tho

[pine forge]

i wanna get command output

[lilac dagger]

Yeah playerprecommand event

You get all the args and label

Ah console

Hmm

[pine forge]

its weird though that consolecommandsender doesnt even seem to capture plugincommands

[lilac dagger]

It depends on server fork you're using

Some versions of paper for example didn't hook into simplecommandmap anymore

Later they added a delegate map for registering commands into brigadier anyway

[pine forge]

ah yes

okay

that makes sense

return new Handler() {
    @Override
    public void publish(LogRecord record) {
        if(record == null) return;
        String message = record.getMessage();
        if(message == null || message.isEmpty()) return;

        if(output.length() > 0) output.append('\n');
        output.append(message);
    }

    @Override
    public void flush() {}

    @Override
    public void close() {}
};
``` using this from JUL now

[lilac dagger]

Yeah probability the best

Capture the logs

[pine forge]

worked well before

Bukkit.getLogger().addHandler(captureHandler);
used log4j before though i believe

this is new

lets test

[lilac dagger]

Might be cool to have a consoleprecommand event

Altho it might be misused

[pine forge]

might be cool to just fake a command sender like in newer versions

i think thats the best way

[lilac dagger]

You might be able to do it still on even older versions

I think paper did the change since 1.20.6 to 1.21.4

Also did you check if brigadier has logging options?

Might be less hacky than fake console senders

[pine forge]

you mean with reflection?

well i want to support from 1.12+

[pine forge]

well you can create your own command source stack

that would work

[lilac dagger]

With Nms modules

Ah 1.12 didn't had brigadier tho

Or at least I know of

[pine forge]

okay the handler doesnt capture anything either

lets do log4j again

guess this was a waste of time

[thorn isle]

did you try rcon yet

[pine forge]

wdym

[thorn isle]

that's existed since like beta or something

[pine forge]

ah

you mean connecting with rcon

[thorn isle]

myes

[pine forge]

well not all servers have that enabled

and its like a network connection

slow

lots of logic

[thorn isle]

it's a completablefuture anyway

[pine forge]

doesnt mean i wouldnt prefer a quick response

[thorn isle]

it's going to be quick on the scope of like under half a millisecond or something since it's a loopback connection

[pine forge]

okay hmm

might make sense

ill try that

ill fall back to console capture if rcon is disabled

[thorn isle]

yeah that seems reasonable enough

[pine forge]

rcon is not enabled by default it looks like right?

[thorn isle]

yeah since it allows anyone with the password and access to the port to run any commands they want

not something you want active by default

[pine forge]

okay well

99% of server swont have it enabled then

[thorn isle]

servers can enable it

[pine forge]

but why would they

[thorn isle]

because your plugin installation guide tells them to

just like some plugin installation guides tell people to set up a mysql database

[pine forge]

well my server actually had it enabled

but no i dont think its important enough to give people extra hustle in installation guide

lots of people dont even know what serve.rproperties is

[pine forge]

so rcon just cuts off command responses?

i dont get the full one

[thorn isle]

there's like a 4k character limit for the response

[pine forge]

not sure thats 4k: Lianecx has the following entity data: {Slot: 0b, id: "minecraft:white_shulker_box", count: 1, components: {"minecraft:block_entity_data": {id: "minecraft:shulker_box", Items...

lets see if its a diff issue

The Minecraft server can fragment responses across multiple packets. There's no simple way to know when the last response packet has been received; approaches include:

stupid

wtf

Wait until we receive a packet with a payload length < 4096 (not 100% reliable!)
Wait for n seconds
Send two command packets; the second command triggers a response from the server with the same Request ID, and from this we know that we've already received the full response to the first command.
The second packet should use a command that will not produce fragmented output
An alternative is for the second C->S packet to use an invalid type (say, 200); the server will respond with a 'Command response' packet with its payload set to 'Unknown request c8'. (200 in hexadecimal)

https://minecraft.wiki/w/RCON#0:_Command_response

[thorn isle]

good shit

[pine forge]

still seems to cutoff

idk

mean it returns these ... itself

i dont think i have that in my code

[quaint basin]

How does DonutsMP allow players to be teleported to coordinates like 100,000? Do they generate the chunks in real time because they have a powerful processor?

[vagrant stratus]

magic

but also good RTP plugins

but also magic

[quaint basin]

I think it's just a good cpu

I don't know if I should put the resource world on other servers; I have the nether, end and overworld

[vagrant stratus]

They're likely just using an optimized teleportation plugin, don't really need a high end CPU just to teleport somewhere lol

[quaint basin]

The problem isn't teleportation itself, but rather the loading of chunks from various players

[vagrant stratus]

Also not difficult. Not much different than regularly playing

Doesn't really take much to pre-load chunks either

[quaint basin]

My point here is that the donut always teleports to coordinates like x=100,000, which means that the donut always doesn't have these pre-generated chunks because nobody in this world has a disk to pre-generate a world of that dimension

[quaint basin]

Pre-generate a 100k * 100k world? That's my point

[thorn isle]

you can pregenerate them on the fly

[vagrant stratus]

^

[thorn isle]

your cpu won't always be at 100% load; you use the idle time to pregenerate some pre-chosen rtp spots as you're able

and then teleport the players to those spots on demand

this is what i was talking about when you first asked help here, with your rtp plugin

i remember spending several hours explaining the concept of a ring buffer

[quaint basin]

I ended up using an RTP plugin from the internet, but I'm scared now xd

[vagrant stratus]

or just generate a world and generate the "starting" chunks. Minecraft handles everything else via the world seed itself

You don't have to load in the entire world lol

Realistically, all you'd have to do is load in the same amount of chunks as the server normally would and you're fine

5x5 chunk area

now if you want to pre-load the entire world you're kinda fucked regardless

[umbral ridge]

https://giphy.com/gifs/dogs-gifmovie-H2Th7oh1EDPIA

[vagrant stratus]

I don't think you have almost 7 TB lying around to pre-gen an entire 100k * 100k world 🙂

[umbral ridge]

meh

who really explores the whole world

has to have a mental thing going on

[vagrant stratus]

exactly. pre-gen 5x5 or even 10x10 and let the server generate the rest as needed lol

[umbral ridge]

or 50k x 50k enough for 20 people

now that there's so many stuff cramped up below and above

[thorn isle]

spoof the player coordinates

keep everyone in a 1000x1000 box but spoof random coordinates in +/-3,000,000 every time they rtp

[umbral ridge]

yes

and add a click button that allows players to teleport to them for free

[thorn isle]

iirc there is or was a "coords obfuscator" plugin some rpg/hardcore servers used to make it harder to cheese with f3

you'd actually have to use maps and learn landmarks

[umbral ridge]

hmmm

is that even possible to do?

to hide F3 coords

[thorn isle]

not hide, but they filtered all packets to change the coordinates with random numbers

[quaint basin]

I think I'll use your solution with the ring buffer

[umbral ridge]

XD

I want that

[thorn isle]

or like apply a random offset to them, which changes any time you teleport

[vagrant stratus]

speaking of chunk stuff. Debating on if I wanna try to re-create the chunking system in C++
Making a 2D ASCII game & it'd be an interesting-ish challenge

[umbral ridge]

what i wanna do soon is my own dynmap

that will be interesting

and a good challenge

[thorn isle]

you know i've had this custom dynamic map idea rattling in my mind of thoughts for a while

earth map servers all mercator projection, which is shit and bad and makes greenland larger than the rest of the world combined

[vagrant stratus]

Needed? No. Not really.
Interesting? Yea, kinda lol

[thorn isle]

so i want to use a Dymaxion projection

[umbral ridge]

that's interesting

[thorn isle]

which essentially splits the world into distinct triangles and other shapes

[vagrant stratus]

https://tenor.com/view/simpsons-nerd-gif-9743791

[umbral ridge]

I already see how im gonna do it

[thorn isle]

it'd have a "portal boundary" in the oceans where you'd get teleported from one shape to the other

[umbral ridge]

i would use a grid system

chunks and blocks

[thorn isle]

and for the dynamic map, you'd map it onto an actual sphere

[umbral ridge]

https://tenor.com/view/don't-make-me-tap-the-sign-simpsons-bus-gif-5399805801037462082

minecraft is racist

[quaint basin]

To get a biome without loading the chunk, I just need to do world.getBiome(x, y, z).getKey().getKey(), right? I saw that ppl suggested me using world#getProvider to do this, but I don't think it's necessary

[thorn isle]

i don't know, check the nms

i remember there are calls that generate/load the chunk and calls that don't

[quaint basin]

The world#getBiome is taking 0.1ms. Would that prove the chunk isn't loading the chunks, I'd say?

[thorn isle]

depends on how fast you can load a chunk

[quaint basin]

In this case, if this loads the chunk, it also generates

Because the coordinates range randomly from -1 million to 1 million

I think it's impossible to generate and load in that time

[young knoll]

It makes a call to WorldGenLevel#getNoiseBiome

So it probably doesn’t load the chunk

[thorn isle]

yeah that won't load it

[quaint basin]

If I use the ring buffer for my RTP plugin, after checking the biome, should I generate/load the chunk to check the valid Y coordinate for that location if there is one; if there isn't, should I not add it to the buffer? Or should the buffer only contain locations that were valid in the first step?

My concern is that if I check everything before adding the buffer and the buffer size is 100, then every time the plugin starts it will generate 100 chunks (in the worst case, which is mainly the case since the locations range from -1 million to 1 million). To solve this, I would have to save the buffer to disk, but I don't think that's a good solution

[sly topaz]

RTP plugins don't need to be complicated

[thorn isle]

the buffer should only contain locations that you can immediately teleport players to; the surrounding chunks should be pregenerated, and the location should be a valid teleport target

[thorn isle]

you can persist them if you really need to

the point of this is to do the work ahead of time, if you just put a bunch of points in a buffer without checking if they're valid, you might as well pick the points on the run and get rid of the buffer

[sly topaz]

it is probably better to persist them given that otherwise you'd be uselessly pregenerating locations at startup when you had already calculated completely valid ones at a point

[quaint basin]

Surrounding chunks should be something like the chunks a player loads when teleporting to a location in the worst case? The worst case would be view-distance value on the client at the maximum the server allows

[quaint basin]

My point in not saving would be due to the high disk usage

[sly topaz]

what high disk usage

why would saving some locations end up with high disk usage

[thorn isle]

it's just a worldname,x,y,z tuple, you can save tens of millions of those per gigabyte

[quaint basin]

If my buffer size is 100, I'll be generating huge chunks; if I don't save the locations of this buffer when the plugin shuts down, I'll have to find another 100 valid positions and generate 100 chunks and their corresponding chunks, and those positions shouldn't even be used anytime soon

That's why I think I should save the buffer positions when the plugin shuts down

[thorn isle]

you should, we agree

it sounded like you were saying you didn't want to save the buffer positions because they'd use a lot of disk space

but perhaps i misread

[quaint basin]

oh no

What I meant was that not saving would use high disk usage

[sly topaz]

no that's what I understood too lol

[thorn isle]

personally i don't bother saving them, i just generate 50ish fresh ones every restart, but i could see how that'd add up over time

[quaint basin]

It also depends on your world size

[sly topaz]

I was tempted at some point to make a random-location based generator for chunky

[quaint basin]

But I think that approach isn't even advantageous if the world is small; if it's small, you can just pre-generate the entire world

[sly topaz]

that way you'd generate just huge spots around random safe locations rather than in a radius, but that'd only really work for some gamemodes like UHC

[quaint basin]

Does UHC have chunk loading?

uhc for me is like a 100x100 map

[thorn isle]

i use it with conjunction with pregeneration, it still does help if your "is a position valid" conditions are stringent and it's difficult to find valid positions

[sly topaz]

can be bigger depending on the player count

[quaint basin]

Is this correct?

[thorn isle]

yeah sounds about right

pregenerating past that will still help when players start exploring around, but there will be some chunks you pregenerate that won't be visited, going to "waste"

[quaint basin]

Perhaps in this case it would be better to deal with the average case and not the worst-case scenario idk

[thorn isle]

depends on what your maximum view distance is i guess

[quaint basin]

yea

[thorn isle]

generating up to i'd say 10 chunks radius is sensible i think, but 32 is probably overkill

[quaint basin]

I don't know how many chunks a survival server should have, but that's something I'll have to research in the future

maybe 8-10

So maybe if this value be the surrounding chunks amount, I don't think it will take up much disk space

[vagrant stratus]

Managed to get the ASCII world & chunk stuff figured out.
Don't necessarily like it though kekw

It'd definitely be better w/ color, but still lol

[thorn isle]

but can you build a fortress in it for your dwarves

[vagrant stratus]

I mean, I can implement that functionality lol

I need to improve world-gen first though

[young knoll]

: is my favourite block

also that island on the right looks like a doggy

[vagrant stratus]

k in my defense I cba to find fitting characters lmfao

[ancient plank]

makes me think of the guy who puts a dog in his games that represents him

[vagrant stratus]

I did slightly improve the world-gen. At least as okay it can get w/ google and ASCII lol

[vagrant stratus]

lol this probably helps more. I'll simplify some stuff when/if I get around to rewriting this as some stuff is rather useless

[young knoll]

T for Tand

[vagrant stratus]

I know reading is hard, but you can do better than that

[ancient plank]

terraria

fkin tand lmao

I love it

[vagrant stratus]

dropped that project for now, currently waiting for something else to run though lol

[ancient plank]

im sitting here thinking about stuff

like

do I want to actually do software engineering full time

[vagrant stratus]

waiting on custom tooling to finish up so I can test stuff lmfai

[ancient plank]

well if I don't, what else do I want to do

[vagrant stratus]

become a Shepard

https://tenor.com/view/dog-between-the-sheep-black-dog-sheeps-dog-meme-funny-dog-gif-1681563161525257436

oh ran out of heap space before that finished KEKW

[mortal vortex]

How do you guys plan things? (like SDLCs)

I wanna rewrite a plugin, from the ground up, and get an idea of everything it needs before i start.

[worldly ingot]

I assure you a total of zero people in this Discord server are using any form of SDLCs for their plugins lol

[vagrant stratus]

^

I personally just update until the code base is a PITA or has enough technical debt.
If it's a private project I'll even go as far as scrapping the github history & starting entirely from scratch, using the old version as a reference for a from the ground up rewrite

[ancient plank]

I took a project management course last semester and used the knowledge from it to design a plugin rewrite

did not use any of the tehcniques I learned other than general skills I picked up during it to better organize my thoughts

[vagrant stratus]

lol

[ancient plank]

for minecraft pplugins it's just not necessary

[vagrant stratus]

^ PM is more for group projects or companies

[ancient plank]

don't get me wrong that class was VERY USEFUL 10/10 would recommend

I learned a lot of really good things in it, but it's not a necessary skill to have when working on something like a minecraft plugin

[vagrant stratus]

I guess if you want the experience, but it's still massively overkill

[ancient plank]

I recommend listing what you currently have, what you want to have, why you want those things, and doing iterations over it to improve the explanations and features

[vagrant stratus]

Worst case, you just end up doing this at some point kekw

[ancient plank]

i.e. this is what my doc ended up looking like

written fresh out of that class

wanted some hands on experience bc I was fkin struggling to understand lmao

[vagrant stratus]

note, probably over kill as fuck

[ancient plank]

for sure

helped me a lot tho

[vagrant stratus]

oh 100% but it's not really required

[ancient plank]

I tend to get lost in the sauce when making new stuff

[vagrant stratus]

Similar to doing game design docs and stuff solo

[ancient plank]

mhm

[vagrant stratus]

not required, but useful lol

[ancient plank]

I wrote a cute lil pipeline to explain the stages of damage calculation

[vagrant stratus]

on another note, praying the changes I've made don't take the entire Java heap again KEKW

Trying to index a few thousand things in memory atm

[ancient plank]

oof

[vagrant stratus]

https://tenor.com/view/kekw-twitch-twitchtv-twitch-emote-kek-gif-19085736

[ancient plank]

we had a server crash and I downloaded the heap dump forgetting that you need like 2x memory to read a heap dump

[vagrant stratus]

I don't need too, but it'd sure as hell make one of the custom tools I'm working on easier lmfao

[ancient plank]

I only have 32GB of ram that thing was like 20GB

[vagrant stratus]

been expanding my reverse engineering & analysis program all day lmao

[buoyant viper]

😭

[ancient plank]

:)

[mortal vortex]

Yeah but the thing is. When we started working on this plugin we underestimated what it would have, and it’s basically on very shaky stilts. Was thinking of consolidating everything we know now, to rewrite the plugin.

[buoyant viper]

@ancient plank probably is

[quaint basin]

But thinking about it: the buffer might have invalid positions if I add the positions to the buffer and then someone leaves that position invalid

[calm dust]

hey, i`ve downloaded quests plugin, do i need to setting my own quest or they alredy have the default quest

[thorn isle]

Incredibly unlikely in a 100k radius world, but if you're worried about it, you can remove the position from the buffer if the chunk is loaded for any reason

[mortal vortex]

Congrats on downloading a plugin

But (a) this is a channel for dealing with developing your own plugins.

and (b) you may want to specify what “quests” plugins. Spigot is a marketplace, not a development team, where individuals may publish their own plugins, several of which may be “quest” plugins.

Move to #help-server

[slender elbow]

or what

[pliant topaz]

At most some basic notes for a broad overview, maybe a few specific implementation details, but most of the improvements happen because I now know what system needs what and can improve it from the ground up (often resulting fully recoding the entire class from scratch as its the easiest in my opinion)

[thorn isle]

my grandfather would, when presented with any large or overwhelming problem, always say "start from the corner", referring to the side of a farm field

in similar fashion i just sit down and start hammering at it in whatever class seems like a decent entry point, writing //TODO's for things as i go

[slender elbow]

i always start with the SpringAbstractedMongoServiceBuilderWebProvider.Factory

and then get off all day long thinking about best practices and clean code and ideal paradigms for java

[quaint basin]

Are you saying that if the player teleports to a loaded location, then try the next location in the buffer?

[thorn isle]

I'm saying listen to chunk load event and if the chunk contains a location that is in the buffer, remove it from the buffer

[quaint basin]

oh good

[thorn isle]

This means it is impossible for points in the buffer to become invalid

[quaint basin]

But verifying whether a location is contained within a chunk, I don't know if that's an easy task

[thorn isle]

It is very simple

Build a chunk coordinate->buffer location map when you populate the buffer

And query it on chunk load

[quaint basin]

oh good

what's the site with minecraft source code?

[vagrant stratus]

There isn't one

[quaint basin]

I know there's a website with the Minecraft nms code

[vagrant stratus]

nope! legally we're not allowed to host the code

gotta use either buildtools or paper's thing to get acess to it

[quaint basin]

Paper members send this website on chats

[vagrant stratus]

That's not source, but documentation

[quaint basin]

There is literally a website online with the Minecraft code

I think Minecraft now offers the code

i just dont remember the url

[vagrant stratus]

You'd still need a decompiler, as they only provide jars

de-obfuscated versions of MC don't even go back to whatever version you're usnig

They're only 1.21.11+ or whatever mojang said

[quaint basin]

I'm on 1.21.11

[vagrant stratus]

https://piston-data.mojang.com/v1/objects/3ca78d5068bf9b422f694d3f0820e289581c0f0d/server.jar

[quaint basin]

That's not what I'm looking for. I'll see if anyone knows

[vagrant stratus]

that's the only way you're getting it afaik

[quaint basin]

i found https://mcsrc.dev/

[vagrant stratus]

yea, uses the same jar I just sent lol

[slender elbow]

if only the huge text block that reads "How It Works" that shows up immediately when you open the website explained how the site does indeed not host and serve the code but rather the binary jar is downloaded and decompiled locally

oh well, i guess the world shall never know

[vagrant stratus]

which is why I've been saying what I've been saying lol

[vagrant stratus]

https://tenor.com/view/nobody-reads-the-manual-sanjay-odd-odd-west-fairly-odd-parents-no-one-reads-the-handbook-gif-23883128

[quaint basin]

Is it possible to generate a chunk without loading it?

[thorn isle]

No, generation happens in memory

Generation also requires accessing adjacent chunks, so even if you wrote your own chunk generator that writes it to disk directly, you'd still need the adjacent chunks

[tender shard]

not really, why would you? you can just instantly unload it

[quaint basin]

It's better to do nothing (just let paper unload them); I just want to generate the chunks without having to load them so that when a player gets close, I don't have to generate the chunks (only load them). Currently, since I need to generate them, I'll also load chunks without players nearby

[quaint basin]

https://pastes.dev/2GX4s5YsH5 If the surrounding chunks are loaded when generating the chunks, then I imagine this code doesn't work 100% because I'm generating the chunks around the location the player will teleport to

How many surrounding chunks are needed to generate a chunk?

[tender shard]

Not sure, but couldn't you just pregenerate the whole world?

[thorn isle]

the server will load the adjacent chunks at the appropriate chunk stage/level transparently

it can't not work

[quaint basin]

it is 1_000_000 x 1_000_000 world

[tender shard]

well okay that's a bit large for pregen lol

[quaint basin]

But then generating the chunks around it like I'm doing might not be necessary

[thorn isle]

for more about chunk stages, read the wiki or this thread where i explain it to someone else

[thorn isle]

it will still be necessary, because they won't be at the "full" stage otherwise

they will be at whatever stage is necessary to get the information for the generating chunk

generally 1 stage lower than the generating chunk

[quaint basin]

i see

[thorn isle]

also this

you can essentially think of it like building a pyramid; you can't just start from the center and pile it 100 stone blocks tall, because it will fall; it needs to be surrounded by adjacent columns of stone blocks, which in turn have to be supported by adjacent stone blocks, forming a step-pyramid shape

this is also why it is much cheaper to look for valid teleport positions in nearby chunks even if you can't find it in the chunk you randomly chosen; it's cheaper to make an already-tall pyramid slightly larger, than build a whole new pyramid somewhere else

[quaint basin]

i see

[quaint basin]

When a location to be placed in the buffer fails, do you recommend trying a nearby location then?

[thorn isle]

depends on the way how it fails; if it fails because it's water, it's very likely that you're in a river or a lake and many surrounding positions will also fail the same condition

[quaint basin]

The issue is that sometimes this might not be advantageous because the plugin might simply try multiple times (and this happens because it tries a nearby location, for example, a player building a huge trap), which ends up not being advantageous to simply "build the pyramid" from scratch

[thorn isle]

if it fails because it's e.g. covered by logs, it's probably a tree and a nearby position is still likely to be valid

[quaint basin]

if is water, will be a river on worst case because the ocean is stopped by the biome

[thorn isle]

it's much cheaper to find a valid position in a loaded chunk than it is to generate the chunk, so you don't want to spend 100ms generating a chunk and then discard it in 1ms

[quaint basin]

world.getHighestBlockAt(location.getBlockX(), location.getBlockZ(), HeightMap.MOTION_BLOCKING_NO_LEAVES)

That would never happen either, I imagine, because I do this

[thorn isle]

well, logs aren't leaves

you can end up being unlucky and picking the actual tree trunk for your location

or a cactus, or any number of other things

[quaint basin]

but getHighestBlock will return the air block

What can happen is that it spawns on top of a cactus, for example

[thorn isle]

no, it will return the highest non-colliding block except for leaves

i.e. it could be inside the leaves

on top of the tree trunk in the tree

or on top of a cactus (you don't want to rtp the player there either)

my point is that you will have this kind of "spurious" failure of the search conditions; just search a few nearby locations if you do

but for other types of failures that are indicative that the whole area might be invalid, it's better to pick a whole new chunk

[quaint basin]

Wouldn't getHighestBlock be the block above the highest X block that is collidable, except for X blocks which are leaves?

Because the tallest non-colliding block is usually the tallest possible Y-coordinate

[thorn isle]

HeightMap.MOTION_BLOCKING_NO_LEAVES indicates a heightmap that accounts for all colliding blocks except leaves

meaning it won't see leaves

meaning if you have a block of grass and 2 blocks of leaves above it, it will return the bottom-most of those leaves blocks; a spot inside the leaves

it can also return the block above the tree trunk of a tree, which will also be occupied by leaves

[quaint basin]

Yes, I agree, but I disagree with the part where you said that getHighestBlock returns the highest non-collidable block

[thorn isle]

it returns the block above the highest collidable block that isn't leaves

which itself could be leaves

[quaint basin]

yea now i agree

In this case, I need to check that the block chosen by getHighestBlock is not cactus, and that the two blocks above are not water, lava, leaves, etc

[thorn isle]

myes

[quaint basin]

actually, an air check is basically for the two blocks above

no need check all type of blocks for the two blocks

[thorn isle]

i'd check for air and some acceptable noncolliding blocks, like tallgrass, flowers, and ferns

[quaint basin]

i see

[thorn isle]

since quite many positions will be occupied by those rather than air

[quaint basin]

If you want to send me this check, I'm accepting xd

It's hardwork because I need to see all types of blocks

[thorn isle]

just eyeball at Material or go looking around plains in f3 and put everything you see that doesn't kill you into a switch block

or maybe material tags

[young knoll]

Isn’t there an isPassable

[thorn isle]

there is, but that might not be right

e.g. i think sweet berries might be passable despite not being nice to spawn on top of

[quaint basin]

i see

[young knoll]

You’d probably have to exclude a few

But that’s easier than explicitly including a bunch

[thorn isle]

i don't know, you can try it; using isPassable and a smaller manual blacklist would probably work as well

[quaint basin]

Why on earth are strawberries considered passable and flowers not?

[thorn isle]

flowers should be passable since they don't have a collision shape

[quaint basin]

oh

How does getChunkAsync work? When does the paper decide: Okay, this tick I'm going to load these chunks that were requested ticks ago?

[thorn isle]

there's a queue

[quaint basin]

I'm asking this because I'm using this method to load adjacent chunks from my buffer, and I never see the paper generating the adjacent chunks

[thorn isle]

well you don't see them, that's what transparent means

the server generates them up to whatever stage they need to be, but in bukkit, only the full stage is considered to be generated or loaded

[quaint basin]

If I generate the chunks with getChunk method, will the chunks that are first in the queue be delayed?

[thorn isle]

getChunkAt will push the chunk to the beginning of the queue, to be loaded "as fast as possible"

[quaint basin]

oh

Has it always been like this ?

[thorn isle]

which generally is not good because now the server is putting everything else it's doing to the side just to generate the chunk

this? as in there existing chunk generation stages? yes

[quaint basin]

I asked if getChunkAt always adds the chunks to be loaded as quickly as possible and not simply load them

In that case, I asked if it had always been this way

[tender shard]

getChunkAt loads it instantly, doesn't it?

[thorn isle]

this kind of "sync" loading has always existed, and was the only way to load chunks before paper made chunk loading asynchronous; there didn't used to be a queue at all, calling the method would simply synchronously get the chunk

[quaint basin]

from what vcs2 said no

[tender shard]

if it does not return a CompletableFuture, but a Chunk object, then it's a blocking call

[thorn isle]

after asynchronous chunk generation/loading was added, this was replaced with a queue that the chunk worker threads watch and fulfill the requests in it as they're able

i'm fairly sure these days calling getChunkAt still uses those async threads (the generation doesn't happen on the main thread) but the main thread will block for it, stopping the game tick loop until it's done

[young knoll]

Isn’t that how it’s always worked

Vanilla use threads for worldgen too

[thorn isle]

as of like 1.13 or something

but back when it was all a single thread, there was no queue or chunk threads

[young knoll]

Yes

That’s like 50 years ago

[thorn isle]

myes

[quaint basin]

If I use getChunkAsync, the difference is that it's added to the queue to be generated/loaded to a later tick, and the game tick loop isn't paused to generate the chunk?

Since in my view, getChunkAsync is asynchronous, it doesn't make sense for the game's tick loop to pause and wait for the chunk to be generated

[thorn isle]

it doesn't care about ticks, it just puts in the queue and the chunk threads will get to it when they're able to; no thread will block for it as the result is a future

getChunkAt does the same thing, but it puts the chunk at the front of the queue and then .joins() the future, which will block the calling thread (which is going to be the main thread)

either way, you basically never ever want to use getChunkAt

[quaint basin]

i see

[thorn isle]

the only case where you would use it is if everything will explode if you don't get the chunk right now instantly

[tender shard]

unless you're in onEnable or similar

[thorn isle]

perhaps

[quaint basin]

https://pastes.dev/2GX4s5YsH5 I was thinking I was overthinking using getChunkAsync and having to add pendingtasks int etc., but then no

[thorn isle]

i'd still do it with the async overload if at all possible, but yeah, it's sort of acceptable in onEnable and onDisable

[quaint basin]

why

[thorn isle]

i have no idea what this code is supposed to do

it looks like you're picking a random location, and then pushing it into the buffer viewDistance*viewDistance times?

[tender shard]

what's the goal of that? Also if you do LocationUtil.getHighestSafeBlockYAt, don't you already have the chunk loaded after that? Why do you need another getChunkAtAsync after that?

[thorn isle]

yeah, accessing the heightmap of a chunk will load/generate the chunk

[tender shard]

the check for isPrimaryThread is also not needed, the javadocs say that the future returned by getChunkAtAsync runs on main thread

[thorn isle]

sort of maybe

[quaint basin]

At each tick; if the buffer size is less than 100, then add a random valid location between -1_000_000 and +1_000_000 to the buffer. Furthermore, upon obtaining the location, the getChunkAsync method is used to generate the adjacent chunks (unfortunately I have to load them, but they will be quickly unloaded because nobody will be nearby in principle)

[thorn isle]

i'm not 100% sure if that's true in practice in every case, with how futures work

[quaint basin]

getChunkAsync is used to generate adjacent chunks

[tender shard]

hm idk but the javadocs for getChunkAtAsync say "The future will always be executed synchronously on the main Server Thread."

[thorn isle]

why are you pushing the same location into the buffer in the "generate adjacent chunks" loop

[quaint basin]

I just need to add an if-check on for: if dx == 0 and dz == 0 then continue the for loop

[tender shard]

are you doing a random TP plugin or what are you even doing?

[thorn isle]

why not add the location into the buffer first and then loop over the nearby chunks to generate them?

[quaint basin]

Note that I have an if-check if (++currentIteration[0] == totalIterations). This only ensures that the push is performed only once. I do this to add this location to the buffer only when adjacent chunks are generated

This way, I prevent the player from using /rtp and getting a position in the buffer where nearby chunks haven't yet been generated

[thorn isle]

there's no guarantee the that the first future completes after/before all the chunks are generated

since it's a queue, it's liable to complete before all the chunks are generated

if you want to make sure the chunks are generated, put all the futures in a list and then do CompletableFuture.allOf on them, and only then put the location in the buffer

[quaint basin]

The problem with this is that if someone enters this chunk before the push and adds blocks, and by chance places a block in a "valid" position that would be added to the buffer, that position will be in the buffer and is no longer valid, and the chunkloadevent will not be called after being added to the buffer. To solve this, simply implement the logic of chunkMap.computeIfAbsent... inside the push, before the for loop inside the RingBufferRunnable, so that if the chunkloadevent is called for a chunk that is in this map but not in the buffer, then add it to a cache structure so that when actually adding it to the buffer (inside for-loop on RingBufferRunnable), check if it is not in this cache structure and if it is, remove it from the chunkMap.

https://pastes.dev/fDzI4Ndgdl
https://pastes.dev/tVxeA9dl5t

[thorn isle]

you're overthinking this a bit

[quaint basin]

I don't understand; if for iterate 200 times and if currentIterations is 200; so I'm guaranteed that the for has already iterated all the chunks

[thorn isle]

since the futures are guaranteed to be completed by the main thread i suppose the counter does work, but allOf is still the better approach

just be aware that this most likely won't work in any other context where you'd use futures

[quaint basin]

I don't understand why you think the counter shouldn't work xd

[thorn isle]

it happens to work in this case because all the futures are guaranteed to work on the same thread

[quaint basin]

Yes, of course; I know this works because what completes the CF is the main thread

[thorn isle]

without such a guarantee, you since you're not incrementing it atomically, it's liable to never reach totalIterations

either way, allOf is a much easier structure to reason about and also allows you to e.g. capture thrown exceptions in a cleaner way

[quaint basin]

I'm not using atomicinteger because I only increment/read it in the main thread

[quaint basin]

never heard about allOf. i'll see that

[thorn isle]

it's a future that completes once all of the passed futures are complete

so if you give it all the chunk futures, you have a "all chunks are done" future

[quaint basin]

https://pastes.dev/Gd2oDGAD66 am I doing well?

this method is powerful

[thorn isle]

yeah that's about right

[quaint basin]

I only forgot pendingTasks-- inside the allOf block

[thorn isle]

also remember to always either .join() or .exceptionally() on any future you run any code in

otherwise, you will not see any exceptions that occur

[quaint basin]

I don't usually add unchecked exceptions in a CF, but it seems like that's bad practice then

Do I see the exceptions this way? https://pastes.dev/IobncntZOh

[thorn isle]

no, that exception also gets swallowed

you need to take the exception and log it/handle it

the thing with unchecked exceptions is that you don't "add" them; they just happen, unexpectedly

e.g. suppose there's a CME on ringBuffer.push(location), or a NPE on location.setY; you will never see them unless you handle the exceptions

it's easy to forget about this because java usually has very reliable exception handling, but with futures, you must always remember to add an exception handler yourself, because java will not log/handle them for you

[quaint basin]

i see

i'm doing this way now:

.exceptionally(e -> {
logger.log(Level.SEVERE, "Exception loading chunks", e);
return null;
});

[thorn isle]

yeah, that's how you'll want to do it

[quaint basin]

I don't really like this because, for example, if I have an exceptionally in a service, every time I use the service, I have to add another exceptionally block

[thorn isle]

yes, this is the downside of futures

[quaint basin]

It ends up being just exceptionally nested blocks

[thorn isle]

well you generally don't need to nest them because of the way how they propagate throughout futures

but yes, it's still ass and very easy to neglect to put an exceptionally where you need it; it isn't always clear where you should or should not put it, either

this is why i personally prefer virtual threads nowadays

[quaint basin]

why generally?

[thorn isle]

as long as you .join or .thenX on a future, you don't need to .exceptionally on it; in case of join, the exception is automatically rethrown, and in case of thenX, that responsibility gets passed to the resulting future

[quaint basin]

My point is:

Let's say I have a service that does CompletableFuture.thenAcceptAsync(code block...).exceptionally(e -> return new QueryErrorResult());

Then, if I use the service's method, upon obtaining this future, and do something like thenAccept again, this thenAccept will be called regardless of whether an exception occurred or not. Therefore, whoever calls the service's future and uses thenAccept should call the exceptionally block after that right?

Sometimes the code inside the second thenaccept block is simple logic that you're sure will never cause an exception, but it would be considered bad practice not to add the exceptionally block ig

[thorn isle]

if it's a service/api method that returns a future, you don't necessarily have to do .exceptionally on it; you can decide that doing that is the caller's responsibility

otherwise these .exceptionally's will proliferate endlessly and your code will just be exception handlers

this is what i meant by "it's not always clear whether you should or should not put it, either"

either way, yes, futures are ass, i'm not arguing in their favor; for best results, look into virtual threads and use them in conjunction with futures

[quaint basin]

i see

[thorn isle]

e.g. your chunk loading example, with virtual threads, can be rewritten as

virtualExecutorService.execute(() -> {
    List<CompletableFuture<Chunk>> futures = new ArrayList<>();
    
    for (int dx = -viewDistance; dx <= viewDistance; dx++) {
    
        for (int dz = -viewDistance; dz <= viewDistance; dz++) {
    
            if (dx == 0 && dz == 0) {
                continue;
            }
    
            int chunkX = cx + dx;
            int chunkZ = cz + dz;
    
            futures.add(world.getChunkAtAsync(chunkX, chunkZ));
        }
        
    }
    
    CompletableFuture.allOf(futures.toArray(new CompletableFuture[0])).join();
    
    location.setY(newY);
    System.out.println("push: " + location + " to the buffer: " + Bukkit.isPrimaryThread());
    ringBuffer.push(location);
    
});

[quaint basin]

What kinds of things make you use completablefuture today?

maybe a database pool and apis already built with completablefuture?

[thorn isle]

except you'd want to use an executorservice in place of a completablefuture for the top-level block

completablefutures have their place, as you can see i'm still using them and the allOf and other future style methods in the above example

the main difference is that instead of runAsync().thenX().thenX().thenAccept().exceptionally(), i.e. chaining futures, I use a virtual thread that blocks on each stage of the "chain"

this means i can use regular flow control and imperative-ish structures like helper methods and even recursion, all combined with futures

so yes, futures still have a place as e.g. the return type of a database query/api, or the chunk loading api in paper

it's just the way how you put them together that benefits from virtual threads

[quaint basin]

In this case, besides avoiding the exceptionally and thenRun blocks, does it also offer better performance?

[thorn isle]

maybe, god knows, probably not measurably

[quaint basin]

The only downside is that I'll need an extra pool for this

[thorn isle]

it's a virtual thread so it doesn't really use a pool

you can also sort of do it without virtual threads if you target java 8 or whatever, but it will be far less efficient than only-futures, since it'll block a platform thread

[quaint basin]

wdym platform thread?

oh i see

[thorn isle]

like an os thread, complete with its own allocation of memory for its call stack and other expensive things

now the os does also schedule os threads independently of hardware threads, so os threads are also a kind of "virtual" thread, but java virtual threads are much more lightweight

[quaint basin]

Currently, I create the executor like this: Executors.newVirtualThreadPerTaskExecutor(). My fear is that if this is called 1000 times, 1000 virtual threads could be created. Perhaps there's a way to limit the number of threads created, like in a fixed thread pool with OS threads, but I don't think that's necessary here because, from what I've seen, virtual threads are designed to support millions of virtual threads

[thorn isle]

the idea with virtual threads is that they block/sleep for most of their lifetime, and so are not mounted on os threads most of the time

meaning you can have 1000 virtual threads, and that is valid, as long as they are doing something like future.join(), io, or sleeping

[slender elbow]

you can create millions of VTs and your computer won't blow up

[thorn isle]

in the case that they all try to do work, yes, you will probably exhaust your platform threads

[slender elbow]

that's pretty much the point why they exist

[quaint basin]

And isn't that a problem here?

[thorn isle]

no, because look at what the virtual thread does

it tells the server to load a bunch of chunks, and then it sleeps until they're loaded; then it puts a location into a queue

what it itself does is very lightweight, so it won't be mounted on an os thread for a meaningful amount of time

what you don't want to do is spawn 1000 virtual threads that all try to compute the 50,000th prime number, because those won't sleep and will hog all the os threads

[slender elbow]

well, only a handful of them will run at a time since they ultimately run on a regular fjp but yeah

most will just sit there waiting for their turn

[thorn isle]

i mean yeah, but the point is it will behave like starting 1000 regular threads computing the 50,000th prime number

so you lose all the advantages of the thread being virtual

[quaint basin]

That's because the operation you mentioned is very time-consuming (it should take minutes or hours), and the virtual thread will have the overhead of constantly switching threads, meaning it will end up being worse. Whereas with CompleteFuture in this case, there won't be this overhead of switching threads and it should execute faster. Now, for other options like the one you used, virtual threads are good because it's an operation that, in the worst case, should take seconds, which is nothing compared to calculating a 50,000th prime number. Did I understand correctly?

[thorn isle]

i probably phrased it a bit poorly; the concern isn't time, but compute

it's fine for a virtual thread to live for minutes; it's not fine for it to do work for minutes

it's fine for it to wait on a chunk being generated; it's not fine for it to do anything computationally expensive, like computing a prime number

[quaint basin]

Isn't an operation that takes time computationally expensive?

Of course, if I use getChunksAsync and only get a response after 100 seconds, that's not computationally expensive even though it's a time-consuming operation. When I talk about operations that take time, I mean sync operations

Obtaining a response from the database is, I believe, faster than obtaining the 50,000th prime number. Therefore, I would say that accessing the database is computationally cheap relative to calculating that number

[slender elbow]

Thread.sleep(999999999999) takes time and is not computationally expensive

[quaint basin]

fair

[slender elbow]

accessing network resources like http call or database queries can take entire seconds and are not inherently computationally expensive

they might just take time due to network latency

and that is where VTs shine

[quaint basin]

I knew that, yeah; I just didn't know that could happen with synchronous methods like thread#sleep

[slender elbow]

they don't sit on a a whole ass OS thread waiting for the resource to become available

[quaint basin]

I think thread#sleep is the only sync case

[slender elbow]

the OS thread that manages the virtual thread is instead tasked with running other virtual threads

[sly topaz]

github actions would like to have a word

[quaint basin]

So you use VTs to connect to the database instead a fixed pool for example?

[slender elbow]

gha can eat tree bark

[lilac dagger]

You can have a fixed pool of vt factory

[slender elbow]

why would you pool VTs?

it makes no sense

[lilac dagger]

It is possible tho

[slender elbow]

sure

it is possible for you to eat tree bark as well

[quaint basin]

But that doesn't make sense xd

[slender elbow]

doesn't mean you should

[sly topaz]

Jakarta has a managed VT pool implementation for whatever reason as well, wonder if anyone uses that

[quaint basin]

You wouldn't say that if you knew the benefits of eating it

[quaint basin]

who is jakarta

[sly topaz]

formerly Java EE, it is an enterprise™ platform/framework

[quaint basin]

Well, if they work on Java enterprise, they're probably useful no?

They're the ones who created virtual threads, they should know in what situations they're useful, lol

[slender elbow]

uhhhhhhhhh

[sly topaz]

not necessarily, on the keynote showcasing them they said and I paraphrase "there's also a ManagedThreadFactory for virtual threads as well, if anyone ever wants something like that"

[thorn isle]

all enterprise shit is snake oil sold to The Suits™ to "increase productivity" that nobody actually doing any meaningful work needs or wants

[sly topaz]

that said, they aren't the same people who work on Java, Jakarta is a separate product

[quaint basin]

Okay. So I should use VTs to do database queries then?

instead of a fixed thread pool

[slender elbow]

to perform the queries, yes

[thorn isle]

anything that does IO or sleeps is natually a good target for VTs

just don't do compute on them

[quaint basin]

and keep using hikari?

[slender elbow]

yes

connection pooling is separate from thread pooling

all you just did was eliminate the necessity for thread pooling

er, wording there is strange

[quaint basin]

Is connection pooling good for using VTS? I know Hikari does that

[slender elbow]

yes, connection pooling solves a different problem from thread pooling

[thorn isle]

much like how creating and destroying a thread is expensive, creating and destroying a connection is expensive

[quaint basin]

And Hikari probably doesn't use vts xd

[thorn isle]

so it makes sense to reuse threads and connections from a pool

[slender elbow]

hikari doesn't "use" VTs because hikari doesn't "use" threads

you use threads

[thorn isle]

except in the case of virtual threads, which are a transparent abstraction of "reusing threads from a pool"

[tender shard]

I recently refactored our base DatabaseProvider to use a virtual thread pool and everything broke

because the common fork pool is pretty smart and forgiving

[slender elbow]

whatever that means anyway

[quaint basin]

For example, Luckperms doesn't use virtual threads because it accepts Java 8 right?

[quaint basin]

The problem with this code is that the push is executed on a virtual thread and modifies an array that is not thread-safe. Should I make the array thread-safe just to use virtual threads here?

[thorn isle]

either that, or run and join a future that runs on the correct thread to do the push

[quaint basin]

I basically need to make everything thread-safe: the pendingTasks variable; the HashSet structure into a chunkMap inside the LocationRingBuffer etc or then do the push and pendingTasks-- on main thread

I don't know if, even with that, the use of virtual threads is still worthwhile on this case

[thorn isle]

in this particular case no, VTs don't help too much; all it saves is one .exceptionally block

but you can make this work with VTs, and they will be invaluable with more complicated threading logic, like flow control and multiple or optional dependencies

i posted the vt example just to give an example on how to go from pure futures to virtual threads

the pure futures one with the .exceptionally block is fine

[quaint basin]

But wouldn't it be less performant with VTs?

on this case obviously

[thorn isle]

perhaps marginally, but not measurably so

[quaint basin]

So here I would use VTs to have perhaps less performance (even if it's not measurable) and it would complicate the code considerably. Therefore, yeah, I don't think it's worthwhile here

But for example, I've already replaced my fixed thread pool with virtual threads and I know how they work, so in any case your code was a great help xd

[mortal hare]

https://tenor.com/view/primeagen-theprimeagen-gippity-chat-gpt-gpt-gif-10694526301644687231

[thorn isle]

VT's excel when there's flow control and complex dependencies, e.g.

//fetch dependencies in parallel
CompletableFuture<PlayerData> dataFuture = dataService.get(player);
CompletableFuture<Waystone> waystoneFuture = dataService.get(waystonePos);
PlayerData data = dataFuture.join();
Waystone waystone = waystoneFuture.join();

//legacy waystone; migrate
if (waystone == null) {
  waystone = new Waystone(...);
  dataService.save(waystone).join();
}

if (data.discoveredWaystones().contains(waystone.getUuid()) {
   //open waystone GUI
}
else {
   data = data.withDiscovery(waystone.getUuid())
   CompletableFuture<Void> dataModifyFuture = dataService.save(data);
   //play discovery animation
   PlayerData owner = dataService.get(waystone.getOwnerId()).join();
   owner = owner.withLog(player.getName() + " has found your waystone")
   CompletableFuture.allOf(dataService.save(owner), dataModifyFuture).join();
}

this would be a nightmare to do with pure completablefutures alone; you'd need several levels of nesting

but with a virtual "orchestrator" thread that fetches data from (possibly separate thread-run) systems and blocks only when necessary, it becomes much simpler

[quaint basin]

yea i see

[young knoll]

Why is the gui being opened inside the data fetching

[thorn isle]

this is the click handler for clicking a waystone

since it runs async, it's free to fetch data, and then act upon it however it likes

[young knoll]

But then there will be a delay before my gui opens

[thorn isle]

there's a cache, but yes, in the worst case it's unavoidable

[young knoll]

Load the data on chunk load :c

[thorn isle]

since the data is on a remote db, and we can't open the gui before we know you can open it, there can be a delay

in most cases i solve this with "spinners" in the gui itself

but in this case since the branch is half-gui half-not-gui, that's not applicable

in practice however the delay isn't noticeable since the db is very close topographically

i opted against any sort of eager caching like prefetching the waystones on chunk load, since the db and the waystones are being modified by multiple backends at the same time

[quaint basin]

I have no idea how to do that xd

[thorn isle]

i honestly wouldn't bother

if you really have a 1 million squared map and are picking positions randomly, if someone places lava on the exact place where you picked a rtp position, and a player teleports to it and dies - god probably wants that player dead and there's nothing you can do

they'll probably be struck by lightning 100 times in a row even if you save them

[quaint basin]

HAHA

https://imgur.com/a/n5yQdsj The problem with my implementation is that getHighestSafeBlockYAt loads the chunk, and in the chunkloadevent I remove the chunkkey loaded in pendingChunks. Therefore, the push will never be executed because my plugin has already loaded the chunk and removed pendingChunks key

If I had a way to know whether the chunk was loaded by my plugin or by a player, maybe I could solve this

But then there's the issue of a chunk being loaded by my plugin, and a player entering that chunk seconds after my plugin loads

[thorn isle]

i mean honestly i wouldn't bother listening to the chunk load

[young knoll]

Surely you can just check if the position is valid again before the actual teleporting

And if it isn’t throw it out and get another one from the buffer

[thorn isle]

that's alright as well, as long as your conditions aren't super strict

i.e. you can check for lava and skip the position if it's covered in lava, or solid blocks

but i wouldn't run more complicated checks like "must be within 32 blocks of a tree", as that'd defeat the point of precomputing the positions

but honestly i wouldn't bother even with the checks on teleport; in a 1 million squared world with say 100 points in the buffer, it's so incredibly unlikely for any of them to be invalidated that the player might as well win the lottery

i suggested this approach before when you originally asked about a rtp plugin like 6 months ago, but that was under the impression that you had a much smaller, like 1k across, world

i wouldn't bother even with a 10k squared world, let alone 1 million squared

[quaint basin]

wdym 32 blocks of a tree

[thorn isle]

well in my rtp plugin i want the player to have easy access to wood

so i require the position to be within 32 block range of a tree

[quaint basin]

+1 thing to my rtp plugin lol

[young knoll]

Pfft that’s no fun

They got legs that ain’t broken

[thorn isle]

spawns in the middle of a mega desert

[quaint basin]

I just have no idea how to implement this

[quaint basin]

1 million world with just desert 😭

[thorn isle]

searching for trees? i just brute force it

that's the advantage of precomputing your rtp points, it doesn't really matter how long it takes

[quaint basin]

And how do you look for trees? Do you look for wood and leaves on top?

[thorn isle]

yeeees

[quaint basin]

If the player builds a vanilla tree, then it's a valid position if near

[thorn isle]

that's fine, a tree is a tree

as long as they get wood

if you want to do it cheaper, just blacklist any biomes without trees

[quaint basin]

What if there is only one tree near that location?

[thorn isle]

4 logs is enough to get started

[quaint basin]

fair enough

And how do you know that the wood/leaves aren't buried?

or inside a house idk

like they can be hidden

[thorn isle]

well i just find a log within a 32 block range, iterate upwards and see if i find blocks following the order log -> leaves -> air -> top of heightmap

it probably doesn't work in some cases and maybe misses some trees but whatever

[quaint basin]

Player making a tree inside a house:

[young knoll]

When would a tree ever be buried anyway

Aside from someone making one

[quaint basin]

But yeah, I don't think I'll worry about it

[quaint basin]

have ppl for evertyhing

[thorn isle]

stochastic correctness 💪

being correct 99% of the time is being right basically all of the time

[quaint basin]

I'll give a rank if the person spawns in an invalid block via rtp

[thorn isle]

that's the spirit

[young knoll]

Imagine someone rtps into lava placed after the location was checked and next to an buried tree

[thorn isle]

that does remind me

lava and "springs", i.e. lone source blocks, don't generate fully-flown, and loading a chunk through the api only keeps it loaded for a few ticks at most

so it's entirely possible for you to rtp underneath a natural lavafall and for the lava to flow into you if you don't move fast enough

[young knoll]

It’s a quick time event

[thorn isle]

i haven't ever seen this but it's theoretically possible

[quaint basin]

lol

[young knoll]

Just give them fire resistance for a bit

[quaint basin]

Do you check the blocks if they are valid when you teleport?

or you just ignore

[thorn isle]

i don't bother

[young knoll]

Alternatively you could load the chunk with a ticket for a while and then check for a safe place

[thorn isle]

i only have like 10k blocks across world and 50 positions in the buffer, but even then its incredibly unlikely for any of them to be invalidated

[quaint basin]

is this resources world?

I have 1 million worlds so I don't have to worry about resetting the world bc is a resources world

[thorn isle]

supposing i had 100 concurrent players that all run the command every 10 minutes and place 1 lava block in the world per second, it takes 5 minutes for the buffer to completely cycle through, in which time 30,000 blocks have been turned into lava - out of the 100,000,000 blocks of the map; so there's a 0.03% chance for any given rtp point to have turned into lava

and that's with 100 players doing nothing but rtp'ing and running around with lava buckets

[quaint basin]

In my mind, prevention is better than cure. If someone loses items because of this, then I'm screwed

[thorn isle]

get IRP and just check the logs when someone does report it

[quaint basin]

Even worse is if someone loses because of this and has no proof, and I don't know if he's telling the truth or not

[thorn isle]

if the player did /rtp and a few seconds later died in lava, and the death pos is lava, they probably died from the rtp

in such a case just /irp restore the player

what we're talking about here is like trying to make sure UUID.randomUuid() doesn't by-chance generate the same uuid twice

if it does happen, be sure to get a lottery ticket

[quaint basin]

IRP u mean inventory rollback plus?

[thorn isle]

myeah

[quaint basin]

u prefer this than coreprotect?

never heard about irp

[thorn isle]

i haven't used coreprotect for a while (i forked prism a few years back), but IRP is complementary to a block logging plugin

irp snapshots the inventory on world changes, joins/quits, and before death

coreprotect back when i used it didn't really have inventory snapshots in any sense

in this case you'd use coreprotect to check if the block was lava at the time of death, and then irp to restore the lost items

[quaint basin]

So you're saying to add the location with timestamp the player teleports to in the RTP logs, and then if someone reports it, I can see that log with the timestamp and view it along with the IRP logs?

oh i see

[thorn isle]

i mean the command usage should be in your latest.log or whatever tar.gz logfile, as will be the death message

from there you can see when it was; then check if the block at that time was lava with coreprotect, and then restore the inventory of that death with irp

[quaint basin]

But if CoreProtect has inventory snapshots, why not use it now?

[thorn isle]

if it does, you can use it instead of irp

back when i used coreprotect, it didn't, but it's been a long time

the point is that you can just restore the inventory from a snapshot

[quaint basin]

In this case, coreprotect allows me to see when and where the player died, and I can then see if the location of the block where the player died was lava

And the locations of the nearby blocks if they are lava too

Anyway, I'll give invincibility for a few seconds when the player teleports to help with this and another things too

How does player#teleportAsync work? For example, if I type a command X that teleports me to location X using player#teleportAsync, but then another plugin, upon typing a different command Y, teleports me to location Y using player#teleport, is it possible that even though command Y was executed later, I might end up last in position X?

I think this results in weird bugs

[thorn isle]

no clue

the basic gist of teleportAsync is that it loads all the surrounding chunks async, before teleporting the player sync

i have no idea if there is any kind of "reservation" system to prevent the player from being teleported while the server waits on the chunk to load

that said i've never run into any issues with either

[echo basalt]

depends on the server impl

solid chance that modern versions of teleport just block teleportAsync

Folia would just throw an error on teleport

but I do think it'd use locks so they'd be sequential

[quaint basin]

So, if teleportAsync is something new, why simply remove it?

[echo basalt]

I said block, not prevent

teleportAsync().join() blocks the thread

[quaint basin]

and with the join is teleportAsync still advantageous?

[echo basalt]

Depends on the impl

[quaint basin]

I'm talking about the implementation of the paper

[echo basalt]

Why would you ever want to join besides maintaining legacy compat

[quaint basin]

im on the latest version

[slender elbow]

joining teleportAsync on main is just going to cause a deadlock

[young knoll]

Yay

[buoyant viper]

Deadlock? like the new Valve game?

[cloud perch]

https://tenor.com/view/tik-tok-our-brains-are-shrinking-please-stop-an-iq-too-high-gif-14451649523511961636

[buoyant viper]

our brains are shrinking becos of chatgpt

[lilac dagger]

nah

[cloud perch]

Yo don't talk about our goat like that

[lilac dagger]

chat gpt just happens to be a solution to our brain rotting away

correlation does not imply causation

[hexed fulcrum]

Hello! How can I get a list of all available Spigot versions programmatically? I know about hub.spigotmc.org/versions/ but is there a proper API for that, or is HTML parsing the only way?

[mortal vortex]

Versions, as in?

Minecraft versions? e.g, 1.21.1, 1.8, etc? or spigot builds?

You can get Minecraft versions in general directly from https://launchermeta.mojang.com/mc/game/version_manifest.json

[hexed fulcrum]

I need to get both. It is preferable to get the mc versions from the spigot portal, in case spigot does not support the new version

[tender shard]

you could check the maven repo, since they're all -SNAPSHOT they all get they own unique ID, I think it's timestamp based or sth, idk https://hub.spigotmc.org/nexus/service/rest/repository/browse/public/org/spigotmc/spigot-api/

but what do you need it for?

[mortal vortex]

Yeah but programmatically infers scraping or something, these are all heavily nested <tr>'s.

[tender shard]

it's nexus, nexus got an API I think

and nexus also provides an easy to parse XML file for snapshot versions

[chrome beacon]

Look at whatever BuildTools does and do that

[mortal vortex]

nah u right @tender shard

GET https://hub.spigotmc.org/nexus/service/rest/v1/search?repository=public&maven.groupId=org.spigotmc&maven.artifactId=spigot-api

this works

[tender shard]

I still wonder why they need that

[last kiln]

Bebop mains

[quaint basin]

I'm doing this on @OnEnable method but this is null (i added WorldManager as soft-depend)

overWorld = Bukkit.getWorld("resources_world");

Why they load the worlds async?

[worldly ingot]

Is that World instance necessary on startup?

I don't imagine you specifically need it in onEnable(). If you're just checking for it in a Listener or something, reevaluate if you even need a World instance. If you want to do something as soon as that world is available, use a WorldLoadEvent listener instead

You may have to set your plugin's load time to STARTUP though because POSTWORLD is the default

I dunno. It depends ¯_(ツ)_/¯

[shadow pumice]

Hey! i've seen some plugins use a custom expiremental menus like this.. how can i read more abt it?

[young knoll]

They’re called Dialogs

[shadow pumice]

omg ty!

[quaint basin]

If I change the plugin logic, no. But I'd like to know why the world manager loads async

It's annoying because multiverse-core probably loads on onEnable, and I have to use two types of logic because worldmanager doesn't load the worlds when the plugin is enabled

[quaint basin]

https://pastes.dev/GaNb7VnfEK location#isGenerated is loads a chunk?

[chrome beacon]

how else would it know if the chunk is generated

It needs to check if the chunk is on disk or not

[quaint basin]

Loading a chunk doesn't cause the chunk to be generated?

[chrome beacon]

Normally it does, but not with that method

Also that is Paper code I'm not sure why you're asking about it here

if you're not running Spigot you should be in the Paper discord

[quaint basin]

So if I use location#isChunkGenerated on a location where a chunk hasn't been generated, this method checks if the chunk is on the disk but doesn't generate the chunk?

[chrome beacon]

yes

[quaint basin]

Using this method at runtime is overkill then

[glad berry]

Hello @sullen marlin, I lost my Spigot account (the 2FA). How do I recover it?

[vagrant stratus]

?support

[undone axle]

[e-mail]

[chrome beacon]

You should have kept your backup codes secure 😭

[chrome beacon]

^ email time. Discord is not the way to contact md

for spigot support*

[vagrant stratus]

pretty sure there's like... a two week ban or some shit too if they get the problem solved lol

[umbral ridge]

can you put pdc on items?

[chrome beacon]

yes

[umbral ridge]

perfect

instant response

[glad berry]

Where do those people end up?

I lost my plugins lmao 🤣

[chrome beacon]

people?

Sent straight to the shadow realm

[glad berry]

Do backup codes arrive in Gmail?

[chrome beacon]

No you got them when you enabled 2fa

[young knoll]

https://tenor.com/view/shadow-realm-jimbo-alexzander-greats-nothing-personal-jimmy-neutron-gif-26475356

[chrome beacon]

with big red text telling you to save them in a secure place and print them out

[glad berry]

Damn, I lost that phone

[vagrant stratus]

welp you're fucked

[glad berry]

lol v:

[young knoll]

If you email support from the email the account is bound to I think you can recover it

[glad berry]

Kk perfect ty

[umbral ridge]

I cant find the Registry for enchantments

where is it

oh its deprecated

[young knoll]

It’s not

[umbral ridge]

what's the best way to parse enchantments by a string?

[chrome beacon]

Paper user spotted

🔫

[umbral ridge]

https://giphy.com/gifs/dogs-gifmovie-H2Th7oh1EDPIA

[chrome beacon]

https://tenor.com/view/bane-no-banned-and-you-are-explode-gif-16047504

[umbral ridge]

https://cdn.discordapp.com/emojis/723936238993801257.gif?size=96&quality=lossless

[chrome beacon]

https://tenor.com/view/namufilms-patrick-star-rifle-gun-gif-25182238

[chrome beacon]

There you go

[umbral ridge]

getByKey is deprecated too, Registry.ENCHANTMENTS is deprecated too

endless trail of deprecation

[chrome beacon]

Read the deprecation note for enchantment registry

[umbral ridge]

god

[chrome beacon]

Same goes for getByKey

[umbral ridge]

all of this crap just for a simple enchantment parse

[young knoll]

On spigot it’s just Registry.ENCHANTMENT.match

[umbral ridge]

yea

cant use that

[young knoll]

¯_(ツ)_/¯

?whereami

[undone axle]

https://cdn.discordapp.com/attachments/864180468680360027/944415049462714469/whereami.mp4

[umbral ridge]

https://giphy.com/gifs/dog-confused-bone-11Xc3nOtJcb8Aw

i like spigot and paper

[quaint basin]

https://pastes.dev/krDLmKAAF3 Am I doing this correctly to check if a location is in a chunk that has already been generated? I noticed that managedBlock is a blocking method and I believe it's not necessary here. Also, I think I can replace the use of getChunkAtImmediately with another method to get the chunk

[chrome beacon]

Why are you copy pasting Paper code in to your project

or did you just post that as reference

https://jd.papermc.io/paper/1.21.11/org/bukkit/World.html#getChunkAtAsync(int,int,boolean) exists if you want to get a chunk async without generating it

Also you really should head over to the Paper discord

[quaint basin]

because the paper doesn't have a location#isGenerated async

[chrome beacon]

Just linked you the method to use

[quaint basin]

So, to check if a chunk was generated, should I use world#getChunkAsync#thenAccept(chunk -> if (chunk.isGenerated() { (...) })); ?

[chrome beacon]

yeah pretty much

[quaint basin]

i got it thanks

[chrome beacon]

or at least if I understand the docs correctly

tbh it's not entirely clear what happens when the chunk doesn't exist

[quaint basin]

It doesn't seem to be working as expected yea

chunk#isGenerated returns true even though it has never been generated (probably getChunkAsync generates it).

[chrome beacon]

Did you pass false to the generate param

[quaint basin]

oh no

[chrome beacon]

if you don't pass false it will generate

[quaint basin]

i see

chunk is returning null when doing this: world.getChunkAtAsync(location, false).thenAccept(chunk -> { (...) }));

If it has never been generated before, then will it be null?

Or could a chunk not be null but not be generated?

[chrome beacon]

Guess it's null when not generated then 🤷‍♂️

You really should ask in the Paper discord though

[quaint basin]

I'm banned there

[tender shard]

lol

[quaint basin]

because "I'm running a cracked server" https://spark.lucko.me/jT9y5Voyfj

[tender shard]

I'm banned there because I sent a soundcloud link

[onyx fjord]

lmao

[umbral ridge]

im banned there too

because i pissed off some mod or something

[quaint basin]

https://pastes.dev/I7OjmQyDQQ I'm using world#getChunkAtAsync before getting the block to prevent the chunk sync loading; am I doing well?

[chrome beacon]

Yeah that's fine

[orchid brook]

https://prnt.sc/5VLqjNSNM44O
https://prnt.sc/_e8Pa_DA0U1h
Hey, does anyone know how to create custom GUIs like these screenshots?
I'm trying to understand how some plugins manage to completely customize the player's inventory view sometimes they fully remove the default inventory interface, and other times they keep the inventory slots but replace all the items with custom GUI elements
What GUI framework (if any) do people typically use to achieve this kind of result?
And technically, how does it work ? are they using packets to hide/modify the inventory, or is there another approach?

[vagrant stratus]

resource packs

[ivory sleet]

with resource packs as optic said, usually people will use nexo, oraxen or itemsadder to glyph the inventory title with a massive sprite

additionally you may use custom item textures for icons etc

[orchid brook]

I know that they use resource pack but im talking about the fact they remove the player inventory part of the gui

[ivory sleet]

u can override vanilla gui

[orchid brook]

So no need to remove the bottom part of generic54 png and display back the bottom part when needed with packet ?

[ivory sleet]

eh I mean you can

but you can also just use a massive gui sprite

[orchid brook]

Yeah but i realy want to only render the top inv

that why

like this one: https://prnt.sc/_e8Pa_DA0U1h

[ivory sleet]

then you might need to remove the bottom part of generic 54

or well, u could use another inventory type for that

doesnt have to be generic 54

[orchid brook]

wdym ? like maybe use a 5 row interface ?

[vagrant stratus]

Most UIs will end up looking like this once you get into resource packs lol

[ivory sleet]

yea for example or lets say shulker box (iirc that has a separate sprite)

but if u want that transparency effect, u likely need to erase it from an existing vanilla texture

[orchid brook]

Yeah that could work but it limits me to only 3 rows of buttons for my custom GUIs. I think I'm going to try a different approach like removing the bottom part of a Generic_XX inside resourcepack, and then manually sending the player's inventory interface back to them via packet outside of my customs GUI by putting the player inv interface inside the title of any gui they open

This way I can support custom GUIs from 1 to 6 rows and only render the top inventory portion. The only downside I can think of is the hover highlight effect on the player inventory slots that I probably can't remove, but I think that's acceptable and can just be ignored

[vagrant stratus]

just look at the OriginRealms resource pack, see how they handle it lol

Obfuscated iirc, but just load it up as a ZipFile or whatever & save the contents to disk

[ivory sleet]

eh I mean a lot of servers just remove generic 54 completely

and then just reapply with a glyphed title

[orchid brook]

Im gonna take a look thank uu

[vagrant stratus]

Easiest way to see how this stuff is done is to just... look at a server that does it already lol

[umbral ridge]

not really

[vagrant stratus]

depends tbf

[ivory sleet]

it can be hard to reverse engineer

[umbral ridge]

its not really the best to look at everyone else and see how its done, best practice is to do it yourself, think things through

[vagrant stratus]

Not too difficult to just load up one or both of these

https://modrinth.com/mod/server-pack-unlocker
https://modrinth.com/mod/nbttooltip

[umbral ridge]

take time

[vagrant stratus]

it is a start though

9/10 times it'll be a glyph or invisible armor stand, or some sort of item with a custom display model

[ivory sleet]

i mean like for this inventory title stuff, its common knowledge reverse engineering it is basically unnecessary

but for more intricate things like player stable models etc, u might wna look into how existing servers do it

[vagrant stratus]

not sure how player stable models would be handled actually.
Though the Display entity certainly helps lol

https://septicuss.notion.site/Basic-Concepts-f93bcf8045714a99886825954608a63c

There's this site too, though I'm not 100% sure how up to date it is kekw

[wet breach]

I usually only do this if I get stuck on something

happens if I am doing something or working with something I am not super familiar with

[tender shard]

InventoryFramework is shit and introduces breaking changes with each release

It also still relies on obfuscated mappings IIRC

For nms stuff like anvil input

[orchid brook]

Yoo is jitpack down ? bruh

[slender elbow]

how unamusing and predictable

[tender shard]

lol

yeah jitpack is shit

[mortal vortex]

who the fuck even owns jitpack

the site isnt explanatory at all