dev-general

[cinder flare]

yeah i love jetbrains man, truly one of the best companies

YouTrack is also like really nice for ticket stuff, never knew that

[brazen ether]

yeah I gave up tryna have neovim match it, making intellij into a nvim like experience is way more realistic for me than vice versa

[cinder flare]

figured i'd just suffer with Jira like everyone else, but no we use YouTrack and it's a super nice UI and experience

[cinder flare]

yea basically same, I just use LazyVim for NeoVim stuff, but I 99.9% of the time just use IJ and IdeaVim

[brazen ether]

yeah I have all the config and file stuff done via nvim/yazi, which is a surprising amount

god bless Russian wifi

5 hours left 💀

[cinder flare]

oh mah gawd lol

[desert horizon]

I think I fixed it, tho the solution wasn't nice.

Basically, I nuked the project from my Desktop and cloned it from my repository... Good thing I didn't had any major changes.

Tho, only time can tell, if that actually fixed things...

[brazen ether]

@cinder flare tysm

got ij profiling for mc plugins working

[cinder flare]

hell yeahhhh

that font is weirdly smooth but i love it

[brazen ether]

fantasque sans mono!

[prisma wave]

I can’t believe measureNanoTime is so expensive!!

Kotlin more like bloatlin

[brazen ether]

top ten moments developers didn't know how to read flame graphs

[prisma wave]

I am reading it correctly aren’t I?

[brazen ether]

or was that a joke 😭

[prisma wave]

It was a joke haha

[potent nest]

Brister Mitten joking??

[cinder flare]

i dunno bro... java.lang.Thread.run seeming pretty chunky right about now

might need to cut it out of our program

like, it's literally Kotlin, I don't need no java.lang in there yk

[prisma wave]

So much bloated code

[prisma wave]

For the first time ever

In history

On a more serious note it is kinda funny how there’s almost zero useful information in the graph when it’s this zoomed out

[surreal garnet]

Wait tf is that?

[brazen ether]

flame graph

[surreal garnet]

Oh interesting

[brazen ether]

hooked up intellij profiler to Minecraft plugin

[surreal garnet]

Interesting lol

[brazen ether]

better than spark reports :>

[surreal garnet]

Mainly to test performance I assume?

[surreal garnet]

😭

Got deets somewhere bout it?

[brazen ether]

yeah, working on physics engine

[surreal garnet]

Nice nice

[brazen ether]

with intellij ultimate you just add a run configuration via jar application and set path to server jar and working sir to server dir

afaik doesn't support remote profiling tho ):

[surreal garnet]

Oh interesting

[surreal garnet]

That sucks tho

Would be interesting if it does

I think it’d be funny to watch the server that eats our mem, just watching the stats on it

😭

[crude cloud]

you do know spark can do flame graphs, yes?

[brazen ether]

I didn't know LMFAO 😭

guess I js never clicked that

unless you mean just producing snapshots and viewing them via smth else

[brazen ether]

oh shit

that's awesome on Spark's end

[pastel imp]

not sure this is the best place but figured some people here did CS

[cinder flare]

i mean i literally just did all the required homework in my classes and that instilled it pretty well iirc

most of my profs also had like extra exam prep worksheets that were mockups of what the exam was, just with slightly different questions, and that really helped cause you could get comfortable with the format and problem process

[sweet cipher]

If you're given example problems do them, do all the homework, look at past exams if you have access, make sure to read and understand all the relevant topics

[brazen ether]

if they're struggling with intuition, go over the proofs

my linear algebra book has proof exercises for every topic and that's way more helpful than anything else

[wind patio]

Well, practice makes perfect, as usual.

"ChatGPT?"?
Well we used Google back in the day 🧓 Obviously if you are stuck on something it will help, one way or another.

I didn't really like maths myself but yeah, doing the homework and a little bit extra is all it took to pass the classes. What also helped was looking for real world applications (if applicable) for different topics.

[potent nest]

ChatGPT wasn't around when I had to learn that stuff 😭

[brazen ether]

I wish more devs would spend 1 week on the geometrically-intuitive side of linear algebra

the amount of times people end up struggling their way through trigonometry when their task is EASILY accomplished by vectors is baffling

[potent nest]

I still don't understand why "devs" would study CS

at least here it's far too theoretical

[brazen ether]

the theoretical part itself is the draw for me

[pastel imp]

Software engineering is uncommon in europe, and here, CS is valued way more when seeking a job. Specially in luxembourg.

[pastel imp]

Tbf I don't even know how to properly use it for studying

[steel heart]

What people are you referring to rn? Just devs in general or who exactly are we talking about

[brazen ether]

devs I've seen ask on spigot or in here

someone was using trigonometry to get a point some distance in the direction of a player's view direction 😭

and countless questions online about how to rotate points

[desert horizon]

CS as in C Sharp?

[potent nest]

Computer Science

[desert horizon]

Ok

Would've had the perfect pic for C#.

[crude cloud]

for the better

[cinder flare]

god I really should go back and try to relearn linear algebra. the one college class i had did not instill it onto me very well, i'm still scared of matrices and dot products lmao

i do understand vectors, at least in the context of games 😄

[brazen ether]

they're very simple once you do enough with them :D

[cinder flare]

yeah, my eyes just kind of glazed over for the pure theoretical stuff we were doing in that class, it was all just pure mathematics, no application

need me some practical exercises to truly understand yk

[pastel imp]

Matrices are pretty chill

[cinder flare]

yeah but like actual practical applications of them and stuff i don't have much experience with

i know just enough about like vectors to be dangerous lol, but not much more

gotta find some cool blogpost someone made or something that will teach it in a revolutionary new way

[brazen ether]

the geometric intuition is pretty simple

neutral networks as well ofc

I liked this video https://youtu.be/Ixl3nykKG9M?si=QEdSgmsSAeeMejUI

guy does a great job except for where he doesn't explain why derivative of a matrix vector operation is A^T with respect to the vector, which is js a property of how Jacobians are laid out. but that's a very minor thing

[brazen ether]

if you did highschool calculus you can understand this

he basically reteaches everything you need to know

[brittle leaf]

i love making little hacky things like this just so i dont have to do return true everytime in my command classes

[wind patio]

[]args just looks odd

so does no spaces in (sender,command,label,args)

and no empty line after CommandExecutor {

and two empty lines after

  }


}

you should probably make it abstract though

and you most likely don't even need Command command and String label params

at least I never really used them

[desert horizon]

IJ would in my case report this as C-style arrays iirc.

[wind patio]

yeah

String[] args makes more sense to me idk

[brittle leaf]

it autogenerated as that

[brittle leaf]

ive never used them either tbh

[ocean quartz]

It's actually correct (with the exception of needing a space before args), it should be @NotNull String @NotNull [] args
You're declaring that the string array is not null @NotNull [] and that each element of the array is also not null @NotNull String
Similar to @NotNull List<@NotNull String>

[brittle leaf]

ive not styled my code to have spaces inbetween each parameter, should i do that?

aswell its not my code that the nonnull annotation came from, its the 1.21.5 version of CommandExecutor thats doing that

[crude cloud]

nah, that'd be String args[], String []args is the normal kind of declaration but it's still really fugly lol

[brittle leaf]

like im not 100% that ive learnt all the best ways to actually present my code yet and its been awhile since ive actually coded something

also something i keep doing is remaking certain bits of code to how i wouldve done it before, feels like im making some sort of progress in how i actually make my code

[ocean quartz]

There are multiple styles you can choose from, you could follow google's guideline, I personally dislike a few things about it though so I mostly mix things up

https://google.github.io/styleguide/javaguide.html

[crude cloud]

nullability annotations were a mistake

[ocean quartz]

So true

[brittle leaf]

something im thinking about rn is for custom items to have an mutable creator class for said custom items, then for the base version with nothing that will change have an immutable version that i can retrieve if i need to reference anything from that item type

since currently i just have 1 class which means that in theory they could be changed afterwards which i dont really want

immutable item class for representing its attributes thatll always be unique for that item.

mutable item class for defining the immutable attributes before its thrown into a hashmap for referencing later.
another mutable item class that uses the immutable version of the item and allowing for defining stuff on an individual item basis such as custom enchants, leather armor colors, etc
basically very similar to how itemmeta works

[pastel imp]

tf first time seeing this

good to know

[ocean quartz]

Wait till you find out about @NotNull String @NotNull ... args

[pastel imp]

Lol

[distant sun]

First annotation is for the first parameter and the second for the rest or smth like this?

[crude cloud]

no, that's the same as @NotNull String @NotNull [] args
a better example to explain is @Nullable String @NotNull [] args, a non-null array of nullable (String) elements, so something like @Nullable String @NotNull ... args is just that

[crude cloud]

my biggest gripe with intellij's static analysis is when I have a @NullMarked class/package/module, and inside that class I have a Map<K, V>, whenever I do Map#get, because the return type is V, it (rightfully) infers it's a @NonNull V when it might not be >:(
I know intellij has some "external annotations" setting or something where you can tell the static analyser to treat certain classes or methods with annotations you can configure, like Contract and such, I need to look into that

[brazen ether]

java -> kotlin

@NotNull String @NotNull [] args -> val args: Array<String>
@Nullable String @NotNull [] args -> val args: Array<String?>
@NotNull String @Nullable [] args -> val args: Array<String>?
@Nullable String @Nullable [] args -> val args: Array<String?>?

[crude cloud]

yes

okay

[cinder flare]

god i love kotlin

someday i'll gain power at my job and force us to write new stuff in kotlin

and life will be good

[ocean quartz]

grok convert this project to kotlin

[crude cloud]

I wish to never see what that grok-created abomination can end up as

[ocean quartz]

I'm hella curious but also terrified

[crude cloud]

just how many racist comments will it contain

[sweet cipher]

Can you annotate it as Map<K, @Nullable V> or is that not what you mean?

[crude cloud]

well, yeah you can, but that is not technically right

i am not storing nullable Vs

but Map.get returns null if there is no mapping

is the point

[sweet cipher]

Oh I see

[ocean quartz]

That's where this would be nice

get -> @NotNull V | Exception
getOrNull -> @Nullable V

[distant sun]

Though what would count as nullable var arg but not null values? method(<no argument = null>), method("Hi", "There")?

[prisma wave]

or simply Optional<V>

[distant sun]

Cool stuff @prisma wave

[prisma wave]

ty!!

the union one is cooler i think but it was probably a bit too complicated to showcase hehe

[crude cloud]

no that'd be an empty array, you need to explicitly pass a null array like method((String[]) null)

that simply means "the array itself can be null, but if it isn't the elements mustn't be"

[desert horizon]

I think the List equivalent would be like @NonNull List<@Nullable String> or something?

Also, since we speak of nullable and nonnull annotations, anyone else has used jspecify for their project?

[crude cloud]

I've been using them here and there in personal projects, it's nothing revolutionary or mind-blowing, we've had nullability annotations before, DefaultQualifier and such

they are overhyped lol

it's just an annotation

the one thing I appreciate is nullability annotations that are only TYPE_USE and not TYPE_USE | METHOD | FIELD | PARAMETER | ..., so they do not duplicate in javadocs like jetbrains annotations

[desert horizon]

I mean. I do like the @NullMarked annotation, allowing me to have everything automatically marked as @NonNull unless explicitly marked as nullable

[crude cloud]

yeah but that's not new

checkerqual has DefaultQualifier

and other annotations also have NonNullByDefault

[desert horizon]

Question here is, if those annotations are supported by IDEs like IJ and recognized properly

[crude cloud]

cq yes

it's been around for the longest

[desert horizon]

And one thing I wonder about NullMarked (Can't recall if I actually send such a question to the people behind it) is, if I mark a class as NullMarked and then f.e. have public @Nullable String method(String value), would this automatically make it have value be considered nonnull, or would it inherit from the Nullable annotation applied to the return type?

[crude cloud]

the value is still non null, Nullable is applying to the return type, not the method

I believe you can @NullUnmarked the method, but that means "no explicit nullability" (like without any nullability annotations at all), not "all nullable"

[desert horizon]

One annotation I would like to have (not sure if CheckerFramework has it) is one that makes IJ shut up avout a "return value of method is unused" for methods of a Builder class

Also, wasn't there one annotation lib that had stuff like @PossiblyNullValue or smth to mark entries of lists, arrays, etc as possibly-null?

[prisma wave]

thats just @Nullable isnt it

[wind patio]

@ThereIsAChanceThatThisValueMayOrNotBeNull

[wind patio]

but yeah, exactly my thought lol

I mean it would only make sense to me that you would use @PossiblyNullValue to indicate that this should not be null, but in some sort edge-case it may, but personally I would just use @Nullable, either way you would need to handle it

[brazen ether]

you would hate Zig 😭

[prisma wave]

Perhaps a @ChanceOfNull(0.25) to indicate a 25% chance of it being null? and then the IDE warns you 25% of the time

[wind patio]

wasn't really talking about IDE warnings, more like for another dev or smt

a comment would suffice too, but as I said, I personally would just use @Nullable

[prisma wave]

@ChanceOfNull(0.00001)

just to troll

@NullableIfTheWeatherIs("rainy")

[prisma wave]

yeah anything more specific is of questionable usage

if you still have to null check then you might as well use Nullable

[prisma wave]

There's also @UnknownNullability which is more or less what you're talking about

[distant sun]

lmao

[cinder flare]

isn't exactly this the point of contracts?

like @Contract("!null->!null, null -> null"), etc.

[prisma wave]

oh yeah that too

[desert horizon]

I still don't get the Contract annotation tbh

[prisma wave]

this means "not null argument implies not null result, null argument implies null result"

[crude cloud]

see for example

@Contract("!null,_ -> param1; null,!null -> param2, null,null -> fail")
public static <T> @NonNull T
requireNonNullElse(final @Nullable T value, final T fallback) {
  if (value != null) {
    return value;
  } else if (fallback != null) {
    return fallback;
  } else {
    throw new NullPointerException();
  }
}

cc andre ig

hm, i suppose i wouldn't necessarily mark fallback as nonnull

but it's basically useless if it is null ig

annotation elitist will also mark with pure = true lol

oh god why do code blocks look like pure garbage on mobile

[ocean quartz]

They used to have highlighting on mobile and it was actually prettier than on desktop, now it doesn't anymore no idea why

Also now there is a copy button

[crude cloud]

that's been around for quite a while lol

[potent nest]

Fractional Permissions but for nullability 😭

[old wyvern]

@SchrodingersNull

[desert horizon]

ItemsAdder plans to implement a new file structure to improve things...
And since I didn't want to spend time updating all my existing furniture to the new format (It's 192 chairs and 12 tables alone) did I instead spent time making a python script to automatically generate the YAML content based on the available wood and wool types.
Later also added automatic lang file creation because why would I want to write 200+ individual entries by hand?

https://paste.helpch.at/aruruviyel.py

I feel surprisingly happy about this tho tbh

Also, lol

Discord detects the URL as "potential harmful download" xD

Probs due to the .py ending

[distant sun]

Neat

[desert horizon]

Ye

[wary kestrel]

first message is the dead channel I didnt know about

[wind patio]

only real og's are here

[wary kestrel]

lmao, I am not an OG but im here now!

its where we can plan to overthrow the #off-topic

[prisma wave]

So true

[dry harbor]

I support that idea let's do it

[oblique heath]

we can overthrow off topic but we still have to stay on topic

so i java propose we python start plotting typescript the revolution php like this

[wary kestrel]

setup a 2gb 2008 laptop as a server running in the other room 🔥 someone put red sharpie all over the screen, thank god for SSH

[brazen ether]

what kind of workplace are you in 😭

[wary kestrel]

my house XD

found old laptops that where gonna be taken to recycling, so i did that....

[abstract osprey]

@brazen ether what was the MSPT / tps like during that test?

#showcase message

looks cool though

[brazen ether]

didn't measure here but for 1000 object pile it's ~30MSPT (so 20 tps)

im aiming for another 2-5x performance improvement :)

[brazen ether]

ah well

the MSPT impact on main thread is like 4

it's all async

but the physics simulation takes 30 Ms per tick

[coral hazel]

So, I've been working on a library plugin for a bit alongside my actual project, every time I need something that I think could be useful for other projects too, I add it to the library, and I wanted to ask the other developers here if there were any tools you need/want? either consistently or just here and there, maybe something you haven't made yourself whether it's because it's too hard, lack of time, or just laziness (I don't judge, been there myself lmao).

So far I've got:
A system for saving code in specific entities to be executed on right click.
Helpers for saving data to Entities and ItemStacks
Line renderers, both using particles and block displays.
JSON file management for human readable save data
Byte file management for saving whole class instances that can't be serialized to JSON
Structure helper, for structure copy pastes, movement, or in my case multiblock recognition.
Some common RNG and miscellaneous helper functions

Any suggestions?

[quiet sierra]

Entity component system, a system for custom itemstack components, abilities / skill systems

Some sort of "value modifier" system for things like custom attributes, damage events etc

They're all things I've done at work and they're all things I wished exist before

[coral hazel]

Since posting it I have actually added a system identical to the entity right click one, but for damage instead haha

[coral hazel]

thank you so much for the suggestions!

Would you be willing to expand on the first 2?

[quiet sierra]

Tools for debugging / messing with PDC

An entity component system (ECS) basically consists of a system where entities have "components"

And you have "systems" that use those components and do something with them

Instead of, for example, an entity consisting of a type, id, location, health data etc, an entity just holds "components" like a HealthComponent, PositionComponent etc

Then systems (HealthSystem, for example) are the ones actually ticking logic for those given components

You can have an entity with all sorts of logic that's invincible because it doesn't have any health to begin with

Click handling can just be a component

https://en.wikipedia.org/wiki/Entity_component_system

For custom item components, the idea is that items already have the data component system, but those components are inflexible because you can't register new ones (as they have to be synchronized to the client)

Instead, your plugin could manage components for server-side use, like, for example, tracking backpack data, right-click rewards, custom food etc

Here's a code snippet from work

As for abilities and "value modifiers", it gets really complex, really fast

[coral hazel]

ah, I see!

thanks for the explainer ^^

Wouldn't it be pretty simple for the itemstack components to be stored in their PDCs?

Though I'm assuming it only seems like that from a surface level

[quiet sierra]

they would but you'd benefit from some sort of pdc codec system

[coral hazel]

codef? Sorry, self taught so if this is industry standard termonology then it hasn't come up in any of previous jobs

[quiet sierra]

Codec my bad

[coral hazel]

ah, ok, phew

haha

[coral hazel]

I believe I already have that come to think of it...

The byte file management system, but just save the output to an itemstack's PDC as a byte array instead of to a file

IIRC byte arrays are a supported data type in PDC

[quiet sierra]

They are but is it really the best way to store data?

[coral hazel]

unsure, would have to look into it for sure

[quiet sierra]

You could have it as a default / fallback but sometimes if all I care about is a string, I might have some overhead

From my understanding you just made an attempt at recreating kryo

[coral hazel]

it's definetely convenient, but people don't download libraries for the library developer's conveniency lmao

[coral hazel]

what's kryo?

[quiet sierra]

https://github.com/EsotericSoftware/kryo

[coral hazel]

but... isn't this in java by default ;-;

I'd have to assume this exists because of an issue with the normal java serialization

[quiet sierra]

If I'm using external helper libraries I might as well use the purpose built ones

[coral hazel]

fair enough

I'mma run some benchmarks

[coral hazel]

ok, yeah, kryo seems to be about 2-3 times faster than base java serialization according to my benchmarks

however, it also seems a lot stricter in what can be serialized, for example I can't serialize UUIDs

Anyways, I'mma head to bed, gn folks 👋

[quiet sierra]

You have to register classes to kryo or configure it to be self-registering

I use kryo at work and it serializes uuids just fine

Worst-case scenario just pass a custom serializer

[coral hazel]

I already tried that, the error on UUIDs is thrown on registration

[coral hazel]

that's true though

[coral hazel]

ah, found the problem!

The automatic serializer didn't correctly choose the default UUIDSerializer when registering UUID, so the fix was to change kryo.register(UUID.class); into kryo.register(UUID.class, new DefaultSerializers.UUIDSerializer());

[brazen ether]

made my tree structure 30x faster by making it stored as a flat array

did not think it would have this big of an impact

[potent nest]

30x sounds odd

[brazen ether]

no recursion and much better cache coherence

flattened its data as well

which were AABBs

gonna do further testing ofc but it looks correct so far

[brazen ether]

(specifically, 30x faster insertion)

[crude cloud]

hm? how did you benchmark it?

[brazen ether]

insert time for 200k bounding boxes

after enough warmups to where times didn't deviate

this is for a clean tree

so no interleaved removals which would impact its cache locality negatively

but that's not my usecase so I don't really care

[pastel imp]

I do recall having to do this in a past project at uni

or heard it in a lecture

idk anymore

[brazen ether]

it's for a dynamic AABB tree with a volume hueristic + local rotation

Box2d's slide deck basically

[brazen ether]

yeah its a cool optimization :)

[crude cloud]

you should look into proper microbenchmarking with jmh

i am assuming this is in java, yes?

[pastel imp]

yeah 30x does sound weird though

I recall it being faster

but not THAT faster

idk the O notation anymore though

[brazen ether]

I've already done that a large amount with other parts

I don't really care

and that's not the purpose of JMH

I'm testing this in the prod environment

[brazen ether]

I'll double check to make sure everything is correct, I agree it is weird

[brazen ether]

it's the same O

[brazen ether]

yes

I used JMH heavily when optimizing my constraint solver

but here I'm testing things where the time to complete drops literal seconds

w functionally identical algorithms

[wind patio]

faster doesn't mean it went from O(log n) to O(n) or similar
Big O notation tells you how fast or slow a program or algorithm gets as the input gets larger

[brazen ether]

Nope it's about right

I'm getting 10-40x faster

40x when N is higher as overhead goes down

[brazen ether]

in real usecase

after warmups

[crude cloud]

microbenchmarking is not the purpose of the Java Microbenchmarking Harness?

in which you can do parameterization of the environment for how the microbenchmark runs to compare different sizes of inputs?

[brazen ether]

JMH is useful for when it's difficult to isolate the code you're using or you need to set up a way to test it with realistic loads and you can't test it with realistic loads easily with an existing setup

• it handles timing stuff well

but this isn't necessarily needed if what you're testing is already easily measurable in your environment

[crude cloud]

difficult to isolate the code you're using
I wholeheartedly disagree

I'd say it shines when you are able to isolate the piece you are benchmarking

it is a __micro__benchmarking tool after all, to benchmark individual aspects and elements of a class or method

@Param my beloved

otherwise you just get a ton of noise

[brazen ether]

this is what I meant

I meant that if you're in prod and your code is not isolated there, you should isolate it and test it with JMH

[crude cloud]

yes, which is a great way to test something as basic as a data structure

[brazen ether]

since JMH let's you do each dummy data setup properly

[brazen ether]

yeah but I don't have a reason to do this

[crude cloud]

lol

okay

[brazen ether]

I'm testing it in its usecase

what benefit would JMH offer

[crude cloud]

apparently nothing, just a waste of time

[brazen ether]

constraint solving required complex data gen which is why I used JMH

I couldn't test it well on a server

but this isn't the same

[brazen ether]

also multi threading reasons

and GC and other stuff

but mostly the data gen part

also multi-second benchmarks don't require complex timing

if you're testing anything <10ms you should definitely use JMH no matter what

[pastel imp]

I am aware of what O is, hence why I questioned the O, those 30x results could've happened only due to low input

[brazen ether]

testing production loads is what matters, which is what I'm doing :>

microbenchmarking is a tool in the toolkit

[potent nest]

so you did many things and one part of that was flattening an array

[brazen ether]

I flattened all the data of a tree

mb I grouped those together because it the same concept of replacing a field access with an array access

but along the way yeah that ofc reduces memory allocations and such

so there's also impact of that

[brazen ether]

btw y'all were right Abt the performance thing not being accurate, it's a much less crazy improvement in reality; I had made a tiny mistake in the cost function causing it to create a linked list instead of a tree, giving a superficially correct output but not being actually useful

mistake was 2 missing characters 😭

[wary kestrel]

sad

[brazen ether]

trying out Claude code properly

to build a testing framework for a battle feature

its made some dumb mistakes at times... and the main thing I've noticed is it's easy to start simply not testing its work

but for menial shit that's not too complicated, along with a very healthy amount of code review, it is good enough

I'm not really letting it make any architecture decisions

[pastel imp]

Recently used lovable to prototype something and I was amazed tbh, it did it almost first try.

[wary kestrel]

I personally love it it’s just handy

I use it all the time it’s great for writing documentation doing small and even semi large things

Terrible at code optimization

[brazen ether]

I also find that it's kind of like very slow for refactora

[quiet sierra]

I don't really use any AI agents to build code

I prefer autocomplete AI

Copilot is ass, I myself use supermaven which has been abandoned but still works

I've heard good of windsurf? or whatever it is

[brazen ether]

maybe just because I use vim motions and intellij and have defent typing speed but it's wayyy slower at for example changing the usage of something at a bunch of places

[brazen ether]

I find this to be much more annoying ngl

its a negligible speedup for me when it does work

and it oftentimes gets things wrong

the time impact of AI auto complete required checking if what it gives you is correct + reaction time delay if you're not waiting for it, or a delay either way if you are waiting for it, + potential modifications. all of this for me is usually much slower than just typing it myself

[half harness]

btw have you tried copilot on VSCode?

it suggests things in other lines or in general places where your cursor currently isn't

really useful for refactoring and stuff

since you change one thing and then it recognizes all the other things that needs to be changed along with it

[brazen ether]

i have not;

I've seen this feature though

I don't know how I'd feel about it

but I do imagine it's a lot better than just auto complete

I've gotten quite good at vim macros and editing so I would say in general this usually isn't my bottleneck

but for example I need a Kotlin DSL for the battle testing code (the only reasonable way to do something like this since the test scenarios are quite complicated) and I absolutely was not going to write that myself when it's mostly bs code

still took a lot of hand holding and Claude doing dumb shit but I've made a fair bit of progress with it, i can confidently say it was better having AI do it and doing review than doing it myself

my passion shit like the physics engine and Comet is what I wouldn't bother trying AI with since it's usually genuinely complicated to do and needs to be done properly... but something like a testing framework for a battle sim does not really need the same level type of critical thinking

[half harness]

oh yeah definitely
I probably wouldn't trust AI for that stuff

but rather more for tedious stuff

although

[half harness]

iirc copilot supports using claude as well, so I wonder how it compares vs asking claude manually

[half harness]

mmmmmmmm vim

I gotta learn that someday

especially for command line - still using nano (and I don't know how to use any of its features either :p)

[brazen ether]

lots of code review :)

[brazen ether]

I'm ngl I'm hoping to move away from work that's less interesting to me in the future so on one hand it's cool but on the other I want to be doing stuff where LLMs are less applicable 😭

[brazen ether]

I just vim motions in Intellij and Neovim for text editing, it's just really nice on my hands

[brazen ether]

I'm ngl

AI is too stupid to design the testing framework

its implementation isn't actually useful because it doesn't use enough of the real code... which it could have easily done since I had already done the prep to make it completely independent of being actually ran in a server

[brazen ether]

been working on my packet optimizer, have gotten a decent result of 6.5x less bandwidth for 30 Comet fire effects while retaining similar visual quality

it works by taking in the data about a display entity over its lifetime, some tolerances for scale, color, etc, and then trying to find minimum number of data that needs to be sent over the server to fit those requirements

open source as Mengshe on my gh, still very bad and very memory-churny code but it's been a cool start

this already works way better than my last attempt at this problem, with weeks less of work

[wind patio]

so what, you trade network bandwidth for increased memory usage and for like 80% accurate representation of the original?

[surreal garnet]

Honestly I’d rather have lower memory usage compared to lower networking, not fully sure good use cases for it.

Sounds like a interesting project tho.

Only thought would be if you’re sending 1000s of packets per tick to the user, but even still they should be able to handle it

[brazen ether]

?

its just marginally higher memory usage?

[brazen ether]

it can be 100% accurate in some cases, for example if the transformations are well modeled linearly

[surreal garnet]

Still, I have actual gamemodes that use 50~ GB of memory each.

And use most of that still lol

[brazen ether]

like there's no way storing position and transformation data for an entity over some ticks is really a big issue

[surreal garnet]

Not saying we aren’t leaking stuff, ye

[brazen ether]

• object pooling anyway

memory usage is not really an issue

and either way it's always solvable

[brazen ether]

6-7x less packet data is quite significant, I don't understand your point about the memory tradeoff, it's not really a real issue

the actual tradeoff is being able to things simulate ahead of time

[surreal garnet]

Yeah ig

[potent nest]

that sounds like something is going wrong ngl

[surreal garnet]

Tbf one of them had 512 dedicated to it back in the day

Just because it had to store so much stuff in memory 24/7 with like

1-3k unique on that server alone

[potent nest]

that still sounds sus

[crude cloud]

what does "6-7 times less" exactly mean here?

do you mean 1/6th / 1/7th ?

because obviously it can't be n - n*7 lol

[brazen ether]

.

[brazen ether]

thats what "6-7x less" means yes

it differs between the effects I tested, some are inherently easier to optimize than others

for example lots of color changes necessitates a lot of metadata updates

I use a ∆E threshold to reduce # of these updates but ofc some is still unavoidable

the optimizer currently works by finding the 100% optimal way to place the teleportation updates to fit the position constraints, and then works off that solution to solve the rest. despite it not being the global optimum it has given really good results

[crude cloud]

that's stupid

[brazen ether]

and for the fire particle specifically it finds it in microseconds, tho I haven't optimized it much yet

[brazen ether]

same way people say "2x less" to mean 1/2

[crude cloud]

yes

that's stupid

[brazen ether]

I feel like it makes sense

[crude cloud]

it doesn't, it's stupid

[brazen ether]

:|

[crude cloud]

half is half, not twice less

english is stupid

[brazen ether]

both work

its not an English exclusive either

[crude cloud]

people are stupid then

[brazen ether]

feels like something most languages would say

[crude cloud]

it's a stupid expression

[brazen ether]

ik Russian and Spanish do

[cinder flare]

wait until you find out how the french counting system works lmao

they start whipping out multiples of 20s and shit lol

[brazen ether]

tbf it's not like they think of it like that

[cinder flare]

yeah but i mean that's how like all the numbers are named after 20s lol

[brazen ether]

kotlin value classes are cool....

have found good use for them in high performance data structures

i have a lot of cases where i thnk of things more as pointers (indices into an array in reality) to store them efficiently in an array, but that means it's quite simple to mix up the value at a location and its index (since they're both of the same type)

however with value classes you can really easily wrap a primitive and bam, it'll be a different type at compile-time and cost you nothing in performance at runtime

    fun find(i: UnionIdx): UnionIdx {
        var idx = i
        var idx2 = idx

        while (idx.value != arr[idx.value]) {
            idx = UnionIdx(arr[idx.value])
        }

        var t: Int
        while (idx2 != idx) {
            t = arr[idx2.value]
            arr[idx2.value] = idx.value
            idx2 = UnionIdx(t)
        }

        return idx
    }

@JvmInline
value class UnionIdx(val value: Int)

example of my usecase

here arr is an Int2Int collection so i have to get its value there, but overall it still improves the overall safety

(no this is not the same as type aliases, the point of this is specifically for them to not be thought of as the same type)

[half harness]

wha
how have I never seen value classes before

[brazen ether]

valhalla shadowing

since these are like

they can only hold 1 field

so they are nowhere near as useful as what Valhalla aims to bring

and ofc jvm is still gonna box you with these if you're using generics

but they can be very useful in some specific cases

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

[brazen ether]

and yes for clarification, these act exactly as primitives here; there is no memory allocations being done despite the kotlin constructor calls, they are compiled down into static methods that return a normal jvm int

no boxing necessary for type safety!

[potent nest]

Yeah it works just like in Scala

[cinder flare]

god i can't wait for valhalla to do this in pure java

[brazen ether]

yes

I really really hope we get like

flat arrays of value objects

I will be majorly disappointed if that part doesn't work well

[brazen ether]

they said this would happen

but they also said that it would be the JIT's decision on when to do it

they said like "there's gonna be no 'flat' keyword, JIT is smarter than programmers about this" and then proceeded to compare it to restrict from C

and uh

I'm ngl to you man the JVM is not that smart

so hopefully it doesn't require jumping through ridiculous hoops for it to flatten value arrays

[potent nest]

Not sure who they is, but generally leaving the decision to the JVM just makes it easier to e.g. support multiple platforms

[potent nest]

It‘s smarter than 99.9% of the programmers

And if you find a missing optimization opportunity, you can just contribute an improvement

[brazen ether]

this is not what i mean

it rarely "misses" and optimization opportunity

it's that developers often do things on accident that make it impossible

also, i do not want to contribute to the JVM

[brazen ether]

or jvm can't know certain things about a situation, and it's not like you can hand the JVM a list of statements for it to use in its proofs

"this will only be accessed from a single thread, i pinky promise!"

[potent nest]

threading is basically the one place where the jvm is allowed to assume that it's only accessed from one thread in certain circumstances :p

[potent nest]

[brazen ether]

i explained that badly, i was referencing a specific example of smth in my codebase but yes that's true, since CPUs do do that

but my point is that it's not a JVM thing

it's that developers must write their code in such a way that the JVM can optimize it

which may be impossible or extremely difficult at times

due to the fact that the JVM cannot know all the things that are true in any given situation, while the developer might have additional information

[potent nest]

yeah that's just general limitations of computing

[brazen ether]

hopefully i don't have to work with java for much of my life 😭

i do not expect that to be the case if i go into physics like i plan to

[potent nest]

python

[brazen ether]

Julia 🤔

i dunno

i'm gonna try gpu programming for another physics sim

i think i would do something in computational physics

so still a whole lot of programming

[potent nest]

Julia is okay-ish from what I've heard

but it probably lacks proper tooling like so many languages

[brazen ether]

god bless FAWE

async block data access 🙏

[brazen ether]

multi threading is awesome

for my physics engine i need to diff the world around bodies, which means a lot of block access; i looked at how FAWE handles async access, implemented that, added multithreading... 0.6mspt to check and compare the states of 16000-27000 blocks

this is crazy fast

it still could get more optimized but probably gonna leave it there

[distant sun]

What newsletter service do you guys recommend that's easy to integrate with a custom website? Tbh I only need a way to send emails to users with latest offers. I've tried kit.com, seems nice, but for some reason I don't receive the confirmation email for my subscription -.-

[prisma wave]

mailgun?

[distant sun]

maybe one that's cheaper? idk how many users I'll have, but 100 mails/day is a bit too little and I don't feel like paying $15 for it.

[prisma wave]

100/day is too little?

it's the only one i know about sorry haha

but 100/day seems like a lot to me

[distant sun]

wait I read it as sending mails to 100 users, is it individual mails sent to however many subscribers you have?

[prisma wave]

i think so

[distant sun]

that doesn't sound right, I think it is number of recipients you can send emails to. If you have 200 subscribers maybe 100 will receive the mail today and the other 100 will be scheduled for tomorrow?

@eager fern you have websites, what do you use?

[eager fern]

sendgrid to send emails

[frigid badge]

i’ve always used sendgrid

[agile galleon]

I'm a fan of brevo

SendGrid's interface was so sluggish to use

[quiet depot]

I use sendgrid cuz I got it for free with github education years and years ago

[quiet depot]

yeah it sucks but I pretty much never have to log into it since I just use their smtp server not their api

[distant sun]

Is hacktoberfest giving t-shirts again? 😱

[pastel imp]

Yup

Already have mine xd

[distant sun]

Damn it

[quiet depot]

o really?

kinda cool

favourite one was the 2019 one

[pastel imp]

I mean u cna still try ur luck

Although only 10k available

[pastel imp]

It does look nice, although last tshirt we had (2023?) Was hella nice too

[quiet depot]

ah I can't remember the last shirt I got

definitely not 2023 I don't think

I think I've got 2019, 2021, and 2022 or something

[pastel imp]

This is the one I meant

(Mild use)

[distant sun]

Probably already out of stock 🙁

[pastel imp]

I'd assume so yeah

[potent nest]

suffering

[desert horizon]

lol

Honestly should've expected it...

But at the same time do I not see participating in it again tbh

[distant sun]

Not really, after 2y of stupid virtual badges

[desert horizon]

Ngl, I don't like the look.

[desert horizon]

Yeah, the Holopin stuff is so stupid

I have yet to find a person who proudly shares their holopins...

At least with stickers you have something to slap onto your laptop or other surfaces... But these pins are useless in pretty much any possible way.

[distant sun]

Ye fr

[desert horizon]

Only good thing is the tree planting... But looking at the HF page, I can't see any mention that even that is still a thing

This also reminds me of Modtoberfest lol

kinda want a variant for plugin development tbh

[potent nest]

plugins can take part there too

at least it was possible before

[pastel imp]

It is

I got a free tree when I claimed the tshirt

But they require me to create an account in tree nation

And I really don't like the sound of that

[desert horizon]

I do have a tree planted and recently got a mail about its birthday... lol

[pastel imp]

lol

[lavish notch]

Hey Folks, has anyone ever worked with Minecraft's Translation Keys in Bukkit/Spigot? (I know Paper/Adventure has better translation support - I'm looking to avoid those as a requirement/compiling with it)

Mojang has some items that utilise translation keys directly on items with NBT like so:
"minecraft:item_name": '{"translate":"block.minecraft.ominous_banner"}'
I'm looking for help correctly parsing the translation key in Inventory Titles as it just appears to display the JSON as a string rather than parsing it.

[onyx loom]

I dont think spigot supports components in titles and will only accept string, atleast that was the case a while ago

[lavish notch]

Yeah, that's what I've read from googling about the place - it's kinda annoying. I am just considering to say fuck it, and force english translation for gui title.

[prisma wave]

it’s probably possible with nms for what it’s worth

[crude cloud]

the lengths people will go to to avoid using paper api

[lavish notch]

It's more I want something that works on Spigot for the idiots that still use Spigot. 🥲

[crude cloud]

for the 10% ?

i mean, if it was up to me today i wouldn't bother, it's too much effort to maintain compatibility with a platform that barely gets any use, especially today after Paper is no longer syncing with Spigot's changes, they are effectively two separate platforms, like neoforge and fabric

but if you wanna go for it, i'm not gonna stop you

[lavish notch]

You do make a fair point.

[pastel imp]

Wait why is it a requirement to not use MM?

[pastel imp]

Actually, it's way less

In latest versions

Spigot is a dying ship

[lavish notch]

MM?

[pastel imp]

The only reason I see devs not getting rid of it is because there is no better alternative for selling plugins

[crude cloud]

because spigot has no component support

[crude cloud]

source?

[pastel imp]

BBB and polymart kinda mhee

[crude cloud]

my source is kenny's secret graph

[pastel imp]

Last time I checked it, it was using mini's?

Idk who was it

But it wasn't kenny's

[lavish notch]

I don't suppose anyone knows off the top of their head what version of Paper added support for Component in Titles?
https://jd.papermc.io/paper/1.21.8/org/bukkit/Bukkit.html#createInventory(org.bukkit.inventory.InventoryHolder,int,net.kyori.adventure.text.Component)

[crude cloud]

since components were added to the api in 1.16 iirc

[lavish notch]

Does Paper impose Java limitations on their server jars? I'm attempting to boot Paper 1.14.4 on Java 17+ and am getting a error message saying to use a maximum of Java 13.

I could have sworn Paper 1.8 runs on Java 17 happily and doesn't have similar limitations seen here with 1.14.4.

Yeah, wtf.

[half harness]

agree to the eula D:

[lavish notch]

The 'wtf' was to Paper 1.8 working with Java 17 (well Java 21 in my case) and not Paper 1.14.4.

[half harness]

oh

[lavish notch]

Accepted the EULA, Paper 1.8 definitely works with Java 17/21.

[potent nest]

there was a time where spigot was overly restrictive

[half harness]

try java -DPaper.IgnoreJavaVersion=true ...

¯_(ツ)_/¯

[potent nest]

paper later changed it to allow loading with that flag

not sure which version though

[lavish notch]

Oh, do you know if Spigot has the same issue? I guess I'll check.

[potent nest]

pretty sure spigot is more restrictive there

[lavish notch]

Give me a sec and I'll download spigot to confirm.

[potent nest]

sounds like a powershell issue lmao, maybe put "" arround the flag

[half harness]

yeah I was confused on why that error was showing

[lavish notch]

Okay, yeah. fml.

[potent nest]

https://stackoverflow.com/questions/26614969/how-can-i-pass-batch-arguments-with-dots-into-a-powershell-script

[lavish notch]

Like this?
java -"DPaper.IgnoreJavaVersion=true" -Xmx5G -jar .\paper-1.14.4-245.jar

Or do you mean the value of the flag?

[potent nest]

that was just a wild guess, see the stackoverflow post for the apparently correct solution

[lavish notch]

I don't really know what I'm looking at there to be honest but here is command prompt instead:

[half harness]

ig it wasn't added in 1.14

[lavish notch]

It has me wondering how other plugin developers are navigating this with Java 17+ requirements... maybe they are just ignoring this 'issue'?

As I've seen some other plugins on Spigot mention "Java 17+" as a requirement.

[potent nest]

plugins compiling to 17+ won't run on these old mc versions anyway because of outdated asm dependencies

[half harness]

o wait actually

so that version check actually does serve a purpose

[potent nest]

well, the server runs fine on newer versions typically, but plugins compiled against newer versions cause problems due to the bytecode rewriting stuff

[lavish notch]

I mean in my instance, the plugins function fine since they aren't doing any crazy shit - it's a shop plugin I'm working on.

As it works on 1.8 and 1.21 just fine - just not 1.14.4 I guess due to this Java limitation.

[potent nest]

the server will still try to rewrite it I guess

1.8 probably doesn't have any rewriting

not sure

[half harness]

wait what is the rewriting for

besides the legacy materials stuff

[potent nest]

mainly that for spigot I guess

sometimes to avoid some breaking changes

e.g. bukkit has its own Function or Consumer interface or something like that and it got replaced?

there were also times when Bukkit.getOnlinePlayers() returned an array I think

[pastel imp]

yeah

I remember that

[potent nest]

that means you're old btw

[pastel imp]

tf you mean

I am just 21 💀

[sweet cipher]

That's past middle age

[pastel imp]

F.

[potent nest]

I don't make the rules

[lavish notch]

Well this is fun. Done some checks and my plugin is incompatible with 1.13 to 1.16 due to these versions having a maximum Java version supported lower than Java 17.

[cinder flare]

It’s okay nobody uses those versions anymore

[lavish notch]

I wish.

It's ShopGUI+ which is old af and has oldies still using versions as old as 1.8 because somehow Factions still has people playing.

[cinder flare]

You can still support 1.8 and 1.16.5+

The only one with a named label is 1.16.5 lol

[lavish notch]

Well yeah, 1.8 to 1.12 has no issues with Java 17+. It's just annoying as there will be atleast one bugger who leaves a 1-star review bewcause of it.

[cinder flare]

Hmm, no bstats? Then we could know for sure how many people would be impacted, sad

[lavish notch]

I got no bstats or similar sadly other than some skewed stats from an addon that was published in recent years.

https://bstats.org/plugin/bukkit/ShopGUIPlus SellGUI/9356

[cinder flare]

Well still, general playerbase-wise, essentially nobody is using those versions anyways

[lavish notch]

I would love it if MC was a 'versionless' game where everyone was forced on the newest version - would make life so much easier. 😄

Unrelated but is the BStats 'Server Location' data broken? All location stats are showing 'no data to display' for me.

[crude cloud]

you should try joining that same server with a minecraft client 😄 you'll find that it does indeed not "run on java 17 happily"

[lavish notch]

I know you have to turn off the native transport feature but it does work.

[brazen ether]

for my kotlin bros

how would you feel about this

(overloaded division operator)

[ocean quartz]

I personally hate it

[pastel imp]

Same

That's scuffed as hdll

[ocean quartz]

Hdll?

[brazen ether]

hell

mispelled

[ocean quartz]

Oh lol

[brazen ether]

this vs File(tmp, "hello.txt")

it's less familiar which makes it a little worse automatically

[half harness]

wait I feel like this would be neat tho 😭

[brazen ether]

i think in an individual case it makes sense too

[ocean quartz]

A plain reason of why I don't like it is that just looking at that I would have no idea it would return a File

[half harness]

oh that's true

[brazen ether]

i don't think that's very fair

[pastel imp]

Yeah hell

Lol

[brazen ether]

unless you have type annotations turned off

[pastel imp]

Typo

[brazen ether]

that's not a real issue

in intellij it would look like val file: File ...

[ocean quartz]

Yeah but it's not as readable, I would rather have something like file(temp, "hello.txt")

[half harness]

but then what about using Path#resolve("hello.txt")?

it'd have the same issue

[brazen ether]

its a tradeoff between just a few characters of concisity and having to remember 1 more little thing

not the type information (ide shows it) but perhaps being unsure of what the function does immediately

[abstract osprey]

don’t get me wrong, I’m a die hard kotlin lover

but there does come a point where it becomes slightly more inconvenient to read

that’s probably crossing the line

[half harness]

I don't understand the hate with that :(

[crude cloud]

why would you want to write code that is illegible

it's confusing to read

[brazen ether]

subjective

[crude cloud]

it isn't clear at all what foo / bar means apart from regular division

if your brain has to accommodate for "oh, actually, it isn't what / is for" then it is confusing to read

[half harness]

although to be fair, / is also the symbol used as path separators (well, I guess windows uses \ but it allows / too)
but yeah I can see how if the variable naming isn't clear, then it can be confusing

otherwise, it'd be like ```kt
val file = directory / "file.txt"
// vs
val file = directory.resolve("file.txt")

but python, for example, does this already: https://docs.python.org/3/library/pathlib.html#operators

and cpp: https://en.cppreference.com/w/cpp/filesystem/path/operator_slash.html

imo as long as it's clear what's being done (e.g. not val result = initial / next) then it shouldn't really be that unreadable

[half harness]

and ruby as well it seems

I was not expecting to find this many languages lol

[brazen ether]

I don't think that's confusing, it's just a different "name" for a function

like after you read it for the first time, it's pretty idiomatic

and it's not as though it's an uncontextualized "foo / bar"

its in the context of IO code

I dislike operator overloading when the operation doesn't make sense in the context or it looks ugly... and true this isn't at all "division" but it is a separator and that still makes sense to me personally

I'm ngl tho I wish Kotlin += *= ... could be expressions

so that you could do like m1 *= m2 *= v2

and it would mutate instead of creating a new object like * implies

and technically I couldnjust make * be a non pure multiplication overload but THAT absolutely oozes confusion

the alternative is named infix functions tho.. which is an option

[wind patio]

Personally I don't really like it. I feel like this is just pointless and you are not gaining anything by using it.

and also several people already said it's confusing

I think its fairly obvious, but there are quite a few books about code clarity. The code should be simple, straightforward and readable.

If you are working alone and on a small scale project, sure, quirks like that maybe make sense, but not if you are working with a big team and on a huge codebase (imo)

[wind patio]

this is also a good point about clarity.
also I feel like if those two lines
val tmp = "/tmp".toPath()
val file = ...
are close by it makes sense, but if one is on the line 23 and other on 89 it's not as obvious

[crude cloud]

yeah we'll just have to disagree, I, as many, find it difficult to read even within context, especially on a team where you have to read hundreds of thousands of lines of code written by other people, I would want it to be very explicit and clear what everything means, writing potentially more confusing code where the mental strain to remember what division means is higher seems like a risk than a reward

you also say "it's just a different name for a function", but that's the problem, it is not a name at all, it is whatever name you give it within the domain of what it does

anyway you're not going to change your perspective but if you want to write hard to read code then more power to you

[ocean quartz]

You could see it as a name, but I'd argue that it'd be just as bad path.dividedBy("file.txt")

[crude cloud]

precisely

[ocean quartz]

That's how path / "file.txt" reads

[crude cloud]

also dkim, saying C++ does it is probably an argument against it

stdout bitshift left string

/hj

[brazen ether]

alright now this is a reach 😭

[brazen ether]

this is valid though

ngl I think if you deal more with math then you're more likely to like operator overloading

because it's a field where people make up new notation constantly if it's sufficiently helpful for the task at hand, oftentimes just to be less verbose

[ocean quartz]

I like operator overloading, I just think it needs to make sense as the operator not as a "cool" syntax sugar to a function

[wind patio]

realistically, what is gained in this case? I mean overloading / for path

[brazen ether]

4 less characters

to write

every once in a while

so not a huge benefit

[wind patio]

exactly

[brazen ether]

I don't think I'd actually do this

[ocean quartz]

listOf("element") + listOf("element2") is a great example of an operator overloading that makes sense

[crude cloud]

it's cool that you can, but like uncle Ben said great power something something

[brazen ether]

any time operator overloading saves you parenthesis spam (math) that's awesome tho

[wind patio]

I mean yeah it might look cleaner/nicer, but imo clarity is more important

[crude cloud]

it can certainly be easily misused so it's easier to forbid it in a large team's rule set than it is to lay down rules as to how and why it should be done

[potent nest]

inb4 the result is listOf("elementelement2")

[crude cloud]

that's some javascript level shit

[potent nest]

that's just how e.g. vector addition works!

[brazen ether]

Heh... true

||sorry for ping 😭||

learning zig... this is so much faster than java

my first ever time spending more than a few minutes

the equivalent java program i wrote in 20 min, this took me 4 hours

didn't optimize the java version but in reality the code is honestly quite similar

then again this is a physics sim so it's not like i was using much of java's features anyway

so it's basically free performance by using a lower level language

[prisma wave]

have you tried "RUST" (it's memory safe btw)

[potent nest]

if you don't use much of java's features there shouldn't be much of a performance difference

[brazen ether]

I'm not sure I care enough to optimize it and test it thoroughly

actually just for that reason I will next week

[pastel imp]

I mean, at the end this is a public thing no?

doubt anyone would use a zig or whatever lib/plugin

[brazen ether]

?

wdym

[brazen ether]

this isn't for minecraft

unless i rewrite my physics engine in zig

which i probably won't

i'm planning to be a computational physics major and i've been wanting to write a sim in a more low level way for a while

after college applications finish i'm gonna start working properly on (probably) a fluid sim and then test out different methods for solving PDE's there (regular numeric vs using neural networks) and both for performance and for learning purposes, i want to write it in a language like Zig or C

[pastel imp]

oh

[pastel imp]

oh that's hella cool then

thought this was mc related

[brazen ether]

naur not directly

@potent nest i'm probably just bad at Zig, but yeah optimized java version is only ~10% slower

[potent nest]

there isn't much reason why it would be significantly worse

[brazen ether]

java has lots of little tiny overheads

but i guess those do turn out to be just tiny

again tho i'm like, very bad at Zig

so this is not a fair comparison

[potent nest]

eh, not sure what these overheads would be

[brazen ether]

worse cache locality

[potent nest]

not really

[brazen ether]

very slightly

even if they're all created together, the object headers still make it worse

damn Java is so fast

[potent nest]

but in contrast the GC is allowed to compact fragmented objects to improve cache locality :p

[brazen ether]

there's no memory usage here

but yes

[brazen ether]

i dont know how much of an impact this has in the real world

cause cache lines aren't that big

what are the chances that some data you wanted together got fragmented, then put back together

i mean like yes that happens but i can't imagine that making a big difference

[potent nest]

I mean it's not just L1 cache

[brazen ether]

oh true

yeah ig with smth as simple as this sim it probably all compiles down to very similar assembly

less memory usage still a win but yeah wow that's not as big of a difference as i thought it would be

like i understood beforehand that it shouldn't be that big a deal, but still

[brazen ether]

zig comptime stuff is NUTTY

https://github.com/markisus/zig-infix-parser

https://github.com/Laremere/alg

alg.math("a * b * c") is parsed at compiletime

meaning that there's no overhead with this

really just so much cool shit you can do with this

[steel heart]

mfw unsafe D:

[frigid badge]

https://tenor.com/view/ferris-rust-rustlang-crab-cute-gif-26396486

[prisma wave]

I don’t understand this last example at all

Why is it multiples of 5

oh

Wait I’m stupid

Sorry!

[brazen ether]

I'm gonna write the same thing as a learning exercise

[brazen ether]

well that was easy:

    const E = try parser.parse("x * y + z");
    std.debug.print("Result: {d}\n", .{E.eval(.{
        .x = 5,
        .y = 5,
        .z = 1,
    })});

didn't add matrix and vectors and stuff

but those are trivial to add

comptime is AWESOME!

the AST is generated completely at compiletime

and there's no overhead with dynamic field access or anything because all the identifiers are known at compiletime

[still portal]

https://www.minecraft.net/en-us/article/removing-obfuscation-in-java-edition

[prisma wave]

wow

[crude cloud]

Blimey, even

[wind patio]

pogers, even

[cinder flare]

wait yo that's actually sick

i remember they released the mappings a while back but they were basically useless cause of legal concerns right? like we still used yarn mappings and stuff for fabric iirc

[potent nest]

mojang mappings exist

I think fabric has legal concerns, but paper doesn't care for a long time already

[cinder flare]

oh yeah you're right, paper started doing mojang mappings a few releases ago right?

i remember they remap plugin jar files on first startup

still nice though! and great to see that paper is already ready to go, as per usual

[pastel imp]

md_5 is gonna have a fun time xD

[wind patio]

this might be a relatively weird question, but is anyone using vscode for java dev?

I am having a weird issue with plugins

The one from Redhat works pretty much perfectly (though missing some refactoring stuff)
https://marketplace.visualstudio.com/items?itemName=redhat.java

The one from Oracle just fucks everything up
https://marketplace.visualstudio.com/items?itemName=Oracle.oracle-java

[potent nest]

the redhat one is based on eclipse, the oracle one is based on netbeans. That answers basically every question

[crude cloud]

have you worn wigs?

[crude cloud]

will you wear wigs?

[wind patio]

common netbeans L

but seriously how do I move from IDEA to VSCode, VSCode is missing so much stuff I can't bruh

[potent nest]

I mean you're probably doing something wrong

[ocean quartz]

but seriously how do I move from IDEA to VSCode
You lower your expectations

[crude cloud]

why are you doing that to begin with?

like, what's causing you to?

[wind patio]

eh I mean, IDEA is basically for java, where vscsode is trying to be for everything

[wind patio]

license is running out

[crude cloud]

why can't you use the community edition?

[ocean quartz]

I do web dev (react with TS) on IJ with no issues at all, IJ can do a lot more than just Java

[crude cloud]

the intellij family can do a wholeeee lot of languages very well but that's a separate conversation

[wind patio]

well, technically I could

honestly I haven't looked at the differences between ultimate and community

[ocean quartz]

Biggest ones are probably things you don't use, like DB integration

[wind patio]

used it briefly, a nice to have, but I guess not mandatory

wait whatever happend to Fleet? is it still relevant or, rather, becoming relevant?

[crude cloud]

having used both ultimate and community editions, currently using CE on the daily, i can't think of anything i'm missing personally

[wind patio]

iirc it's basically just vscode copy, but from JB

[crude cloud]

there was one thing but i can't remember

[ocean quartz]

Fleet is getting the google treatment

[crude cloud]

must not be important enough to be remembered

[wind patio]

I remember all the hype around it but it kind of died out

[frigid badge]

when will you wear wigs?

[wind patio]

I guess there are quite a few differences, question is which ones do I care about lol
https://www.jetbrains.com/products/compare/?product=idea&product=idea-ce

sucks community doesn't support Spring

[frigid badge]

honestly just paying the 170 a year is cheap for the products you get

[wind patio]

kind of saving money rn and it's not like I need it that much for personal use

I have all products pack on my work laptop, but that has a cloud license or whatever, so I can't use it on my pc

[pastel imp]

Fy WebStorm exists too for webdev 🙂

[frigid badge]

I only have rider and webstorm on work, but all products for personal

[wind patio]

I've only used PhpStorm, since I pretty much only use Laravel lol

[frigid badge]

a little less than 15 euro a month is cheap for it all

[pastel imp]

well it sucked

Only good thing about Fleet was it's simplistic UI

[ocean quartz]

Didn't really find any real benefit between using WebStorm or just using IJ

[pastel imp]

which they are kinda bringing to IJ with their new islands theme

[wind patio]

All of the functionality in WebStorm is available in IntelliJ IDEA Ultimate, either out of the box or via free plugins. There's no need to purchase licenses for both IntelliJ IDEA Ultimate and WebStorm unless you like having a dedicated tool for each language. Then WebStorm may be the right choice for focused JavaScript development.

[ocean quartz]

Yeah that

[wind patio]

basically WebStorm is stripped down version of IDEA (Ultimate? not sure about community)

[pastel imp]

It has some small extra builtin stuff, but I mostly the same yes. I only used it cause it did use less resources due to being only focused on web dev

[ocean quartz]

What small extra stuff?

[pastel imp]

But nowadays, specially since I am actively working on hangar, I use Intellij since hangar's backend is in java anyways

[ocean quartz]

I couldn't find a single one

[pastel imp]

eh been a while but I remember that some plugins were automatically installed or smt

[ocean quartz]

Automatically installing plugins is hardly a reason to pay :')

[pastel imp]

yeah

I agree xD

I just have it for free

so I didn't really consider that part

[pastel imp]

But nowadays I use Intellij for everything as stated here

it's just sad that intellij kinda sucks lately

performance has been worse, more bugs, etc.

specially with the 2025 edition

I thought it was just skill issue on my part but apparently a lot people are reporting having issues with IJ

[wind patio]

haven't noticed anything tbh

[crude cloud]

can't say it happens to me

[ocean quartz]

Same

[crude cloud]

and i use intellij on 3 different environments

[pastel imp]

I haven't had any issues since the latest update though

so, we shall see

but I do wish it was less resource intensive

But well, the IDEA engine is old as shit I assume

[wind patio]

post the pc specs

[pastel imp]

less resource intensive is mainly a comment for my laptop, but it's not really a big issue

the issue is (or well... was? maybe?) bugs and issues.

But fy, laptop is an Lenovo 7i Aura Edition 16GB ram, I7U 256V Lunar Lake and desktop has a 7900XTX, 32GB ddr5, R7 7800X3D

[wind patio]

seems fine

[cinder flare]

don't you get a perpetual license? so you just wouldn't get updates?

and if you have/had the student pack, you get like a 40% discount on buying them outright for a few years iirc, so that could be pretty nice!

[wind patio]

uh I'm not sure tbh, I read the terms or whatever and fairly certain it meant I was not covered

[forest pecan]

do you still have your college email

[cinder flare]

oh weird, i swear you just had a perpetual license for whatever versions you were paying for, so you can just use them forever and not update

[cinder flare]

As of November 2, 2015, JetBrains ceased the sale of perpetual licenses for IDEs and .NET tools. However, you're eligible for a perpetual fallback license with an annual subscription or after completing 12 months of on-time payments with a monthly subscription. Licenses bought before November 2, 2015, remain valid indefinitely.

ah I see, you didn't have 12 months of subscription?

[wind patio]

I suppose so yeah

[wind patio]

the emails are deactivated after a year or so after you graduate

so sadly no

[forest pecan]

💀

[wind patio]

yep

I remember my GH account primary email was my university one

I'm glad I didn't realize too late to switch lol

[forest pecan]

OSS license, or beg boss to make a deal with jetbrains 🥲

or switch to community edition

[cinder flare]

i think they have it at their work

this is just for personal use

[forest pecan]

who's stopping you from using it

/s

[wind patio]

well I used the one I had at my previous workplace since it was tied to an account

and I could login on my personal computer

and now the new workplace has a license server that is only accessible under a vpn

[pastel imp]

ah that sucks

well, honestly, I just use ultimate cause I have it for free

I'd otherwise use Community, the difference is not big

[wind patio]

student pack?

[pastel imp]

I have only used the "Code with me" thingy like once or twice in my life xD

[pastel imp]

yes

[forest pecan]

i have my school email permanently

so i could theoretically have the student pack forever ig lol

[pastel imp]

th, lucky

[ocean quartz]

Used it once at work to go over some code and explain but then we were like "why tf are we doing this" and called and shared screen instead lol
(forgor to reply to the message, was talking about code with me)

[pastel imp]

on a completely different note, has anyone ever done a component library (specifically for vue)?

[forest pecan]

Code with me is something you touch once because it's cool and never again lmao

[ocean quartz]

Yeah lmao

[forest pecan]

nobody wants a google docs ide

[pastel imp]

only time I used it was during a project at uni, we were using PlantUML and for some reason only the PC hosting showed the graph so we hooked his PC to the projector and used mine to code lmao

[forest pecan]

😭

[wind patio]

it would be nice if it was faster and easier to setup (not saying it's difficult, but some people are very not used to it)

I kind of like it tbh

but yeah, not something that you use often

[forest pecan]

unrelated but i literally hate my college class rn

my professor covered python in a single lecture

react in a single lecture, shell scripting in a single lec

[pastel imp]

so uh, I assume no one did?

[forest pecan]

and now there's a midterm 💀

[pastel imp]

that's typical tbh

my professor of prog2 covered RiskV, C and Java in 3 lectures lol

you still think you fucked? xD

[forest pecan]

Yes lmao

[pastel imp]

yeah, I think so too

[forest pecan]

whatever

who cares about school

😭

[pastel imp]

exactly

lets hope pray that we get jobs like matt

the legend got a job in a block game

[forest pecan]

im prolly not going to go into game dev

but i'm not sure

[wind patio]

@ocean quartz where r u working bruv

[forest pecan]

where are you working

[wind patio]

well, in short terms banking sector

worked in energy sector at my previous workplace

[forest pecan]

ic

[pastel imp]

fy lots of math and physics

but you also get lots of math if you go into CS

[pastel imp]

MCCI

[wind patio]

I work in banking and I don't have to deal with any math

https://tenor.com/view/grimace-cat-grimace-cat-frozen-cat-froze-cat-not-moving-gif-17585113

[pastel imp]

in other words, matt works for NoxCrew

[forest pecan]

afonso have you done internships?

[pastel imp]

oh yeah, in your actual job chances are you don't use math at all, but for some reason a CS degree requires 20 stages of cancerous math

[wind patio]

yep

[pastel imp]

wellllll, according to my university, yes

in reality no

[wind patio]

calculus 1/2, probability & statistics, yadayada

[forest pecan]

i have to take physics 💀

maxwell's equations

[wind patio]

I had physics too, was probability the easiest class ever lol

[pastel imp]

At my university, instead of having an internship, they have a full intensive 7 week full blown project that mimics the process of working at a company, with design phase, implementation, etc.

so no, I officially did not do an internship

[forest pecan]

Ic

[pastel imp]

I have done several workshops at Amazon (AWS) though

[wind patio]

yeah same here actually lol (maybe not exactly the same), though we also did have interships

[forest pecan]

Yeah

[pastel imp]

and got an internship offer there actually, but couldn't take it due to other reasons