💀 Cortex Engine

[void condor]

if the externs are optimized, yeah, significant improvements. https://github.com/dazKind/linc_meow/blob/main/linc/meow_hash_x64_aesni.h#L271

I have yet to see someone use SIMD in their haxe-code 😄

[grizzled laurel]

I mean

Kha does SIMD iirc?

[void condor]

kha & nme are like 80% extern code

[grizzled laurel]

yeah i guess

[void condor]

published my clay externs here: https://github.com/cortex-engine/linc_clay

[grizzled laurel]

'nicee

lovely

[void condor]

mkdir linc_yoga 😄

[grizzled laurel]

lol

what would poor linc_clay think

this is abandonment

[void condor]

Nature will have its way

[grizzled laurel]

xD

[void condor]

wtf, we just reach 5k messages in this thread...

[vapid quail]

This thread has 5079 messages

[grizzled laurel]

we talk too much

lmfao

[ionic badger]

I think you're fine

[void condor]

xD

[void condor]

yoga is smooth sailing so far

[grizzled laurel]

How is yoga compared to clay in terms of usage?

Which API do you prefer?

[void condor]

a bit too early to praise/hate anything, i wanna make it functional first. but what is clear, is that yoga is a lot more mature and the apis focus on what matters. clay is too young and is lacking in certain spots + tries to do more ui-library stuff like dealing with text

[grizzled laurel]

I see

[void condor]

like, you advertise a layout-lib and it emits batched rendercommands? just gimme the layout

[grizzled laurel]

isnt that just an UI lib at that point xDD

[void condor]

less than 4 hours and yoga is working

compare that with clay: 3 days just to get all the boilerplate externs going

now all that is left is wiring all the styles and shit

[green breach]

Good job Facebook?

[grizzled laurel]

Facebook making something good?

[void condor]

for sure

[grizzled laurel]

Wait i wonder

did they make this for React Native

[void condor]

yep

[grizzled laurel]

yeah i just looked it up

sick

yeah no wonder it is mature

[void condor]

Here is a frame that adds the same UI. bottom is my old algo in haxe, top is yoga

[grizzled laurel]

oh shit

the old method was expensive

the new method is still quite expensive but wayyyyyy better

[void condor]

If I can shortcut my way to increased performance by not writing code myself, totally worth it

it still needs some more love but there is enough room for improvements everywhere

now I need to rebuild my css to use proper flexboxing everywhere

[grizzled laurel]

good luck

[clear dome]

Making clay externs was cursed from the beginning, you know it

[void condor]

not a total loss. Improved the cppia generator macros quite bit to handle structs better 🙂

also a lot more fluent writing externs now

[grizzled laurel]

you know, the macro stuff you do should really be in haxe itself imo..

[clear dome]

Btw, do you have more info about that macro extern stuff? Like where should I start if I wanted to do the same for my libs

[grizzled laurel]

like there is no reason haxe can't automatically generate stubs for cppia

[void condor]

boxes ❤️

[void condor]

See

• https://github.com/cortex-engine/linc_clay/blob/0843aa9e982aec83d1c318b87544dab2bad3878b/clay/HxClay.hx#L337
• https://github.com/cortex-engine/linc_clay/blob/0843aa9e982aec83d1c318b87544dab2bad3878b/clay/HxClay.hx#L327
• https://github.com/cortex-engine/linc_clay/blob/main/linc/LincCppia.hx

[void condor]

ok, setting up and building new apps apparently is super easy now

guess I found a pattern

[void condor]

hmmm, I think the issues people are seeing with hxcpp-debugger and cppia is due to the fact that cppia-builds dont retain their stackframes files & positions. Did debug stackframes ever work completely with cppia?

starting the debugger on the script I can see the classes but there are no debuggable files for them

this might also be the cause why people think you need to compile your scripts into the host, making working cppia kinda pointless for debugging

cppia left, native right

all being debug-builds

@obtuse burrow what do you think?

[obtuse burrow]

I think cppia scripts do retain the stack frames. I threw together a quick script and if I get it to print out the callstack it appears works correctly

Foo is the class defined in the script

if I change it to throw an exception the host also correctly shows the line and file of the cppia script which threw

but if I run the program through vscode with the hxcpp debugger cppia stack frames are shown, but it cannot find the source files so I can't navigate / step / set breakpoints, etc, etc

I tried defining absolute-path when building the cppia script but that doesn't appear to change that debugger issue

the cppia scripts do clearly contain absolute paths now though

[obtuse burrow]

looking at this I've also found the cause of the issue with host interfaces and cppia implementations you came across on the callables branch, it's a regression in nightlies in general

[void condor]

oh this is all pretty interesting

[void condor]

not sure if the vscode extension obfuscates the issue, you might wanna try the embedded local debugger, you can check the included sourcefiles vs. included classes that way directly

[obtuse burrow]

that was my initial attempt but there's very little docs on the debugger and cli debuggers always make me feel like I've gone back in time several decades. So I switch to the vscode one

just gone back to it and figured it out, seems to work fine

stepping into cppia functions and showing the callstack works fine

Foo / Foo.hx is also shown in the outputs of the files, filepath, and classes commands

[void condor]

so you compiled the host without the script code?

[obtuse burrow]

https://gist.github.com/Aidan63/d4ae6c1e2c6b4db3f6967d1c3a37b221 thats the gist of the small sample I've been testing with

IFoo is in the host and exports that to the dll-export file

the script is compiled importing that and defines an implementation of that host interface

hangon... I've been using dll-export in the client too 🤦‍♂️

ok, cppia script is much smaller now. very large script should have caused me to realise something was wrong...

all seems to still work though, local debugger can step around fine, stack traces print fine, and the files / classes commands show the class in the script

[void condor]

ok, I need to test this. maybe this is on my end :/

[obtuse burrow]

this is with haxe 4.3.x and git hxcpp

just tested with a nightly which I know doesn't have the cppia issue and that sample works fine as well

[void condor]

yeah that sample works for me as well. though I have to test some more stuff

im such a fucking idiot

Im building cortex cppia without -debug and never noticed

more than a year

ok, stepping around works, but breakpoints dont seem to work

[void condor]

i guess my mystery has been solved, not sure If I care about the debugger. But this in particular is a win

[ionic badger]

https://media.discordapp.net/attachments/956928971927339018/1346954448223604828/image.png?ex=67ca1095&is=67c8bf15&hm=179320c1d2db7666ab505abe4654c4019ccf171cbff9f87562ba41842d3aa6b2&=&format=webp&quality=lossless&width=1440&height=297

i'm testing compiling a project with tracy on win 11 and it is also failing

is it just win 10 that's compatible?

[void condor]

no, not at all

sometimes there is a problem with the version of the windows sdk but that is about processor features not linker errors

[grizzled laurel]

Try removing your output directory

I got that error sometime ago as well and building from scratch fixes it for me

[ionic badger]

i did that twice

[void condor]

linker errors like that can also happen with hxcpp cache when switching between release/debug builds

haxelib run hxcpp cache clear

[ionic badger]

oww never knew about that command

booted back into linux

wanted to see if the error was the same or not

[ionic badger]

cache isn't defined

deleted all output directories and trying again

yeah, same issue

[obtuse burrow]

Issue should now be fixed in nighties, also pulled in that change to the callables branch. Hopefully thats the cause of what you were seeing.

[void condor]

ah nice! Will give it another shot tonight

[void condor]

oh, you should try to work with the cache when using hxcpp, speeds up native compiles significantly

in any case I think you have an issue in your setup, be sure you have the right versions of everything setup

[ionic badger]

What should I be looking for? Project compiles without the tracy flag

I thought the cache thing was on by default lol

[void condor]

a link error means, it doesnt find the cpp file containing the function

[ionic badger]

So tracy isn't findable? I don't recall needing to do anything specific for that 🤔

oh... maybe haxe version

[void condor]

haxe and hxcpp version

[ionic badger]

hxcpp version is fine, just trying to address the haxe version problem

there's something i have to do to get ceramic to play nice with haxelib version change and I can't recall what it was 😅

[ionic badger]

latest haxe nightly fails as well :/

[void condor]

just pulled everything. testing is getting harder since I switched to c++20

[void condor]

yeah, ok I know why I hate cpp

[void condor]

ok, I monkey patched your branch to compile under c++20, native builds work and run fine, cppia builds explode (cppia closures cant be called)

it's a closure that gets added to a callback list in the host

not sure I understand whats going on there

[void condor]

@obtuse burrow here is a minimal sample illustrating this crash behavior: https://gist.github.com/dazKind/d5c159581350aad2625f4cca37a18348

[void condor]

@lethal talon More background/reference info on w-questions: https://gist.github.com/dazKind/0a86a513793ac866a8c299bd4b30c86f

[lethal talon]

Oh thanks ❤️ this is very helpful

[void condor]

There is no explicit "How", we tackle that as part of the "who" & solution concept

really depends on the scope of stuff how we setup that up

[lethal talon]

Makes sense, it's a great template, it's hard for me to sit down and work on the conception pieces because to me it just always sits in my head, but there is power in expressing the ideas.

[obtuse burrow]

thanks, I'll take a look tomorrow

[void condor]

almost done with the flexbox grind

[molten sleet]

i fear css

[void condor]

you are not alone

I was concerned the flexbox model wouldnt cope with my text-layouting. but turns out it's good enough

[green breach]

speaking like a true web developer

[void condor]

word. The fact that this shit just works is hilarious

[clear dome]

My brain never managed to assimilate the flexbox model

Of course I can always manage to do what I want at the end, but it doesn't feel natural

[void condor]

same here. It never felt right since I grew up with box-model and tables. But working on this + some cool docs finally made it make sense for me https://css-tricks.com/snippets/css/a-guide-to-flexbox/

[clear dome]

Ah yeah that guide is good I also ran into it 😄

And totally get it, also grew up with flexbox-free CSS (for the better and the worse lol)

Could just be the wording that also makes it harder to grasp I don't know

Btw I find it kinda funny that you are somehow reproducing the "stylesheet" separated from the "content" whereas in the web ecosystem, Tailwind is becoming more and more popular and is actually getting closer to more traditional approach to styling elements individually like in application development 😄

[void condor]

You just wait till they re-invent themes

[clear dome]

https://tailwindcss.com/docs/theme

[void condor]

hahaha

in any case, I totally dig the separation between visuals, logic and data

I just dont worry how shit looks when building UI, I just make sure I have a clean hierarchy and then I slap some visuals over it

I guess I wrote too much component specific rendering code in the past

(which cortex' ui still allows)

[clear dome]

Yeah, there is probably not a once-fit-all solution to that anyway, both options have their pros and cons

https://adamwathan.me/css-utility-classes-and-separation-of-concerns/ that article from Tailwind author is interesting imo

In practice, I have found myself making the graphics/ui and data model separation quite clear

But my way of building UI is kinda close to tailwind's, in the sense that creating the hierarchy of layout components is part of the graphics as much as how they are stylized

But the data model is strictly separated from that (thanks to tracker library)

That's also not very far from how you make a react + tailwind app connected to some reactive data model

[void condor]

ok, rant incoming

to be honst, I dont give a fuck what any of those fancy web-frameworks attempt to do. It's always the same, "re-thinking" some fundamental concept by adding personal preference and hyping it into mass-adoption. Then some new web3c standards are drafted but ultimately abandoned since people jumped to the next hype thing.... like wtf people, we have serious well functioning patterns and tech that run the fucking web right now, use that shit and keep it basic

[void condor]

again, not critizing you in any way, just pissed at the web-folks

[clear dome]

Oh I'm definitely sick of all the new things being reinvented over and over in web for sure

[clear dome]

No worries 😄

Like, you build an app with node, nextjs or I don't know what (and don't really care)

and six months later somehow that's not how it should be done anymore

But there is one pattern that did strongly stick to me

It's the reactive data model (before React)

[void condor]

the other day I look at this react lynx stuff and seriously, we need another world war so we can assign all these people as the first wave

[clear dome]

More precisely, when I talk about reactive data model, is automatic implicit bindings, the "autorun()" pattern that re-execute code when a depending value changes

This is highly compatible with how I think and build apps

And that's basically why I made tracker

[void condor]

react was cool when its object-lifecycle was the way to setup components

[green breach]

i hate js, but svelte is the least awful thing in the ecosystem

[clear dome]

(my first contact with that pattern wasn't react, it was Meteor, but anyway that's just the first time I ran into that)

[void condor]

I built apps with svelte and meteor back in the day. it was ok

[clear dome]

Now it's definitely a mess

[void condor]

in the end I wasnt really more productive compared to creating and sending html serverside

[clear dome]

But that doesn't mean there haven't been a few good things from it. Need to discard 99% of the trash and keep the good parts I guess 😄

[clear dome]

Yeah that's the network part of Meteor, it had it load of problems anyway

And now it's struggling becose node-fibers are not a thing anymore

[void condor]

but then, dont take me seriously on any of these topics. I like jquery

[clear dome]

But like, now I'm allergic to most code that does explicit UI bindings

like you create some UI

[void condor]

im with you there. that's why I stick to the basics

[clear dome]

And you have 40 lines of "addEventListener(myButton, CLICK, handlePayButtonClick)" and so on

[void condor]

I actually dont mind that at all

makes things very explicit

[clear dome]

(I mean, maybe that's not the right example)

But sometimes the code becomes very hard to navigate because you need to dig through 3 or 4 different files to understand what is happening with the UI

[void condor]

hmm, I think that is just the nature of the beast. That shit can never be easy for free

[clear dome]

Probably, but then that's where thing like Svelte or React make you increase expresiveness and conciseness of the code (if done correctly, you can make a mess with any tool)

[void condor]

hierarchies that dynamically change, you cant(shouldnt) really hide the complexity

ah I dunno

[clear dome]

The fact that you can make a scrollable list of element just by writing a template-like xml (like in React)

VS having to manually setup a UIScrollView, a UIScrollViewDataSource, a UIScrollViewLayout and so on

Well, I see the advantages there

of course that does have a cost and sometimes you need more control depending on the situation

But sometimes if I just want a scrollable list of elements, I'm very happy if I can just describe it in a few lines, which also include what data I'm displaying

Anyway, my attempt to address that in Haxe world is wisdom library

Other options are ImGui-style utilities that also help a lot with conciseness and expressiveness imo

[void condor]

true

[clear dome]

Setup of ImGui-style is simpler

Reactive-data-model stack is more complex, but, imo, can be more flexible, given that it gives you enough control via its API

UI is hard anyway, good luck, regardless of opinion, cortex stuff in that part looks very cool 😄

[void condor]

we gotta kill UI with fire. wisdom is still on my list to checkout

[clear dome]

For now the focus is web/html so pretty much similar use as React, but the idea is to later plug it to native UI like Ceramic's one

[void condor]

gaaah, basis, grow and shrink done

gap & min/max width/height left to be done

[grizzled laurel]

i've taken a look at you guys' rant

i fully agree lol

i like to keep things simple too, simple templating

works well and doesnt deprecate because of some new fundamental change requiring a rewrite of all code

[ionic badger]

i was reading yesterday as well, agree completely, i liked early day react stuff and then it became a massive mess

i don't understand why everything needs to be new and constantly changing

[lethal talon]

Personally I love declarative ui as code, I also found that to let me just build things quickly

Other stuff just gives me a headache tbh

[obtuse burrow]

The problem is the generic Signal class. Since haxe performs type erasure Signal<T> becomes Signal<Dynamic> at runtime, so when the cppia script adds it's CppiaFunction instance to the array it performs no conversion to turn it into one of the new callables.
Some knowledge of the original type parameter is retained though, and when the host iterates through the connected listeners it will reinterpret the dynamic elements into the new callable type before invoking it. But this array element isn't actually a callable, its a cppia function, so this leads to a access violation error.

that StaticCast is the problem, it performs no type checking. Quickly changing it to Cast which does perform type checking seems to get this to work.

Might not be ideal to always generate Cast though, as in 99% of the cases the type checking at runtime is unnecessary. This just so happens to be the 1% edge case where it is needed...

[void condor]

hmmm

that's a nasty one

means that cppia gets a regression with the new callables and the answer is paying extra cast cost for 100% of these cases?

[obtuse burrow]

I think there might be a way around it, seems like the StaticCast function is already aware that some types need special conversion when going from dynamic

Adding the callable type to that DynamicConvertType tag should mean that just callables go through the proper conversion from dynamic, I think...

[obtuse burrow]

yes, that seems to do the trick, just pushed a new commit to the hxcpp branch which should fix that issue

there is now going to be a bit of extra cost to going from dynamic to a haxe function since it will do a dynamic_cast, but only for functions, everything else is unaffected.

it may also still end up faster than before, since previously all function calling was treated as dynamic / resulted in boxed objects

[void condor]

you could limit that fix to scriptable 😄

[obtuse burrow]

yeah, I might wrap it in a #ifdef HXCPP_SCRIPTABLE

[void condor]

then it's win-win

coz this really only triggers when dealing with cppia. didnt have any issues with pure native builds of my projects

looking forward to do some nice before/after testing tonight

[obtuse burrow]

The best solution would be for the createClosure and createMemberClosure cppia functions to return callables instead of instances of CppiaClosure, but I'm not sure how to do that

the callables are heavily templated and need to know the parameters at compile time, but cppia is all runtime. I haven't yet thought of a nice way of going from a runtime cppia function to a templated callable

aside from a ungodly switch statement... which I'm not doing

[void condor]

hmm, cppia is all Dynamic anyway, so I would go for the lowest barrier fix for the moment

[void condor]

ok, linc_yoga is pretty much complete now

smoothest extern experience I had in a while

[void condor]

some more worst case testing between the old layout & style handling vs. new yoga layout + style optimizations

now lets check what @obtuse burrow has been cooking up

[void condor]

Ok, that fix did the trick. cortex cppia runs without issue on the typed-funcs branch now

In general the typed-funcs branch looks like a solid 5% performance win when comparing the frame-times

note, that this depends on how I work with callbacks in cortex. other projects might see a bigger boost

@obtuse burrow Ill continue running that nightly+typed-funcs branch and let you know if I notice any further issues

[obtuse burrow]

great, good to see all the immediately obvious issues aresolved

[ionic badger]

where should I look to figure out how to get tracy to work on linux?

just some basic keywords and locations to aid some google searches

I imagine i shouldn't need to be doing any extern work to get it working

[void condor]

You might, again I dont think I added the necessary stuff for linux compiles. I'll review it this week and see if I can make it work

[ionic badger]

ah cool, thanks!

[void condor]

@obtuse burrow Did you test your branch with hxcpp cache?

I notice that most cpp files get recompiled all the time and I need to clear the cache from time to time since the linker locks up in an inf-loop

not sure what im observing here, but it defo changed to this behavior once I switched to your nightly+branch

[obtuse burrow]

I didn't, no

I wonder if its due to the fact that because it's all templated it's implementation is also in the header, so that causes cascading recompilation

I'll enable the cache and give it a play around

[void condor]

could be one reason, another idea could be you are like regenerating ids/var names with every compile, havent checked though

[void condor]

ok, all the ui stuff is back where it was before functionality-wise. Now lets see if we can actually build an editor 😛

[void condor]

oh, we so game!

hss is lovely. just realized you can even embed small images directly into the compiled css

[molten sleet]

like

straight up images
in css?

[void condor]

yep, css becomes a little assetpack

[void condor]

watching as the deepseek model in my basement is writing documentation for cortex

actually not half bad

[void condor]

@ionic badger Is is just me or did the activity across the whole discord decrease over the last few weeks?

[ionic badger]

I'll check stats, but personally I haven't been around much because i'm trying to get some stuff IRL in order

so can't say for sure

[void condor]

oh dont worry then, hope all is good on your end

[void condor]

say whaaaat

[ionic badger]

Its nothing "bad" just trying to get organised and improve workflows and focus

[void condor]

so im not crazy and something happened end of january

[ionic badger]

not necessarily, things have been super active in general i think

sometimes everyone needs a break

daily view

can see things somewhat starting to trend up

[void condor]

yeah, could be

[grizzled laurel]

But surely that doesn't happen in a day

It dropped like crazy in 1 day

[ionic badger]

first chart is a weekly view

[grizzled laurel]

I know

I mean on the daily chart

[ionic badger]

you also get spikes of high "abnormal" activity

that crash could be related to some hyper interesting HRR session

yeah, almost defo was the roundup roundups, sunday 😄

[void condor]

we need to overlay crypto charts, LLM releases and trump becoming president

[grizzled laurel]

xD

[void condor]

then we have an anwer

[grizzled laurel]

Lmfao

[void condor]

from my perspective, since I parse the channel activity passively everyday, amount of conversations dropped in all main channels

[ionic badger]

you're probably right, i've mostly just been letting go some attention for discord. Just paying attention to interactions and otherwise just avoiding it where possible

[clear dome]

Any national holiday changing the numbers maybe?

[ionic badger]

i think last year summer haxe jam i remember saying i was gonna take a break from discord for a bit

and then i just never did 🤣

[void condor]

xD

[clear dome]

Sometimes I want to take a break from my project, then I realize 10 other projects are waiting behind

[void condor]

You ever need a stand-in, lemme know. Im a good benevolent dictator with a proven track record! 🧙‍♂️

[ionic badger]

i don't necessarily see low activity bad here tho, i like to think of it as people are focused on their projects or their real life instead of discord 😄

[void condor]

taking a break never works for me. things haunt me and I need to continue, no matter how stupid it is

[ionic badger]

you kind of have to "learn" how to "break" 😆

[lethal talon]

Yeah I would imagine a lot of the activity came from the co worker discord voice chats, those dropped off a few days after the round up round ups

[lethal talon]

I am the same way

[ionic badger]

yeah we don't get access to a massive history of stats so its hard to say if the HRR skewed the scales a bit

those 24hr hrr sessions were a bit crazy lol

[grizzled laurel]

AHAHAHAHAHA, this is way too relatable

[grizzled laurel]

same, but then i will be forced to take a break at some point

usually due to lost motivation

which is just unfortunate because if i would be able to take breaks

then lost motivation would be less of a problem

[void condor]

yeah

I know that

[grizzled laurel]

i think the worst kind there is when you have general lost motivation for programming in general

has happened to me once so far and it was for a good year or so. couldnt touch any code, all i could do was stare at my IDE

[void condor]

i have like 2-5 active projects / experiments that I rotate between

I never lost it for coding though

[grizzled laurel]

yeah same haha

[void condor]

that's crazy

hmm

[grizzled laurel]

yeah i cant imagine that either today, programming is quite literally my life at this point. It is what calms me from a busy day 🤣

But yet it happened that i lost motivation, i hope i never get there again

[ionic badger]

for coding and stuff i find mid day breaks are best at managing things, a few hours where i'm just outside doing nothing or something like that. For discord, taking breaks is easier because i don't really like social media (lol) and its interesting to see how things go when you're not around what changes etc

i've taken months off from here before and the place is still the "same" but also very different

[grizzled laurel]

i hate social media too, with the only exception for discord maybe

[grizzled laurel]

dang that is really impressive

it's fair to say the haxe community is a strong one

[ionic badger]

i like seeing it doesn't need me 😄

[void condor]

😄

it's hard to realize what's driving our mental health

[grizzled laurel]

yeah for sure.

[void condor]

But Im happy that you guys are ok

I also noticed that there is activity / engagement when stuff happens. Just look who woke up and engaged here 😄

So people are watching, maybe not enough good shit is happening

[ionic badger]

these things will always have more lurkers than otherwise 😆

i think we have some stats for that

[grizzled laurel]

the lurkers stat

xD

[grizzled laurel]

those ratios are quite insane

especially go2hx

[ionic badger]

talking is effort and time consuming 🤣

[grizzled laurel]

lmfao

[void condor]

@clear dome Q: why you settled for https://gitlab.com/eidheim/tiny-process-library vs. https://github.com/sheredom/subprocess.h in linc_process?

trying to understand what your thoughts / options were

[void condor]

in other news, I think the whole AI IDE/tooling stuff for development is cursed. I have literally never seen soo many crypto-bros or web-devs suddenly building IDEs with signup-gatekeeping/dialogs behind every fucking setting (e.g. cursor)

[ionic badger]

more stats dump

[ionic badger]

yeahhhh everyone is spamming AI everywhere

need zed to get strong 😄

but even that has some AI stuff in it lol

[void condor]

dont get me wrong, AI stuff is cool if you can use it on your own

but people discourage local models

so setup / integration is something you need to hack yourself

[ionic badger]

i think everyone is just trying to data mine to hit the next lottery product

[void condor]

looking at it, programming has been unlocked for a lot more people through LLMs now and the tools/products reflect that

[ionic badger]

will probably make good programmers standout a lot easier

interesting times

soon computers will do everything for us and we'll get to be cavemen again x)

[void condor]

dude, there are stats out there, like 90% of the code currently produced is AI code

esp in webdevelopment

[green breach]

99% of code isn’t very good.

[void condor]

stuff like lovable generate 25k projects A DAY

[clear dome]

One of my requirements was to be able to set env values in the subprocess (which is not possible with haxe Process afaik). Didn’t really do a thorough analysis of all the options. Just sticked to the one I found ticked all my checkboxes. Maybe subprocess.h is better, who knows lol

[void condor]

sadly true

[void condor]

ah i see

yeah, haxe process has problems esp when messing with stdout/stderr from multiple threads

shit just locks up

[clear dome]

Actually, a good summary of how I work on personal projects is a cycle: I have moments when I can be super productive and creative, which can last weeks, or months… then at some point I need to take a break from that and enter a « consuming » mood for a while (watching movies, playing games, reading etc…), which can also last weeks or months, until I feel like I need to get into creative activity again naturally. Learnt to accept this

Right now I’m probable near the end of a creative/productive time that lasted several months

[void condor]

same pattern for me. I found that playing games where I can enter a flowstate for 1-2 days really speed-runs recovery of my brain drain

stuff like counterstrike or rivals, that are heavily reaction & intuition based

it's like you need to have different chemical cocktails rotated in and out of your brain

[grizzled laurel]

This is also very relatable...

It never came to me that this is normal behaviour?

I thought my ADHD brain was just a bit funky, but many people have this?

[clear dome]

Yeah, as long as you accept it, and you are confident that it's a cycle, there is no need to feel guilty of improductiveness or whatsoever

[grizzled laurel]

Interesting

[clear dome]

(of course when we are talking about a day job that's a bit different lol)

[ionic badger]

how is it even possible to figure that out lol

[clear dome]

cursor -> cursed. That was a sign

[ionic badger]

it's "human behaviour" everyone is like this, see the discussion about server activity, people ebb and flow

[clear dome]

Imo there are different degrees of creativeness too. I know some people who are absolutely fine staying in the "consuming" loop

[ionic badger]

99% of people aren't "on" 100% of the time despite what they may act like

[grizzled laurel]

i swear people are slowly going braindead because of AI. At our school we have a subject "Personal Leadership" which is about becoming the best version of yourself... BUT... people use AI to finish the assignments.......

[ionic badger]

we just don't encourage that around here

[clear dome]

Personally I don't produce/create thing because I think it's better/worse. I just need to do it

[grizzled laurel]

even then im not even talking about all other braindead uses of AI

[ionic badger]

i don't think its a degree but more just some people take longer to realise what they're doing or not doing. I've been in a consumption loop for 10+ years but once I figured it out I basically have no consumption loop at all now comparatively

[grizzled laurel]

Hot take here, but imho the only "valid" use of LLMs is mock data or "transformation/mutation" of data (translating code to docs as an example)

it should never be used to create actual serious data itself

when it comes to AI outside all of the LLM stuff i think that there are alot of potential good uses

like healthcare

[clear dome]

The limit doesn't really lie into how far you go with AI imo. It's more the loss of skill that is concerning

[void condor]

I know people that live their whole life like that

[clear dome]

If you did learn a specific skill, it can be ok to delegate, because you still know what you are doing, you can review it etc...

(same for medical healthcare assisted with AI)

[grizzled laurel]

this is very true, but i doubt that people can be trusted with this

[clear dome]

But skipping the learning phase, replacing it with AI, this is the dangerous part

[grizzled laurel]

whenever your job is to review and not investigate

people WILL slowly get more relaxed regarding the job at hand

[clear dome]

Like, you could even use AI as a tool to learn how to draw, in addition to more "classic" ways of learning

But if you are just generating an image, this is definitely not drawing

[ionic badger]

well... its kind of how things go. I think my schooling was in the middle where calculators were kind of being accepted but we were encouraged to do without. We had some times where in emergency situations we were allowed to use computers for exams but they were heavily monitored.

Now, everything is "we must get pass grades forget about the learning, good grades means good things"

I'm now 31 and basically have forgotten 90% of my ability to mental math lol

[clear dome]

AI is very double-sided. A super powerful learning tool, if you use it to learn, and an incredibly cursed tool if you skip learning

[void condor]

true

[grizzled laurel]

i agree with this

and unfortunately, many many many (if not most) people try to use it to skip learning

[void condor]

in any case, the fat years are over

low-profile programming jobs are basically dead

[ionic badger]

AI is seemingly being used to compensate for effort which is really bad, because now people will "accept" even more than they already do that "they aren't capable"

[grizzled laurel]

few days ago i was in the train, and a couple came sit next to me in the train. the guy was talking to his girlfriend about how AI did his homework assignment and was even trying to get his girlfriend to use AI for her school work. I had to really try my best to hold in the laughter...

[clear dome]

I guess you will have a lot of "that person in a 10-employees company who is project manager and creating various tools with LLMs for it"

[ionic badger]

One of my friends can't really write good english so in his 60k per year job, he uses AI to write his reports for him

[clear dome]

Yeah, I can't imagine the number of people who are already skipping learning so many skills because "AI can do it". It's only going to get worse

[void condor]

I already see it at work. We have the first requests to takeover/transform/integrate some AI generated stuff from Ops/Field folks

[ionic badger]

its dangerous man, we're heading towards a time with a ton of incompetent people

[grizzled laurel]

for sure

[void condor]

and lemme tell you, it's hilarious how wrong that shit is when someone disassembles it and takes a closer look

[grizzled laurel]

lmfao

[clear dome]

LLMs have a tendency to find a solution to the problem that is 5 times more complicated than what it could be

specialists in spaghetti code

[void condor]

It's stochastic in nature, not sure why people think it would ever be deterministic or well-defined in terms of outcome

this is what really gets me. Shit is just "good enough"

[clear dome]

That being said, I'm still impressed by Claude output from time to time

[void condor]

You are impressed bc you dont know what you will get. It's messing with your brain

[ionic badger]

because its often being judged by "does it do the intended task" so it gets conflated

[clear dome]

It's just that it's much better to work with that output if you know your subject

[ionic badger]

people don't care 😅

[clear dome]

I mean, Claude has a project feature, where you can give it a bigger context

I tried it with loreline code base

With enough input in the context, it can really give interesting output

But again, if you just take its output as it, you gonna end really bad

But if you see it as some assistant that always get reviewed by you

At least in my case, it has saved me a ton of time, without any compromise to the final code

[void condor]

sure, the more patterns you supply the higher the probability of tokens chosen that fit.

[clear dome]

What I can't use, however, is cursor automatic suggestion

It's just noise for me

[void condor]

true, cant argue with that. This is what im looking at as well

[clear dome]

Cognitive load of decyphering what cursor is giving you when you didn't even ask for it is as high as just thinking about the problem yourself

[void condor]

but dude, I wanna get better at my job, not becoming the best code-reviewer on the planet 😄

[clear dome]

Same here. My point is LLMs are not always working against that goal if used correctly

And I'm not saying we should go 100% LLM (I'm very far from that)

My number one rule is that anything I'm doing with the LLM, I should be able to do it myself if the LLM is down

(And that remains true for me until now)

And sometimes the LLM just can't solve your problem, this is also true, and that's ok

[void condor]

What I would love to have in sublimetext today:
have a shortcut for a prompt of what I wanna do that takes the current selection/method and lets me paste output somewhere

[clear dome]

You can do that with CMD/CTRL + I and github copilot, no ?

(in vscode)

[void condor]

sublimetext

[clear dome]

Ah you want that in sublimetext, not vscode?

[void condor]

another thing is, Im offline with this, it has to use my LLM

yep

maybe I need to use AI to set that up 😄

[clear dome]

in vscode, I'd say it's probably not that difficult to make an extension that does that, with enough effort, maybe some exists already "local-copilot-like"

sublimetext, well, good luck 😂

[void condor]

my deepseek slave is already on it

[clear dome]

great

[clear dome]

What made you prefer sublime over vscode? non-electron or other reasons?

[void condor]

it used to be best crossplatform editor that didnt require an embedded browser-engine

[clear dome]

indeed

[void condor]

it's very performant and minimalist, i like that

[clear dome]

I recall moving from sublime to atom because you could just do whatever you want with atom editor (back when making haxe extension with ruby), but atom was super slow

Then vscode came out

And for me it's the best ratio between... huh... bloat and hackability 🤣

Also much more fluent with node/js than python

[void condor]

I think vscode is horrible. Never seen such a convoluted UI

I guess if you stick to the basics it could be fun, but there is soo much stuff being used everywhere

[clear dome]

I guess we are all wired differently. It's far from perfect and a few things have been annoying, but so far it's ok for me

[void condor]

whatever floats your boat

I think I can make this extension work

90% ai-code

[clear dome]

Oh I'm sure about that

[void condor]

😄

[clear dome]

You are just doing the opposite of what you complained about lol 😉

No but seriously, that's a good use of LLM imo. You can clean it up, but it kinda does the work of telling you what API you should use to make that extension

Interactive docs somehow, where you don't need to dig through pages and pages of things you don't need

[void condor]

yeah I know

[clear dome]

Like, I need to work with Unity for my day job

Unity docs are such a mess

which make LLMs pretty useful to help you navigate it

(well... until it hallucinates some Unity API that doesn't exist 😅)

[ionic badger]

have you seen zed?

no browser engine

[void condor]

hahaha, and the confidence it has when you try to correct it 😄

[void condor]

nope, the only Zed I know is a league champion I used to play

[lethal talon]

Of course you would play Zed...

[ionic badger]

https://zed.dev/

I remember zed, he was a cool champ

[clear dome]

That intro video showing "hey, see how the tab contents are switching so fast" is a bit silly to me. I mean, that's really not how you are using a code editor anyway

[green breach]

Minus the AI stuff Zed is pretty good.

The team was real friendly when I used it in private alpha

[ionic badger]

that video, when i first saw it, really put me off zed lmao

[clear dome]

That's gonna be a hot take I guess, but having a tab content need a quarter-second to render in a code editor is... just fine

[ionic badger]

i had to overcome that a bit

[clear dome]

That's not what I expect from a code editor, to be good

[ionic badger]

yeaaa

[clear dome]

It has to be responsive for sure, but this is silly imo

[ionic badger]

it actually runs better than vscode

the video is just rediculous and not worth watching

[green breach]

The cool stuff is all the collaborative code things

It has a built in voice chat for some reason

But you can share a project with someone and work in real time like a Google Doc

[ionic badger]

I couldn't get it to work for me but it works for some people around here

(completion, ide works)

[void condor]

xD I used to be a way better Pyke, though https://www.youtube.com/watch?v=NkRaXTEIR1k

[green breach]

Yeah the Haxe plugin doesn’t work for me

[clear dome]

I'm just going to stick to vscode sorry guys lol

Now I'm too much invested in haxe and vscode, that's the hard truth

[green breach]

No hard feelings (minus wisdom not working on my editor)

[ionic badger]

lol

[clear dome]

If you managed to make tmLanguage syntax files works I guess you'd be good to go

[ionic badger]

i aint trying to move anyone, just throwing it on the table 😆

[green breach]

Does VSCode support treesitter?

[ionic badger]

lol

[clear dome]

Not officially for sure

The default way of making syntax highlighting is with tmLanguage syntax files

[green breach]

Aw sad.

I think lsp highlighting might be a thing?

[void condor]

need to step out for a while, thanks for the lovely exchange guys 😄

[grizzled laurel]

cya again soon

[clear dome]

You can for more contextual things that are resolved by intellisense yeah, but that's overkill for the rest

[grizzled laurel]

also since we are speaking about lsp stuff, jeremy, how did you make the language server for loreline? Any resources you used?

[clear dome]

To be fair, since lookahead and lookbehind are now working fine with tmLanguage syntax regexes, you can go pretty far in the details (Loreline use case is much harder than wisdom's in that sense)

But for maintainability, it's probably best to work with some intermediate format that is "converted to tmLanguage syntax"

Like, I'd like a syntax that allows to split regular expressions into multiple parts that I can name/comment separately, and then "compile" that into the final format. Maybe I'll try it some day

(maybe that's already a think idk)

[green breach]

vshaxe does that iirc

[clear dome]

Official spec, but also Claude for some boilerplate code that got me started

[grizzled laurel]

i see

[ionic badger]

can probably feed some ai vshaxe and then ask for some guidance

[clear dome]

In what language do you want to write the lsp server?

[grizzled laurel]

ideally the same approach as you

using haxe to bring the lsp server to js

to then run in vs

[clear dome]

Then you can take a look at vscode-loreline repository

The boilerplate is pretty generic

In loreline repository, there is the actual server code in src/loreline/lsp

[grizzled laurel]

gotcha

so if i understand correctly, vscode-loreline is communications and loreline itself also contains the business logic for the LSP server?

[clear dome]

And the Protocol.hx file is completely generic, you can use it as is if you want

(Minus the package name)

[clear dome]

Yeah, lsp implementation is in loreline repo itself

I know vshaxe has its own language-server-protocol repository, but I prefered rolling my own as a single haxe file

[grizzled laurel]

yeah that is nicer

especially if you dont know all ins-and-outs of LSP

[clear dome]

(Also used claude to make that Protocol file from official spec, pretty useful for that kind of stuff)

[clear dome]

Last time I checked it was just a tmLanguage json/yaml file too in vshaxe

[grizzled laurel]

yeah defo

that is also what i ment with "transformations/mutations to data" as usage

well guess ill give it a try

if i have any questions, would you mind if asked them to you? if so, where would you prefer me to ask you said questions?

[clear dome]

https://discord.com/channels/162395145352904705/1297978372529586207 you know my territory 🤣

Another thing critical for the lsp of loreline has been creating a Lens class which is basically providing all the static analysis of the ast for the lsp

[grizzled laurel]

xD

[grizzled laurel]

i see, its like a bunch of helper functions to allow for completion

finding things, resovling things, etc

[void condor]

hunting bugs is fun

esp when you can relicate them by randomly hammering inputs on keyboard and mouse 😄

[obtuse burrow]

I had a play around today and didn't immediately see anything out of the ordinary, but I also don't really have anything large to test it in anger (I haven't touched most of my larger projects in years and have long since forgotten what I was doing with them...)

I did notice though that haxe class header files were now including the header files of the interface glue classes where as previously they did not, so I've fixed that

might have been part of the issue since those interface class header files contain implementations, not just definitions (or rather, they use to, I've now changed that as well)

[void condor]

oh, nice. gonna give it a go

compiles and runs. I notice that the compile time per cpp-file is better now. But the cache is still trashed with every compile

gonna record some vids for you so you can see the same workflow in both versions

@obtuse burrow Here are your branches (haxe+hxcpp). first I do a full compile with cleared cache and no intermediates for the cppia host, it is followed by a 2sec compile for the app in the script and then the app starts.
Once that is complete I do another compile with hot-cache and intermediate files from the previous compile. Notice how it does a full compile of all the code and then it locks up the linker (this linker lock-up is usually happening for me when conflicting versions of an obj are found in the cache, e.g. debug vs. release).

This is my current default setup. I switched branches and followed the same workflow: clear cache and intermediates, full compile host+cppia (~40secs) and app run. Then I do a second compile with hot cache and intermediates that lasts about 4secs + 2secs cppia.

you notice how hxcpp doesnt print any cpp-file when it finds a cache entry

lemme know if you have any questions

in my user I use .hxcpp_config.xml to configure the cache:

[obtuse burrow]

thanks, I'll take a look and see if I can glean anything from the files which are changed

[obtuse burrow]

that is odd behaviour and I'm not seeing it for my project

clear bin / cache, build which then populates cache, rebuild with no changes and nothing is re-compiled

this doesn't use cppia though (I just added the scriptable define in case that was related), so I'll play around with some host and client stuff tomorrow

[void condor]

hmm weird, so it might be in my end

gaahh, I hate nothing more than anything related to time-steps

added some test animations to the UI and noticed a stutter. turns out that I have the effect of global time for 2 consecutive frames being the same. which makes no sense. Im specifically using kha's scheduler so I dont have to mess with this shit

[void condor]

Turns out im the idiot. I was updating components the wrong way

2 systems updated the same components in a single frame

[obtuse burrow]

when you're not using builds from my branch what haxe version are you using?

latest nightly or something a bit older

[void condor]

something a bit older

[void condor]

@clear dome quicky hacked this together https://github.com/cortex-engine/linc_subprocess

just a simple blocking run with output, no streaming pipes

[clear dome]

// dynamic to the rescue!

Anyway, nice! Do you plan to handle streaming std pipes?

[void condor]

atm, not really.

I might if need be, but I dont have that

at least for the moment

[clear dome]

I guess the subprocess just inherits stdio at the moment?

I needed it for Ceramic CLI anyway, because I'm transforming the output before displaying it on the console in some situations

[void condor]

man, you are way too user-friendly

😄

[clear dome]

https://tenor.com/view/matrix-online-gif-9800727944098993984

Lol, dazKind be like

[void condor]

blondes everywhere

[clear dome]

What made you chose that subprocess lib over tiny-process-lib btw?

Simpler implementation?

[molten sleet]

im totally not gonna use linc_process for haxelumina's tools

[clear dome]

That's alright, I'm still interested to know why though

[molten sleet]

(Totally)

-# (this is a lie i am gonna use it)

ceramic's output when compiling is pretty
Its not just hxcpp yapping

[grizzled laurel]

This actually reminds me of what dazs setup looks like

[void condor]

single header and simpler api

that's about it

[clear dome]

This is mine btw, don’t ask me why I read the GIEC thing, we are all doomed anyway

[clear dome]

Oh I actually forked tiny-process-lib to make some improvements

I added the option to create a detached process

And I did a few fixes regarding process killing, because it had some edge case issues

(GIEC = IPCC)

[void condor]

way toooo organized!

[clear dome]

Just after tidying 😁 (it was waaaaaay messier a day before lol)

[clear dome]

Too organized, Too user-friendly, what else xD

[void condor]

we gotta talk about the collection of books though, "guide des champignons" ?!! Not what I would have imagined to find in your library 😄

[clear dome]

Haha, it's just big so we put it there, "family book" I guess

[void condor]

here is my bookshelf

[clear dome]

I didn't expect any less from you ha ha

[void condor]

I have one of those as well, 2 secs

[clear dome]

To be honest, I don't have much technical/programming books

[void condor]

shit, I dont know where I put my phone

left it in the car, here we go

talking about "family books" 😄

[clear dome]

xD I prefer my mushroom book

[void condor]

it's a super interesting biography i decided to hold on and read every few years

[clear dome]

That book in the shelf is actually a classic. One of the references in terms of mushroom knowledge, before all the apps and more modern ways of identifying them

[void condor]

oh dude, yeah, such books contain undeluted essence

not comparable with the attention & subscription seeking mess we have these days

same reason I love reading old flipcode entries

https://flipcode.com

[clear dome]

Damn, I was 12 at the climax of this website

[void condor]

this, gamasutra & gamedev.net + irc were the gamedev-place-to-be back then

[clear dome]

gamedeveloper.com is such heartless now

[void condor]

yeah 😦

[clear dome]

I liked the old school design of gamasutra

made it much more gamedev-indie-friendly than this "we-are-an-industry-now" look

anyway just ranting lol

[void condor]

im with you there. the community blogs + spotlight articles were awesome

[clear dome]

What would be the spiritual successor of gamasutra now?

[void condor]

I have no fucking clue. Shit is all over the place. There are a shitton fragmented communities & discords now

there are some corners of reddit that are cool

but nothing I would call a place where I wanna involve myself

[clear dome]

I recall gamedev channel on reddit being a bit too hostile to anything that looks like autopromo

But yeah, Discord communities I guess, even if it's different

[void condor]

https://80.lv feels prolly closest to original gamasutra atm

[clear dome]

Right, but they need a dark mode ffs

[void condor]

xDDDD

[clear dome]

https://tenor.com/view/eyes-burning-my-eyes-spongebob-gif-5183364

[void condor]

I might need some help in the coming months. I wanna do linc_jolt but it's gonna be a shitty grind externing everything

jolt and steam audio are like the last big missing pieces that i need to add to cortex

[clear dome]

whoa

[void condor]

aside from a particle system

but im gonna hack that. Yeah so externs jolt and steamaudio

[clear dome]

Good luck. I'm drowning in my own projects already

Some day I'd like to work on that C# reflaxe target too

[void condor]

nobody needs C#

[void condor]

Thanks. But remember, drowning together is more fun! 😄

[clear dome]

Speak for yourself, I want to plug haxe to many things that work with C#

@tall veldt lurking around I see 👀

[void condor]

Im pretty sure he has a reflaxe bot alert

😄

[clear dome]

reflaxe reflaxe reflaxe

[tall veldt]

what what whats happening

i am only here because i was just ping

[tall veldt]

screw unity, lets just go all into reflaxe/gdscript for ceramic godot backend

[clear dome]

But regarding C#, the truth is Haxe 4's C# is still working ok at the moment, I'm exporting Loreline to C# just fine

So there is no imminent urgency to ship Reflaxe/C# yet, even though that will be the way going forward

[clear dome]

I'm afraid running the whole Ceramic 2D renderer via GDScript is going to be too wild

[void condor]

I think you can just batch into the renderserver with a few calls

[clear dome]

Yeah but if I want to have the whole rendering stack of Ceramic available within Godot (like I do with Unity), having the whole visual layout/transform system and batching system running on top of gdscript seems... huh... a bad idea 😄

[void condor]

xD

[clear dome]

Wondering how optimized is the C# usage in godot though

[void condor]

it's a shitshow, functional though

[clear dome]

Within Unity, if you use IL2CPP, you can get quite fast C#, so Ceramic runs pretty fast on it

[tall veldt]

Apparently Godot has support for .net aot

[clear dome]

Didn't dig into how it is integrated with Godot

[void condor]

pretty sure ceramic->C#->godot would work

[tall veldt]

According to some random reddit comment

[clear dome]

Ah cool, then yeah that could definitely work pretty well

[clear dome]

Yep, best bet so far

[tall veldt]

That would be pretty cool to see

[void condor]

as much as I hate godot, I agree this would be cool

[green breach]

godot console support seems pretty bad

[clear dome]

console support is not the only reason to target such engine (including unity)

It can also give access to 3D scenes from a Ceramic project

[green breach]

ah fair

[clear dome]

I don't intend to make a 3D engine myself, because, well, I only have one like heh

But being able to plug Ceramic and all its utilities to an existing 3D engine can definitely work

You could mix 2D and 3D too

render 2D parts with "regular Ceramic code" onto render textures that you then use in your 3D scene, for instance...

If some day I want to make a 3D game, I'll probably try something like that

[void condor]

make sense, integration part sounds nasty though

[clear dome]

What do you mean by integration part?

[void condor]

making 2 engines work together, figuring out authority etc

[clear dome]

Ah, that's not very different from what I'm doing with Ceramic running inside Unity

I just plugged it to the render loop of Unity

But Unity is the "authority" in some way

you are inside a Unity scene, that runs Ceramic

It wouldn't be much difference if you have actual 3D elements in that same scene

Ceramic main thread runs in the same main thread as Unity

So you can update objects in a monohebaviour style from Ceramic mostly the same as how you'd do it with C# components

[void condor]

I did something like that in the past when I hijacked opengl/lime with my engine. But it was hard to make sure both sides didnt interfere with things in like renderstate in unexpected ways, etc.

[void condor]

ah ok, that sounds like a clean cut

[clear dome]

Yeah, that's not a problem so far. Basically Ceramic rendering is just a "render step" within an unity frame

As a matter of fact, when using URP, Ceramic is running within a RenderPass added to the URP pipeline

So all clean yeah

I guess correct setup would be "render ceramic stuff first", then do the regular rendering steps of unity

So that you can use anything rendered in 2D by Ceramic within your 3D stuff

[void condor]

you leverage unity resource handling for ceramic or you keep that separated as well?

[clear dome]

Ceramic textures are actual Unity textures loaded via unity asset system, if that's what you're asking

[void condor]

yeah

[clear dome]

Audio is using unity AudioSource + AudioPlayer too

etc...

The rendering backend of Ceramic for Unity is creating (and reusing) unity Materials on the fly from the rendering parameters

Then uses CommandBuffer to send a Mesh data with the correct material.

Basically using commandbuffer like a GL-like API

It works great

[void condor]

makes sense. Unity as a backend always struck me as odd, but this way you gain everything

[clear dome]

Skipping all the higher level API, there is just one single MonoBehaviour to plug Ceramic onto the Unity scene

When I started it I wasn't sure at all it would work, but I have been surprised by how well it works

And 100% of Ceramic 2D API is supported

[clear dome]

For me Unity is the mainstream « backup » plan in case I can’t do something with plain Ceramic

Some small game studio decide to use Unity because they can’t afford to risk « making their own tech ». Ceramic is some kind of middle ground where you make your own stuff without being tied to Unity, but you can still take advantage of all that work within Unity too if needed

So many things like tracker library, state machine system etc… provided by Ceramic can be useful for non-Ceramic graphics too within a unity project or similar

And Haxe > C# of course 😛

What I wanted to avoid at all cost is start making a framework on top of Unity, which doesn’t work on anything else. Many indie studio are making their own kinda framework within Unity, but then they are locked up within that stack

[crystal grove]

I forget which thread is which as I skim through messages lol

"This is a lot of side convo for Jeremy's thr... oh I see"

[lethal talon]

Building a Haxe target especially one that wants to pass the Haxe tests can will likely become it's own goal in of itself. Making a c# unity backend is great but will likely take a while before that is possible.

[grizzled laurel]

This has been a very cool convo to read

[grizzled laurel]

Altho won't you get RSI from that keyboard or something..?

It's minimalist but doesn't look comfortable

Otherwise nice setup

Also before I forget

"Reflaxe"

[molten sleet]

the ritual to summon srd

[grizzled laurel]

yeah truly magic

the bot alert

[clear dome]

That keyboard is fine. Did change the mouse for a more ergonomic one though, since I took that picture

[clear dome]

For sure, thankfully we have the haxe4 version as reference, especially the std lib implementation

[lethal talon]

From my limited understand of the c# situation, the haxe4 version as ref may be much less useful then the stdlib, because of poor quality implementation of the target, I hope it's not the case as well for the stdlib.

I am really looking forward to the c# target, and will try my best to knowledge share to give it the best chance, if/when the c# target becomes mature, it would open the door for this sort of hybrid approach for other languages too.

Right now the targets written in OCAML have an obvious advantage of shared code and utility functions that deal with many common cases, plus when new Haxe features get added, they are added across all of the targets. If reflaxe based targets can bridge the gap and allow Haxe targets to be written in Haxe, it almost solves the reasons for wanting to do Haxe in Haxe, and keeps both worlds intact.

[clear dome]

I recall that "shared utility" logic between java and cs targets (gencommons) was part of the issue, so I guess sharing too much utilities is not always the best option

But yeah, Reflaxe is providing some helpers already anyway

[tall veldt]

what the hell, the quotes made me not get notified

Wait

reflaxe

Reflaxe

I think it's the capitalisation

@ionic badger plz fix

[ionic badger]

really?

[tall veldt]

Yes

[ionic badger]

i did make a change recently

what i did was basically make a keyword have to have a space char before and after

[tall veldt]

Ahhh

[ionic badger]

to avoid something like hl triggering in a word like highlight

[tall veldt]

But I'm saying I also didn't get notified for thiss

[ionic badger]

but it should be case insensitive

[tall veldt]

Hmmmm...

[ionic badger]

oh wait

you can't trigger your own tracker

reflaxe

Reflaxe

[clear dome]

btw, how can we setup a notification, like how can I be notified if someone mentions Ceramic

[tall veldt]

Only lowercase

[ionic badger]

yeah, the /tracker command i dm'd you a while back

[tall veldt]

hahah yeah i was doing it for dramatic effect

[clear dome]

reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe reflaxe

[tall veldt]

[ionic badger]

can you do it with a capital now

[clear dome]

Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe Reflaxe

[tall veldt]

Oof nope, sorry

[ionic badger]

huh no dm

[clear dome]

It's ok I can use lowercase if I want to summon SRD

[tall veldt]

Theyve found a way to hack the system

[ionic badger]

so many missed Reflaxe conversations

[tall veldt]

They gonna talk about reflaxe without me!!

[clear dome]

FOMO at it's climax

[lethal talon]

I feel like every discussion I don't join is more interesting to read soooooo

Reverse FOMO

[ionic badger]

impossible

[tall veldt]

nah this is so relatable. i feel like everyone knows so much more than me and i just distract from good conversations

[ionic badger]

said by the 2/7 of the smartest people on the server

[tall veldt]

NAH IM TIRED OF THIS

go2hx is so outragously more amazing than anything ive ever done

how dare you imply im anywhere near shadow

[lethal talon]

That's true but I was thinking when the bot tracker notifies me, I think the discussion gets shutdown when I give definitive answers, because there's not much else to discuss :p

[ionic badger]

#controlling false information

[tall veldt]

oh thats also true and relatable

[lethal talon]

I consider you right next to me, so I dare thee

[lethal talon]

It's more a conversation is a lot like an exploration process and talking back and forth, it feels shitty to just come in an kill off that discovery because likely if it's about go2hx I've spent considerable more time thinking about any detail someone could think of

It's why I muted the bot, because it's more interesting for me to see after :p

[tall veldt]

Yea that true too. I just end up seeing most stuff anyway

I mainly like it for when the fnf channel, the one channel i dont follow, mentions it and i can quickly jump in there as if im some crazy huge fnf fan

[ionic badger]

heh, use it how you want, sometimes people want to ask and just don't want to ping

the point of the conversation is to find an answer at the end of the day

[lethal talon]

Yeah I also felt bad, one time when I instantly jumped in a Convo about go2hx with NotBilly and Logo in casual chat, and it's not my place to have to be in every Convo about my own project

[ionic badger]

errrr i don't have any memory of this

[lethal talon]

I'll find it one sec

#casual-chat message

And my delayed reply #casual-chat message

[void condor]

feeling bad for something like that? come on

[ionic badger]

I would say you did nothing wrong

You care about your project and things related to it

feeling bad about enquiring about what interests you is something to get over imo

[void condor]

I use trackers for all kinds of shit in order to keep my finger on the pulse of certain topics

[ionic badger]

you can obviously pick and choose where you insert yourself, there's no hard and fast rule. But, as far as i'm concerned this is a public server and anyone can chime in regardless

[void condor]

You gotta reflaxe

[ionic badger]

lol

[void condor]

ups typo, relax

btw, how do I list my active trackers?

[ionic badger]

by reminding me that such a thing does not exist

[void condor]

you are welcome

[ionic badger]

i can get you a screenshot from the db if you want 😂

[void condor]

lol

[green breach]

just send the db credentials instead

[void condor]

I forgot what trackers I setup. I think I have at least 2 for variations on my nickname, but not sure

[ionic badger]

i did that by accident when i sent the entire source code for my project

[green breach]

ono

[ionic badger]

its cool, i realised and changed everything 🤣

[ionic badger]

never checked what trackers people have lol

[void condor]

Thanks for making it public, now @tall veldt can take revenge

xD

[ionic badger]

i did think about dming

[lethal talon]

I don't think it's wrong, but I do think it's sub optimal. I want my project to be mature and make Haxe a general purpose language, my perspective is known and pretty well understod. Me jumping into a discussion uninvited to argue that the project can do that versus only crypto libraries or a more limited subset, doesn't change anything in reality, until in reality it can reach that level.

Why discuss hypotheticals rather then just accept okay, people think it's limited and can never reach said level, that's fine, I'll just keep building.

[clear dome]

I'm gonna setup a tracker to send raids of Ceramic police 👮‍♂️

[green breach]

lol i felt bad about that conversation too. i feel like i phrased things poorly

[ionic badger]

well yeah, that's the skill element of it. You don't necessarily have to jump into a discussion to say "go2hx can do this", you can enquire why people think it can't or gleam different kind of ideas

[void condor]

The black leather coated CERAPO!

[clear dome]

With their prisons made of clay

[void condor]

and their endless wisdom using tracker

[ionic badger]

also, gotta remember 90% of the server is unlikely to know your affiliation with go2hx and its more likely just the active 10% who do

[lethal talon]

No you're fine, what you said was fair, it's my bad for pushing you to defend a position that is true now.

[ionic badger]

and the 10% of us who do, don't care about where you insert yourself lol

[clear dome]

They bind you in no time

[lethal talon]

Yeah but I do personally, and I don't want to make the 10% of people that are interested in the project like logo, get put off from my omnipresence

[void condor]

he better learn to deal with it

[green breach]

I AM THE 10 PERCENT

[lethal talon]

Yeah that's true

[ionic badger]

logo has tracker access as well, he has his own omnipresence

[lethal talon]

I HOPE SO haha

[void condor]

You should self-insert yourself any time you think you can meaningfully contribute or help

[ionic badger]

[tall veldt]

omg i have a great idea. it came from my frontal cortex, daz the kind of thing i need. what if instead of trace haxe added a tracy statement to maximize the amount of traces you do. godot

[clear dome]

Meanwhile I'm punching my head against wall to figure out how to link some unity project with an iOS dynamic .xcframework, enjoy life guys

[void condor]

peak prompting!

[grizzled laurel]

reflaxe reflaxe reflaxe

[tall veldt]

damn in retrospect daz really does just show up whenever tf godot gets mentioned XDDDD

[ionic badger]

ya'll haven't realised that you only get 1 notification per message*? 😄

[void condor]

MY SECRET HAS BEEN DISCOVERED

IM A HIDDEN GODOT SHILL

#exposed

[green breach]

damn tbf it is a good idea

trace("hello");
tracy("hello");
tracier("hello");
extratracy("hello");
traciest("hello");

[clear dome]

"godot reflaxe", and then you can summon a new discussion between two characters of your choosing

[lethal talon]

That is kind of badass, and the best part is very few of the specators know you are doing that.

[ionic badger]

lmao

[tall veldt]

as in, if one person gets notified from one message, another person cannot??

[lethal talon]

So it looks magical

[void condor]

sure, lets make sure we dont break any ceramic while we are at it

[green breach]

i think i only have notifs for haxeget

[clear dome]

the chief writing secretary Loreline is watching

[ionic badger]

ahhh, no, that message would have only notified daz "once" but others would have been notified if they had the trackers as well

[lethal talon]

Challenge: write a message with the most amount of pings without using the @ symbol

[tall veldt]

HAHAHAHA

[clear dome]

now we need a bot to show scores

[ionic badger]

lol i didn't even do that on purpose

[void condor]

xDDDDD

[tall veldt]

it's cause it's two different keywords

[clear dome]

"congrats, you mentioned X persons with this message"

[tall veldt]

Billy means spamming reflaxe 20 times in a message only sends me once notification

[ionic badger]

good thing i didn't screenshot the full db 😂

[tall veldt]

But I havent said anything cause thats hilarious

[void condor]

thats why you gotta distribute them

reflaxe

or

reflaxe

[ionic badger]

that's how you gettem

[tall veldt]

anyway Billy, I mentioned the exclusive notification thing earlier because I didnt get notified for this message.

[ionic badger]

errrrrrrrrr

[tall veldt]

I figured the bot was just down for maintenence, but seeing that Daz got the notification... figured I should mention

[lethal talon]

Because it doesn't have a space at the end of reflaxe

[tall veldt]

😅

[ionic badger]

crap 😄

[lethal talon]

it has to be padded with spacing on both sides

[void condor]

oh

really

[ionic badger]

it's a very recent change

like, in the last 2 weeks change

[tall veldt]

No rush to fix, just trying to give the feedback. Not a vital feature for haxebot XD

[ionic badger]

i'll change that to an or instead of an and

[green breach]

daz

[tall veldt]

daz

[void condor]

xD

I feel important now!

[ionic badger]

its just to prevent things like words with otherREFLAXEwords in the middle

but realistically it can just be an or

didn't think about special characters

[tall veldt]

Why not just good old /\bsomething\b/ regex?

Too slow?

[ionic badger]

i didn't think about regex

[lethal talon]

Wait we should pick a phrase that we all get summoed to join in on

[ionic badger]

seemed too simple

[green breach]

ok kinda random, but i wonder how confusing it is to new people in the server who don't know that all the fun conversations happen in threads

[lethal talon]

Almost like a bat signal

[tall veldt]

\b is shockingly powerful for the exact case you're struggling with imo

[ionic badger]

much appreciated 😄

[lethal talon]

Those poor people

[ionic badger]

i had an idea to solve this

discord's management of server discoverability sucks ass

[green breach]

make linux thread its own channel when

[ionic badger]

i would have already done it if i could transfer the history easily

[green breach]

we could do a re-enactment

[void condor]

brb

[tall veldt]

what channel is this in lol

[green breach]

https://discord.com/channels/162395145352904705/956928971927339018

[void condor]

shit is fragmented

[tall veldt]

Ahhh casual chat

No wonder I missed out

[ionic badger]

but i also don't want to necessarily put tooooo much attention on threads

seems kinda cool that people have some "private spaces"

[green breach]

i think we should disable threads entirely and just move the popular stuff to their own channels

[void condor]

all the threads pull attention away from the main channels, making them less engaging for people not following the details

[green breach]

projects stuff is fine ofc

[lethal talon]

I think this server should become portal, where channels can be linked together in a spagetti fashion, and do crazy things

[tall veldt]

i think there should just be only voice channels. only extroverts allowed here

[ionic badger]

you'd think that but the linux thread is pretty popular lol

but i guess that's also to the same point lol

[void condor]

you gotta be kinding me

[tall veldt]

goddamn daz

[lethal talon]

+1 it would be a 24/7 haxe round up round up

[tall veldt]

Just release cortex already so it can get a channel

[ionic badger]

it already has a channel

[void condor]

no wonder I need to farm your engagement here to keep up with postcount!

[ionic badger]

#cortex

[clear dome]

Ceramic channel is orders of magnitude more dead than my current "personal" channel

[green breach]

imagine not having archive role

[void condor]

we cant go #cortex . WE WOULD LOSE THE 💀

[ionic badger]

need people to spearhead the activity otherwise these things happen 😆

[green breach]

gotta just rename it to #jeremy

[lethal talon]

Because you have been posting so much about loreline!

[clear dome]

I was wondering if, is it because ceramic docs are good, people don't ask questions anymore, or is it just "nobody uses Ceramic" lol

[tall veldt]

No Access, why live

[ionic badger]

like, haxeui was a really popular channel but cause ian is busy with work stuff its quietened down a lot

its fine tho

[green breach]

uhh the docs are good but probably more the latter 😅

[lethal talon]

I think it's 1 and more people would probably use it if they knew about it.

[clear dome]

ono

[lethal talon]

I set it up ceramic 2 weeks ago and I liked the setup, and demos, I'll likely use it for my next sample project.

[green breach]

ono the onos are contagious

[ionic badger]

/archive

[vapid quail]

Archives are shown

[lethal talon]

/ban SomeRanDev

[tall veldt]

[void condor]

I'll release soon but it'll be a shitshow

[green breach]

speaking of cortex, does it still support haxeui?

[void condor]

too much bleeding edge shit

[clear dome]

That looks like haxe story in general

[void condor]

the backend needs some work, but with some love, yeah

[clear dome]

The question is "can Ceramic run on top of it"

[ionic badger]

"someone" has to be talking about it and showing stuff off

[lethal talon]

Yes, but also Haxe likely has a lower rention rate then ceramic users by a large margin

[ionic badger]

or "showing" things being done

ceramic is basically so far into development that there's just not much to talk about but nobody knows that 🤣

[green breach]

if cortex ends up being open, it might be a good backend for my loreline thing

[void condor]

sure, why not

though it'll be super experimental till the end of the year. Im still bringing stuff online that requires refactoring stuff

[lethal talon]

Honestly a consolidation/collaberation between frameworks would be a welcome change

[clear dome]

That's unfortunately very difficult to do and very time consuming

[ionic badger]

it'll probably happen naturally

i don't think collabs would be done necessarily, but, consolidation likely will

[clear dome]

But there could still be some kind of collab with for example, externs/linc libraries

[ionic badger]

collab could happen with heaps backend for ceramic