🌊 NB's Riverside View | Haxe | Page 2

plain vine Oct 24, 2025, 4:22 PM

#

i'll give that a try when i'm okay with restarting my pc 😆

#

I'm actually quite happy with kha's scheduler though, seems way less jank than the haxe timer and everything seems to be running quite smoothly at 60fps

#

to the eye, everything here seems to be running completely fine

#

ctrl+alt+esc for instant restart cinnamon 😄 no need to reboot pc at all

plain vine Oct 24, 2025, 5:50 PM

#

cleaned up my shutdown sequence thanks to the easy priority api from the scheduler 😄

plain vine Oct 24, 2025, 8:36 PM

#

this really is a cool project

#

whilst using ceramic i browsed the codebase quite a lot building up my project, and noticed various things

#

never really paid much attention to it

#

but now i'm doing it myself, the "tidyness" of ceramic really stands out

#

my codebase is an absolute mess 🤣

#

things like when to destroy visuals was seperated in ceramic, but i never thought that was that important

#

but actually, its super important... you can end up with a random reference left over, a crash because you timed the destruction improperly etc

#

dialogs more or less working 😄

#

#

the fucking upside down dropdown was due to the vbox dimensions 🤦‍♂️

carmine path Oct 24, 2025, 9:14 PM

#

xD

plain vine Oct 24, 2025, 9:31 PM

#

... I think i may have a fix for the infinite loop thing, but it is a change in haxeui core

plain vine Oct 24, 2025, 9:49 PM

#

@dense elbow https://github.com/haxeui/haxeui-core/pull/691 If you could review this change at some point it would be cool, i have no idea why no other backend doesn't have issues with scrollviews being performance limited but this change solves the issues on my side

#

at least i suspect so, i should test more, but will leave the pr there incase i forget x)

#

I don't feel or notice any difference with deferring the updates with call later in the ui, you notice it if you have traces though

#

will keep on testing it either way

#

maybe skip call later altogether and just insert at the end of the queue... wil try that 🤔

#

nah call later is better 😆

#

fuck tables seem to break with the change i made :/

broken python Oct 24, 2025, 10:08 PM

#

plain vine things like when to destroy visuals was seperated in ceramic, but i never though...

wdym by "separated destroy"?

plain vine Oct 24, 2025, 10:08 PM

#

broken python wdym by "separated destroy"?

you mark visuals as destroyed and destroy them at the end of the frame

broken python Oct 24, 2025, 10:10 PM

#

Ah, that's not necessarily true, but what is 100% true is that as soon as you call destroy() on an entity/visual, its destroyed field is true, preventing anything from binding back to it (like event handlers, autoruns...)

#

dispose() is doing what you describe

plain vine Oct 24, 2025, 10:11 PM

#

ah memory might be a bit muddled

broken python Oct 24, 2025, 10:12 PM

#

broken python Ah, that's not necessarily true, but what is 100% true is that as soon as you ca...

What I mean here is that, let's say you have a stupid code listening to the onDestroy() event of a visual, and from that handler does do "visual.onSomethingSomething(...)`. That code will simply not work, it will explicitly tell you that you can't bind an event to a destroyed entity

#

There are subtleties in destroy() method that are handled with the entity macro

#

Like, imagine your base Entity class has the base destroy() method, super call of all subclass overrides of that destroy() method

#

This base destroy() method set's destroyed field to false (i'm simplifying a bit)

#

But what if you override destroy() in a subclass like this:

#

override function destroy() {
    // destroyed is false here?
    ... do stuff specific to subclass ...

    // destroyed still false?
    super.destroy();
    // destroyed is true now?
}

#

Ceramic destroy() method will become this:

override function destroy() {
    destroyed = true; // Injected by macro
    // destroyed already true right from the beginning
    // even when super.destroy() hasn't been called.
    ... do stuff specific to subclass ...

    super.destroy();
}

#

Well, I'm oversimplifying because actually destroyed field is backed by an int, not a bool, so that "currently destroying" and "destroyed and not destroying anymore" can be distinct states

#

Anyway, the idea is "as soon as you call destroy() on an entity, it's destroyed field will return true to prevent any post-destroy reaction to "bind things to it" which will leak forever

dense elbow Oct 25, 2025, 5:52 AM

#

plain vine <@456226577798135808> <https://github.com/haxeui/haxeui-core/pull/691> If you co...

no chance im changing the invalidation loop for a backend... and im not even sure the change makes sense, it will delay all invalidation by one frame (and its already trigger on "the next frame")

plain vine Oct 25, 2025, 11:11 AM

#

dense elbow no chance im changing the invalidation loop for a backend... and im not even sur...

yeah its fine the change only partially solved the problem and not completely :/

plain vine Oct 25, 2025, 11:43 AM

#

I forgot about daz's pr https://github.com/haxeui/haxeui-core/pull/665 so i'm gonna run this for a while and see if i run into the same issues

plain vine Oct 25, 2025, 11:46 AM

#

broken python Well, I'm oversimplifying because actually `destroyed` field is backed by an `in...

wow that's a super clever system, your system even prevents the silly actuality of forgetting to call super on destroy xD

broken python Oct 25, 2025, 11:59 AM

#

plain vine wow that's a super clever system, your system even prevents the silly actuality ...

Right, the macro insults you if you forget 😄

#

The gist of it: your biggest enemy is yourself xD

plain vine Oct 27, 2025, 12:49 AM

#

    public static function getDpi() {
        var display = SDL.getWindowDisplayIndex(App.window.window);
        if (display >= 0) {
            var arr = [];
            #if scriptable
            var dpi = SDL.getDisplayDPI(display, arr);
            if (arr.length > 1) {
                return arr[1];
            }
            #end
            trace('DPI FALLBACK');
            return arr[arr.length - 1];
        }
        return 96;
    }

#

a trick for cppia client incompatible code that is defo going on the host

final field Oct 27, 2025, 12:50 AM

#

#if scriptable?

plain vine Oct 27, 2025, 12:50 AM

#

the scriptable define just simply lets you compile stuff for the client and the net outcome shouldn't actually be any difference

final field Oct 27, 2025, 12:50 AM

#

is that host or clien-

plain vine Oct 27, 2025, 12:51 AM

#

final field `#if scriptable`?

scriptable refers to the host

final field Oct 27, 2025, 12:51 AM

#

ah

plain vine Oct 27, 2025, 12:51 AM

#

cppia define refers to the client

final field Oct 27, 2025, 12:51 AM

#

examples of bad define names #1

#

scriptable and cppia as opposed to cppia (global) cppia_host and cppia_client

plain vine Oct 27, 2025, 12:52 AM

#

I should make host/hxcpp define combo so I don't make silly mistakes and forget at some point x)

plain vine Oct 27, 2025, 4:31 PM

#

ohohohohoh we got clipping working gooooood

plain vine Oct 27, 2025, 5:56 PM

#

#

muahahaha now i've got some power

plain vine Oct 27, 2025, 11:32 PM

#

but I'm still too dumb to parse this so I think I'll try and read the Tracy manual before bed 😂

plain vine Oct 28, 2025, 12:09 AM

#

bruh, pdf text can't really be changed on mobile or kindle so this is a terrible experience

dense elbow Oct 28, 2025, 7:27 AM

#

plain vine

pretty fast progress, you're becoming an old hand at haxeui backend creation now 😄

plain vine Oct 28, 2025, 11:25 AM

#

dense elbow pretty fast progress, you're becoming an old hand at haxeui backend creation now...

Its coming along, there are some real baffling things going on in other places though

#

#

look how broken treeviews are 🤣

#

but clipping was probably my biggest "concern" and even though there are majorly confusing issues like above, i don't think they're going to necessarily be big deals?

#

I could be wrong there, but, we'll see

plain vine Oct 29, 2025, 11:35 AM

#

#

I redid the depth system to work a bit better with the way haxeui works >:)

#

I don't think it has any negative side effects either 🤞

plain vine Oct 29, 2025, 12:02 PM

#

side effect 😭

broken python Oct 29, 2025, 12:09 PM

#

Seeing how fast you display haxeui stuff makes me want to try to make my own version of haxeui-ceramic

agile snow Oct 29, 2025, 12:21 PM

#

YES

plain vine Oct 29, 2025, 12:26 PM

#

broken python Seeing how fast you display haxeui stuff makes me want to try to make my own ver...

dude its so cool, you map a few things and 10 components suddenly work

#

just a lil slide show of some of the test pages I have currently

#

depth is not perfect (as always) but, i think the harder parts are mostly done. then its gonna be about figuring out why components are not initiatialising with the correct width/height and then mapping styles

broken python Oct 29, 2025, 1:31 PM

#

Are there resources/docs that tell what should be implemented in a backend in details?

plain vine Oct 29, 2025, 1:33 PM

#

broken python Are there resources/docs that tell what should be implemented in a backend in de...

I have a casual little thing I've been writing whilst working on the backend, but other than that just other backends

#

you'll probably pick up the concepts decently just from reading the haxeui-ceramic backend

agile snow Oct 29, 2025, 1:34 PM

#

is haxeui-blank still up to date?

plain vine Oct 29, 2025, 1:35 PM

#

agile snow is haxeui-blank still up to date?

ya, that's what i used for haxeui-ceramic and the current one i'm working on

dense elbow Oct 29, 2025, 2:14 PM

#

broken python Are there resources/docs that tell what should be implemented in a backend in de...

"docs" ... ... i keep hearing this word... ... one day ill have to look into its meaning 😄

plain vine Oct 29, 2025, 3:07 PM

#

#

ohohoho

#

I'm not sure what's going on with that bottom tabbar, maybe a different issue

#

but i think i've got depth good????

agile snow Oct 29, 2025, 3:23 PM

#

plain vine ohohoho

you sound like rob lol

plain vine Oct 29, 2025, 3:27 PM

#

i'm learning a lot with this in general

#

things that were confusing to me in ceramic land are now less confusing even tho i'm not doing anything there atm

#

i recon i could do stuff better in ceramic now

broken python Oct 29, 2025, 3:34 PM

#

What did you find confusing specifically ?

plain vine Oct 29, 2025, 3:36 PM

#

broken python What did you find confusing specifically ?

simple things like clipping, depth, i got a decent done in there but these are things i've never dealt with much outside

#

depth in other haxe frameworks either didn't exist, or was very simple

#

i don't think i was that confused with ceramic in particular

#

more just had gaps in concepts ceramic used

#

i spent a while on kha many years ago, and things that I did over there are clicking for me better now as well

broken python Oct 29, 2025, 3:46 PM

#

I see. Yeah the idea of depth range in Ceramic can feel a bit unnatural at first for sure

plain vine Oct 29, 2025, 3:48 PM

#

broken python I see. Yeah the idea of depth range in Ceramic can feel a bit unnatural at first...

I couldn't ever figure out whether I could apply it to the haxeui component system or not

#

also, from previous conversations with you and ian, i've got my pointer events at a component level rather than screen level now

#

never knew you could actually map events in a haxeui backend either so i've mapped cancel propagation for haxeui - no idea if it uses it but its there in the html5 backend

plain vine Oct 29, 2025, 8:40 PM

#

i can't test haxeui windows because of a cppia thing - i think :/

#

Called from haxe.ui.containers.windows.WindowManager::bringToFront haxe/ui/containers/windows/WindowManager.hx line 132
Called from haxe.ui.containers.windows.Window::get_maximized haxe/ui/macros/Macros.hx line 643
Called from haxe.ui.behaviours.Behaviours::get haxe/ui/behaviours/Behaviours.hx line 306
Called from haxe.ui.behaviours.Behaviours::find haxe/ui/behaviours/Behaviours.hx line 152
Error : behaviour maximized not found
I'm getting this error the Maximized behaviour is on the host, so i'm wondering if there's some macro getting booped away somewhere

plain vine Oct 29, 2025, 9:05 PM

#

Called from haxe.ui.containers.properties._PropertyGroup.Builder::onInitialize haxe/ui/containers/properties/PropertyGroup.hx line 37
Called from haxe.ui.containers.Collapsible::set_collapsed haxe/ui/macros/Macros.hx line 694
Called from haxe.ui.behaviours.Behaviours::set haxe/ui/behaviours/Behaviours.hx line 236
Called from haxe.ui.behaviours.Behaviours::find haxe/ui/behaviours/Behaviours.hx line 152
Error : behaviour collapsed not found
Same issues with property grids as well :/

plain vine Oct 30, 2025, 12:05 PM

#

get the same issue with ceramic + haxeui

plain vine Oct 30, 2025, 3:27 PM

#

@dense elbow Any idea why this is happening? I thought it might be because of not setting component ready, but i've played a decent amount with that and couldn't solve it

#

issue seems to be components aren't being sized correctly on start

dense elbow Oct 30, 2025, 3:34 PM

#

no clue, if i had to guess i would say that the text size is being reported back incorrectly

#

maybe you are applying the text style "to late" so there is a mismatch... ...

plain vine Oct 30, 2025, 4:21 PM

#

I did try adding call later to measure text but that made everything worse 😂

plain vine Oct 30, 2025, 5:52 PM

#

as for text style i'm just overwriting validateStyle like the other backends

#

oh no... could this be another cppia issue??

#

simple/Visual.hx:230: Depth changed
haxe/ui/backend/TextDisplayImpl.hx:77: 153,false
haxe/ui/core/Screen.hx:316: WARNING: Screen event "rightmousedown" not supported
haxe/ui/core/Screen.hx:316: WARNING: Screen event "rightmousedown" not supported
haxe/ui/backend/TextDisplayImpl.hx:62: set fontsize

#

last thing that occurs is set font size which should make it cause a remeasurement - right?

plain vine Oct 30, 2025, 6:22 PM

#

dense elbow maybe you are applying the text style "to late" so there is a mismatch... ...

#

It's the opposite! the timer was too fast, adding a longer delay has the above effect but that's also not ideal

#

and it also seems to not apply for dropdowns :/

#

maybe i'll try using haxe timers instead of kha's scheduler

#

lol bad idea... infinite loop somehow

#

ahh nothing to do with cppia, commented out trace and forgot about that

#

didn't fix the other window issue, but was able to at least test it works if cppia has all the references to things

#

dragging is laggy because of the high call later delay for the text measurement

#

i also think pointer events are working as expected here

plain vine Oct 31, 2025, 12:10 PM

#

it was a really subtle issue

#

#

so close 😭

#

haxe/ui/backend/TextDisplayImpl.hx:76: Autosized Panels (Height),Label,284,284,true,0,18,18
haxe/ui/backend/TextDisplayImpl.hx:76: Autosized Panels (Height),Label,284,284,true,18,18,18
haxe/ui/backend/TextDisplayImpl.hx:76: Panel Title,Label,283,0,false,0,0,18
haxe/ui/backend/TextDisplayImpl.hx:76: Panel Title,Label,283,116,false,18,18,18
haxe/ui/backend/TextDisplayImpl.hx:62: set fontsize
haxe/ui/backend/TextDisplayImpl.hx:76: ,Label,0,0,true,18,18,18
haxe/ui/backend/TextDisplayImpl.hx:76: Panel Title,Label,243,116,false,18,18,18
haxe/ui/backend/TextDisplayImpl.hx:62: set fontsize
haxe/ui/backend/TextDisplayImpl.hx:76: Action,Label,71,71,true,0,18,18
haxe/ui/backend/TextDisplayImpl.hx:76: Action,Label,71,71,true,18,18,18

#

every value seems to look fine, wonder why the switch isn't in the correct location

plain vine Oct 31, 2025, 11:38 PM

#

#

we got properly aligned text 😄

#

another odd one to take some time getting right

plain vine Nov 1, 2025, 12:20 PM

#

#

got alpha going 😄

plain vine Nov 1, 2025, 2:18 PM

#

da fuck is meant to be hiding this checkmark

#

visibility is mapped and so is alpha, so why can i still see the checkmark 🤔

#

I need to map removeImageDisplay :D

#

hm.. i already did nvm x)

#

ahhh forgot to remove it from a child list which meant it got collected for rendering indirectly

#

#

😭

broken python Nov 1, 2025, 3:26 PM

#

@dense elbow is there some kind of exhaustive list of what needs to be implemented by a haxeui backend?

broken python Nov 1, 2025, 5:24 PM

#

(maybe I should ask that in #haxeui )

plain vine Nov 1, 2025, 5:25 PM

#

broken python <@456226577798135808> is there some kind of exhaustive list of what needs to be ...

no such thing exists

#

just existing backends code

broken python Nov 1, 2025, 5:27 PM

#

Like, for instance, in Ceramic, there is some sort of specification: all API that needs to be implemented as backend is provided as interface in a spec package.

#

(even though the interface isn't used when building the final app)

plain vine Nov 1, 2025, 8:04 PM

#

broken python (even though the interface isn't used when building the final app)

Ahh!! I was wondering what that stuff was for as i was browsing around the repo

plain vine Nov 1, 2025, 8:52 PM

#

can't seem to get images back on screen in the tabs... no idea where they've gone

plain vine Nov 2, 2025, 1:12 PM

#

plain vine

note to self - commit more 🙃

plain vine Nov 2, 2025, 6:21 PM

#

I'm setting up a base template project so i can easily spin up new projects when i want and i think i've fixed that weird fallback issue i was having with haxeui components

#

some behaviours still seem to not work though, guessing some kind of cppia issue there

#

can repro that on ceramic + cppia

#

so not unique to my stack at least

#

but i think I have enough done to at least start porting over some of my app

plain vine Nov 2, 2025, 6:54 PM

#

need to figure out why this particular behaviour isn't working but others must be working

#

so a quick look... it seems like clonable behaviours are fine and non clonable behaviours are not 🤔

#

unfortunately added clonable doesn't fix it 😄

plain vine Nov 2, 2025, 7:48 PM

#

plain vine I'm setting up a base template project so i can easily spin up new projects when...

CallMember haxe.ui.containers.windows._WindowList.WindowListItemLayout (0x5c32279e6164 0x5c322a4c3c60) 'get_paddingRight' fallback
CallMember haxe.ui.containers.windows._WindowList.WindowListItemLayout (0x5c32279e6164 0x5c322a4c3c60) 'marginLeft' fallback
CallMember haxe.ui.containers.windows._WindowList.WindowListItemLayout (0x5c32279e6164 0x5c322a4c3c60) 'marginRight' fallback
CallMember haxe.ui.containers.windows._WindowList.WindowListItemLayout (0x5c32279e6164 0x5c322a4c3c60) 'get_horizontalSpacing' fallback
nvm 😄

plain vine Nov 2, 2025, 8:48 PM

#

i'm very lucky, want my base project template to have git submodules so i can use it with gsm easily

#

but creating a project on gitea wasn't carrying over the submodules

#

they fixed it, so just a simple upgrade process and gitea continues to be completely cool

#

also haxe files now have the haxe logo 😄

plain vine Nov 2, 2025, 9:09 PM

#

@dense elbow https://github.com/haxeui/haxeui-core/pull/692 lemme know if this change breaks something for you, it works fine for me but i'm only testing in cpp land atm

final field Nov 2, 2025, 9:19 PM

#

2 imports jumpscare

plain vine Nov 2, 2025, 9:28 PM

#

yea, haxe 5 don't like the implicit import situation x)

#

it's a small change in theory

#

so it should all be fine...

#

but you never know

plain vine Nov 4, 2025, 11:41 AM

#

Gonna see if I can get a page from my app running in my framework

#

will use the moment to potentially refactor a bunch of stuff

#

just want UI transfered... no behaviour to begin with

#

#

😄

#

getting traces i never hit in my haxeui backend

#

this is going to be a fun thing

plain vine Nov 4, 2025, 12:10 PM

#

@nocturne adder Heh, here's an example of something that i was talking about yesterday. Trying to port my app over to my new setup, went to use the mouse wheel and realised i never mapped it - wouldn't have noticed otherwise x)

nocturne adder Nov 4, 2025, 1:01 PM

#

xD

#

I can fully imagine you being like
" why is this scroll wheel not working ??? "
" oh . . . "

#

happens to all of us ahahahaha

plain vine Nov 4, 2025, 1:03 PM

#

yeah man, with framework building hat on you're thinking is different

#

something inconsequential like scroll wheel is not even on the rader

plain vine Nov 4, 2025, 4:33 PM

#

@agile snow I have the opposite of a load bearing trace 😄

#

a trace that breaks everything :D

agile snow Nov 4, 2025, 4:51 PM

#

that somehow makes less sense

#

great job

plain vine Nov 4, 2025, 7:00 PM

#

gaaaah, having issues with haxeui and cppia

#

not sure what's going on... but if i reload the script i can't seem to load certain things again

#

#

buuut that's still pretty damn cool

#

Its working :D

#

if i reload the script i can't see the pages again :/

#

however it seems like the navbar itself can be reloaded fine... maybe something to do with navigationmanagers state...

#

yep... that's it

#

@:privateAccess haxe.ui.navigation.NavigationManager?._instance = null; added to the cleanup process

#

i'm not sure why it takes a bit of time to load up the orderpage, maybe its just a big xml so it takes a bit? but hey, the poc is all working blobcheer

#

ceramic is a decent bit faster at loading the page than i am

#

so something bottlenecking me i guess?

carmine path Nov 4, 2025, 8:03 PM

#

plain vine `@:privateAccess haxe.ui.navigation.NavigationManager?._instance = null;` added ...

a classic

plain vine Nov 4, 2025, 9:23 PM

#

this whole trace breaking shit issue is getting a little annoying :D

#

var val = Sys.getCwd();
trace(val);

....... white screen

#

remove trace? completely fine!

nocturne adder Nov 4, 2025, 9:27 PM

#

plain vine ```hx var val = Sys.getCwd(); trace(val); ``` ....... white screen

wtf

#

what does your debugger say?

plain vine Nov 4, 2025, 9:28 PM

#

nocturne adder what does your debugger say?

no errors or anything, not sure what the debugger should report

#

#

all traces run as expected, but i just get a big old white screen

nocturne adder Nov 4, 2025, 9:30 PM

#

what if you open it up with your debugger and set a breakpoint at your mainloop

#

or perhaps just make it immediately break, and see what it is at

#

perhaps it's stuck in a loop somewhere?

plain vine Nov 4, 2025, 9:30 PM

#

trace('[haxeui-simple] Error - Could not find image'); <--- fine
trace('[haxeui-simple] Error - Could not find image($filename)'); <--- white screen

nocturne adder Nov 4, 2025, 9:30 PM

#

if that doesn't give you any clue, it might be something memory related?

#

unpredictable stuff can happen if you have any misusage

plain vine Nov 4, 2025, 9:31 PM

#

very likely

plain vine Nov 4, 2025, 9:32 PM

#

nocturne adder what if you open it up with your debugger and set a breakpoint at your mainloop

pausing gives me an error

nocturne adder Nov 4, 2025, 9:32 PM

#

I see LLDB, so I assume this is native debugging?

plain vine Nov 4, 2025, 9:32 PM

#

yeah

nocturne adder Nov 4, 2025, 9:33 PM

#

Perhaps try another debugger that gives you a litttleee more information

#

you want to see the callstack, what each thread is doing and that kind of stuff

plain vine Nov 4, 2025, 9:33 PM

#

what should I use? I do get thread info here

nocturne adder Nov 4, 2025, 9:33 PM

#

oh you do

#

hmmm

plain vine Nov 4, 2025, 9:33 PM

#

nocturne adder Nov 4, 2025, 9:34 PM

#

where is __GI__clock_nanosleep called from?

plain vine Nov 4, 2025, 9:35 PM

#

HXLINE( 157) SDL_Delay(::Math_obj::floor(((_gthis->targetTime - elapsed1) * ( (Float)(1000) )))); seems like here's the closest i can get to it

#

that's expected stuff though, however commenting it out doesn't make any difference

#

wonder what the compiled code differences are

#

HXLINE(  32)            ::haxe::Log_obj::trace(((HX_("[haxeui-simple] Error - Could not find image(",10,b5,75,95) + filename) + HX_(")",29,00,00,00)),::hx::SourceInfo(HX_("haxe/ui/backend/AssetsImpl.hx",0e,18,11,48),32,HX_("haxe.ui.backend.AssetsImpl",cf,ee,b0,8f),HX_("imageFromFile",81,cb,29,81)));

So that's with the $filename variable

HXLINE(  32)            ::haxe::Log_obj::trace(HX_("[haxeui-simple] Error - Could not find image()",19,b9,88,31),::hx::SourceInfo(HX_("haxe/ui/backend/AssetsImpl.hx",0e,18,11,48),32,HX_("haxe.ui.backend.AssetsImpl",cf,ee,b0,8f),HX_("imageFromFile",81,cb,29,81)));

that's without

nocturne adder Nov 4, 2025, 9:39 PM

#

and filename is a String?

plain vine Nov 4, 2025, 9:39 PM

#

yes

nocturne adder Nov 4, 2025, 9:41 PM

#

weird...

plain vine Nov 4, 2025, 9:42 PM

#

yeah, how the hell do i figure this shit out

nocturne adder Nov 4, 2025, 9:42 PM

#

I'd love to help but I am not on linux atm

#

So that would be difficult unfortunately

#

Can't debug it XD

#

but this kind of RNG behaviour sounds like memory issues to me...

plain vine Nov 4, 2025, 9:43 PM

#

how does one investigate memory issues

nocturne adder Nov 4, 2025, 9:43 PM

#

try valgrind I suppose?

#

I do think it won't be smooth sailing

#

considering the GC and stuff, wouldn't be surprised if valgrind flags everything as misuse XD

plain vine Nov 4, 2025, 9:45 PM

#

blobmeltsob

plain vine Nov 4, 2025, 9:46 PM

#

nocturne adder I do think it won't be smooth sailing

#

🤣

#

oh very interesting, that only occurs when i DONT have the trace

#

==71571== LEAK SUMMARY:
==71571== definitely lost: 658,123 bytes in 3,138 blocks
==71571== indirectly lost: 7,884,372 bytes in 113,580 blocks
==71571== possibly lost: 25,455,557 bytes in 53,330 blocks
==71571== still reachable: 26,163,496 bytes in 29,507 blocks
==71571== of which reachable via heuristic:
==71571== newarray : 4,456 bytes in 13 blocks
==71571== suppressed: 0 bytes in 0 blocks
==71571==
==71571== Use --track-origins=yes to see where uninitialised values come from
==71571== For lists of detected and suppressed errors, rerun with: -s
==71571== ERROR SUMMARY: 434425 errors from 8386 contexts (suppressed: 0 from 0)

#

Well, i'm guessing that aint great

#

i wonder to what degree cppia is involved here

#

might need to use hxcpp and also face whatever issues occur on that side of the coin 😄

nocturne adder Nov 4, 2025, 10:23 PM

#

plain vine

I'm not surprised XD

nocturne adder Nov 4, 2025, 10:24 PM

#

plain vine oh very interesting, that only occurs when i DONT have the trace

But I will say, this is highly suspicious.

#

That makes me think that your program is frozen or something?

#

Like truly frozen

#

But you did get traces?

#

Something weird is going on for sure...

plain vine Nov 4, 2025, 10:25 PM

#

nocturne adder That makes me think that your program is frozen or something?

it could be, its just hard to tell because the screen is flat white

#

also, daz has mentioned in the past that there's a way to extern things like cpp structs and still use them with haxe semantics

#

like null checking etc

#

do you have any idea on this? I might take a quick detour and redo my skia externs now that i'm not so blind

nocturne adder Nov 4, 2025, 10:27 PM

#

I don't know what that would be exactly.

#

cpp.Struct<T> sounds close to that. But it has some gotchas

#

Is that what you are referring to?

plain vine Nov 4, 2025, 10:28 PM

#

i'm not sure tbh

#

I just want to be able to null check some cpp types (or null them) in haxe land that makes hxcpp unhappy

#

    public var image(default, set):Image;
    function set_image(value) {
        imageSet = true;
        return image = value;
    }

#

@:keep
@:unreflective
@:include("linc_skia.h")
@:native("SkImage")
extern class NativeImage {}

#

In file included from ./src/simple/Rectangle.cpp:15:
include/simple/Rectangle.h:63:26: error: cannot declare field ‘simple::Rectangle_obj::image’ to be of abstract type ‘SkImage’
   63 |                  SkImage image;
      |                          ^~~~~
In file included from /home/j/Documents/Projects/cpp/TradingProject/haxe_modules/linc_skia/linc/linc_skia.h:12:
/home/j/Documents/Projects/cpp/TradingProject/haxe_modules/linc_skia/lib/include/core/SkImage.h:270:14: note:   because the following virtual functions are pure within ‘SkImage’:
  270 | class SK_API SkImage : public SkRefCnt {
      |              ^~~~~~~
/home/j/Documents/Projects/cpp/TradingProject/haxe_modules/linc_skia/lib/include/core/SkImage.h:427:18: note:     ‘virtual bool SkImage::isTextureBacked() const’
  427 |     virtual bool isTextureBacked() const = 0;
      |                  ^~~~~~~~~~~~~~~

//more errors

final field Nov 4, 2025, 10:35 PM

#

Smells like either a race condition or hxcpp shitting itself

plain vine Nov 4, 2025, 10:39 PM

#

nah, if i understand things properly this is because i'm trying to use a cpp type by value that shouldn't be used that way

#

but in haxe land, we use things by value

#

gaaaah, i think what i'm gonna do is new project, and start from like step 2/10 for skia externs and just try and see if i can build it up closer to what the actual api is

#

which means saying bye bye to my project for a bit again 😭

#

maybe i'll hop back and forth a little

#

I think technically there's like 3 targets here that i'm trying to compile for.................. what a massive pain in the ass

final field Nov 4, 2025, 10:51 PM

#

We need a modern ammer

#

XML->JVM, HXCPP, reflaxe.cpp, HL and Neko trolley

nocturne adder Nov 4, 2025, 10:52 PM

#

give @:native("::cpp::Struct< SkImage >") a try

#

that should work

plain vine Nov 5, 2025, 3:08 PM

#

==34546== LEAK SUMMARY:
==34546== definitely lost: 7,072 bytes in 159 blocks
==34546== indirectly lost: 5,232 bytes in 6 blocks
==34546== possibly lost: 6,065,029 bytes in 2,822 blocks
==34546== still reachable: 11,016,742 bytes in 23,521 blocks
==34546== suppressed: 0 bytes in 0 blocks
==34546== Reachable blocks (those to which a pointer was found) are not shown.
==34546== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==34546==
==34546== For lists of detected and suppressed errors, rerun with: -s
==34546== ERROR SUMMARY: 541 errors from 541 contexts (suppressed: 0 from 0)
Okay, with a completely simple setup, none of my old extern code and running on hxcpp instead of cppia, seems like we still have a bunch of errors

nocturne adder Nov 5, 2025, 3:09 PM

#

Don't worry about it

#

That is normal probably

#

It is very likely falsly flagging a bunch of stuff as "leaks" because it hasn't been collected by the GC

plain vine Nov 5, 2025, 3:10 PM

#

yeah, just don't have any frame of reference for this kind of debugging

nocturne adder Nov 5, 2025, 3:11 PM

#

my intent for using valgrind was not for checking leaks, but checking other memory related issues

plain vine Nov 5, 2025, 3:11 PM

#

my externs did need redoing anyway, so i'm not gonna stop just was curious about the state of this side of things

plain vine Nov 5, 2025, 11:24 PM

#

#

so far, only one cpp wrapper function :D

#

and everything is compatible between both hxcpp and cppia so far

plain vine Nov 6, 2025, 12:28 AM

#

#

basically like 80% through the rewrite 😄

plain vine Nov 6, 2025, 10:36 PM

#

@dense elbow ```hx
typedef ImageData = Image;

For a backend, is it possible to wrap `ImageData` in anyway? or does it have to be a typedef?

#

all backends seems to just pass the type to it directly can't find an example of one otherwise

#

i guess ic an just wrap the image into a class right here, can't I? 😂

#

i can be very silly sometimes

plain vine Nov 6, 2025, 11:42 PM

#

its funny, i just noticed how my goal of "keeping to the skia api" was leeching a bit in my framework on top of it

#

its completely unnecessary :D

plain vine Nov 7, 2025, 1:56 AM

#

#

Its fucking alive

#

Hxcpp and cppia both working

#

okay some really important information, if you're going to be constantly going back and forth between hxcpp and cppia. TURN OFF THE COMPILATION SERVER

#

I went on a wild goose chase for the last 2-3 hours because of that shit

final field Nov 7, 2025, 2:07 AM

#

plain vine

hxcppia

plain vine Nov 7, 2025, 2:10 AM

#

#

I can hopefully drastically reduce the time i stare at the shitty out of date skia docs

#

well, we'll see, i know there's some things i'm probably gonna have to look at

dense elbow Nov 7, 2025, 6:46 AM

#

plain vine <@456226577798135808> ```hx typedef ImageData = Image; ``` For a backend, is it ...

any type is fine

#

class ImageData {} is also perfectly alright

carmine path Nov 7, 2025, 8:56 AM

#

plain vine okay some really important information, if you're going to be constantly going b...

oh, that explains a lot. I dont use the server at all

broken python Nov 7, 2025, 10:50 AM

#

carmine path oh, that explains a lot. I dont use the server at all

not even for code completion?

plain vine Nov 7, 2025, 10:50 AM

#

i never did for code completion

#

ohhh i typed it wrong

broken python Nov 7, 2025, 10:50 AM

#

plain vine i never did for code completion

Well, that's the default in vscode

plain vine Nov 7, 2025, 10:50 AM

#

i meant compilation server

broken python Nov 7, 2025, 10:51 AM

#

Yeah, I usually avoid compilation server. Got enough bugs already, no need to add more

plain vine Nov 7, 2025, 10:51 AM

#

I use it for cppia, drastically speeds up the recompiles

#

especially with haxeui in the mix

broken python Nov 7, 2025, 10:52 AM

#

For sure you must restart it if you change your build params

#

But even when doing that, sometimes it does weird thing

#

Like I get a compile error on some type not being valid or resolved

#

Or even something that compiles but gives different output than if server was off

#

It's definitely not bug-free

#

(unfortunately I didn't managed to get a reproducible pattern, pretty random)

plain vine Nov 7, 2025, 10:53 AM

#

i really never considered it, it makes sense now that i'm looking at it in hindsight how i'd be able to relatively decently understand what issues were for cppia, but when i'd move to hxcpp and try to address the same errors in a similar way the solution i was applying wasn't working

#

but i was too unskilled to think that was a setup problem :/

carmine path Nov 7, 2025, 11:00 AM

#

broken python not even for code completion?

nah

broken python Nov 7, 2025, 11:04 AM

#

carmine path nah

Curious, what code editor do you use,

plain vine Nov 7, 2025, 11:21 AM

#

Feel a bit empty/relieved now that i've got the hxcpp and cppia concept working together

#

could almost say i feel like i'm done, but aint no where close 😂

#

next up is starting to abstract away some of the cpp stuff of skia in my framework now and i think i'm still having shutdown issues with cppia so gonna tackle those next

#

I do not want to switch over to hxcpp one day and get silly can't compare to dynamic errors x)

#

sometimes these errors don't give a clear origin point

plain vine Nov 7, 2025, 11:46 AM

#

hm... NavigationManager in haxeui doesn't recollect its state on reload, so it doesn't understand that certain pages exist

carmine path Nov 7, 2025, 11:47 AM

#

broken python Curious, what code editor do you use,

sublime text

plain vine Nov 7, 2025, 11:47 AM

#

i basically need to either rebuild the correct state or make my own navigation manager

plain vine Nov 7, 2025, 12:16 PM

#

oh nice, navigationmanager just cares about the constructor so i think rebuilding this will be easy :D

#

oh, maybe not, it stores the reference to the item created in the last reload 😂

plain vine Nov 7, 2025, 1:36 PM

#

        if (@:privateAccess nav.registeredRoutes.length == 0) {
            trace("no routes registered");
            var route = null;
            var routes = ['blank', 'calendar', 'home', 'orderpage'];

            for (page in routes) {
                var val:Void->INavigatableView = switch (page) {
                    case 'blank':
                        Blank.new;
                    case 'calendar':
                        Calendar.new;
                    case 'home':
                        Home.new;
                    case 'orderpage':
                        Orderpage.new;
                    default:
                        null;
                }
                var route:RouteDetails = {viewCtor: val, error: false, preserveView: false};
                if (page == 'home') {
                    route.initial = true;
                }

                nav.registerRoute('/$page', route);
            }

#

oh man this is crazy even by remaking the routes in the scripts entry point, the items it seems to reference here is the old version?

#

Not sure how to address that at all

#

hm... might still have other cleanup issues

plain vine Nov 7, 2025, 8:08 PM

#

managed to sort it out

#

project is in a good place to remember to commit :D

#

basically everything in the foundation appears to be working now

#

hxcpp, cppia and hotreloading

plain vine Nov 8, 2025, 12:44 AM

#

#

property grids work now :)

#

still seem to have the weird invalidation loop bug but the behaviour bug has gone now with the rewrite

#

as long as i don't open a dropdown or click a scrollbar i'm okay xD

plain vine Nov 8, 2025, 11:48 AM

#

added a little convenience feature and playing with cppia a bit

#

if i do a hotreload, it will go back to the last page i was looking at x)

#

I'd like to figure out how to get it to instantly go back to the same page instead of having to play with timings, but, for now this is still nice

plain vine Nov 8, 2025, 12:29 PM

#

hahaha got it 😄 no more ever needing to faff about what page i'm working on

#

think i managed to shave off around 0.1s of the recompile time by turning dce on full for the script side only

plain vine Nov 8, 2025, 4:15 PM

#

[71640:1108/161431.461908:ERROR:gpu/command_buffer/service/shared_image/shared_image_manager.cc:389] SharedImageManager::ProduceSkia: Trying to Produce a Skia representation from a non-existent mailbox.

#

Running my appp in ceramic to check things out and notice this lil error message referring to skia 😄

#

project isn't fully transferred over but with cppia my reload times are on average 50% less

plain vine Nov 8, 2025, 5:32 PM

#

#

managed to get tracy working properly and give me haxe code readings and... mother of god

#

why the hell is this a thing for me but not in other backends

#

🤔

nocturne adder Nov 8, 2025, 6:19 PM

#

ohhhh the event loop

#

yeah....

#

i despise the event loop

carmine path Nov 8, 2025, 7:31 PM

#

That is horrible

#

You are creating soo many Timers

plain vine Nov 8, 2025, 9:20 PM

#

Will have to look when I get home, maybe I'll see if I can queue up haxeui events instead but I have no idea why there's so many

plain vine Nov 9, 2025, 12:17 AM

#

#

that's better

#

also see there's some low hanging fruit for things to optimise in here

#

#

madness tho

#

i wonder if queuing these events up would make much difference, it would basically be the same just moved... right?

plain vine Nov 9, 2025, 9:23 AM

#

#

lol such a small change and big difference

nocturne adder Nov 9, 2025, 9:28 AM

#

yeah that's the fun part of tracy imo

#

you can very quickly tell (and also see) how small changes can have huge implications

plain vine Nov 9, 2025, 9:34 AM

#

i like it much more than chrome's profiler

#

wish we had a decent hxcpp debugger for the haxe side tho

nocturne adder Nov 9, 2025, 9:42 AM

#

plain vine wish we had a decent hxcpp debugger for the haxe side tho

I personally use remedybg

#

I really like it

plain vine Nov 9, 2025, 9:43 AM

#

i have a cpp debugger that works fine

#

its just i want to skip the whole find cpp file, search through the cpp code to find the corresponding haxe code dance

nocturne adder Nov 9, 2025, 9:44 AM

#

I remember Aiden used to work on mixed debugging for hxcpp

#

that would probably be ideal for what you want, but I think it is still very experimental (and possibly abandoned?)

plain vine Nov 9, 2025, 9:47 AM

#

yeah, i checked it out recently

#

gonna retry the hxcpp debugger and see if its okay, maybe i can just improve the workflow a bit

nocturne adder Nov 9, 2025, 9:49 AM

#

good luck 🫡

plain vine Nov 9, 2025, 9:58 AM

#

if i setup its own hxml i think its pretty okay

#

yeah, it seems to work fine if i don't use the same config as the rest of my project

plain vine Nov 9, 2025, 9:44 PM

#

okay, so i think i have a theory

#

what might be happening is i'm treating this invalidation cascade like "next..next...next...next"

#

if i were to just skip or batch the things somehow then maybe it would be better? 🤔

#

batch events instead.... 🤔

#

also, funnily enough those changes i made earlier and yesterday from tracy, has made a difference in hxcpp and made a drastic difference in cppia

#

still very hangy, buuuut its really cool

plain vine Nov 10, 2025, 12:38 AM

#

Before

#

Don't know how obvious that is but the page loads in a lot faster now

#

also, I seem to have gotten rid of the hanging completely

#

The page loading in DOES actually still hinder frames according to tracy, but the only thing i'm doing differently here is just simply managing it better, which inturn seems to also speed things up a bit

agile snow Nov 10, 2025, 12:40 AM

#

it feels like a normal speed now

#

it was really noticable when you first posted it

plain vine Nov 10, 2025, 12:41 AM

#

yeah, it was quite awful

#

still is not great, but certainly usable and app never seems to stop responding now

#

There is actually a change that i have pegged to do that i might push sooner rather than later

#

which might substantially help but for another time

#

release build feels pretty nice all things considering

#

not as efficient as i'd like but feel good to use regardless

carmine path Nov 10, 2025, 12:55 AM

#

pretty good stuff!

plain vine Nov 10, 2025, 12:58 AM

#

right now everything is very much immediate mode as well 😄

plain vine Nov 10, 2025, 3:24 PM

#

@dense elbow what's going on here, does anyone have any insights on what my framework isn't handling well in this chain of events

#

What i'm doing is, i'm opening a new page

#

#

that's the page^

#

There's something i'm not handling well and i cant see what it is

#

i can bare with it, but, its bugging me x)

#

maybe its something silly

dense elbow Nov 10, 2025, 3:58 PM

#

is it doing that over and over?

#

i mean, the cloneComponent part... that should only happen once

#

well, i mean, in one chunk

plain vine Nov 10, 2025, 4:00 PM

#

dense elbow is it doing that over and over?

just once from what i can tell, each pic is a frame

#

first and last frame just seems to be component validation

#

#

these blocks all seem to be ComponentBase.validateComponent calls

dense elbow Nov 10, 2025, 4:10 PM

#

is validcomponent being called when the UI is idle (because it shouldnt be)

plain vine Nov 10, 2025, 4:10 PM

#

nah, the issue only occurs when haxeui "actions" take place

#

so changing a page

dense elbow Nov 10, 2025, 4:11 PM

#

so i dont understand, whats the issue? I would expect a bunch of invalidations if things are.... .... well.... invalidated

plain vine Nov 10, 2025, 4:11 PM

#

it "locks up" during the load in

#

i don't get the same effect on other backends, so i'm wondering what i'm not handling properly but i can't seem to identify much in my framework

dense elbow Nov 10, 2025, 4:12 PM

#

oh, so too many invalidations maybe... are you batching things? I mean, the whole point of callLater is you save up a bunch of callbacks and then execute them "at once"

plain vine Nov 10, 2025, 4:13 PM

#

dense elbow Nov 10, 2025, 4:13 PM

#

its a way of syncing validations basically so if you invalidate a component 100 times, you should only get a single invalidation

plain vine Nov 10, 2025, 4:13 PM

#

class CallLaterImpl {
    static var queue:List<Void->Void> = new List();
    static var id:Int = -1;
    public function new(fn:Void->Void) {
        if (id == -1) {
            id = Scheduler.addFrameTask(processQueue, 1);
        }
        queue.add(fn);
    }
    public static function processQueue() {
        for (i in 0...queue.length) {
            queue.pop()();
        }
    }
}
``` that's what i currently have for calllaterimpl

dense elbow Nov 10, 2025, 4:14 PM

#

looks ok i think

plain vine Nov 10, 2025, 4:14 PM

#

its the best working version of call later i have

#

Schedule.addTimeTask(fn, 0,001) (like in kha) was a lot worse

#

#

but for that frame, this is all my framework really contributes

#

so i'm not sure what exactly i'm not handling very well here

dense elbow Nov 10, 2025, 4:17 PM

#

me neither, scheduler looks alright, but its hard to know fully... usually a lot of nuance in these things

plain vine Nov 10, 2025, 4:17 PM

#

yeah, i suspect its something really simple and obvious

dense elbow Nov 10, 2025, 4:17 PM

#

(i hate that calllater is an allocation to 🙁 )

plain vine Nov 10, 2025, 4:18 PM

#

but i don't have the full context of haxeui internals to easily understand

dense elbow Nov 10, 2025, 4:20 PM

#

it feels weird that something like get_parentComponent is so expensive

#

is this cppia?

plain vine Nov 10, 2025, 4:21 PM

#

hxcpp

#

cppia funnily enough runs better 😂

dense elbow Nov 10, 2025, 4:21 PM

#

does it run better / different in release?

plain vine Nov 10, 2025, 4:21 PM

#

things do run better in release but i still feel that "chug" when changing the page

dense elbow Nov 10, 2025, 4:21 PM

#

weird

#

im not fully conviced its the calllater tbh

#

it doesnt look wrong or anything

plain vine Nov 10, 2025, 4:23 PM

#

maybe i have some extra calls to parentcomponent, will investigate that one

dense elbow Nov 10, 2025, 4:23 PM

#

or maybe you are invalidating things in the backend? Its quite rare to have to do that

plain vine Nov 10, 2025, 4:23 PM

#

i did do that early on but i removed them.. will double check that one

#

i call resizeComponent when adding one in screenimpl and i call comp.ready() when an item is added to screen in componentimpl

dense elbow Nov 10, 2025, 4:24 PM

#

that seems fine / normal

plain vine Nov 10, 2025, 4:25 PM

#

only reference to parent is this:

    public function hasChildRecursive(parent:Component, child:Component):Bool {
        if (parent == child) {
            return true;
        }
        var r = false;
        for (t in parent.childComponents) {
            if (t == child) {
                r = true;
                break;
            }

            r = hasChildRecursive(t, child);
            if (r == true) {
                break;
            }
        }

        return r;
    }
``` which i took from another backend to check if components were over

dense elbow Nov 10, 2025, 4:27 PM

#

defo could be expensive, but thats not even calling get_parentComponent

plain vine Nov 10, 2025, 4:28 PM

#

this is using daz's fork of haxeui but i get basically the same issue on normal core

dense elbow Nov 10, 2025, 4:28 PM

#

btw, may or may not help, but this is the new way in v3, which i might backport into v2, its defo better:

    public override function onNextTick(fn:Void->Void) {
        queue.push(fn);
        if (!scheduled) {
            scheduled = true;
            if (tickDelay == -1) {
                Browser.window.requestAnimationFrame((_) -> {
                    scheduled = false;
                    var copy = queueTickStart;
                    queueTickStart = [];
                    for (f in copy) {
                        f();
                    }
                    copy = queue;
                    queue = [];
                    for (f in copy) {
                        f();
                    }
                    copy = queueTickComplete;
                    queueTickComplete = [];
                    for (f in copy) {
                        f();
                    }
                });
            } else {
                haxe.Timer.delay(() -> {
                    scheduled = false;
                    var copy = queueTickStart;
                    queueTickStart = [];
                    for (f in copy) {
                        f();
                    }
                    copy = queue;
                    queue = [];
                    for (f in copy) {
                        f();
                    }
                    copy = queueTickComplete;
                    queueTickComplete = [];
                    for (f in copy) {
                        f();
                    }
                }, tickDelay);
            }
        }
    }

you can ignore the queueTickComplete, queueTickStart and tickDelay

plain vine Nov 10, 2025, 4:29 PM

#

ah, call later change?

dense elbow Nov 10, 2025, 4:30 PM

#

so basically:

        queue.push(fn);
        if (!scheduled) {
            scheduled = true;
            Browser.window.requestAnimationFrame((_) -> {
                scheduled = false;
                copy = queue;
                queue = [];
                for (f in copy) {
                    f();
                }
            });
        }

dense elbow Nov 10, 2025, 4:30 PM

#

plain vine ah, call later change?

calllater doesnt exist in v3... its Platform.instance.eventLoop.onNextTick(...)

plain vine Nov 10, 2025, 4:30 PM

#

oh interesting

plain vine Nov 10, 2025, 4:59 PM

#

So in those 3 frames ruleMatch is getting called ~1.5million times

#

could that be reduced some how 🤔

dense elbow Nov 10, 2025, 5:01 PM

#

i mean, the css engine is pretty not optimized (i think Daz's fork is better) but 1.5M sounds like too many

plain vine Nov 10, 2025, 5:01 PM

#

ahh sorry that's not right

#

yeah, tracy doesn't automatically reconfigure the stats if you change frame with different settings

#

its around 300k

dense elbow Nov 10, 2025, 5:03 PM

#

better but still seems high

#

hard to know though really

#

are invalidating anything in your backend?

#

can you zip it up, i dont need to run it, just have a quick scan

plain vine Nov 10, 2025, 5:05 PM

#

dense elbow can you zip it up, i dont need to run it, just have a quick scan

📎 haxeui-simple.zip

#

i think the backend is quite clean

#

nothing too complicated going on in this one

#

you can also take a peak at the readme 🤣

dense elbow Nov 10, 2025, 5:08 PM

#

can you send simple.Scheduler also

#

why List also?

plain vine Nov 10, 2025, 5:11 PM

#

dense elbow can you send `simple.Scheduler` also

https://github.com/Kode/Kha/blob/main/Sources/kha/Scheduler.hx

plain vine Nov 10, 2025, 5:12 PM

#

dense elbow why List also?

did a simple test on try haxe array vs list and for some cases list was faster, i did question if list was a problem here

dense elbow Nov 10, 2025, 5:12 PM

#

i think list is slower of modifcations...

plain vine Nov 10, 2025, 5:12 PM

#

i switched it to an array and it made little impact so i just fellback to list based on the little benchmark

#

i can move it back to array, its no biggy

plain vine Nov 10, 2025, 5:13 PM

#

dense elbow i think list is slower of modifcations...

list works better when you remove/modify from the ends

dense elbow Nov 10, 2025, 5:14 PM

#

but costs more to add? Anyway, i doubt its the bottleneck

#

so looking, i do think its the call later impl... it feels smelly for some reason

plain vine Nov 10, 2025, 5:15 PM

#

is it the deferring to a time in the frame?

dense elbow Nov 10, 2025, 5:16 PM

#

dunno, maybe its fine... its does mean that the frame task is always running, but i guess thats fine... it could just be with an empty queue

plain vine Nov 10, 2025, 5:18 PM

#

the evaluation of roughly 3 average fps over those 3 frames, time task vs that call later approach look the same in tracy

#

but the frametask "feels nicer"

dense elbow Nov 10, 2025, 5:20 PM

#

hang on... i think you are running these callbacks in the wrong order

#

you are running them in the opposite order they have come in at

#

so that could mean you are causing a bunch of invalidations because things arent being validated in the order they were expecting (haxeui validation loop sorts things based in a validation depth)

plain vine Nov 10, 2025, 5:22 PM

#

add to the end and take from the top, isn't that the expected order? fifo

dense elbow Nov 10, 2025, 5:22 PM

#

no, the order is the order they are added

#

add to the end and take from the end (which is what you are doing) isnt fifo

plain vine Nov 10, 2025, 5:23 PM

#

lilo* 😂

dense elbow Nov 10, 2025, 5:24 PM

#

it should be fifo... they should be executed in the order they have been given to you

plain vine Nov 10, 2025, 5:24 PM

#

oh huh

#

lemme try that

#

#

😅

dense elbow Nov 10, 2025, 5:26 PM

#

... better i take it?

plain vine Nov 10, 2025, 5:26 PM

#

lol

dense elbow Nov 10, 2025, 5:27 PM

#

it makes sense... haxeui validation loop goes to some lengths to minimize validation by validiting, iirc, children first, then parents (or the other way round, i honestly dont remember now)... so that the children are "done" and wont cause revalidation

plain vine Nov 10, 2025, 5:27 PM

#

dense elbow so that _could_ mean you are causing a bunch of invalidations because things are...

i think that change caused this

dense elbow Nov 10, 2025, 5:28 PM

#

are you saying its worse, or better?

plain vine Nov 10, 2025, 5:28 PM

#

worse, red means < 30 fps

#

average fps for those frames are around 11fps

dense elbow Nov 10, 2025, 5:29 PM

#

ok, well, its defo the right way to be doing things

#

maybe try array now

#

you shouldnt be reversing the invalidation order... defo not

plain vine Nov 10, 2025, 5:30 PM

#

so push and then pop?

dense elbow Nov 10, 2025, 5:30 PM

#

i would just do this:

#

class CallLaterImpl {
    static var queue:Array<Void->Void> = [];
    static var id:Int = -1;
    public function new(fn:Void->Void) {
        if (id == -1) {
            id = Scheduler.addFrameTask(processQueue, 1);
        }
        queue.push(fn);
    }
    public static function processQueue() {
        if (queue.length == 0) {
            return;
        }
        var copy = queue;
        queue = [];
        for (f in copy) {
            f();
        }
    }
}

(written blind)

plain vine Nov 10, 2025, 5:32 PM

#

same outcome

dense elbow Nov 10, 2025, 5:32 PM

#

¯_(ツ)_/¯

plain vine Nov 10, 2025, 5:32 PM

#

yeeaaa 🤣

dense elbow Nov 10, 2025, 5:32 PM

#

no clue then, but almost certainly reversing the order isnt good...

#

if its worse, thats probably something else to look at

plain vine Nov 10, 2025, 5:33 PM

#

order wasn't getting reversed

dense elbow Nov 10, 2025, 5:33 PM

#

it was

#

queue.add(fn1);
queue.add(fn2);
queue.add(fn3);
queue.add(fn4);
queue.add(fn5);

...
...
...

queue.pop() <--- fn5
queue.pop() <--- fn4
queue.pop() <--- fn3
queue.pop() <--- fn2
queue.pop() <--- fn1

plain vine Nov 10, 2025, 5:34 PM

#

pop on list pops from the start

#

not the end

dense elbow Nov 10, 2025, 5:34 PM

#

oh?

plain vine Nov 10, 2025, 5:34 PM

#

so the result here is the same as the list

dense elbow Nov 10, 2025, 5:34 PM

#

really bad naming...?

plain vine Nov 10, 2025, 5:34 PM

#

yeaaaa

#

confused me as well 😂

#

but it works in terms of code language

dense elbow Nov 10, 2025, 5:35 PM

#

ok, then fine, wast that either... my bad... i assumed pop was the end

plain vine Nov 10, 2025, 5:35 PM

#

less verbose

dense elbow Nov 10, 2025, 5:36 PM

#

A linked-list of elements. The list is composed of element container objects that are chained together. It is optimized so that adding or removing an element does not imply copying the whole list content every time.
yeah, it actually looks like a much nicer choice

#

anyway, as to the actual issue, no clue im afraid

#

i would put like one component there and a dummy counter and see how many invalidations you get

#

maybe there is something in the sheduler that is calling that task a bunch of times... hard to tell

plain vine Nov 10, 2025, 5:39 PM

#

#

it would appear here

#

very confusing place to be 😄

#

will try a simple page and see what occurs

#

but the only time i'm noticing the issues is when I load up a page with stuff

#

text measuring is getting called a decent amount of times as well

#

once the page has loaded in, its solid

#

its just that loading in phase

#

i'm not batching the objects, but could that make such an effect here?

#

draws*

dense elbow Nov 10, 2025, 5:46 PM

#

honestly, no clue, it might not even be the calllater

#

if it ends up "fine" then i suspect its not about the render

plain vine Nov 10, 2025, 5:48 PM

#

that's what i thought

plain vine Nov 11, 2025, 5:22 PM

#

got my cppia reloads down to around 0.35secs max

#

:D

plain vine Nov 12, 2025, 11:02 AM

#

now to decide, what to work on next, the chart stuff or rewrite my exchange api

#

exchange api stuff is a bit boring

#

chart stuff is what i want to work on 😄

plain vine Nov 12, 2025, 12:14 PM

#

i should probably see if i can get windows to build

#

at least verify the cross platformness of it all

plain vine Nov 13, 2025, 11:27 AM

#

struggling to get it to compile for windows to match the linux setup, every successful compile i get, the lib file is massive or its missing apis that i need :/

#

worst part is, i have to hang around in windows to keep testing it x)

#

#

oh shit!!!

#

moment of truth 🤞

#

#

everything but text 😄

#

path = '/home/j/Documents/Projects/cpp/TradingProject/assets/fonts/Roboto-Regular.ttf';

#

#

cross platform is working!!!

#

I wish i could verify macos

#

text quality seems to be worse on windows than it does on linux here, maybe need some settings

#

i'm very happy with that, proof of concept works :D

#

the detail like text not being 100% is probably a simple setting issue either in my code or in the windows build process

topaz stump Nov 14, 2025, 12:46 AM

#

yo billy, that trading app looks so good tbh

plain vine Nov 14, 2025, 10:43 AM

#

topaz stump yo billy, that trading app looks so good tbh

Thanks!

plain vine Nov 14, 2025, 11:31 AM

#

#

i think i may have found some kind of hint to the haxeui validation bug

#

this project is clean and just has my lib, as i wanted to start trying to build up the chart stuff and have 0.06 hotreloading is crazy af

#

seems like setwidth is getting called every frame for some odd reason

plain vine Nov 14, 2025, 12:22 PM

#

public var width(default, set):Float = 0;
This calls the setter every frame 🤔

#

switched it to @:isVar public var width(get, set):Float = 0; and that helped with that one

plain vine Nov 16, 2025, 10:02 AM

#

I'm getting eerie temptations to try and see if I can setup android but I may resist that temptation 😅

plain vine Nov 17, 2025, 1:05 PM

#

#

simple bar chart blobcheer

plain vine Nov 21, 2025, 11:09 AM

#

#

the part I hate the most about chart stuff is the axis labels

#

this shit is annoying as hell

plain vine Nov 24, 2025, 5:29 PM

#

early stages, but, it appears to me that my last attempt at charts was grossly overcomplicated 😅

#

it could also just be the case that this is the second or third time i've done this so its just easier

#

its also humourous that I asked claude: "can i get a 100 candle demo with natural movement" and it generated the above which is absolutely not natural market movement

#

but in a funny way, it could be perceived as natural from a computer pespective

plain vine Dec 1, 2025, 4:56 PM

#

its coming along 😈

nocturne adder Dec 1, 2025, 5:07 PM

#

That looks sick!

plain vine Dec 1, 2025, 5:07 PM

#

thanks 😄

#

its still crazy to me how this is running on my own little framework

nocturne adder Dec 1, 2025, 5:08 PM

#

Also man you have balls of steel, I am pretty confident in my skill but I'd practically shit rocks if I had to build an application for stock trading. I just imagine some kind of error causing data to be shifted by like a tinyyy bit causing bad purchases / sales.

nocturne adder Dec 1, 2025, 5:08 PM

#

plain vine its still crazy to me how this is running on my own little framework

Oh this is skia? Sick!

plain vine Dec 1, 2025, 5:09 PM

#

cppia in action 😄

plain vine Dec 1, 2025, 5:09 PM

#

nocturne adder Also man you have balls of steel, I am pretty confident in my skill but I'd prac...

well, the primary use case for me is going to be journaling, but I do want a terminal and I did have to spend quite a bit of time getting the maths right over there

nocturne adder Dec 1, 2025, 5:10 PM

#

I can only imagine xd

plain vine Dec 1, 2025, 5:10 PM

#

the maths isn't necessarily hard, but, my maths skill is low and finding fees for things is scarce

#

from the ceramic version i had to make some assumptions to keep things flowing, so i actually appreciate the little reminder on that lol

nocturne adder Dec 1, 2025, 5:11 PM

#

i SEE

#

oops caps

#

*I see

#

It looks really good though

plain vine Dec 1, 2025, 5:11 PM

#

the assumptions aren't massive, its more just assuming what the quote currency is going to be

#

but that's not flexible if you want to trade whatevers available

nocturne adder Dec 1, 2025, 5:12 PM

#

I mean do you have any issues with pricing

#

I've not gone in depth on stock trading but I know that the APIs to provide the data you might need are usually stupid expensive - people want to make trading bots and companies see a way to gain a lot of profit there...

plain vine Dec 1, 2025, 5:13 PM

#

ceramic version worked well enough for an mvp, but, the order pad was a bit odd with handling balance for posting orders. However, examination of trade completion, that was all accurate in comparison to the exchange I was using

#

and from a journalling and log perspective if the records are accurate that's not a terrible place to start from xD

plain vine Dec 1, 2025, 5:14 PM

#

nocturne adder I've not gone in depth on stock trading but I know that the APIs to provide the ...

crypto is still baby land so it's the time to jump on board

#

the window might be closing though, governments are starting to want to dig their claws in with regulatory measures 😄

#

every year it seems like there's some new policy that exchanges need to comply with that limits access to some degree

#

laws in a way hinder newer people, like take all the data regulatary things going on now

#

the people who can take the biggest advantage of this, already has the data, the law will prevent easy access for newer competitors, but i can almost guarantee you, that google has a way to ID you without you even needing to be emitting any data

nocturne adder Dec 1, 2025, 5:19 PM

#

yeaaaaa....

#

google knows more about me than I know about myself

plain vine Dec 1, 2025, 5:19 PM

#

that's a bad place to be 😄

#

pay attention to yourself

nocturne adder Dec 1, 2025, 5:19 PM

#

I do have a good VPN though XD

#

mullvad my beloved

nocturne adder Dec 1, 2025, 5:19 PM

#

plain vine pay attention to yourself

Oh I do

#

but google probably does so more XDDD

plain vine Dec 1, 2025, 5:20 PM

#

lol

#

been listening to some business stuff and its very interesting to hear how marketing has been evolving

#

over there, obviously so, advertising is brilliant

#

if you have a niche, you can hammer into it seemingly quite simply

#

places like IG/Youtube/TikTok aren't considered as regular social media platforms but "interest media"

#

and the term just makes absolute sense

agile snow Dec 1, 2025, 6:18 PM

#

plain vine places like IG/Youtube/TikTok aren't considered as regular social media platform...

discord is kinda the last bastion of social media that isn't consumption based

nocturne adder Dec 1, 2025, 7:07 PM

#

plain vine places like IG/Youtube/TikTok aren't considered as regular social media platform...

that term makes alot of sense

plain vine Dec 1, 2025, 7:10 PM

#

agile snow discord is kinda the last bastion of social media that isn't consumption based

I think of discord like old forums

plain vine Dec 1, 2025, 7:12 PM

#

nocturne adder that term makes alot of sense

it does, doesn't it! And if you have a thing you want to get out there, it really is a solid vehicle. Listening to the way its used (when the objective isn't to extort) I really can't see much problems with the model

#

because in practice there's no major difference between Google telling me about a product and a friend

nocturne adder Dec 1, 2025, 7:14 PM

#

That is true

#

I guess the main difference is intent

broken python Dec 1, 2025, 7:16 PM

#

I just wished Discord wasn’t a closed source platform that you can’t host yourself, or that at least the server data would be accessible more easily

plain vine Dec 1, 2025, 8:03 PM

#

broken python I just wished Discord wasn’t a closed source platform that you can’t host yourse...

What server data?

broken python Dec 1, 2025, 8:07 PM

#

plain vine What server data?

Just the messages, just like a forum is crawlable by anything that can fetch webpages

topaz stump Dec 1, 2025, 8:27 PM

#

plain vine

omg that looks so gooood

plain vine Dec 1, 2025, 8:54 PM

#

broken python Just the messages, just like a forum is crawlable by anything that can fetch web...

I'll probably do something like that one day if I ever get hint of something I don't like by discord

#

but no motivation for such a thing otherwise

plain vine Dec 5, 2025, 10:21 PM

#

Screenshot_2025-12-05-22-21-34-732_com.google.android.gm.jpg

#

I'm not sure I really did 24k requests 😂

plain vine Dec 9, 2025, 12:04 PM

#

ohhhh i remember the last time I got to this stage :D

#

managing the axis is a lot easier for components like this

#

still got some refinement to do

broken python Dec 9, 2025, 3:50 PM

#

Looks clean!

plain vine Dec 9, 2025, 3:51 PM

#

much thanks! 90% of the cleanness is hidden away

#

but i'm very happy to have gotten back to this point

plain vine Dec 10, 2025, 11:30 PM

#

    public var onDown:(x:Float, y:Float, button:MouseButton) -> Void;

    function set_onDown(value) {
        trace(visual == null);
        if (value != null) {
            visual.on(MouseDown, onPointerDown);
        } else {
            visual.off(MouseDown, onPointerDown);
        }
        return onDown = value;
    }

"why isn't onDown getting called!?"

#

Took me too long to figure this one out 😅

topaz stump Dec 11, 2025, 1:07 AM

#

uuh

#

why?

#

are you using onpointerdown as the event callback instead of ondown?

plain vine Dec 11, 2025, 1:43 AM

#

topaz stump are you using onpointerdown as the event callback instead of ondown?

didn't set the property definition 🤣

topaz stump Dec 11, 2025, 2:05 AM

#

OOOOOOOOOOOOH

#

im so dumb

plain vine Dec 12, 2025, 11:47 PM

#

just used grok for an evaluation on a multivitamin

#

and got one of those "which response do you prefer" things

#

bruh, option 1 said it was great and option 2 said it was pointless

#

da fuck am i meant to do with that

plain vine Dec 13, 2025, 3:15 PM

#

fml its been like 2 or 3 months since i did the event system

#

I completely forgot about implementing propagation control

#

i've had like 3 use cases for this, first 2 i added some booleans to remotely control

#

on the third one, i was like

right.. gotta sort this out..
wait a second?
didn't i set up some mechanism for this?
why isn't it working....
...
I wasn't fucking using it

#

it works so damn well

#

i can implement events that overlap each other in terms of input interaction but the child can block all parents from ever receiving it

#

fleshing out the setup for component behaviours

plain vine Dec 13, 2025, 4:00 PM

#

and with very little effort, we have a new horizontal line component with all the same behaviours :D

plain vine Dec 16, 2025, 12:28 PM

#

what's interesting is, i find the component behaviour setup messier than my last method, but waaaaaaay nicer to build with

#

hopefully I get to figure out some way to make it nicer the more I use it x)

plain vine Dec 16, 2025, 2:23 PM

#

#

I forgot to handle newlines in the framework 🥲

plain vine Dec 19, 2025, 3:00 PM

#

i swear AI gets progressively worse every time it updates

#

questions that used to be fine now just do silly things

#

maybe its just degredation on free tiers

broken python Dec 19, 2025, 5:00 PM

#

Yeah I don't know, I've been using only Opus 4.5 lately, and I get pretty good output from it most of the time

plain vine Dec 20, 2025, 5:32 PM

#

came across a situation where event system tripped up, had a visual that had 2 events on it. But I wanted the 2nd event to fire and prevent the first from activating

#

so added a priority system and a simple stopImmediatePropagation() setup

#

works nicely :D

plain vine Jan 30, 2026, 11:30 PM

#

I'm a bit stuck x)

#

For whatever reason I can't seem to get haxeui to work with my stack anymore

#

It either won't compile or I'm getting into infinite loops at runtime

#

I'm contemplating making my own ui setup :/

#

leveraging daz's already cppia compatible yogalayout externs

#

I don't wanna make my own ui setup tho 😭

broken python Jan 30, 2026, 11:35 PM

#

the plot thickens

plain vine Jan 30, 2026, 11:48 PM

#

i'll have to sit on it for a bit, but, i basically can't work on my frontend currently

#

it's a bit of a big decision, will see if i can work on other stuff

broken python Jan 31, 2026, 6:00 AM

#

I think regardless of what you decide to do, you should investigate those infinite loops, if you don’t want those to bite you again later

plain vine Jan 31, 2026, 10:53 AM

#

I've defo tried 😄

carmine path Jan 31, 2026, 10:54 AM

#

oh boy

#

Yeah I think the compiler folks lost the plot a bit when it comes to hxcpp / cppia with hx5

plain vine Jan 31, 2026, 10:56 AM

#

That's what I'm suspecting a little bit, it might be a haxe 5 thing in the mix as well

carmine path Jan 31, 2026, 11:06 AM

#

yep, I think the cppia codegen might have some issue

plain vine Jan 31, 2026, 11:26 AM

#

The infinite loop situation I can repro on hxcpp

#

but the compilation issues seem to be macro issues that were defo not there before

broken python Jan 31, 2026, 3:37 PM

#

plain vine The infinite loop situation I can repro on hxcpp

When you say infinite loop, it's infinite loop when running or when compiling?

plain vine Jan 31, 2026, 3:41 PM

#

when running, it gets into a loop of registering events

broken python Jan 31, 2026, 3:42 PM

#

I see, then that's easier to debug than if that was on the compiler

#

You should debug your binary with visual studio

#

and pause it when running

plain vine Jan 31, 2026, 3:42 PM

#

I don't think this is a problem I can solve without Ian's context but I also don't understand where the source of the problem lies

broken python Jan 31, 2026, 3:42 PM

#

that will tell you where it loops

#

Ah maybe you already identified the loop

plain vine Jan 31, 2026, 3:43 PM

#

yeah

#

haxeui's very big with a lot of behaviour hidden behind macros so its hard for me to dive deep into it

#

possible for sure

broken python Jan 31, 2026, 3:54 PM

#

That's the main risk when you use a large library you didn't make yourself for sure

plain vine Jan 31, 2026, 3:54 PM

#

and tbh, i'm not sure how much of these issues are even issues with haxeui

#

a fair few, if not all of them might be constraints to do with haxe 5, cppia and ⁨⁨⁨--macro include('haxe.ui')⁩⁩⁩stuff

#

I don't really wanna stop using haxeui tho

#

so i'm a bit conflicted here

#

it's like ceramic vs skia all over again 😭

agile snow Jan 31, 2026, 4:10 PM

#

I’m confused on how it just magically broke. Did you update anything?

carmine path Jan 31, 2026, 4:11 PM

#

it's usually dependency hell pulling some random ass module you shouldnt pull

#

some rogue import you shouldnt have somewhere

#

I remember I noticed the imports behaved a bit differently with hx5. maybe this coupled to macros is screwing you here

plain vine Jan 31, 2026, 4:45 PM

#

agile snow I’m confused on how it just magically broke. Did you update anything?

haxe, hxcpp, haxeui

#

but reverting to my last working version in commit isn't working because for some reason it can't find the commit

#

which is fine, because I have a fork https://github.com/jarrio/haxeui-core with my previously working cppia/haxe 5 stuff but it no longer works as well

#

so i have no idea what broke

#

I know something i did broke it, it wasn't magical, but that isn't useful

plain vine Jan 31, 2026, 4:52 PM

#

carmine path some rogue import you shouldnt have somewhere

could that be it? I've tried a pretty barebones project and have similar issues

#

but yeah, its hard to debug when you don't know the internals of the library very well or macros

carmine path Jan 31, 2026, 4:52 PM

#

most likely haxeui core macro finds shit it didnt before

plain vine Jan 31, 2026, 4:54 PM

#

problematic eitherway

plain vine Feb 16, 2026, 9:04 PM

#

Playing around with making my own ORM

#

connection.find(table).where(age > 10 && age < 21).one().then(function(result) {
  trace('id=${result.result.id} name=${result.result.name} age=${result.result.age}');
});

#

surprisingly little amount of macros & code to get this to work fine

#

age is a field on table

#

no type checking or anything like that, got a simple field check to catch spelling errors though

#

var foo = 30;
connection.find(table).where(age > 10 && age < foo).one()```
works but is a bit odd looking though

#

connection.find(table).where(age == 22 || age == 90).one();
//src/Main.hx:30: id=3 name=l age=90
``` this is cool

agile snow Feb 16, 2026, 9:11 PM

#

what does the object definition part look like?

plain vine Feb 16, 2026, 9:17 PM

#

agile snow what does the object definition part look like?

what part?

#

the table definition?

agile snow Feb 16, 2026, 9:18 PM

#

possibly

plain vine Feb 16, 2026, 9:19 PM

#

agile snow possibly

table definition is simple atm

import db.types.ITable;

class Users implements ITable {
    public var id:Int;
    public var name:String;
    public var age:Int;

    public function new() {}
}

#

interface has no macros on it currently

agile snow Feb 16, 2026, 9:19 PM

#

ah cool

plain vine Feb 16, 2026, 9:20 PM

#

There's only one expression macro atm lol

#

and i'm not sure what else i'd need a macro for 😂

#

besides adding some type checking and some table creation stuff

plain vine Feb 16, 2026, 9:44 PM

#

#

that's not bad right

plain vine Feb 16, 2026, 9:59 PM

#

#

ohoh, we did better

plain vine Feb 16, 2026, 10:43 PM

#

    public function new(table:T, connection:IConnection, builder:IQueryBuilder) {
        this.connection = connection;
        this.builder = builder;
        switch (Type.typeof(table)) {
            case TClass(c):
                trace(Type.getClassName(c));
            default:
        }
    }

oooooo this works?!

#

it traces src/db/TypedQuery.hx:32: Users

#

I thought that wasn't a thing

#

    function objToClass(dbResult:Dynamic):T {
        if (dbResult == null) {
            return null;
        }
        var inst = Type.createInstance(tblClass, null);
        for (f in Reflect.fields(dbResult)) {
            Reflect.setField(inst, f, Reflect.field(dbResult, f));
        }
        return inst;
    }

I'm guessing this kind of magic is unavoidable with database related things - right?

#

#

there's no way that's a valid key

#

Reflect.field(data.ogResult[0], 'COUNT(*)') is 3....

#

madness

plain vine Feb 17, 2026, 11:54 AM

#

con.find(table).update(name = "Billy").where(id == 1).execute().then(function(resp) {
  trace(resp);
});

#

What do we think about this kind of query?

plain vine Feb 17, 2026, 4:33 PM

#

con.find(table).orderBy(table.age, DESC).get().then(function(resp) {
    for (item in resp.result) {
        trace('id=${item.id} name=${item.name} age=${item.age}');
    }
});

id=3 name=l age=90
id=2 name=o age=35
id=1 name=Sam age=10
😄

#

This is quite fun to make

#

macros are pretty fun to use

plain vine Feb 17, 2026, 4:55 PM

#

#

I'm enjoying making errors for the macros

plain vine Feb 17, 2026, 10:06 PM

#

var obj = new Users();
obj.age = 28;
obj.name = "Logo";

con.find(table).insert(obj).execute().then(function(resp) {
    trace(resp);
    trace(resp.success);
});

#

Took longer than expected to figure out the insert setup but there we go

#

with that i can now read, write, create and delete

agile snow Feb 17, 2026, 10:07 PM

#

Me and Bob o are best friends

plain vine Feb 17, 2026, 10:08 PM

#

The mighty three

agile snow Feb 17, 2026, 10:09 PM

#

What is the purpose of execute? Could you chain insert and update for instance?

plain vine Feb 17, 2026, 10:12 PM

#

execute "runs" the query otherwise it continues to chain

#

currently you can chain insert/update

#

and that is defo undesirable, might have to fix that with a macro but not sure how to address that just yet

#

might just end up doing another class separation

#

currently its:
Database -> TypedQuery<T:ITable>
probably can do

`Database` -> `Action` -> TypedQuery<T:ITable>`
                 - insert
                 - delete
                 - update...

#

But i was more interested in getting meat of it done, i'm not gonna do that mistake myself

#

What's next to do for orms 🤔

#

maybe table creation 🤔

agile snow Feb 17, 2026, 10:32 PM

#

One thing none of the Haxe ORMs have is migrations

#

Obviously get the basics first, but it would be nice to finally get that

plain vine Feb 17, 2026, 10:51 PM

#

agile snow One thing none of the Haxe ORMs have is migrations

I think it would be doable because I have got it designed so that it can have additional backends, it would probably be a case of just mapping the db type back to haxe types

#

nothing fancy with my backend setup either, just an interface with build functions

plain vine Feb 18, 2026, 9:20 AM

#

#

Macros really are pretty fun 😅

#

it's the rapid feedback you get from them

plain vine Feb 18, 2026, 11:17 AM

#

agile snow One thing none of the Haxe ORMs have is migrations

when you say migrations could you expand on what you're talking about? Are you referring to things like altered tables?

plain vine Feb 18, 2026, 8:54 PM

#

#

we created a table from a haxe class definition blobcheer

plain vine Feb 19, 2026, 11:28 AM

#

Table alterations!!

#

I'm not sure how to go about handling deletions yet

#

I could just assume and boop it away, but I don't like that... its so easy to change these things in code and in turn very easy to just vanish data

#

maybe its okay, but i'll think on that one

#

maybe an explicit @remove meta on the field

plain vine Feb 19, 2026, 1:56 PM

#

got nullable fields marked with wrapping items with Null 😄

plain vine Feb 19, 2026, 4:20 PM

#

class Users implements ITable {
    @size(100) public var name:String;
    public var age:Null<Int>;
    public var password:String;
    @delete public var email:String;
}

opt in field deletion works

plain vine Feb 19, 2026, 5:33 PM

#

@rename("password") public var pass:String;
renaming done, a bit awkward but I'm sure the explicitness of it will have its upsides, flow here is:

apply the meta -> rerun -> rename variable to the value in the meta

plain vine Feb 20, 2026, 4:32 PM

#

#

ohoh step 1 for relationships 👀

plain vine Feb 20, 2026, 5:18 PM

#

forgot to build insert queries 😅

#

orr i have bad memory

plain vine Feb 21, 2026, 9:33 AM

#

#

we got the obj -> int <- obj setup done

#

I guess all I need to do is figure out the relationship query and then I should have a poc of a one to one relationship 👀

#

also, i think the syntax to using a query might not be ideal, i'll have to keep using it but i don't like something about it

#

it might make more sense to do something like

var user = new Users();
user.insert();
user.find();

#

I wonder if i coulld abstract away the promises in some way

agile snow Feb 21, 2026, 11:48 AM

#

plain vine it might make more sense to do something like ```hx var user = new Users(); user...

I feel like you would be basically reinventing entities2 at this point

plain vine Feb 21, 2026, 11:49 AM

#

agile snow I feel like you would be basically reinventing entities2 at this point

I am a bit, but e2 doesn't work on my haxe version and I am gonna making improvements where I can

#

I have an idea for the promise stuff but I'm trying to tackle the relationship stuff atm. Having some complications in figuring out how to evaluate the query

#

but if you have alternative ideas feel free to say, i'm not quite sure exactly what I want just yet so i just wanna play with things

plain vine Feb 21, 2026, 5:05 PM

#

#

I got the relationship query built but this is a new problem

#

mysql dumps the result into a flattened obj

#

I could match fields on the corresponding classes but then you wouldn't be able to have identical field names across objects which is a bit lame...

plain vine Feb 24, 2026, 1:59 PM

#

#

👀

#

have I finally got it mapped

plain vine Mar 1, 2026, 12:59 PM

#

My project is alive 🥹

plain vine Mar 2, 2026, 4:10 PM

#

// Select
db.from(User).where(age >= 20).get();
db.from(User).withAll().then(...);   // joins all relations
db.from(User).orderBy(user.age, ASC).get();
db.from(User).count().then(...);

// Insert
db.from(User).insert(user).one();   // insert from instance
db.from(User).insert(name = "J", age = 23).execute();

// Update
db.from(User).update(name = "Bob").where(usersId == 1).execute();

// Delete
db.from(User).where(usersId == 3).delete();

#

mvp of orm all done, gonna use it and see how it feels

plain vine Mar 2, 2026, 7:43 PM

#

how reasonable is it, to allow multiple database "types" at runtime? So being able to say, use both mysql and mongodb at runtime

#

Is that something people do or has any particular purpose?

#

conceptually I could see nosql db's covering sql and vice versa, just not sure whether I should actually consider it down the road

agile snow Mar 2, 2026, 7:53 PM

#

plain vine how reasonable is it, to allow multiple database "types" at runtime? So being ab...

ecto (from elixir) lets you do that, it makes a lot of sense

#

it is pretty common to use multiple databases for cache, search, content, etc

#

you have a separate "repo" for each db type

#

a repo is basically the equivalent to your db

#

the repos are controlled by an adapter which would be the backend

#

you can also configure pooling and stuff there

plain vine Mar 2, 2026, 8:11 PM

#

agile snow it is pretty common to use multiple databases for cache, search, content, etc

oh, interesting, i'll not go down a path that would make that kind of setup harder then

plain vine Mar 9, 2026, 8:24 AM

#

I added a message counter to haxebot around a month ago, for multiple reasons, but the initial prompt was scammer related, I kind of wanted to build a list regular people who use the server

#

its really quite cool, In the last month I now know that 214 people have communicated on server

#

server has lots going on these days so its hard to just pick that up mentally

plain vine Mar 9, 2026, 12:40 PM

#

test

plain vine Mar 11, 2026, 4:09 PM

#

should text input cursor always be black?

#

seems wrong here

#

not sure how to handle that in haxeui though

#

text inputs properly working, haven't done double click highlights but I got a very solid foundation here

#

even got multiline working

#

no cut, copy, paste but I think for now they're more luxury until I actually find a moment where I want it

plain vine Mar 11, 2026, 9:20 PM

#

we got cut/copy/paste x)

#

text input feels good, gonna start bringing things together again

plain vine Mar 12, 2026, 11:16 AM

#

forking so many things to get haxe 5 things to work

plain vine Mar 12, 2026, 4:35 PM

#

#

as expected things aren't working perfectly, but, chart rework is presentable, exchange api is working and we can actually see things

#

its been so long since i've been able to see the app running, its very motivating and it also makes it seem like there's way less I need to do

plain vine Mar 19, 2026, 1:26 PM

#

#

who can see the crappy bug

carmine path Mar 19, 2026, 4:10 PM

#

centering stuff is hard

#

Not sure if you have seen it, but there is https://www.pencil.dev. Quite crazy if you need to cook designs/layouts and get a format that can be easily adapted into others

#

I just fed it 35 user-stories and it created all the necessary screens for that UX

plain vine Mar 19, 2026, 5:49 PM

#

carmine path Not sure if you have seen it, but there is https://www.pencil.dev. Quite crazy i...

oh wow, that looks pretty cool

#

we got some app flow going with the backend 😄

plain vine Apr 6, 2026, 12:25 PM

#

#

lol, how to make this look good

#

if it expands to the screen size it looks too big

plain vine Apr 11, 2026, 9:30 AM

#

I think i've managed to finally iron out all of my cppia hotreloading loose ends

#

can get a bit messy when working with threads and sockets

#

But I had a session yesterday with around 10+ recompiles and i didn't have any weird issues in app

#🌊 NB's Riverside View