arma3_feedback_tracker

[dreamy bane]

would have to use "" strings for code though

[sinful kettle]

compile

[untold sky]

yeah.
Or just have "funcname" and the script could detect it and resolve that through to the real function

[dreamy bane]

no syntax highlight 😄

[untold sky]

Well if you want to deal with config, you'll have to deal with config no matter what

[sinful kettle]

btw about the destruction, does the game destroy the missionNamespace first then objects, etc.?

or if it wants to destroy an object, ctrl, display, etc., does it destroy its variables list first?

[untold sky]

I don't think that is.. properly defined

We added a.. destructor eventhandler for displays? or controls? smth like that?
recently and it made stuff crash quite a bit because of the ordering
So in general, the order of destruction will be
Also because of ref counting. A hashmap might be referenced in missionnamespace, there might also be a reference in a spawned script or on an object or on a control.
Even if you know that namespace is destructed before objects. You don't know if there is an object that also holds a reference

Simple solution, don't call destructor at all on mission end.
Only call it when
_value = nil makes it be deleted
But again, you'll have that references mess, unlike object/location/config there is no way to have a weak reference (without introducing a whole new datatype)

if you take care where you store references to the map, and how you null it out, it'll be fine.
But if you, or any mod/script grabbing your map doesn't... welp

Now you know why I said destructor will be hard 😄

[sinful kettle]

_value call ["destructor", []] and remove anything you need in the destructor

[untold sky]

Well.. tru you could "null" it out, by just doing a "delete" by calling destructor manually :x
But that kinda defeats the purpose of having the cool "ah this went out of scope" thing

I will try to implement destructor like you imagined, we will see where it dies.
Worst case I disable the destruct on mission end thing, and you'll have to live with everything collapsing if other scripts grab and store a reference to your object. Already added stuff to do the same with that control/display whateveritwas

destructor will be unscheduled
Well it'll be documented so gonna be fine

[gray wharf]

will it now

[untold sky]

_class = [  
  ["Method", {diag_log "Oh I'm a method!"; _thisObj set ["identifier", _this];}],
  ["destructor", {diag_log ["goodbye", _thisObj get "identifier"]}]
]; 

call {
  _myObject = createHashMapFromArray _class;
  _myObject call ["Method", "hello"];
}

14:47:13 "Oh I'm a method!"
14:47:22 ["goodbye","hello"]

This was already chaos XD
When a refcount goes to zero it gets destructed.
Inside destructor, I call the destructor script, which gets the object as argument, meaning the refcount increases.
When the destructor call is done, the refcount decreases again, tries to delete again... 😄

Also, after the destructor call, the map will be gone.
If you save it into a variable, from inside the destructor call, and that variable survives outside the destructor.. omg

I'll probably have to workaround that, end result being that destructor can run multiple times, if you let the object value escape

[sinful kettle]

If you save it into a variable, from inside the destructor call
force it to no copy before calling destructor?

[untold sky]

_x = _y
is not a copy

Like... 🖕 if anyone does this.. but.. it works.. ish..

_class = [  
  ["Method", {diag_log "Oh I'm a method!"; _thisObj set ["identifier", _this];}],
  ["destructor", {diag_log ["goodbye", _thisObj get "identifier"]; _dangerEscape = _thisObj;}]
]; 

_dangerEscape = 0;
call {
  _myObject = createHashMapFromArray _class;
  _myObject call ["Method", "hello"];
}

14:57:10 "Oh I'm a method!"
14:57:10 ["goodbye","hello"]
14:57:10 ["goodbye","hello"]

Also thinking about it. Spawning a new copy from destructor should be valid right? 🤔
I don't think that does any harm (besides on mission end. But I now disabled destructors on mission end)

I'm trying to think of a way to get away from this without having to add a new command to create object 🤣
But probably bad idea anyway, also its easier to add hashmap commands :3

[sinful kettle]

you still need one to call constructor :\

[untold sky]

yes that later

Giv description for call command
"Calls hashmap value with CODE type."
sounds sad
my brain can't

[sinful kettle]

Calls the code value that corresponds to the provided hashmap key. The magic variable _thisObj can be used to access the hashmap inside the code
hashmap call [key, args]

idk I can't think of anything good either

[sinful kettle]

btw it will work with any key type right? not just string? _hashmap call [0, []]?

[untold sky]

Well its just for the ingame. So the first sentence looks good.

[untold sky]

yes.. if you really wish to do that xD

oh nvm it doesn't, I did a string type check.
Fixed now though

[uncut briar]

Wouldn't it be more in line with what we already have to do _args call [_hashmap, _key]?

[sinful kettle]

I suggested the other one because it's like set and get

and thus looks more consistent

[untold sky]

true.. but.. we're building a new hot thing but..
"object call X with args"
makes more sense to me than
"args call object's x"

it doesn't fit to old call usage, but it fits to general OOP'ness and hashmap usage

[uncut briar]

Yeah maybe. Somehow I don't like the whole idea of this.

[untold sky]

_class = [  
  ["Method", {diag_log "Oh I'm a method!"; _thisObj set ["identifier", _this];}],
  ["constructor", {diag_log ["Hello!", _this];}],
  ["destructor", {diag_log ["goodbye", _thisObj get "identifier"]; _dangerEscape = _thisObj;}]
]; 

_dangerEscape = 0;
call {
  _myObject = createHashMapObject [_class, "args"]; 
  _myObject call ["Method", "hello"];
}

16:33:45 ["Hello!","args"]
16:33:45 "Oh I'm a method!"
16:33:45 ["goodbye","hello"]
16:33:45 ["goodbye","hello"]

I'm done for today so... Discuss!

[sinful kettle]

is constructor scheduled?

[uncut briar]

I guess inherits current environment.

[sinful kettle]

imo constr and destr should be unschd

[sinful kettle]

btw _thisObj is private right?! 😅

[untold sky]

same as _this

constructor is currently unscheduled. But I can easily make it scheduled
call supports scheduled
destructor must be unscheduled

[sinful kettle]

no unschd constructor is fine imo

[untold sky]

as long as the c.....
the create function must return a value, so I cannot make it scheduled (unless I do way more work that I don't want to do)
all scheduled commands return nothing.. though I maybe can leave something on the stack and make the call discard the return value.. huff

[sinful kettle]

if anyone wants schd they can call a method afterwards

I cannot make it scheduled
by scheduled, I meant same as current environment, like call

[untold sky]

no I mean I cannot

unless the unless i wrote

because commands that support scheduled (like then, call, while) all return nil and cannot return a value. And need extra stuff that pushes result on stack and bla bla bla

I have to implement alot more code to support that. I will probably do that for constructor, we'll see

[sinful kettle]

nah just leave the constructor unschd

[sinful kettle]

^

[untold sky]

that makes it ugly. And that means the caller has to consider a "internal" implementation detail inside the class

though tbf if it requires scheduled, calling from unscheduled would also mean you need to consider that detail.. so eh
I'm fine with unscheduled too, but more support would always be nicer

[sinful kettle]

createHashMapObject [_class, "args", _isScheduled];

[untold sky]

if i support scheduled.
You can have the class be done initializing after the creation (which may suspend). And you can handle call from unscheduled by checking for it.
If i don't support it, you'd need to have everyone basically manually call the real constructor after creating the object.

Mh you could cheat it...

_this pushBack [] spawn {...};

WaitUntilDone (_args select 3)

but again, implementation detail leaking into usage is crusty

My brain is confident that scheduled support is better

[dreamy bane]

what if you have two scheduled scripts setting same variables at the same time, is that possible to happen?

[untold sky]

No. Scheduled scripts don't run in parallel, there is always ever only one script executing

[uncut briar]

If i don't support it, you'd need to have everyone basically manually call the real constructor after creating the object.
I do not understand this part. Why not supporting scheduled means that.

[untold sky]

You expect that construction is finished after object is created.

Constructor //1
CreateObject //2
//3

You expect 2/1/3 ordering.

But if you need scheduled...
Constructor: [] spawn {//1}
createObject //2
//3

You will get 2/3/1

At 3 you will use the object and expect it to be initialized

If you need scheduled you still need to check for canSuspend in case you're created in unscheduled.
But at least you can use scheduled at all, and let's be frank, most code runs scheduled, which is generally a good thing

[uncut briar]

But if you support only unscheduled the construction will be finished once object is created, right? Did I understood it backwards?

[untold sky]

Yes, but you loose the option of being able to do scheduled wait things

Like.. hey i know it's probably stoop but.
What if you wanted to implement a mutex in scheduled, it has to wait in constructor for slot to be free

[uncut briar]

_constructor = {
    _thisObj set ["constructScript", _this spawn {}];
};

waitUntil {scriptDone (_obj get "constructScript")};

sleeping in like an initialization of an object seems weird.

[untold sky]

It doesn't do any harm to support it besides annoying my brain that i have to write code.
And it makes it more versatile so no contra arguments allowed

[uncut briar]

But then somebody does sleep 1 in object constructor.

I run this in unsched.

What happens?

[untold sky]

Yeah again, then you need to respect implementation detail on every usage.
Might as well not use an object if you manually call a func anyway

[untold sky]

Same as a call in unsched, that's not new

[uncut briar]

Error then, so I need to know about the object implementation details.

¯_(ツ)_/¯

[untold sky]

There is a difference between "you can only use this object in scheduled" and "you need to call this thing every time you create this object because we couldn't put it in the constructor oh and btw you can only use this object in scheduled"

[uncut briar]

Both sucks.

[untold sky]

But one sucks less

[uncut briar]

TBH I'm bigger fan of your first idea about the "EH" for variable destruction. Feeling like it would be enough, if somebody needs OOP he will create one with this.

This current idea... I feel like it will only result in even more spaghetti code being around.

This whole OOP feels weird in SQF with scheduled/unscheduled and all its quirks.

[untold sky]

🫂

I mean..
If we already have a flag for NoCopy, i can also add a flag to force the constructor to run unscheduled

If you want to be unscheduled and don't want to pay for a canSuspend check

[uncut briar]

isNil {} is cheap ;D

[untold sky]

But it's cheaper to not do it at all :3
And it could also handle all methods too

Maybe we also need a disable serialization flag or so :3
I have 62 bits to spare so.. whatever you can come up with that might make sense

[tropic jewel]

canDeployWeapon exists. by chance is the deployment position already calculated, and hopefully easily made accessible?

[west onyx]

not a bad request @tropic jewel , even if its just to know how its calculated so we can run our own "lineIntersectsSurfaces"

[west onyx]

Is possible to have an alt syntax of "setvehiclecargo" which doesnt apply the after-effect scripting? my use is to unload a "vehicle cargo" child object directly from the parent object. if we just use "detach" then there are mass issues on the parent vehicle (setvehiclecargo applies mass)

so for instance to unload, we need to go objnull setvehiclecargo <child> first

but doing that applies some scripting, like setting the position and if above terrain level, adding a parachute

i want a syntax for setvehiclecargo to detach the object, do its mass calc, and then nothing else (no pos setting and no parachute)

also an "attach/detach" event handler would be great

[desert trench]

spam: https://feedback.bistudio.com/T171041

[untold sky]

Yes! FT ticket please
with script command syntax, bonus points if you can fit it without creating a new command 😄

[gray wharf]

inb4 setDammage overload

[alpine tulip]

u seem to love that jok

[gray wharf]

it's still a strong one inside BI

[alpine tulip]

Can imagine

[untold sky]

Scheduled scripts are fun

_a = [];
_b = _a;
_a pushBack 1;
_a // [1]
_b // [1]
```right?
Well. Unless
```sqf
_a = [];
_b = _a;
// Save game, and reload
_a pushBack 1;
_a // [1]
_b // []

[gray wharf]

oh no

[untold sky]

_class = [  
  ["Method", {diag_log "Oh I'm a method!"; _thisObj set ["identifier", _this];}],
  ["constructor", {diag_log ["Hello!", _this]; sleep 2; diag_log "done constructing";}],
  ["destructor", {diag_log ["goodbye", _thisObj get "identifier"];}]
]; 

call {
  _myObject = createHashMapObject [_class, "args"]; 
  _myObject call ["Method", "hello"];
}
``` ```11:55:09 ["Hello!","args"]
11:55:11 "done constructing"
11:55:11 "Oh I'm a method!"
11:55:11 ["goodbye","hello"]

🫂

[untold sky]

About the nocopy thing, I'd say +_object throws an error (that doesn't kill the script) and returns a reference instead 🤔

[sinful kettle]

I'd say kill the script 😠
anyone who wants to copy something probably wants to do something bad

[untold sky]

ClassDef = [  
  ["Method", {diag_log ["Method", canSuspend];}],
  ["constructor", {diag_log ["Hello!", canSuspend];}],
  ["destructor", {diag_log ["goodbye", canSuspend];}]
]; 

[] spawn {
  _myObject = createHashMapObject [ClassDef, "args"]; 
  _myObject call ["Method", "hello"];
}
````12:44:15 ["Hello!",true]`
`12:44:15 ["Method",true]`
`12:44:15 ["goodbye",false]`

```sqf
ClassDef = [  
  ["Method", {diag_log ["Method", canSuspend];}],
  ["constructor", {diag_log ["Hello!", canSuspend];}],
  ["destructor", {diag_log ["goodbye", canSuspend];}],
  ["flags", ["unscheduled"]]
]; 

[] spawn {
  _myObject = createHashMapObject [ClassDef, "args"]; 
  _myObject call ["Method", "hello"];
}
````12:45:06 ["Hello!",false]`
`12:45:06 ["Method",false]`
`12:45:06 ["goodbye",false]`

[sinful kettle]

why did you make everything unschd?!

[untold sky]

because people may want that

For example ACE/TFAR have quite a few places where they need to check if call is scheduled, and then transform it into unscheduled.
So if you know you want unscheduled, you can get it with simple flag.
If you only want specific methods to be unscheduled, you can use isNil in there, but if you need alot you can solve it with flag and don't need to add code everywhere

[sinful kettle]

ACE/TFAR have quite a few places where they need to check if call is scheduled, and then transform it into unscheduled.
putting a simple isNil is faster than checking

[untold sky]

yes, but you can also not do it at all if you need it everywhere

So there, flag

Don't have to use it if you don't want, we have 60 flags remaining so doesn't hurt

[near cedar]

Interesting, i wouldn't have expected _b to also update when _a is updated (first case)

[untold sky]

When we create a readOnly hashmap.. That has arrays or other hashmaps as values.
Need to deepcopy these values too and make them readOnly too
That's why I didn't do it back then

[sinful kettle]

I think readOnly should only apply to its keys, i.e. not adding anything new to it or removing any keys

[untold sky]

But when you have
["value", [1,2,3]]

you can
(_map get "value") pushBack 4

even on a readonly map, you can modify the contents of a readonly map then

[sinful kettle]

texts can be modified by ref too 🥲

[untold sky]

Deepcopy doesn't actually copy text types.
Not for array deepcopy either

the engine can't even do it 😄
Text copy method, returns reference to same text, it doesn't even copy
I shall leave that bug be bugged

[sinful kettle]

wait but does it mean the hashmap class can't modify itself anymore?!

what would be the point?

[sinful kettle]

anyway I think a new flag can be added called "sealed" that does this

btw do you think it's possible to prevent member access from outside?!

or I guess it should be done via wrapper functions...

[untold sky]

In theory yes, but I don't want to pay for that check on all get's

although that check is really cheap

[sinful kettle]

well for example if you prefix a member name: ~method it becomes private

or #method

[untold sky]

nope.
the _thisObj call... doesn't know if it was called from inside a method inside the same class, or from anywhere else

[sinful kettle]

yeah I figured

[untold sky]

it feels so ugly to reuse compileFinal hashMap but.. you're literally making it final so that command keeps doing what it has always done, just the compile name is a bit unfitting

[sinful kettle]

makeFinal 😛

finalize?

[untold sky]

I don't want to add a new command for a thing that we already have a command for that does the same thing

[sinful kettle]

will it work on arrays too?!

[untold sky]

For now I only want to add CODE and HashMap

[untold sky]

Read-only error messages are crap....

Or atleast the old ones were

But do I want to update all the old non-hashmap ones.. eh

[gray wharf]

yes, yes you want

[untold sky]

I should add a proper new error message type but that is soo much extra work :u

[untold sky]

So then you'd have to pre-define all member variables at the start and cannot add new ones or remove them.
Right? Sounds good

[sinful kettle]

yeah

[untold sky]

Now.
When you
+_object

what flags do get copied and re-applied to the resulting object..
Unscheduled flag 100%
ReadOnly flag 0%
Sealed? I think yes

[sinful kettle]

all?

[untold sky]

If you have a read-only object, I'd like a way to make a non-read-only copy that can then be used

another way to make data structure "templates" of something, you have to copy before you can sue

[sinful kettle]

you have to copy before you can sue
should I show the copy to my lawyer?

[untold sky]

Mh...
_var = compileFinal "..."
actually locks the _var so you cannot reassign anything to it.

_var = compileFinal hashMap
Should it lock the variable too or just be a read-only instance where the variable can be re-assigned, I think it should lock
I'm pretty sure that same as with compileFinal, if you make a read-only thing and put it somewhere, you probably want it to stay somewhere

And if you want it to be re-assignable... hashMap set ignores whether value is final (I should fix that for objects to block code though...) so you can store your read-only hashmap in a hashmap, and then you can replace it 🤣

[tropic jewel]

https://feedback.bistudio.com/T171128 n_n

[untold sky]

How do we deal with merge/insert into a sealed instance.

currently, I insert all the values, until the first violation (trying to add new key or overwriting a compileFinal'ed method) and then throw error and abort.

That means a insert or merge may half-complete. The half-complete insert already existed, if you tried to insert with some invalid key type in the middle, it would insert up to it and then abort.

Alternative would be to throw error, but then continue inserting and potentially repeat-spam errors

[sinful kettle]

I prefer the errorless version

[untold sky]

Well have to somehow tell the user that what they tried to do was invalid and didn't happen.
So both version will throw error

keys are case sensitive so
constructor
copyconstructor
destructor
flags

or

Contructor
CopyConstructor
Destructor
Flags

? I prefer the second, mostly because CopyConstructor being two words

[gray wharf]

or camelCasing

[sinful kettle]

why not shorten copyConstructor? 😓

e.g onCopy 😓

cc

[untold sky]

cconstructor i was thinking about

but eh 🤔

[sinful kettle]

why not use symbols?
constructor -> "()"
destructor -> "~"
copy -> "="

[untold sky]

might be too unreadable?

[sinful kettle]

well words are error prone, like you could get the casing or spelling wrong

[untold sky]

true that 🤔

emoji 🤣

# for flags then

[gray wharf]

don't you dare!!! 👀

[sinful kettle]

how about:
"class"
"!class" (or "~class")
"+class"?

"flags" (or "#class")

(or "obj" instead of "class")

[untold sky]

Also, compileFinal prevents from replacing methods. But, deleteAt can delete method and then you can re-add a new one.
so compileFinal'ed methods should also be un-deletable?

[sinful kettle]

I guess so?

[untold sky]

I guess it has to be.. welp

Other idea for flag, throw error when calling unkown method, instead of silent ignore and return nil on call?

or silentErrors flag, to do the opposite and also turn off errors when you insert and try to overwrite final or sealed stuff 🤔

So collecting some feedback.
_thisObj -> _self
sounds reasonable, but relation to thisTrigger, thisList and such

[gray wharf]

I get PHP vibes here 😄

[untold sky]

copyconstructor -> clone
Though it would then be post-clone, not actual clone method that returns a new instance
or maybe just simple copy or copied

https://feedback.bistudio.com/T170647#2421993 Here's all the current documentation I have

tostring method, custom implementation for str _object

reserved names should have special naming to not conflict.
#constructor
#destructor
#flags
? I like #, but people may want to implement "private" methods by prefixing them with #, but really i don't know what people want 😄

or _ prefix makes most sense to me for private stuff

[gray wharf]

aren't some BIS_ or BIN_ "private" variables prefixed with #?

I would say "go for it" and if people want their own stuff they should use _ or TAG_ 😊

[sinful kettle]

making them longer is not a good idea tho 😓
I'm all for making them shorter

also constructor and destructor are hard to type

maybe create and destroy? tho destroy is also hard to type

create and discard?

[gray wharf]

destroy is also hard to type
wat?

[untold sky]

well if you typoed them you'd probably notice quick

[dreamy bane]

delete? 😄

[untold sky]

create and destroy sounds better

[sinful kettle]

make and trash 😛

[sinful kettle]

for me at least

[daring wagon]

destory...

[sinful kettle]

for some reason I always mess up things that have oy

[sinful kettle]

delete is easier to type and almost impossible to get wrong tho
destroy may become destory like R3vo said 😅

[untold sky]

I don't want to make crusty because people no wanna type correctly :sad:

if you add imaginary "on" prefix, they all make sense
onCreate
onClone
onDelete
onToString (eh)

delete is better, we already have precedence with Deleted EH, and deleteVehicle

[sinful kettle]

onToString is weird

onString at least

or onStr

[sinful kettle]

actually now that they're shorter # makes sense

#create #delete #str #clone #flags

[untold sky]

tostring sounded so nice 😢
but str is better I guess ye

https://feedback.bistudio.com/T170647#2421993 Updated and with example.

[gray wharf]

https://tenor.com/view/ok-ill-allow-it-okay-alright-gif-15634749

[untold sky]

the parameter to clone doesn't have to be _this
it would make sense to name it _source for example. But function parameters are always _this so that may be too bad of an idea

[sinful kettle]

noticed a couple of errors but sounds good

[untold sky]

if its just typo's then doesn't matter that can be fixed later on wiki 😄

[sinful kettle]

btw does it make sense to define the class it was created from? (I meant a new key #class)

actually no nvm

[untold sky]

could
All the keys are copied though so its kinda there

[sinful kettle]

yeah that's what I meant by nvm 😅

[untold sky]

It would be useful for non-sealed classes, to see what they started with.
You could do it yourself in constructor if you wanted (if copying is forbidden it'll be a bit more annoying :D)

[regal nimbus]

When you're searching for similar issues and the closest one is from 2013 :/

[uncut briar]

@untold sky do you think it would be feasible to add third array element to getMissionLayerEntities with all groups in the layer? (or dedicated command but expanding existing commands is preffered iirc)

Currently it needs scripted handling, something like this:
https://gist.github.com/veteran29/4735d4d96add5cdf0134a834cd2f12d3#file-getmissionlayergroups-sqf

[untold sky]

yes, ticket, if you get ticket tonight I'll do it tomorrow 🫂

[uncut briar]

https://feedback.bistudio.com/T171135 ❤️

[desert trench]

would it be considered a bug that killing setCaptive units doesnt reduce the rating? (but continues to increase it if side relation of both groups are enemy)

in simple terms should shooting captive/"surrendered" units get punished or not

(or would such a change/fix be too risky due to setCaptive use also in other context)

[gray wharf]

I'd wager too risky for backward compatibility - use an event handler?

[desert trench]

We dealt with it already. Asking if its worth to report or not

[gray wharf]

nay

I will add the information to the setCaptive page

Tankbuster's 2019 note has been integrated to https://community.bistudio.com/wiki/setCaptive

[untold sky]

Would you be ok with exact same implementation (find unique groups of the units) in engine.
Or.. what about empty groups placed inside the layer in Eden, is it important to return these too? As that would need completely different solution (I already made that different solution, but lots of extra code and serialization, would be easier to just run that "script" in engine if that's sufficient
Are empty groups a thing? 🤔 I guess if the entities all got deleted, and the empty groups are still left behind...

[[B Alpha 1-1:1,B Alpha 1-1:2,B Alpha 1-2:1 (dedmen),B Alpha 1-2:2,B Alpha 1-2:3,B Alpha 1-2:4],[],[B Alpha 1-1,B Alpha 1-2]]
turns into
[[<NULL-object>,<NULL-object>,B Alpha 1-2:1 (dedmen),<NULL-object>,B Alpha 1-2:3,<NULL-object>],[],[B Alpha 1-2]]
when units die.
But the 1-1 group still exists, with 0 members inside it.

[uncut briar]

Would be good to be able to reach them... But on the other hand you can grab all your groups into gvar on mission init.

[west onyx]

do we have a function that selects unique elements in an array

[alpine tulip]

Removing duplicated? I guess arrayIntersect will do

[west onyx]

hmm

i must have been looking at the wrong wiki example

[gray wharf]

reformatted the description to make it clearer

[fringe python]

will arma devs be hanging around here

[fathom slate]

This is (probably) a channel dedicated to the feedback tracker, not for direct feedback.

[fringe python]

oh ok

[mighty violet]

no its for Feeeeeedback ;)

[west onyx]

<parent> modelPos <attached object> // [0,-3,1.56]

instead of <parent> worldToModel (getPosWorldVisual <attached object>)

is there demand for that?

[solid marten]

make a ticket, I will overload getRelPos

[west onyx]

@solid marten https://feedback.bistudio.com/T171193

[daring wagon]

I am wondering. Since the introduction of regex. Would it be feasible to implement wildcard support to Eden Editor search?

[sinful kettle]

this is wrong tho

getPosWorld is ASL

worldToModel takes AGL

please gib worldWorldToModel

[sinful kettle]

Potential bug in append:
https://feedback.bistudio.com/T171200
it's slower than insert and +

it probably doesn't preallocate or something?

[turbid terrace]

I think this glitch is new.
Anyone else has it? It might be my FOV but I think I defaulted it.
You need a CSAT grenadier rifle, pull up grenade sights and crouch.
https://youtu.be/8NsC5gJ09SE

[alpine tulip]

Shouldn't be new since is a model issue

[turbid terrace]

Yup, but I think I would've noticed it before, maybe not 🤷‍♂️

[west onyx]

yes

[west onyx]

is this a thing where you choose between render scope and sim scope?

i would probably prefer render scope

also doing a lot of logistics scripting these days. an “AttachTo/Detach” mission event handler would be extremely useful

[gray wharf]

worldWorldToModel & worldWorldToModelVisual 😆 😅

[west onyx]

and a render scope variant for this function origin getPos [distance, heading]

origin getPosVisual [distance,heading]

unless it doesnt need one :\

[untold sky]

even without regex we had wildcard support in some places.
yes, probably

[untold sky]

it does.
append is actually even simpler than insert or +..
especially + should be slower, because that copies ALL elements to new array

[sinful kettle]

But why is it slower than insert and +?

#arma3_scripting message

The benchmarks

There's not much difference tho

But still it is slower

[solid marten]

@sinful kettle when you make a ticket, just pretend that we are all idiots and add steps to reproduce, please

[solid marten]

Thats why you should include repro so we can VERIFY

[sinful kettle]

So you mean I was wrong and it's not slower?

[solid marten]

No I mean what I said

[sinful kettle]

Well I'll do later today

I need to make a more reliable repro

[daring wagon]

Want a ticket?

and which "some places"? 🤔

[untold sky]

yes, dunno but I've seen it

[daring wagon]

I see

I'll make a ticket when I am back home.

[desert trench]

Got a question on Zeus not showing assets - if to report or not:
AFAIK it also checks the scope of the crew/driver for vehicles and only shows >2. This is not documented as far as I could see, but only a community understanding from trail and error.
with hasDriver=-1; there is no driver - this seems not considered by the Zeus coding and thus you cannot have "driverless/one man" vehicles exposed in Zeus as far as i can tell - can someone confirm? worth to make a ticket?

[daring wagon]

@untold sky Wildcard ticket https://feedback.bistudio.com/T171218

If someone has better/more example feel free to add them 🙏

[gray wharf]

B_*_F

[sinful kettle]

done

[sacred river]

Can I also post here a FT of a possible useful feature request?

[outer frigate]

you can

[sacred river]

https://feedback.bistudio.com/T166882

[untold sky]

At the moment there is a function called

setVehicleArmour

Uh what? https://community.bistudio.com/wiki/setVehicleArmour

https://community.bistudio.com/wiki/setVehicleArmor ah

no :harold:

[alpine tulip]

Excuse me OFP Elite devs, what is this 😓

[west onyx]

can someone else test this in dedi multiplayer... exec on server, ensure there are some server-side spawned units

private _unit = objNull;
{
    _unit = _x;
    {
        _unit removeItems _x;
    } forEach [
        'chemlight_blue',
        'chemlight_green',
        'chemlight_red',
        'chemlight_yellow'
    ];
} forEach allUnits;```

is giving me massive RPT log spam

19:07:32 Warning Message: No entry '.scope'.
19:07:32 Warning Message: '/' is not a value
19:07:32 Warning Message: Error: creating weapon chemlight_blue with scope=private```

[solid marten]

Ask yourself this, is chemlight a weapon?

[alpine tulip]

I think is a magazine since is a grenade

[west onyx]

hmm so it is. for some reason I had treated them as items. thanks for checking

must have been fiddling with "uniqueUnitItems" prior

[gray wharf]

I see it as a way to not mix up "set damage" and "set health", (1 - damage) or (1 - health), this kind of "negative thinking" etc

[untold sky]

@gaunt depot ref https://feedback.bistudio.com/T170754 UI2Texture aspect ratio.
If I accidentally change the size of all map drawIcon's, you would notice that right?

instead of hardcoded 640x480 I now do 640x(640*aspectRatio) which should keep the size of the icons the same, but not fully confident on that

[gaunt depot]

Well it won't be gamebreaking but might break visuals in a lot of places

[untold sky]

Yeah i hope it doesn't, which why asking so you can notify me if that happens

[gaunt depot]

Probably BI missions too as they probably use drawIcon too?

[gaunt depot]

Sure, I can compare my stuff with the dev branch, but more tests will be needed

[untold sky]

the icons drawing should be good-ish now.
I can't really see the squished text that you mentioned

[gaunt depot]

Didn't ACE have some extensions to map tools, ruler and all that, doesn't it use drawIcon for it too?

[untold sky]

https://github.com/acemod/ACE3/blob/76676eee462cb0bbe400a482561c148d8652b550/addons/maptools/functions/fnc_updateMapToolMarkers.sqf#L37 seems ACE handles aspect ratio issue by itself
so they'll have to remove that fix then
Thanks I'll check that

[gaunt depot]

Map texts are squished the same way as other display texts due to always rendering in 1:1 despite target texture not being 1:1. It all looks weird compared to proper map lines and grid squares in the background.

So its kind of a same issue of ui2texture not taking target resolution aspect ratio into account I guess

Also what's strange on your last screenshot is that drawIcons are squished vertically while texts are stretched vertically

Original bugged drawing has both drawIcons and texts stretched vertically

[untold sky]

Yep it breaks ACE's thing that tries to workaround the bug by doing their own stretching

[untold sky]

Can you adjust your repro to give me some text that you can properly see?

[gaunt depot]

on it

[untold sky]

Seems like images inside structured text also have the 640x480 hardcoded scaling 🤔

[gaunt depot]

Added big visible text in the repro in comments

                    _map drawIcon [
                         "#(argb,8,8,3)color(0,0,0,0)"
                        ,_map getVariable "color"
                        ,_map ctrlMapScreenToWorld [0.3, 0.3]
                        ,1e-6
                        ,1e-6
                        ,0
                        ,"drawIcon text"
                        ,2
                        ,0.08
                        ,"PuristaMedium"
                        ,"right"
                    ];
```alongside other `drawIcon`s

[untold sky]

Players own position marker (while inside vehicle with GPS), Tactical Ping marker on map, HTML (structured text?) images, focused marker (marker: html link in structured text, interesting didn't know that was a thing)
all have the same scaling issue, but I assume most of these are generally too small to notice

Doesn't seem too badly stretched now

Freaky Arma resolutions

Could it be that text is already fixed on dev branch?

[gaunt depot]

Map text is distorted the same way as the rest of display texts

[untold sky]

My eyes cannot see that

[gaunt depot]

I guess I explain it wrong

[untold sky]

but the box around it is also stretched the same

[gaunt depot]

Both numbers outside of map and "drawIcon" text are both stretched vertically

[gaunt depot]

Box is 0.1x0.1 in UI coordinates

That's issue of ui2texture aspect ratios, there is no easy way to draw 1:1 UI on non-1:1 texture

Box is stretched vertically because texture is rendered into 1:2 texture (#(rgb,1024,2048,1)ui('RscDisplayEmpty','%1'))

1024x2048

If you ask me, it should just cut off right half of the UI when drawing

[untold sky]

If you ask me, you shouldn't try to draw 1:1 UI onto 1:2 texture and expect it to look "correct" in whatever your imagination of correct is

Because different people will have different ideas of "correct"

[gaunt depot]

That's why I suggested having this behaviour optional and configurable

[gaunt depot]

Even if you divide all your heights by 2 to be appropriate for 1:2 aspect, texts will still be drawn in 1:1 aspect ratio

[untold sky]

Font handling is very special and hard to influence

[gaunt depot]

Basically if I make my squares 0.1 x 0.05, text inside them will still be stretched

Repro from related ticket

Texture is 2:1 here so text is squished

Proper text is achieved by having 1:1 texture stretched vertically inside 2:1 texture

[gaunt depot]

That's why I suggested to cutting off part of display instead of trying to fix aspect handling inside each UI element

This way 0.1x0.1 UI square stays square, text is drawn properly

[untold sky]

can do, separate ticket.
probably different syntax
like
#(rgb,1024,2048,1)uisquare('RscDisplayEmpty','%1')
maybe or smth
which will internally render as 2048x2048 no matter what you said

[gaunt depot]

Another way of doing it, but having displays with UI height = 2 to cover whole texture is pretty weird

[gaunt depot]

Add another parameter to current procedural texture?
#(rgb,1024,2048,1)ui('RscDisplayEmpty','ca',XX)
stretch: current behaviour
cut: cut off part of display to fill texture (pic 1)
fit: resize display to fit into texture (pic 2)

[untold sky]

I don't like adding more parameters, thats not very extendible.
Maybe uiext that has designated naming for parameters
#(rgb,1024,2048,1)uiext(displayclass:'RscDisplayEmpty',fillbehaviour:'stretch',texsuffix:'ca')
I hate this current syntax and its limitations

[gaunt depot]

How about:

"myproceduraltexture" createTexture ANYTHING_WE_WANT;
_obj setObjectTexture [0, "myproceduraltexture"];

we can even have it take hashmaps for named parameters

"myproceduraltexture" createTexture createHashMapFromArray [
     ["procedure", "ui2texure"]
    ,["type", "ca"]
    ,["width", 1024]
    ,["height", 2048]
    ,["format", "argb"]
    ,["mipmaps", 4]
    ,["display", "RscDisplayEmpty"]
    ,["fill", "stretch"]
];

Who needs array as arguments when we can use hashmaps!

I'm looking at it from mission maker standpoint though, not sure how problematic it will be to use with mods.

[untold sky]

texture must be a string because 20 years of legacy

[gaunt depot]

Gonna make sure textures are generated before they're used

[gaunt depot]

"myproceduraltexture" won't fit as texture name?

[untold sky]

Then you can't pass the arguments

texture including all parameters leading to its creation must be a string

[gaunt depot]

You first script create it, then use by setting your string

Have createTexture do all the creation by right operand hashmap properties

[untold sky]

Can't

[gaunt depot]

instead of parsing the string

🤔

"myproceduraltexture" createTexture createHashMapFromArray [
     ["procedure", "ui2texure"]
    ,["type", "ca"]
    ,["width", 1024]
    ,["height", 2048]
    ,["format", "argb"]
    ,["mipmaps", 4]
    ,["display", "RscDisplayEmpty"]
    ,["fill", "stretch"]
];
obj setObjectTexture [0, "#(argb,8,8,3)custom(myproceduraltexture)"];

If its a matter of having this procedural texture syntax.

[untold sky]

Would have to network sync

and serialize

[gaunt depot]

Network sync? Why?

Textures are local, have each client do their own texture

[untold sky]

Because player2 will try to load "myproceduraltexture" and won't find it

[untold sky]

Yes exactly, that's why

[gaunt depot]

setObjectTexture is local

[untold sky]

Ah local variant

[gaunt depot]

_rsc_picture ctrlSetText "#(argb,8,8,3)custom(myproceduraltexture)"

[untold sky]

ok that'd work, but then still serialization, and the network sync for when you use global variant

[gaunt depot]

Leave that to mod\mission designer, have them init their textures before using them

[untold sky]

And won't work for textures set in p3d's, they are string only

[devout wave]

Textures as a mission config class? 🤔

• and normal config too

[gaunt depot]

Yeah, that what I was thinking about earlier, I'm looking at it from mission design view point, not taking mods into account

If you'll have P3D with that texture baked in, your mod will have to call some preInit function to initialize the texture before P3D is used somewhere

[untold sky]

much easier to just create a proper syntax for these strings, rather than build a whole extra system that doesn't work in all cases

[vocal abyss]

What about generating the string?

private _textureString = createTextureString ...;
MyObj setObjectTexture [0, _textureString];
```Then you can easily build complex (not necessarily human-readable) strings.

[untold sky]

yeah that'd be easy

but.. why extra command for something you only ever do once and can do by hand just as well

[gaunt depot]

To be honest I long dreamed for more advanced procedural textures

[untold sky]

Can make a ~10 line script function that does the same and plop it on wiki

[gaunt depot]

Like applying masks to texture, hue shifts, etc.

And I had this idea for procedural texture command for a while, as a convenient way of providing arguments into procedural texture rather than cramming it all into a string

I'm dreaming again of course

[untold sky]

well I maybe add that new syntax, then you can make a script function that converts hashmap into texture string if you like

actually you can make that script function right now already

its one format and a bunch of hahmap getOrDefault's

[gaunt depot]

Yeah, its called format, the issue is that if we will ever have anything complex, having it all in the string is a bad idea

Back to original issue, not sure if we will ever need more arguments, so I'd just make 2 new proc texture names and that's it:

#(rgb,1024,2048,1)ui('RscDisplayEmpty','%1') - current
#(rgb,1024,2048,1)uicut('RscDisplayEmpty','%1') - cut off
#(rgb,1024,2048,1)uifit('RscDisplayEmpty','%1') - shink to fit

like you suggested with uisquare

What I always dreamed about with procedural texture is having some kind of toolset to operate textures in the game. For example taking existing asset textures and hue shifting them to have different car colors. And to avoid hue shifting unneeded details have a black and white mask to provide for effect.

That's why I went so much into ui2display because its a step towards it

Instead of having huge premade textures with repeated patterns bloating the mission file, now you can have separate elements and procedurally arrange them on display and render it into the texture

Yeah I know its perfect time to implement all that when A3 is EOL

[gaunt depot]

Edit current ticket as its about "Aspect ratio issues with UI2Texture" or create a new one for the fit\cut option?

[untold sky]

new

[gaunt depot]

https://feedback.bistudio.com/T171238

[gaunt depot]

Btw, what is default texture type if texType is not specified? Wiki says its "ca"? but its definitely not ca

Left is default (not specified), right is "ca"

Took me some time to notice laptop screen not being shiny anymore to realise its not "ca"

[gaunt depot]

Going back to that, if you ask me, ui2texture should behave how I suggested it with "cut" behaviour. Since displays are often text, I can't imagine anyone wanting their text to be squished or stretched when they use it on non-1:1 texture, can't call it "correct" what so ever.

So I'd just make it simple and change current ui2texture to always do "cut", I doubt there is a lot of backwards compatibility to support

no need for new proc texture names, no need for new parameters

[near cedar]

@untold sky is the bug where amphibious vehicles fail at driving around houses ( https://feedback.bistudio.com/T167742 ) related to tracked amphibious vehicles failing to cross rivers? Like they don't get out of the water again and just float. ( https://feedback.bistudio.com/T170421 )

Do they get treated as boats and then are unable to pathfind their way out on to land?

(it's weird because this doesn't affect wheeled amphibious vehicles)

[regal nimbus]

Do they delete their waypoint or just not move towards it?

[near cedar]

From what I remember they just don't move towards it after reaching the middle of the river but it's been a while since I tested. Engine seems on (see images in ticket) but speed is 0.

[solid marten]

render yes

[untold sky]

co

[untold sky]

In my test they had no issue going across water.
But they would sometimes decide its a better idea to take a huge detour over terrain, and drive upwards through bridges (which they can't) instead of taking a much shorter, direct water route

no the code that was changed is same for tracked and wheeled

[gaunt depot]

I guess no reflections are caused by something else rather than it being "co" or "ca"?

[near cedar]

Interesting, I could reliably get them stuck in the water on Livonia 🤔 I don't think i tested near a bridge though

[untold sky]

co is diffuse, and default is diffuse

dunnö

reflections are in the material

[ornate marlin]

feedback for the feedbacktracker. little incetion taste :D

[true solar]

Upvote for glory m8s!

http://feedback.arma3.com/view.php?id=26370

[trim acorn]

Attack Helicopter Pilots please upvote: http://feedback.arma3.com/view.php?id=25388 :-)

[west onyx]

@sterile nova gib ther vehicle cargo variants of these

[alpine tulip]

ViV can be done in a config side

No longer model tweak needed

[west onyx]

hmm

do we have access to do in vanilla or mod required?

[alpine tulip]

Mod required though unfortunately

But reyhard is not necessary needed

[west onyx]

who else is there to do vanilla configs for the models

[alpine tulip]

Ded or KK both can do config works no?

[west onyx]

not sure, usually they toss config work over to reyhard

[tropic jewel]

Is it intentional that the left custom info panel cannot be resized, but the right side can?

[west onyx]

probably they did not anticipate people still modding the game in 2023 😄

"theyll never know"

[west onyx]

@solid marten chatGPT is asking for a "findMax" command

_index = findMax [4,5,9,6]; // 2

_index = _arr find (selectMax _arr);  // 2```

not a joke btw 😄 it actually wanted to use a "findMax" function and when i told it was non existent it requested

[alpine tulip]

Sounds good to have

[west onyx]

I am using to find the longest sides of multiple bounding boxes to align

[near cedar]

huh i thought we already had min/max functions

guess not

[alpine tulip]

Only selectMax/Min?

Wait a minute doesn't select suit for you?

[west onyx]

well select returns the element, but doesnt return the index of the element

[alpine tulip]

True

selectMax and findIf does for you BTW?

[west onyx]

well im just using _arr find (selectMax _arr);

just thought it cute the bot wants a new command added to the library

[alpine tulip]

Fair

[gaunt depot]

Obey the AI overlord

What would be really useful is ARRAY find [ANYTHING, NUMBER] so we can do finds after certain index

Also had a need for forEachReverse or something like that, had to use ugly for "_i" from count _array - 1 to 0 step -1 instead

reverse modifies original array and making a copy just to reverse to use with forEach is even uglier

[solid marten]

The overloads for find are basically all used it has to be a different name

[solid marten]

if changing dir is a matter of flicking a flag then it could be added

[gaunt depot]

forEachReverse will also be useful for quick array cleanup

You can't do:

{
    if(isNull _x) then {_array deleteAt _forEachIndex};
} forEach _array;

but you could with:

{
    if(isNull _x) then {_array deleteAt _forEachIndex};
} forEachReverse _array;

Yeah you can do - [objNull], but its for primitives only, won't work with more complex conditions

for "_i" ... step -1 is works, but uglier

Ideas also went as far as imagining:

_dead_units = _array deleteAt {!alive _x};

select + deleteAt

6 matches for step -1 in my project

None in really performance critical places though

[mint hedge]

I have just submitted a ticket for a potential bug with the Config MFD system that I have found. Is there anything else that I can do to improve this? https://feedback.bistudio.com/T171305

[torn bear]

Not anything big, but if you would please :) http://feedback.arma3.com/view.php?id=24589

[solid marten]

adding foreachreversed is not a big job however might need a separate class to keep performance intact

[devout wave]

Is there any chance of getting an alt syntax for the magazineDetail commands that returns array (and ideally classname) instead of that horrific string format?

[sinful kettle]

yeah what on earth is that anyway?! 😅

[gaunt depot]

Probably some dev command to debug new inventory that somehow got into release

I wonder if other items like uniforms, vests, items and weapons have this network id and creator id like mags do

[devout wave]

https://feedback.bistudio.com/T79477

[tropic jewel]

Ah, because they resize to the same size.

[rugged nova]

any chance to get createHashMapObject to act like a namespace?

Maybe more flags, like:
"namespace" -- acts like a namespace
"alwaysBroadcast" -- all changes to the hashmap are broadcasted to every client like the third parameter in setVariable

[sinful kettle]

a hashmap already acts as a namespace tho?

[rugged nova]

yeah, but no sync

[sinful kettle]

all changes to the hashmap are broadcasted to every client like the third parameter in setVariable
this doesn't seem like a good idea

[rugged nova]

and no persistence saving of data

[sinful kettle]

might as well poll changes then send at once.
otherwise every change would attempt to sync, e.g. even if you did a simple set on a member var

[rugged nova]

yeah, I was gonna ask for a force broadcast cmd and a pause broadcast cmd

[sinful kettle]

idk what you mean by this one. namespaces are not persistent either

[rugged nova]

I should have said, a lifetime of Mission duration like missionNamespace

[sinful kettle]

but if you save the hashmap in missionNamespace you do get that result already

[rugged nova]

yes, but I already do that.

I just want a different mechanism to do so for hashmaps

https://sqfbin.com/ezadupiguhihewiyupez

I have to set on the hashmap then setVariable again to broadcast the change

[west onyx]

the AI is now requesting an "Eject" event handler

i didnt write this ```myPlane addEventHandler ["Eject", {
params ["_vehicle", "_pilot"];

// Call the eject_handler.sqf script
[_pilot] execVM "eject_handler.sqf";

}];```

[sinful kettle]

what's the diff between eject and getout?

also who uses execVM anymore?

[near cedar]

I think it's common

Especially when you want sleep

[sinful kettle]

you shouldn't use execVM in EHs

[sinful kettle]

use spawn instead

[near cedar]

Doesn't execVM do that anyway?

[near cedar]

What issues does that cause?

[devout wave]

The GetOut/GetOutMan EHs detect eject actions, but they don't tell you it was an ejection, so you have to kinda guess if you're wanting to e.g. do stuff related to an ejection seat

[sinful kettle]

slow because it recompiles the script every time

[sinful kettle]

then just add a bool to the GetOut/GetOutMan EH
params [...., "_isEject"]

[near cedar]

Does the speed matter much if the EH is triggered only sometimes?

[devout wave]

that's what I was going to say ¯_(ツ)_/¯

[sinful kettle]

are you saying it's impossible for it to trigger multiple times per frame?

[sinful kettle]

like in here, what if multiple units eject at once?

[near cedar]

I mean, if the EH isn't onEachFrame then it shouldn't, i would hope?

And if regular EHs can unintentionally fire multiple times in a frame then maybe this should be stated somewhere prominently in the wiki...

[sinful kettle]

I didn't mean one EH firing multiple times. I meant multiple EHs

[sinful kettle]

as I stated here

[near cedar]

Oh i thought that was a reply to QS

[west onyx]

i mean my use case is to detect when a pilot ejects in order to trigger a pilot rescue task

we can use logic in a getoutman event like "getpos > 100"

so there is no great need

[solid marten]

I like this

Havent looked in it but without the ticket there is a chance I would overlook it

[solid marten]

forEachReversed Revision: 150434 please make a wiki page @gaunt depot

[gaunt depot]

It just forEachReversed ARRAY or forEachReversed HASHMAP too?

forEachReversed for hashmaps makes little sense due lack of order in the first place, but I still wonder

[solid marten]

no hashmap has no order

[gaunt depot]

So just command is just for arrays?

[solid marten]

yes

[gaunt depot]

👌

[gaunt depot]

https://community.bistudio.com/wiki/forEachReversed

[hallow sun]

That will help a lot of noobs/beginners. Deleting in a foreach is a common mistake (also did it) 😆

[solid marten]

Might be worth to make a note in forEach to encourage to use forEachReversed for deletions

[dreamy bane]

has alt syntax for groups <array> been considered?

[gaunt depot]

added

[trim acorn]

http://feedback.arma3.com/view.php?id=24027 last one from me. Thanks.

[regal nimbus]

I don't suppose this is worth writing up:
moveInAny somehow breaks driving AI when used to move an AI into a gunner seat, such that they'll refuse to drive anywhere, or occasionally act like they have a brick wall in front. setEffectiveCommander to the driver will allow them to drive again, but setting it back to gunner (which is default) will block them again. You can do moveInDriver + moveInGunner and it's fine, while moveInDriver + moveInAny will break. Tested with Hunter HMG, Marshall and a couple of GM armed trucks.

[gaunt depot]

Might be worth it, probably incorrectly sets effective commander by default

[wind light]

@trim acorn regarding http://feedback.arma3.com/view.php?id=25388 the controls that appear in Profile files and not in official controls menu (The one in game) are not officially supported. That said the switching weapons using keys is one of the systems we are currently working on and planning on releasing with the EDEN Update

[patent trail]

@wind light will that include switching weapons while moving for things other than launchers?

That is something that someone in the community already created a script for but would be nice to have in-engine

or at least switching to pistol while moving

[wind light]

The system is gonna support Vehicles and Infantry. So for example if you bind switching to pistol to X key and press the key. Your character will switch to the pistol. Is that what were you asking about?

[patent trail]

I am talking more about the animation side

[dusky wigeon]

That's already in @wind light

[patent trail]

When you switch weapon's currently the your character stands still and switches to another weapon and then you can once again move.

[wind light]

Yeah, now i understand.

[trim acorn]

Thanks adam. But the weapon selection worked flawlessly, why was it removed again? (Vehicles)

[patent trail]

Seems like it would be really easy to implement becuase it is just controlling the upper half for the reload animation, so the lower half could still preform the other stuff.

[wind light]

@trim acorn because it sneaked out accidentaly. THe system is sitll being refined and improved :)

[patent trail]

Other stuff as in running or walking.

[gaunt depot]

@untold sky Might this also be a case of non-4:3 aspect after messing with FOV settings?
https://www.reddit.com/r/arma/comments/12a6y8u/is_this_normal_how_to_fix/

Seen some random video on reddit, thought about you recently fixing that elsewhere

[alpine tulip]

ctrlSetAngle you mean?

[gaunt depot]

Hmm, I thought scopes behave like this by the engine 🤔

Just tested vanilla MOS scope and it doesn't rotate when leaning

Nevermind then

[dusky wigeon]

@patent trail wants some of this https://www.youtube.com/watch?v=wouoN9e9fM4 ^_^

[patent trail]

was just about to link that

[dusky wigeon]

Things discord needs #12312: forcing a link not to show the preview.

[wind light]

I'm pretty sure we already gave statement on the "Switching weapons while moving" issue multiple times. I might as well be mistaken.

[dusky wigeon]

Yep, doesn't stop people from wanting/dreaming about it tho. :P

[patent trail]

I don't remember seeing that attached to any of the feedback trackers

[dusky wigeon]

Check the Smookies posts in my thread about that addon in the video.

[patent trail]

Link?

Or is it the addon thread for that

ah..... I see its in the description

[wind light]

https://forums.bistudio.com/topic/148904-switching-weapons-on-the-move/page-10 is this the one Sniper?

[dusky wigeon]

Yep.

[wind light]

Ok :)

[dusky wigeon]

Smookie makes a few posts in there why it's hard for them to do properly.

TLDR: Anim system is :skype_puke:

Page 3 has Smookies posts.

[patent trail]

yeah after I look through 7 pages for him you mention that ;)

I kind of figured that was the reason being with regards to the animations starting and stopping... feel bad for devs having to work with Arma engine ;)

[trim acorn]

Could it accidental sneak back in until the refined version comes back?

[dusky wigeon]

@patent trail I've gained a lot of sympathy for the BI employees in general recently.

[patent trail]

I can only imagine how much work the HR department has to do

I spend a few days trying to get some weapons in-game and working right and I basically want to shoot my own foot :(

[dreamy bane]

should I make ticket for this, unless already made? will devs want to look?

[untold sky]

And it creates a copy. If you're fine with copy might aswell just use select

{
    if(isNull _x) then {_array deleteAt _forEachIndex};
} forEachReverse _array;
```==
`_array = _array select {!isNull _x}`
The problem is just if you HAVE to work with references
If you need reference you could
```sqf
_newArray = _array select {!isNull _x};
_array resize 0;
_array append _newArray;

[untold sky]

That is _newArray = _array select {alive _x}; _dead_units = _array-_newArray; _array=_newArray
which yeah a bit ugly, but do you really do that often

[untold sky]

No. A namespace needs a global object that it can sync with.
If you want a namespace, just create a location and use that

[untold sky]

Maybe, messing with FOV settings is not supported

[wind light]

@trim acorn I dont think so :/

[uncut briar]

It's CBA scopes framework that do the rotating ones.

[uncut briar]

Location is local, simple objects should be used for global custom namespaces.

[vagrant charm]

soo feedback tracker huh? played almost 2400h, Im happy customer :D

[desert trench]

why is custom FOV not supported - too much engine work needed to have it working "everywhere"?

[alpine tulip]

IIRC Ded ever told us why it won't have an official implementation

And the reason is some kind of “pretty much everywhere” I recall

[untold sky]

ye

[gaunt depot]

Lets have a mission option\script command to force reset the FOV

Very annoying that people use insane FOV in multiplayer to peek over big obstacles in 3rd person

(yeah I know 3rd person sucks but plebs won't play 1st person only games)

[untold sky]

I forgot that in the graphics script command thingy i guess. That should probably also return FOV

[gaunt depot]

Lets draw thick black border around their screen when they're in 3rd person with a huge FOV

[uncut briar]

getResolution has fov in the output already

[uncut briar]

The fov can change between resolution\aspect ratio tho, 21:9 will have bigger one than 16:9 won't it?

But resolution can be found out too so probably doable anyway.

[gaunt depot]

Not sure what fovTop and fovLeft when you're on ultrawide 🤔

[uncut briar]

fovTop=0.75;
fovLeft=1.7777778;

for 3440x1440

[gaunt depot]

I wonder if this messes up drawIcons on the map on stable? 🤔

Unless only fovTop matters for it

[vagrant charm]

I wish tho, that arma4 comes with ACE (especially the ballistics) and ShackTac by default, like when chenging difficulty to veteran/expert. there you go, it be long time to a4, but thats what I like to see in it ontop whats already in a3 :)

[desert trench]

is there a workaround of sorts for ctrlSetAngle when people use a custom fov?

[uncut briar]

no

[regal nimbus]

Nah, moveInGunner sets the effective commander to the gunner too. The only apparent difference with moveInAny is that they can't drive.

[desert trench]

we had also issues with moveInAny - need to be careful to get AI mixed with player into the vehicle in the right order, or you cannot command it (workaround is to do setEffectiveCommander as mentioned)

---

we are getting some locality issues(?) with mine/explosive boxes in DS env

they are defined as TransportMagazines as with vanilla A3 ammoboxes with them

even more weird some mine types work, some dont - is that known/a workaround/not TransportMagazines to be used?

[uncut briar]

Effective commander can get fucky too, AI commanding in general is weird.
https://feedback.bistudio.com/T170979

[desert trench]

couldnt find anything via google than some locality problems with ammoboxes in the past - not sure if this is still actual tho

[devout wave]

Could this be why Zeus-placed vehicles sometimes refuse to move if you place the vic and crew separately and then drag the crew into the vic? I don't know whether Zeus drag-and-drop uses moveInAny, but it sounds suspiciously plausible

[gaunt depot]

I might have a repro for position failing to transfer alongside locality

Been working with backpack UAVs lately and had them disappear twice on me on a local server

Made a repro, out of 100 assemblying, 1-2 are usually fucked

[regal nimbus]

The behaviour's 100% consistent with the commands so I'd be surprised. When I was testing the bridge bug I remember ~1/10 zeus placed vehicles refused to move anyway for some reason.

Also seen another no-movement bug that I think boiled down to createVehicle + setPos in the same frame.

IIRC in both the latter two you can move the vehicle slightly and then they regain the ability to path.

While the moveInAny bug doesn't care.

[gaunt depot]

https://feedback.bistudio.com/T171344

Repro made

Repro is unreliable when reporting failed vehicles, but you can see visually that some drones didn't update their position

1-2% failure rate

[uncut briar]

Sounds like it will be quite fun to debug

[regal nimbus]

man, that's an effort repro :D

[gaunt depot]

Nice position, 7km under ground

[uncut briar]

Wasn't there an engine thingy that's supposed to prevent that?

[gaunt depot]

What was it?

[uncut briar]

I mean going underground, there was a change that's supposed to move physx objects to ATL 0 if they go underground and their simulation is enabled.

[gaunt depot]

Hm, maybe its only when its first activated?

[regal nimbus]

no no, that can definitely spam :P

[gaunt depot]

In this case client has the drone in a proper position for a few frames, then (i guess) usual vehicle update message finally reaches them and puts it back into "out" position?

[regal nimbus]

had multi-gigabyte logs full of moving the same can back to z=0

Might only apply to uncontrolled objects though?

[gaunt depot]

🤔

Disassembling back and forth also messes things up quite a bit

This might be due to assemble\disassemble mess, vehicle reverting position after locality change is a global issue

Actually might be due to groups limit

[regal nimbus]

Any chance of an alt syntax for assignedItems that doesn't omit missing items? Because that command is an absolute arse to use for checks like "do we have a radio".

[sinful kettle]

what about getUnitLoadout?

is it too slow?

[regal nimbus]

ah yeah, meant to check that.

[sinful kettle]

I use that in my mods but it's definitely slower than assignedItems

having an alt syntax as you said would be neat

[regal nimbus]

~8x slower than assignedItems

[sinful kettle]

well if you have something like ACE that fills you with bandages and stuff I'm sure it'll get much worse 😅

[regal nimbus]

This is a pretty clean vanilla loadout :P

just too many arrays to construct I guess.

[sinful kettle]

yeah. just make a ticket 😅

it's reasonable

[daring wagon]

Absolutely agree.

[dusky wigeon]

Feedback tracker as in http://feedback.arma3.com/ :P

[gaunt depot]

1. Preload all radios from config
2. assignedItems player findAny _all_possible_radios >= 0

Not sure how much faster it is though

[regal nimbus]

About 7x slower than if it had the alt syntax :P

but I didn't know about findAny and it looks generally useful.

[gaunt depot]

getNumber(configFile >> "CfgWeapons" >> "ItemRadio" >> "type") => 131072
getNumber(configFile >> "CfgWeapons" >> "RadioItem" >> "type") => 611

Thought maybe we can have something like player getAssignedItemType NUMBER

to return item from a specific slot

But these types confuse me

131072 looks like pre-A3 bitwise inventory type

Oh yeah it is, but A2 didn't have any other number to distinguish these items from each other

only simulation

[desert trench]

https://community.bistudio.com/wiki/Arma_3:_Vehicle_Weapon_Switching

states you can define the type also per ammo

however this still doesnt seem to work for ammo of the same weapon - is that hardcoded to work only with helicopter and planes? can this be fixed to allow switching between tank ammo for example without the dreaded action menu

[gaunt depot]

What would be approximate ETA for 2.14 to go stable? Summer? Autumn?

[solid marten]

That could be improved

[azure crane]

I'll answer for the devs -

[solid marten]

https://community.bistudio.com/wiki/getSlotItemName

[gaunt depot]

nice

[uncut briar]

👍 Using numbers for slots is bit meh but I can live with it.

[gaunt depot]

#define SLOT_RADIO 611
player getSlotItemName SLOT_RADIO
```looks nice

[devout wave]

<#dev_rc_branch message> "late summer"

[gaunt depot]

https://feedback.bistudio.com/T171357

Made a ticket for missing UAV Terminal commands

If either KK or Dedmen are willing to implement these, I'll make repro scripts to test the commands.

UAV Terminal bugs deserve another ticket (mostly MP issues), but at least having BOOL = disconnectUAV ENTITY will let me bypass engine UAV issues

[neon swan]

#dev_rc_branch message

[solid marten]

parsing string and comparing string is meh in comparison with converting int to enum

[tropic jewel]

Could setTurretOpticsMode and setTurretOpticsMode perhaps be extended to work with pilot's turret path [-1], for Pilot Camera zoom?

[neon swan]

Turret optics mode does not necessarily include zoom, I'll double check but I doubt that will be easy to do without affecting zoom outside vehicles.

[neon swan]

Do you mean the pilot/driver cam that you can put where the GPS would be in the HUD? Or is there some pilot camera I am missing?

[devout wave]

I think they mean like the targeting pod camera some jets have

[neon swan]

Ah okay, I missed it was that view but it is the same cam you get in the GPS HUD.

[fathom slate]

read the header below the channel name, that helps

[outer spear]

prepare thy selves.. previously ignored FPS warnings about ShipX vehicles and roadways, is about to get real.

[fathom slate]

why hatchet ?

[outer spear]

gonna release the Mk.V to the masses. :D

[fathom slate]

oh....

just read the tickets, this is going to be fun :D

[outer spear]

i expect a fair bit of screaming..

[tropic jewel]

Nikko has it, PilotCamera class modes. The command currently forces a mode in Turret\OpticsIn, but that doesn't exist for drivers. Purpose is to support shenanigans like using the camera from the internal view, sim-lite style 😅

[neon swan]

I just need to test it a bit, but I worked on get/setPilotCameraOpticsMode commands

It was easier than cannibalising the turret version, and it makes more sense.

[tropic jewel]

Ft ticket was actually made ages ago xD https://feedback.bistudio.com/T169615

[rugged nova]

Hey Dedmen what ever happened to the attachTo disableCollision parameter?

#community_wiki message

[untold sky]

nothing

[stone tree]

https://feedback.bistudio.com/p/violentlesson/

Spam account

[unborn flame]

http://feedback.arma3.com/view.php?id=26380 export option for config viewer

[unborn acorn]

https://feedback.bistudio.com/T171375#2427082
ho no spammer

[stone tree]

https://feedback.bistudio.com/p/jhonallen/

Spammer

[gray wharf]

all 3 removed, thanks

[alpine tulip]

17:55:28 Warning Message: Error: creating weapon CUP_arifle_Mk17_Base with scope=private
Result of compatibleMagazines for scope = 0 weapons, even though the array is returning correctly

[gaunt depot]

Just realized how useful it would be to have a unit event handler that would fire each time unit wears a new item

including the slot

[alpine tulip]

eg InventoryChanged?

[gaunt depot]

InventoryChanged sounds more broad, something like InventorySlotChanged

params ["_unit", "_slotNumber", "_oldItem", "_newItem"];

Something like that

[alpine tulip]

Oh I like

Not sure _slotNumber is mandatory

[gaunt depot]

Well it could save looking up slot type in config

[alpine tulip]

True

[gaunt depot]

Plus some items like ItemRadio have incorrect type in config

Gonna save us some additional switch cases

With _slotNumber you'll be able to easily run code against item types you're interested in

[uncut briar]

yes pls so we can stop polling player inventory in CBA all the time.

[gaunt depot]

Performance gains! 💪

[sinful kettle]

is it supposed to trigger every time you fire?

[gaunt depot]

Are magazines even inventory slot items? 🤔

[uncut briar]

why it would fire

[gaunt depot]

If they are then probably on reload?

[uncut briar]

yes on reload.

[sinful kettle]

oh you're talking about those?
I thought you mean getUnitLoadout

[gaunt depot]

Ticket done: https://feedback.bistudio.com/T171418

[uncut briar]

I've linked CBA implementation, it's limited and works only on player for performance reasons.

https://github.com/CBATeam/CBA_A3/blob/master/addons/events/fnc_playerEvent.sqf#L43-L48 it needs to do this every frame 💀 and then a bunch of comparisons.

[gaunt depot]

Personally I need it for checking when UAV Terminals were swapped. Since UAV Terminals store forbidden UAVs (disableUAVConnectability) and current connected UAV and they're not entity, its a bit of a hassle to know when your terminal changed.

[gaunt depot]

oof, it builds and compares entire getUnitLoadout each frame

[uncut briar]

[uncut briar]

I see you want just the slot changes, imho whole inventory support would be better.

[gaunt depot]

https://community.bistudio.com/wiki/getSlotItemName
Weapons are not part of this list so I guess either new EH gonna have to be expanded in its scope, or another one introduced if engine is not spaghetti enough to call EHs on weapon\attachment\magazine changes

[gaunt depot]

Yeah, just thought of that, either expand this suggested EH or have a new EH for different kind of inventory items

InventoryWeaponChanged ?

Unless you also need each and every item in the uniform, vest and backpack tracked

I only thought about equipped items

I see that you only use equipped weapons in getUnitLoadout so I guess we thought of the same

Weapons are types 1, 2 and 4 and don't seem to collide with slot items so perhaps EH could also include swapping the weapons?

Weapon attachment types doesn't seem to collide either, but more info is needed for them for EH to make sense

[untold sky]

much requested over the last years, pretty sure we still have a internal ticket open for it
Sadly inventory system is a mess

[gray wharf]

plz fix 2.14 kthxbai

[untold sky]

Also wouldn't help with the CBA problem, it still needs to check container contents

Easiest way might be to move CBA's each frame unit loadout check, into engine
But even that would be crappy

[uncut briar]

Still a lot faster

[regal nimbus]

Any chance of a command to disable specific vanilla actions from appearing? There are hacks that can be done in some cases but most of those are horrible, and some actions are hard (impossible?) to clear without config.

[desert trench]

Ticket is here: https://feedback.bistudio.com/T158795

technically this should be quite simple i would assume as its already determined via show parameter in cfgActions - only if the engine caches the data on game start, it might be a little hassle

nice to have would be the ability to manipulate all parameters for actions in there, yet probably just show on/off would be good enough

[gaunt depot]

Even if its limited to one kind of inventory types, items (headgear, radio, gps, uniform, goggles etc.), it would still be very useful

A week ago somebody was asking for a way to find a moment when goggles(mask) was taken off a dead body, such event handler would've solved it

My requested parameters are theoretical, even if its not possible to know last item's class it would still be very useful

[gaunt depot]

#arma3_scripting message

[gaunt depot]

Here is how I see disabled and enabled actions could work:

BOOL = ENTITY disableEngineAction STRING; (CfgActions: Assemble, GetInDriver, etc.)
BOOL = ENTITY disableConfigAction STRING; (UserActions: OpenDoor_1, etc.)
BOOL = ENTITY disableUserAction NUMBER; (addAction command)
BOOL = disableEngineActionUI STRING; (CfgActions: Assemble, GetInDriver, etc.)
BOOL = disableConfigActionUI STRING; (UserActions: OpenDoor_1, etc.)

disableEngineAction stops entity from giving units its actions
disableEngineActionUI stops UI from displaying all actions of this kind all together

BOOL = disableConfigActionUI STRING;
```sounds strange though, not sure if this is worth having at all, you can't disable all door openings for example, some door open action might have config name completely different from BI standard

And opposite enable*Action commands

Or one command with a flag in parameters: [STRING, BOOL (Enable/Disable)]

BOOL returned would be command success - such engine action existing, etc.

I thought about suggesting it to be even more complex by disabling actions in a format of [CALLER, TARGET, ACTION] but this might be too complicated if you say want to disable all actions all together

player disableEngineAction [jet, "GetInPilot"];

But you'll have to re-add this disabling for player each respawn, have to add it against all vehicles, etc.

Can't supply objNull as a mask for "every vehicle", since you might provide a variable that went null and it would disable all actions by accident

So I'd keep it simple, as most usual usage for these commands would be disabling bullshit actions and implementing your own handling

Alt name for disableEngineAction would be disableGameAction

Game, Config, User

[west onyx]

how about just a uinamespace var for each of the cfguseractions or something

I also agree a way to hide engine actions would be useful for advanced scripting

not everyone uses huge mods

[sleek scaffold]

Lots of modders have been asking for this, hopefully it gets implemented one day 🙏

[west onyx]

just something simple like uinamespace setvariable [format ['BIS_hideaction_%1',configname ...]];

[stone swift]

I created a feature request for marked SensorTarget but maybe it is a bit excessive https://feedback.bistudio.com/T170978. Maybe i could get an additional return to getSensorTargets instead something like Null, Marked, Locked?

[stone swift]

Ahh... Nevermind i found out that cursorTarget provides that functionality.

[desert trench]

seems some spam wave happening

https://feedback.bistudio.com/feed/query/sOoh4cB96RJf/

[stone tree]

https://feedback.bistudio.com/p/Servis_shoes/

[solid marten]

Just realized how useful it would be to

[regal nimbus]

compatibleMagazines is case-sensitive on the muzzle name. Modders are incapable of getting cases right for their own classes. Whose fault is this? :P

Given that compatibleMagazines actually fixes cases for the magazines & magazinewells, this seems like an oversight?

[gaunt depot]

Script fix until its fixed in the engine:

compatibleMagazines ["arifle_Katiba_GL_F", configName(configFile >> "CfgWeapons" >> "arifle_Katiba_GL_F" >> "EGLm")];

[regal nimbus]

yeah, already did it

[desert trench]

wasnt there some fix or workaround in 2.12 or 2.10 already to have AI no longer shoot at empty static weapons?

[devout wave]

Spam:
https://feedback.bistudio.com/T171439
https://feedback.bistudio.com/T171424

[devout wave]

At the moment, enableIRLasers doesn't seem to work with optic slot lasers. Is it reasonably possible that it could be made to do so?

• this might also apply to the gunLights command and optic slot flashlights, but I haven't had the chance to test that just yet

[devout wave]

I did some more investigation and it turns out the truth is actually much stranger: visible-light laser beams don't glow in the dark and become completely invisible at night

[unborn acorn]

_myBuilding = (nearestBuilding position my_bestPlayer);
{
    _myBuilding setHitPointDamage [_x, 0.5];
}
forEach ["glass_1_hitpoint","glass_2_hitpoint","glass_3_hitpoint","glass_4_hitpoint","glass_5_hitpoint","glass_6_hitpoint","glass_7_hitpoint","glass_8_hitpoint"];

Dedicated Server
if _myBuilding is Eden Editor placed building this code will be exec successfully in initServer.sqf

but if _myBuilding is Terrain Object code will no any effects if exec it in initServer.sqf
Although if code will be in init.sqf the glasses will damaged

I'm not sure if this behavior is normal?

If this is a problem, it might be related to
https://feedback.bistudio.com/T133265

[alpine tulip]

getText (configfile >> "CfgWeapons" >> "arifle_AK12_GL_lush_F" >> "baseWeapon")->"arifle_AK12_GL_khaki_F"
Sanity check, am I the only one?

[gaunt depot]

Yep, I also have it like this

Looks like arifle_AK12_GL_lush_F was arifle_AK12_GL_khaki_F at some point

[solid marten]

@gaunt depot how do you remove goggles and headgear?

or to be precise what mass remove command removes them? Because they are not part of removeAllAssignedItems

[rapid badge]

@patent trail : From what I've read apparently Enfusion does (or is supposed to) support "upper half separate from lower half" better... or is this PurePassion not the BIF's? https://www.reddit.com/r/dayz/comments/3lmk4p/can_you_please_remove_the_thing_which_makes_your/cv7r4ea

[gaunt depot]

Looks like I ended up using removeGoggles

Can't find a command that mass removes the googles

[solid marten]

what a mess

[alpine tulip]

What is this window after all? It is very frustrated to close this whenever it opens while I tweak the changelog

[solid marten]

unification is incoming

[solid marten]

apparently addWeapon removeweapon works with all assigned items, always worked

what a mess

[solid marten]

removeAllWeapons has nothing to do with removing weapons, and the actual weapons removal looks like side effect, but the command is geared towards removing all magazines, no idea who what why

[regal nimbus]

Fun with sendSimpleCommand: key down + right + key up + key up results in you turning left, and vice versa :P

Wondering if this is partly why the AI is so bad at turning, but it's probably five other reasons.

[sinful kettle]

no. that command is pretty much useless. it also has a lot of delay

[devout wave]

Maybe it was supposed to be a complete disarm, i.e. remove all weapons [and all ammo]

[solid marten]

judging by version history it looks like that

[gray wharf]

it was that in OFP at least - setCaptive + removeAllWeapons

[limpid rune]

Yeah I tried to use it to micromanage tank facing a bit and it basically meant AI would be turning left and right constantly because the delay on the 'stop' was too high

was extremely frustrating to have something that was almost useful

[unborn acorn]

https://feedback.bistudio.com/p/pretzelsalienated/
spammer

[dreamy bane]

wouldn't it be cool to have real try catch system that is triggered on script errors? Unless I'm wrong sometimes script errors are not reported at all like in eachFrame EH for example

[gray wharf]

you are right about them not being reported in some unscheduled environment

having a try-catch system is not the way to go for performance so no (and I believe the way SQF is done you can't do that (easily?))

[dreamy bane]

yes it probably would have some effect on performance. depends on how its implemented though.

but you could use macros to enable the try & catch only for testing 🙂

[gray wharf]

just don't make mistakes

[dreamy bane]

lol

[gaunt depot]

Using nils as operands is not reported in unscheduled environment

I actually use that as feature in few places

_object = hashmap get "object";
if(!isNull _object) then {
    // something with the _object
};
// This line also works

Ends up working as both isNil and isNull check

instead of doing hashmap getOrDefault ["object", _objNull]

saves a grain of performance

[solid marten]

is there event handler for adding removing uniform, vest, backpack containers? The SlotItemChanged is firing for those because getSlotItemName supports those, but no reason to duplicate if there is EH already

[alpine tulip]

I think no, never heard of

[gaunt depot]

Nope

[tropic jewel]

What are the chances snapping info can be added by config or global var hashmap, if provided by community?
Would need something like:

snapPoints[] = {
    {-1, 0, 0},
    {1, 0, 0},
    "snap_mempt_1"
};
snapClass = "BaseClassName"; // used to look for compatible objects.
snapAngleIncrement = 90;

https://youtu.be/Ial9o7p6q00

I really feel this would be ideal in base game. And if mod, would want it to be in both 3Den and Zeus Enhanced xD

[alpine tulip]

How the cargo platform/museum parts configured? I'm interested

[tropic jewel]

they have mem points, and the function is hard coded for those. \a3\functions_f_enoch\CargoPlatform\fn_CargoPlatform_01_update.sqf
This is just numbers.

[alpine tulip]

Hmm

Softcoded solution is nice indeed

[daring wagon]

Oh no. Stop giving me ideas

We just need a hash map with classes as keys and snap data as values

Just checked the code. That's already the way you do it.

[tropic jewel]

For a lot of things, if the bounding box is tight, just the corner or mid points can be snap points

[tropic jewel]

Posting here was more about how to best make the data and functionality available.

[desert trench]

Any chance to get these picked up?

1. Performance optimization - we use the sound object for various audio features but the supposed "search free area" makes the local variant quite heavy.

[Feature Request] createVehicleLocal with newer "createVehicle array" syntax
https://feedback.bistudio.com/T170615

2. For better usability we are providing various custom (advanced) hints, but also missions or 3rd party use hint. The getter would allow a flexible system (as the Hint key action already allows you to display the last hint, this should be easy to read and expose the last/current hint)

[Feature Request] currentHint
https://feedback.bistudio.com/T170257

3. Performance optimization - often features are group based instead of unit based. If you need to build the group list manually in sqf frequently (ie per frame), this gets fairly expensive fast with enough groups/units.

[Feature Request] SQF command nearGroups
https://feedback.bistudio.com/T170144

[desert trench]

Dedmen is it possible your fix for "[BUG] Unknown enum value spammed in rpt for waypoints" is leading to this reordering? is that a problem/something that needs attension at all?
ref: https://feedback.bistudio.com/T171125

[untold sky]

Previously there were two type entries, the later one overwrote the earlier one.
The earlier one was bogus and was removed. So yes, now it appears lower down.

[desert trench]

👌

[gaunt depot]

A thought about nearGroups and nearLeaders: How about making inArea\inAreaArray\inAreaIndexes work with groups?

[untold sky]

probably easiest idea. Just need to add parameter whether any group member, or only leader or average center position of group or all members n such

[gaunt depot]

Then you could do:
allGroups inAreaArray _trigger

[untold sky]

When I saw that request I thought that was already a thing

[gaunt depot]

leaders ARRAY to return leaders of provided groups

allGroups inAreaArray _trigger => groups if at least one unit is in area
leaders allGroups inAreaArray _trigger => groups if leader is in area

of provided units too, perhaps

leaders allUnits => all leaders

[untold sky]

leaders allGroups inAreaArray _trigger => groups if leader is in area
won't work, leaders would return array of units, so inAreaArray cannot return groups if it got units

[gaunt depot]

Is there allGroups SIDE like units SIDE? 🤔

[devout wave]

groups side

[gaunt depot]

True, won't work

Thought leaders could be useful to turn array of units into array of group leaders

So results of say nearEntities can be quickly turned into leaders

groups ARRAY, to turn list of units into unique groups

groups(leaders allGroups inAreaArray _trigger) will make leaders into groups in area then

[untold sky]

alt syntax for "leader" sounds easy.
adding new leaders command feels too feature creep

[gaunt depot]

Is there a monthly quota for new commands provided by BI? 🤔

Thought seriously leader ARRAY is fine too

[untold sky]

next we'll have drivers and gunners and commanders with array of vehicles and such more stuff. If we work like that

[gaunt depot]

drivers actually sounds useful, gunners and commanders not so much. effectiveGunners maybe.

I'm fine with alt syntax for existing commands

Question is, does it satisfy @desert trench ?

[untold sky]

If performance like that was the main factor, we could apply optimizations like I did on configClasses

If you run
_x apply {leader _x}

Don't... re-run that script for every element.
Know on engine side what the script wants to do, and do it right there and then without invoking actual script function.

I would much prefer that over dozens or hundreds of new tiny special usecase commands who's only purpose is to do something slightly faster than what you could already do.
These commands don't add any new functionality, they just do stuff you can already do, faster.

Actually I think it would be extremely easy to optimize the

_x apply {command _x} case, and just evaluate it directly in engine, without running script
I can try that in a bit

[gaunt depot]

🤔

[uncut briar]

[gaunt depot]

How about both commands and hardcoded optimisations for applying single commands?

[gaunt depot]

Speaking of average center position, having a command for that sounds useful too averagePosition GROUP/ARRAY

allGroups apply {averagePosition _x} inArea...

[untold sky]

input apply {command _x}

in engine is
foreach element in input
Create new script stack level (memory allocation)
Add local variables (_x, foreachindex, ...) (memory allocation each)
Return to interpreter
grab next instruction
getVariable instruction
find _x variable, push on stack
return to interpreter
grab next instruction
command instruction
grab value from stack
run command
put result back on stack
return to interpreter
get next instruction
end of expression instruction
clean up stack
return to interpreter
get next instruction
none found, break out of stack level
clean up stack level and local variables
newArray append value returned from stack level

I can optimize that into

if script is single command
foreach element in input
newArray append (singleCommand(element));

[uncut briar]

I guess something similar could be done for select?

[gaunt depot]

Sounds great, maybe have it for other _x commands too?

[untold sky]

yeah sure

[gaunt depot]

count, select

[untold sky]

I'll try and benchmark that when I'm done with the texture stuff

Btw I'll trust that you'll find issues if I broke something with that UIToTexture stuff.
The shared display between multiple different textures is now in

[gaunt depot]

I'll have a good playing around with it

[gaunt depot]

I actually thought about something similar some time ago, thought it could work like: ARRAY applySingleCommand STRING with STRING being SQF command

[untold sky]

We can parse assembly no need to be string

string would actually be more expensive, than passing code and parsing assembly

because we don't have to lookup the command, that's done on compiletime (except for nular commands on non-profiling branch)

[gaunt depot]

🤔

[untold sky]

aw no dev branch this week actually

[untold sky]

Actually in that listing I forgot type checking, and different command versions for argument types.
But that will just be added on both ends, we still save all that stuff listed there

Most important part, we need to find a name for this process 😄
How about short-circuiting 🤔 Or direct evaluation

[gaunt depot]

Blast Processing

The Real Virtuality has blast processing, the Enfusion engine doesn't!

[regal nimbus]

Any chance of extending this optimisation to binary commands?

You could do pretty much anything with it by chaining applys then :P

[untold sky]

I'll probably support any command, but just one command deep.
combining with the bytecode optimizer I may also support const array arguments to commands

or stuff like
_x apply {_x command [_x, ...]}
but that might get dangerous if command is modifying _x

huuuuuuh wtf

omfg how can I be so stupid

Back in january I did Tweaked: Improved performance of nular commands and variables
that changed that stuff like true would be looked up at runtime if its a variable or a command, to look up once at compiletime.
But..
The feature flag for it is
#define _A3_NULAR_INSTRUCTION (0 && _EXPERIMENT_BUILD)
It has been off this entire time while I thought we were testing it on prof

Result: 0.0013 ms
Code: true;true;true;true;true;true;true;true;
->
Result: 0.0009 ms
Code: true;true;true;true;true;true;true;true;

[gaunt depot]

🤔

[untold sky]

I'm trying to test the thing with select {true} and wondering why I get a getVariable instruction 😢

Question is also.
Can i turn

input select {command}
into
_result = command;
input select {_result}

without any problems 🤔 There is probably some weirdo edge case, like diag_tickTime slightly increasing between each, even though its unscheduled

[gaunt depot]

This is beginning to look like SQF dark magic

[untold sky]

I think this looks promising
in debug build with select {true}
[2.12979,470] short circuited
[33.6,30] normal

Ignore the 1; I just use that as a marker to enable short-circuiting, its not executed.

_bigArray = [];
_bigArray resize 10000;

[
diag_codePerformance [{_this select {true}}, _bigArray],
diag_codePerformance [{_this select {1;true}}, _bigArray]
]
``` `[[2.86819,349],[0.158003,6329]]`

```sqf
[
diag_codePerformance [{_this select {false}}, _bigArray],
diag_codePerformance [{_this select {1;false}}, _bigArray]
]
``` `[[2.73973,365],[0.127259,7858]]`

Interesting that returning 10k elements, doesn't actually take much more time than returning 0 elements.

That above is
```cs
    foreach (value in input)
      if (executeCommand())
        result.Add(value);

Now if apply dark magic

if (executeCommand())
    foreach (value in input)
        result.Add(value);

We get

[
diag_codePerformance [{_this select {true}}, _bigArray],
diag_codePerformance [{_this select {1;true}}, _bigArray]
]
``` `[[2.68097,373],[0.0367539,27208]]`

```sqf
[
diag_codePerformance [{_this select {false}}, _bigArray],
diag_codePerformance [{_this select {1;false}}, _bigArray]
]
``` `[[2.66223,376],[0.00038,100000]]`

I would say.....

That might have been a brilliant idea

What is that.. like 90x faster?
Its pretty rare that I can apply dark magic though
You don't really do select when the condition is always the same 😄 (And if you do, SHAME!!! 😠 )

[rose cloak]

inb4 chaining 484 applies because it's faster than single 4-instruction one

[outer frigate]

dedmen chanting around his computer with a massive cloak on sacrificing miller to the sqf gods to get 90x better code performance

[untold sky]

One thing here is also. Every script command executed, checks against CfgDisabledCommands, every time, before every execution.
With this, I can do this check once before the loop, instead of repeating every loop iteration.

For bigger pieces of code, I can also move out all the local variable stack (if you do not use local variables) and just run the commands without a stack just passing values between each.
For big pieces of code (more than one command) I probably can't do such extreme optimziation, but I can still get it down quite a bit

So what was the other thing
arrayOfGroups apply {leader _x} ?
I'll try that

[untold sky]

_bigArray = [];
_bigArray resize 10000;
_bigArray = _bigArray apply {group player};

[
diag_codePerformance [{_this apply {leader _x}}, _bigArray],
diag_codePerformance [{_this apply {1;leader _x}}, _bigArray]
]
``` `[[3.60072,278],[0.677048,1477]]`
yup

Compared to this

[
diag_codePerformance [{_this apply {leader _x}}, _bigArray],
diag_codePerformance [{leaders _this}, _bigArray]
]
``` `[[3.5914,279],[0.523286,1911]]`
Yeah, not really worth it building special case commands, if I can just solve the general case.
And I think I can make that general a bit faster still

[untold sky]

The performance difference to leaders command is, it doesn't do type-check on each element (though a real implementation should, I didn't do t hat here)
but any unary command that has multiple overloads, needs to run typecheck to figure out which overload to call (That's why I prefer to merge things into existing commands, without creating overload)

I think this is the max I can reasonably get out of that

[
diag_codePerformance [{_this apply {leader _x}}, _bigArray],
diag_codePerformance [{_this apply {1;leader _x}}, _bigArray]
]
``` `[[3.71111,270],[0.635324,1574]]`
I optimized the type checking a bit, but there's only so much to be done.
If I remove the type checking and just hardcode call the correct overload

`[[3.7037,270],[0.593472,1685]]`
But I can only do that if I know all elements in input array have the same type

I think its fast enough, close enough to having a "leaders" command

Even at array size 1 its good [[0.00089,100000],[0.00051,100000]]

[desert trench]

Outstanding 👏

[untold sky]

looks like no binary commands and no arrays for now

[untold sky]

neat.
This stuff also knows before it goes into the loop if _forEachIndex is used, so it doesn't need to spend time setting it if you're not using it

huh what? apply has no forEachIndex? :u

select also doesn't.. huh.. didn't know that

findIf also doesn't? can't be :u I must be missing something

wow no they really don't.
Welp, learned something new I guess

[regal nimbus]

lack of forEachIndex in apply is occasionally annoying.

So if we could have that without the perf cost when you're not using it then that would be good.

[sinful kettle]

yeah having a magic var for that would be nice

can't be _forEachIndex tho due to backward compat

[regal nimbus]

oh, because you might be using apply inside a forEach?

[untold sky]

Nice.
First test with the more generalized version
It decided to optimize a lookup in CBA's preStart

private _cfgPatches = configFile >> "CfgPatches";
private _allComponents = "true" configClasses _cfgPatches apply {configName _x};

Without/With

[
diag_codePerformance [{private _cfgPatches = configFile >> "CfgPatches";
private _allComponents = "true" configClasses _cfgPatches apply {1;configName _x};}],
diag_codePerformance [{private _cfgPatches = configFile >> "CfgPatches";
private _allComponents = "true" configClasses _cfgPatches apply {configName _x};}]
]

[[13.9167,72],[13.1429,77]] 😢

Oh I forgot to apply a optimization in my old configClasses thing

[[7.65649,131],[6.86986,146]]
ohhh crap. Welp. That's what adding a reserve(n) and a std::move does to you....

Moving out the configClasses, and only testing the apply itself
[[0.962464,1039],[0.142045,7040]]

[untold sky]

[[3.82509,263],[0.757576,1320]]
The generalized version performs a bit worse than my special initial implementation 🤔

[untold sky]

Buuut. It can do fancy stuff like apply {leader group leader _x}!

[[3.8876,258],[1.60514,623]]

or leader group leader group leader group leader group leader _x
[[9.50943,106],[4.07317,246]]
Mh its only 2x tho 😢 Needs to be better 😠

yeah unaries are expensiv due to the type checks...

apply {true}
[[2.87931,348],[0.244499,4090]]
nular's get you the 10x

Ah I know how to optimize the typechecks.. tomorrow

[sinful kettle]

yeah

[untold sky]

old type comparison check

Are we any?
Are they any?
Are we null? (should never happen, could be checked once at game start, but is instead re-checked every time on every type check on every unary/binary command execution and also often inside commands if the argument is array)
Are they null? (should never happen)
If we are single type
Are they single type? Quick compare
They are multitype, look up if their typelist contains our type, in an unsorted array that we just O(N)
If we are multitype
Are they single type? Check if their type is in our unsorted typelist
They are multitype, O(n²) iterate over all elements to see if any match together

New type comparison check

(int & int) != 0

That's gonna be interesting.

[regal nimbus]

raises eyebrow

maybe one day SQF will be fast enough to write stuff in it :P

[uncut briar]

Add PBR renderer to RV, let dedmen work on it 1-2 years more and it will be viable for Arma 4

|| and we can forget about that Refunder thingy ||

[near cedar]

So uh to steer us back into feedback tracker territory for a second 😛

There was a discussion that came up in the GM discord a week or two ago regarding the behaviour of AI fired SACLOS missiles when the gunner of such a missile dies

Since AI-fired SACLOS missiles are really "fire and forget", killing the gunner means the missile still tracks and hits the target when it ideally shouldn't

[untold sky]

AI related stuff
fun

[uncut briar]

make a FT ticket with repro.

[untold sky]

AI related stuff is unlikely to be fixed

[near cedar]

i am aware how to give feedback, hold on 😄

[untold sky]

sometimes does happen though for major enough things. Which I think this is not

[uncut briar]

What you want would need most likely an CfgAmmo flag that tells the game to stop the missile from tracking the target after the firer dies.

[near cedar]

(I couldn't think of a proper name for the ticket anyway, which i also wanted to ask about)

[uncut briar]

this behaviour should be quite easily scriptable too.

[near cedar]

my idea was somewhat related, i was wondering if it could effectively trigger the "countermeasure" response

i.e. behave as if the target had popped smoke

(now i don't know what exactly that mechanism is, maybe it's the cfgammo flag that you're talking about)

So should i bother making a ticket about it, and if so what should I even call the ticket.
"AI fired SACLOS missiles don't lose track after gunner dies"?

[uncut briar]

"AI fired SACLOS missiles don't lose target track after gunner dies"
sounds good, IMHO worth a ticket.

Might turn out it's actually simple.

Ideally attach a mission that makes it easy to debug for the devs.

[near cedar]

I don't have the ability to attach a mission sadly, my gaming pc is on the other side of the planet

i'll add the repro in summer, or maybe someone else volunteers (perhaps the person that brought it up originally in GM discord)

[near cedar]

Also yeah probably not the most important thing but was hoping it turns out somewhat simple to do

would improve immersion noticeably i think

[gaunt depot]

Should be _findIfIndex, _applyIndex, _countIndex

[desert trench]

As someone (Sa-Matra/Dedmen?) had asked for/suggested to request yesterday specific requirements in regards to performance

Obviously in general anything with perFrame, especially UI stuff, 2d map or 3d visualization where you need/want realtime to avoid lag or false information
Combine that with groups (50+) or units (100+) - even basic sqf commands add up if you have to do checks if that group/unit is relevant
(for custom HUD, target indicator, nametags, etc)

Hence any prefiltering the engine can do, or even did already, can save a ton of computation
Or to be able to use already present engine native systems (aka "only" getter&setter needed)

Samples:

• [Feature Request] SQF command to return visible targets for an unit: https://feedback.bistudio.com/T168404
• [Feature Request] Extended drawIcon/drawIcon3d with isVisible parameter: https://feedback.bistudio.com/T169233
• [Feature Request] SQF command to return only vehicles for a given side: https://feedback.bistudio.com/T168217
• [Features Request] SQF command to toggle marker scaling on and off per marker: https://feedback.bistudio.com/T169166
• [Feature Request] Add drawUnitIcon and family of SQF commands akin to GroupIcon: https://feedback.bistudio.com/T164042
  (or that nearGroups/nearLeaders)

[west onyx]

“players [east,west]” to return list of sided players

[gaunt depot]

sides of units and vehicles is another can of worms

Unit side can differ from unit's group side for example, plus there is unit's config side

Which should it return?

[gaunt depot]

drawIcons can be be optimised quite well with new inAreaArray command

You need to update a location each frame to fit map's size then you run all your potential draw candidates against it with inAreaArray

[untold sky]

wtf did I just find.
When game tries to create value of type Array.
It goes

Ok you want array?
Good, I'll iterate over all types in this unsorted array, where each element is a pointer reference to a location that is most likely not in CPU cache and needs to go to ram, create a temporary copy of each type!?!?!?!?!?!?, find the one that == the type you passed in.
Then when I found it, use that to create the value.

DUUUUUDE, The type is already there! Just use it! OMFG what the hell.

[gaunt depot]

SQF performance gains month?

[alpine tulip]

Dedmen of the day I guess

[gray wharf]

Enfusion is cancelled guys, Arma 3 just got +9001% perf!

[alpine tulip]

Over 9000?!

[gaunt depot]

👆

[alpine tulip]

Can't we just invest Dedmen DLC?

[gaunt depot]

I'll buy 3

[gray wharf]

Sega®

[gaunt depot]

SEGA!

[gray wharf]

what have I done now

[untold sky]

wow OFP had script types for Vector, Matrix3, Matrix4 🤔
Seems they were never used tho

[gaunt depot]

Deemed too hard for us SQF plebs I guess

[untold sky]

// this return same result as (bool)(this & with)
Followed by a 60 line function

[gray wharf]

"why make it simple when…"

[untold sky]

So this again.
Before that type comparison change: [[9.87255,102],[4.21941,237]]
After (Only applied to the short-circuited): [[9.87255,102],[3.82824,262]] (lol I reran it till the first value was close to tthe old one, and I hit it exactly)
10% neat

Applied to both sides [[9.90099,101],[3.89494,257]]

[daring wagon]

@gray wharf We should document these findings on the biki. For historical reasons and to have something to smile About in the future

[gray wharf]

I want to fight depression, not feed it