Broadcasters best practices
-
@d-healey Since we're talking about function pointers, I'm guessing that if you don't clean up after a child panel is removed that you will crash (not when it's removed, but when called).
-
@Christoph-Hart said in Broadcasters best practices:
having listeners attach to at various places.
Sure, but where do you declare the broadcasters?
Let's say I have a namespace called Downloader and I have another namespace called Uninstaller.
From the Uninstaller namespace I want to know if a download is in progress - because I don't want to be uninstalling things while other things are downloading.
Do I create an isDownloading broadcaster within the Downloader namespace? - this would mean the Uninstaller namespace has to be "aware" of the Downloader namespace. Or do I create a middle-man namespace where the broadcaster is declared and let both the Downloader and Uninstaller namespaces talk to that? This would keep the namespaces independent of each other. Or is there another solution?
-
@d-healey The whole idea is that it's a sort of dynamic callback that you can add to from wherever.
I have a control helper class that has its own value change broadcaster, which is attached to all controls by default.
As I add functionality, if I want things to happen on value change, I just call .addListener and add a function. If I remove that class, no worries, the broadcaster is still there and all other classes are still benefiting from it.
You just gotta watch the order of declarations.
I first include my script library, then it's ColorPalette, TransportHandler, Animation, PaintRoutines, and then it's ControlHelpers.
Getting on top of broadcasters would be my first suggestion to anyone doing code in HISE, as it's the class that makes it all come together.
-
@d-healey said in Broadcasters best practices:
Do I create an isDownloading broadcaster within the Downloader namespace? - this would mean the Uninstaller namespace has to be "aware" of the Downloader namespace. Or do I create a middle-man namespace where the broadcaster is declared and let both the Downloader and Uninstaller namespaces talk to that? This would keep the namespaces independent of each other. Or is there another solution?
You declare the isDownloading broadcaster in your Downloader namespace. It simply sends out isDownloading to all of its listeners whenever the download changes. You can even pass that broadcaster object into the function for a callback.
Then in your Uninstaller, you can check if it's defined, and if so, attach functionality to it. That way your modules are still independent but get "upgraded" if other modules are present.
I don't know if broadcasters can be defined dynamically, but you can add listeners dynamically. So unless you have something important to do in on init, you can always add user functionality listeners with a delay (Content.callWithDelay) of 500ms or so, that way you don't really care about module import order as all base broadcasters will already be defined by then.
-
@aaronventure said in Broadcasters best practices:
The whole idea is that it's a sort of dynamic callback that you can add to from wherever.
Yeah I know this, I've been using broadcasters since they were a thing, but I'm not happy with how intertwined my code becomes.
@aaronventure said in Broadcasters best practices:
You declare the isDownloading broadcaster in your Downloader namespace. It simply sends out isDownloading to all of its listeners whenever the download changes. You can even pass that broadcaster object into the function for a callback.
Then in your Uninstaller, you can check if it's defined, and if so, attach functionality to it. That way your modules are still independent
This is the part I'm missing. How is the Uninstaller independent of the Downloader in this scenario? It still needs to be aware that there is such a thing as a Downloader namespace, even if just to check that it exists.
I think what I need is a script wide broadcaster handler. So I just say "Hey HISE, does a broadcaster with this ID exist? I don't care where it is, just tell me if it exists or not and give me a reference so I can listen to it". That would allow all scripts to be completely independent of each other. @Christoph-Hart This should be possible, no? Since the broadcaster map has this info.
In my current project I have a namespace called Header which has a logo on it. When the logo is clicked I want an about page to popup. The about page is handled in an About namespace.
What I've been struggling with is do I just put a callback in the Header namespace and call
About.show()
- which is a nasty solution. Do I create a middle-man broadcaster - which seems like over kill, or do I create a broadcaster in the Header namespace purely for the About namespace to listen to - which seems wierd. -
@d-healey said in Broadcasters best practices:
What I've been struggling with is do I just put a callback in the Header namespace and call About.show() - which is a nasty solution. Do I create a middle-man broadcaster - which seems like over kill, or do I create a broadcaster in the Header namespace purely for the About namespace to listen to - which seems wierd.
The header gets a mouse click broadcaster. About says if (isDefined(Header.clickBroadcaster)) Header.clickBroadcaster.addListener().
If you don't want to worry about include order, call this with a delay, because it's user functionality.
-
@aaronventure said in Broadcasters best practices:
About says if (isDefined(Header.clickBroadcaster))
Right this is the issue. I might as well just use About.show() in the button's callback in the Header than go to the trouble of creating a broadcaster.
-
@d-healey said in Broadcasters best practices:
I think what I need is a script wide broadcaster handler.
I guess this would be functionally the same as using a dedicated broadcaster namespace actually.
-
@d-healey But why? Your About namespace is a cool thing on its own, but if it's added to the same script that has the header, it'll pop up when the header is clicked.
The only thing you'll have to lock down is how you name your namespaces.
-
@aaronventure But why should the about namespace know or care if there is a thing called Header? The only thing it needs to know is should it show the pop-up. And perhaps in the future I will want other triggers to show the About pop-up that aren't in the header.
I really like code to be as portable and self-contained as possible. It makes future extension and maintenance easier - usually.
-
@d-healey said in [Broadcasters
I really like code to be as portable and self-contained as possible. It makes future extension and maintenance easier - usually.
Makes sense but if you write two modules which need to listen to each otherโs broadcaster then they are not self-contained and blaming the broadcaster is just shooting the messenger - literally :)
-
@Christoph-Hart Sure, but let's say both namespaces are aware of a third namespace called Broadcasters. They can both check if there is an isDownload, or showAboutPopup broadcaster there and do whatever they need to do. The implementation of the broadcaster is totally independent and is therefore easily transferred to other projects or used in different contexts.
If the Header is looking for a specific About namespace or vice versa then they are more coupled than if they are looking for a generic Broadcasters namespace.
I think what I'm closing in on is having a dedicated Broadcaster namespace. I'll have to play with it and see if I like the workflow.
-
@d-healey said in Broadcasters best practices:
I think what I'm closing in on is having a dedicated Broadcaster namespace. I'll have to play with it and see if I like the workflow.
Actually that's what I've been doing, a namespace called
DataGlobal
which houses reference arrays, broadcasters & constants that can be picked up by other scripts. I started out with local broadcasters and then decided to refactor it midway, so what I did was keep the old variable name and just point to the broadcaster defined in the DataGlobal namespace.Before:
namespace Something { const var somethingBroadcaster = Engine.createBroadcaster({}); somethingBroadcaster.addListener(...); }
After:
namespace DataGlobal { const var somethingBroadcaster = Engine.createBroadcaster({}); } namespace Something { const var somethingBroadcaster = DataGlobal.somethingBroadcaster; // This line (and everything that uses it doesn't need to change). somethingBroadcaster.addListener(...); }
You could even go as far as use the globally defined broadcaster (if it exists) or create a local one if you want to modularize the codebase even further:
namespace Something { const var someBroadcaster = isDefined(DataGlobal.someBroadcaster) ? DataGlobal.someBroadcaster : Engine.createBroadcaster({}); }
-
Just to sorta blunder in and ask possibly a stupid question... namespaces are just containers for related functions and variables right? You can't actually instance a namespace IE: use it as a kind of class ?
-
-
Right I see. So an Object Factory is what I would want if I wanted to create a class.
-
@Orvillain not really. A object factory creates a data structure with a densely packed memory layout that can be used for a neat implementation of MIDI logic, but it lacks most of the features that you would expect from a object-oriented class object (like methods or polymorphism).
What are you trying to achieve?
-
@Christoph-Hart said in Broadcasters best practices:
@Orvillain not really. A object factory creates a data structure with a densely packed memory layout that can be used for a neat implementation of MIDI logic, but it lacks most of the features that you would expect from a object-oriented class object (like methods or polymorphism).
What are you trying to achieve?
I'm mainly pondering what would be the cleanest way to create an object prototype and then instantiate however many of them I'd like - and each instance automatically creates the relevant DSP nodes, UI widgets, and various methods and callbacks required.
In Python you'd do something like:
class WhateverTheClassIsCalled(inheritance here) def __init__(self, and then any initialisation arguments here) self.blah = blah def do_a_thing(self, blahblahblah) thing_is_done = blahblahblah(bloo bloo bloo) return thing_is_done for i in range(0, 32): instance_var = WhateverTheClassIsCalled(blahblahblah, bloloooooloooo)
That sort of thing. In Python it is fairly trivial to put together a class this way, and tbh, syntax aside it looks like the Factory Object approach might do the same sort of thing???
BTW - the above is probably the best code I've ever written. :face_with_tears_of_joy:
-
@Orvillain One cool thing about namespaces is that each comes with 32 high-speed reg variables.
-
You can do this kind of simple object oriented stuff by creating a JSON object and attaching functions as member, which will then access the JSON properties through the
this.property
syntax:// This as close as you'll get to object oriented programming in HISE Script function createInstance(valueToUse) { var prototype = { create: function() { // Access properties of the "class" object with this Console.print(this.value); }, value: valueToUse // store the "constructor argument" as "class member" }; return prototype; } // And now let's add polymorphism: function createSubClassInstance(valueToUse) { var prototype = createInstance(valueToUse); prototype.create = function() { Console.print(this.value + ": SUBCLASS"); }; return prototype; } // Create two instances of the "class" var x1 = createInstance(90); var x2 = createInstance(120); var x3 = createSubClassInstance(500); // call the method of each instance x1.create(); x2.create(); x3.create();