Library design quandary - actionscript-3

Ok so I am writing an open source library. A section of this library deals with moving an entity in a two and three dimensional space so it will have functions that manipulate the rotation, position etc.
Now ideally I would like my library to work well with other libraries, in particular things like Papervision3D and other Flash 3D engines, but not forgotting basic image objects like a Sprite or Movieclip.
So this is my quandary. The functions that will manipulate the entity will also need to modify the values of the underlying model data (so either a Sprite, Papervision object etc). What is the best way to make my library flexible so that it can support multiple data models. Performance is also important aspect too.
Currently I am thinking of something like this:
//this is the public function that I expose in my library
public function rotate(val:Number,func:Function,objData:*):void
{
func(val,objData);
}
//example of a function that could be passed in
//this one will rotate a MovieClip
private function modelFunction1(rot:Number,objData:*):void
{
var myMov:MovieClip = objData as MovieClip;
myMov.rotation = rot;
}
//second example of a function that could be pass in
//this one will rotate a point
private function modelFunction2(rot:Number,objData:*):void
{
//yes I know this piece of code makes no sense :P
var p:Point = objData as Point;
p.x = Math.cos(rot);
p.y = Math.sin(rot);
}
so then it could be used like:
rotate(4,modelFunction2,myPoint)
//or
rotate(4,modelFunction1,mySprite);
I should add that in reality I, as the client code, won't be able to directly call the rotate function. Instead the rotate function that I want to pass in would need to be stored somewhere as a class member and then be called by the rotate function. Its just less code for me to write it out like above.
This to me seems quite flexible although the performance implications of casting and passing functions concerns me (but might be ok). Can anyone else suggest an alternative or is what I have the most logical solution. Thanks :)

I suggest the adapter pattern.
In your case you could define interfaces which offer type safe definitions for what your library expects instead of having function arguments.
then you need to write adapter classes which implement your librarys interfaces and wrap for instance a papervision object and delegate the function calls to your interface methods to the papervision object.
interface IRotatatable {
function rotate(deg : Number) : void
}
class YourLibraryClass {
public function rotate(r : IRotatatable, val : Number):void {
r.rotate(val)
}
}
class P3DAdapter implements IRotatable {
public function P3DAdapter(p3d : SomePaperVisionObject) {
_p3d = p3d;
}
public function rotate(r :Number):void {
p3d.rot = r;
}
}
function someClientCode():void {
var adapter : IRotatable = new P3DAdapter(p3d)
new SomeLibraryClass().rotate(adapter, val));
}

Related

AS3 Super Class Issue

I have a problem and I'm not too sure the best way to resolve it.
Scenario:
I have a Super Class called 'food' and I have 20 different foods and extends 'food' like Pizza, Curry, Fish and Chip etc.
When I remove a 'food' I keep a record of it so I can reuse (for performance purposes). Can I make a new Pizza class that uses an old 'food'?
E.g.
public class Pizza extends food
{
public function Pizza()
{
super = FOOD.returnUsedFoodClass();
}
}
Is this possible or would I need to save the extending Class as well?
Hope this all make sense.
EDIT:
When I say remove I mean I no longer need it - so I would normally remove all references to it. But instead, I have placed all 'food' classes that I no longer need in a static vector so I can reuse them later.
You misunderstand the basic OOP principles here.
First: a constructor runs only once and only when the object is created so any attempt to stop the creation of the object or replace the object from within its constructor is illogical and cannot succeed since the object at that moment is already created.
Second: Classic misunderstanding of the super keyword. No super doesn't point to any other instance of any other object, super in constructor points to the super class implementation of the constructor. Trying to assign an object to super cannot work and is also illogical. I'm guessing you meant to use 'this' which would also not work anyway.
What you are trying to achieve cannot be done that way and this in any OOP language. There's no way to run a constructor (meaning creating the object) and make this object point to something else within its own constructor. What you are looking for is a classic object pooling system via static methods like this:
var pizza:Pizza = Food.getFood("pizza") as Pizza;
Where the static method checks if any Pizza instance (from the pool) is available and if it is it returns it and if it's not it creates a new one and returns it.
Pooling can be implemented loosely or explicitly, I prefer the more solid and flexible explicit version. Here's an example:
Food class pooling additions:
static private var recycledInstances:Vector.<Food> = new Vector.<Food>();
//hold our reclycled instances
public function recycle():void
{
var index:int = recycledInstances.indexOf(this);
if(index >= 0)
{
return;
}
recycledInstances.push(this);
}
//explicitly add this instance to recycle
private function reset():void
{
}
//to run in constructor and when the instance is retreived from recycle
//this method purpose is to reset all values to default.
Now when an instance is no longer used you call the instance recycle() method to place it in recycle. Then when you need a new instance you do:
var food:Food = Food.getFood();
And this is implemented that way in Food class:
static public function getFood():Food
{
if(recycledInstances.length)
{
var totalInstances:uint = recycledInstances.length;
var instance:Food = recycledInstances[totalInstances - 1];
instance.reset();
recycledInstances.length -= 1;//faster than splice
return instance;
}
return new Food();
}
You can extend this easily to descendant of food class by adding a type member variable to Food for example and check the type of recycled instances before returning them.

Can I still create Global variables in AS3

Following the answer here, I have created a file called MyGlobals.as and placed some global variables and functions so that I can access it from anywhere within my project just like AS3 buil-in functions such as trace() method.
This is MyGlobals.as which is located in the src folder (top level folder)
package {
public var MessageQueue:Array = new Array();
public var main:Main;
public var BOOKING_STATUS_DATA:Object;
public function postMessage(msg:Object):void {
MessageQueue.push(msg);
}
public function processMessage():void {
var msg:Object = MessageQueue.pop();
if (msg) {
switch (msg.type) {
}
}
}
Looks like my IDE (FD4) is also recognizing all these functions and variables and also highlighting the varibles and functions just like any other built-in global functions. However, I am getting compilation errors "Accessing possibly undefined variable xxx". The code is as simple as trace(MessageQueue) inside my Main (or another classe).
I am wondering if there was any change Adboe has done recently that it can't be done now or am I missing something? I am not sure if I need to give any special instructions to FD to include this MyGlobals.as?
I am using FD4, Flex SKD 3.1, FP12.0
I am aware of the best practices which suggests to avoid using this type of method for creating global variables but I really need it for my project for my comfort which I feel best way (right now) when compared to take any other path which involves daunting task of code refactoring. I just want do something which can be done in AS3 which I guess is not a hack.
I've done some playing around; it looks like you can only define one (1) property or method at package level per .as file. It must be the same name (case-sensitive) as the .as file it is contained in.
So no, nothing has changed since the older Flash Versions.
In your case that would mean you need five separate ActionScript files along the lines of:
MessageQueue.as:
package
{
public var MessageQueue:Array;
}
main.as:
package
{
public var main:Main;
}
...etc. As you can see this is very cumbersome, another downside to the many others when using this approach. I suggest using the singleton pattern in this scenario instead.
package{
public class Singleton{
private static var _instance:Singleton=null;
private var _score:Number=0;
public function Singleton(e:SingletonEnforcer){
trace(‘new instance of singleton created’);
}
public static function getInstance():Singleton{
if(_instance==null){
_instance=new Singleton(new SingletonEnforcer());
}
return _instance;
}
public function get score():Number{
return _score;
}
public function set score(newScore:Number):void{
_score=newScore;
}
}
}
then iin your any as3 class if you import the singleton class
import Singleton
thn where u need to update the global var_score
use for example
var s:Singleton=Singleton.getInstance();
s.score=50;
trace(s.score);
same thing to display the 50 from another class
var wawa:Singleton=Singleton.getInstance();
trace(wawa.score)

AS3 MVC Design Multiple Controllers

So I'm reading a book about MVC and the author create a controller (PlayerController) and put some functions for keyboard use and some for mouse use. But he just comment out every keyboard use.
It gave me a idea to create 2 controllers, PlayerMouseController and PlayerKeyboardController so I can decide how to control the player changing one line. And if I can design this way, later I can add a AIController for monsters that use the same view and model but are controlled by AI and so on...
I have my model Player and it do the physics stuff. Now I want two controller, one for mouse and other for keyboard. So I create a PlayerMouseController and PlayerKeyboardController.
The PlayerMouseController has 2 functions: processUpdate() and processMouseDown()
The PlayerKeyboardController has 2 functions: processKeyDown() and processKeyUp()
I create the object like this:
_player = new Player();
_playerController = new PlayerMouseController(_player);
_playerView = new PlayerView(_player, _playerController, stage);
addChild(_playerView);
If I want to change the controller I can just change the _playerController line for this:
_playerController = new PlayerKeyboardController(_player);
And it works fine... But I dont know if the design I use is fine for a large project
To make this work I have to create a Controller class with nothing so I can extends the others controllers and my view can call all methods.
public class Controller
{
public function processKeyDown(e:KeyboardEvent):void
{
}
public function processKeyUp(e:KeyboardEvent):void
{
}
public function processUpdate(stage:Stage):void
{
}
public function processMouseDown(e:MouseEvent):void
{
}
}
In my view (PlayerView) I accept any Controller:
public function PlayerView(model:Player, controller:Controller, stage:Stage)
{
_model = model;
_controller = controller;
_stage = stage;
}
and I decide what to use based on its type:
if (_controller is PlayerKeyboardController)
{
_stage.addEventListener(KeyboardEvent.KEY_DOWN, onKeyDown);
_stage.addEventListener(KeyboardEvent.KEY_UP, onKeyUp);
}
else if (_controller is PlayerMouseController)
{
_model.addEventListener(Model.UPDATE, onUpdate);
_stage.addEventListener(MouseEvent.MOUSE_DOWN, onMouseDown);
}
// EVENTS PlayerKeyboardController
private function onKeyDown(e:KeyboardEvent):void
{
_controller.processKeyDown(e);
}
private function onKeyUp(e:KeyboardEvent):void
{
_controller.processKeyUp(e);
}
// EVENTS PlayerMouseController
private function onUpdate(e:Event):void
{
_controller.processUpdate(_stage);
}
private function onMouseDown(e:MouseEvent):void
{
_controller.processMouseDown(e);
}
So... this is a good idea? How should I improve??
I think as your project evolves you'll soon hit a bottleneck with this kind of approach. I'd suggest creating a general IEntityController interface (don't be afraid of the word 'entity', it just shows that this is a game object controller, something that is parent for example of hero, enemy, etc.).
Then I'd create two separate implementations of this interface: BasicMouseController and BasicKeyboardController, so that I'd have two branches of these controllers with basic logic. If you need additional functionality for the Hero, you'd create a HeroMouseController class which would extend the BasicMouseController class and would have the advantage of calling super methods and adding the extended functionality easily.
You'd still have the benefit of passing different controllers to the PlayerView class as it's constructor would receive an IEntityController as a parameter, meaning anything implementing this class can be passed in.
There are many approaches for problems like this and StackOverflow is usually not meant to give these kind of answers, as every answer to these kind of questions is subjective and this website more fit for Problem/Solution kind of posts.

Enumerating class properties

I'm trying to iterate through the properties of a custom class, however, the methods provided by Adobe appear to not work. I'm not receiving compile or run-time errors.
Class
package {
public dynamic class enum {
public var foo:Number = 123;
public function enum() {
this.setPropertyIsEnumerable("foo", true);
if (this.propertyIsEnumerable("foo") == false) {
trace("foo:" + foo + " is not enumerable.")
}
}
}
}
// outputs "foo:bar is not enumerable."
Implementaiton
var test:enum = new enum();
for (var property:String in test) {
trace(property);
}
// outputs nothing
I try to keep my code fast and flexible, so it's really frustrating when you must change the class to Dynamic just to be able to use for ... in on the properties. Jackson Dunstan's testing confirms that this can be 400x slower than static class properties, but those have to be explicitly referenced (impractical for property agnostic methods), or use reflection of the class (computationally expensive) to be accessible.
The only way I've found to sidestep the whole issue is to use dynamically declared variables... which is pointless since at that point using setPropertyIsEnumerable(prop, true) is superfluous; all dynamically created properties already are enumerable. Additionally, dynamic variables cannot be strongly datatyped, and performance goes out the window.
For example...
Class
package {
public dynamic class enum {
public var foo:String = "apple";
public function enum(){
this.dynamicVar = "orange";
this.dynamicProp = "banana";
this.setPropertyIsEnumerable("foo", true);
this.setPropertyIsEnumerable("dynamicProp", false);
}
}
}
Implementation
var test:enum = new enum();
for (var key:String in test) {
trace(key + ": " + test[key]); // dynamicVar: 1
}
// outputs "dynamicVar: orange"
Now that the class is dynamic, we see that only one of our 3 test properties are being iterated. There should be 2.
It almost feels like Adobe wants us to adopt bad programming habits. Words fail me...
Non-dynamic classes do not provide enumerable properties or methods.
As stated in the description of the link you provided.
Sets the availability of a dynamic property for loop operations.
I think you might want to refactor your code on this approach.
I have never had to loop over a classes properties like you are doing here.
If you want to track items dynamically you should use an associative array and track them that way not at the class level like you are doing.
And if you want strong data typing then use a vector.

Application Design Patterns AS3

just thought I would share something I have found to help delivering data across an application I am wondering what others think about this I wanted to have a way to capture event bubbling up back down to other components but in a way that it would make it easy to use anywhere in may app so this is what i came up with.
I Extend the Application class and wrap in an abstract function registering a function of any component anywhere and capture it at the top most level and pass to where ever i chose to.
public class AxApplication extends Application
{
public var ___registeredEvents:Array = new Array();
public var ___registeredFunctions:Array = new Array();
function AxApplication()
{
super();
}
public function localRegisterForEvent(e:Event,func:*,caller:*):void
{
caller.addEventListener(e.type,localCallerEventHandler,true,3);
caller.addEventListener(e.type,localCallerEventHandler,false,3);
___registeredEvents.push(e);
___registeredFunctions.push(func);
}
public function localCallerEventHandler(e:*):void
{
if(e!=null)
{
for(var i:int = 0 ; i< ___registeredEvents.length; i++)
{
if(e.type == ___registeredEvents[i].type)
{
___registeredFunctions[i](e);
//the registered function gets called
//there no garbage collection implemented!
}
}
}
}
}
I think that is not a very useful solution. Why? Because you scatter AxApplication references around the application. Views and Model instance don't need any references to the application at all. It would be better to to implement a controller layer which uses a simple eventBus property, which could look like:
private static const _EVENT_BUS:IEventDispatcher = FlexGlobals.topLevelApplication;
protected final function eventBus():IEventDispatcher {
return _EVENT_BUS;
}
If you implement a base view controller/mediator (depending from which framework you're coming), you don't have any reference to non-framework classes at all, which makes it highly reusable. It is just a simple reuse of the Application singleton which you use to dispatch system wide events. You register listeners in the view controller/mediator and update the views or models accordingly. RobotLegs for example uses a system wide event dispatcher as well.
Why not just using the parentApplication approach? Because you can't implement tests (the generated test-runner of IDEs won't extend your AxApplication) or just yank the components/models in a different application - that is basically not possible.