I've always found this interesting, and haven't managed to fully understand it yet. Take this class:
package
{
public class SomeClass
{
private var _myvar:String = "hello";
public static function sayHello():void
{
trace(_myvar);
}
}
}
As we know, this will throw an error unless I declare _myvar as private static var _myvar
1120: Access of undefined property _myvar.
I don't completely understand why _myvar isn't in scope unless it's static. Can anyone provide a easy to understand explanation of this?
As _myvar is not static it's a property that is attached to an instance of SomeClass, i.e. it has only meaning each time you create a new SomeClass.
In contrary a static property is attached to the Class object, it exists only once and is not dependent of each instance of SomeClass created.
So when you try to reach _myvar from your static function sayHello it's the same as calling this._myvar but you have only one static property and can have multiple instance created over the timelife of your application, what should be the value of this in that case ?
What I am trying to do is kind of odd, but I am wondering if anyone can come up with a clever way to do what I want to do. Basically, I want to re-define a named function at runtime. I can do this with anonymous functions, but I can't figure out a way to do it for named functions. I want to do this so that I can implement a "spy" functionality on an object for a testing framework (a port of Jasmine to Flex).
Take, for instance, this class:
public class TestClass
{
public var anonymous:Function = function():void {
trace("original anonymous");
};
public function named():void {
trace("original named");
}
}
I can easily re-define the anonymous function because it is just a variable. Javascript uses this idiom a lot.
var testClass:TestClass = new TestClass();
testClass.anonymous = function():void { trace("overridden anonymous"); }
BUT, when I do the same thing for named functions, you get a compile-time error:
// Does not compile
testClass.named = function():void { trace("overridden named"); }
I tried to make it a bit more "squishy" but this leads to a runtime failure "Cannot assign to a method named on TestClass".
// Compiles with runtime failure
testClass["named"] = function():void { trace("overridden named"); }
Can anyone more clever than I come up with a way to hack this? Can the bytecode be hijacked? Something?
I want to modify an object, not a
class
But object doesn't contain functions, only non-static variables. I tried to use prototype property and replace method there, but original method still gets called instead of injected one.
About "hack" bytecode, do you mean "hack" already loaded SWF in runtime? I think it's not possible. I'm sure, though, you can parse SWF with something like as3swf, find method in bytecode, replace it and save result in new SWF.
I had an idea bout making a function "cache" . This might work with what you need.
Let's say you have a class "Car" with a method you need to redefine at runtime:
public class Car extends Sprite
{
private var functionCache:Function;
public function Car()
{
super();
}
public function flexibleFunction(functionBody:*=null):void{
if(functionBody is Function){
functionBody.call();
functionCache=functionBody;
} else {
functionCache(functionBody);
}
}
}
Usage:
public class Main extends Sprite
{
private var car:Car;
public function Main()
{
car = new Car();
car.flexibleFunction(function(){trace("redefine test #1")});
car.flexibleFunction();
car.flexibleFunction(function(doParametersWork:String="let's see"){trace("redefine test #2: " + doParametersWork);});
car.flexibleFunction("yes they do");
car.flexibleFunction();
}
}
an easy way to accomplish what you want is to simply pass a new function to the original function and execute it from there:
package
{
//Imports
import flash.display.Sprite;
//Class
public class RedefineFunction extends Sprite
{
//Constructor
public function RedefineFunction()
{
originalFunction();
originalFunction(redefinedFunction);
}
//Original Function
public function originalFunction(redefinition:Function = null):void
{
if (redefinition != null)
redefinition();
else
trace("Original Function Definition");
}
//Redefined Function
private function redefinedFunction():void
{
trace("Redefined Function Definition")
}
}
}
traces:
Original Function Definition
Redefined Function Definition
I've asked this same question with Python.
Now I like to know if this can be done in AS3.
If I have something like this:
package
{
public class SomeClass
{
private function A():void { C() }
private function B():void { C() }
private function C():void
{
// who is the caller, A or B ???
}
public function SomeClass()
{
A()
B()
}
}
}
Despite the design or other issues, this is only a question of an inquiring mind.
Note: I like to have an access to an instance of the caller function so I can call that caller function (if I want to)
Note 2 : This has to be done without changing function C() signature
"Unlike previous versions of ActionScript, ActionScript 3.0 has no arguments.caller property. To get a reference to the function that called the current function, you must pass a reference to that function as an argument."
From http://livedocs.adobe.com/flash/9.0/ActionScriptLangRefV3/arguments.html
That's the only way you can do that, otherwise you'll need to make a global variable to tell what function is calling C
Sure it can be done. You can do something like
private function C():void
{
var e:Error = new Error();
var stack:String = e.getStackTrace();
//analyze stack and find out which function called it.
}
this is ugly but it would work.
I've done this before, but i can't quite remember the syntax.
What i've got (simplified):
function createText(clazz:Class)
{
var font:Font = new clazz(); //throws Instantiation attempted on a non-constructor.
}
I believe this can be done without using getQualifiedClassName, but its been a long time. Any help appreciated.
You are probably passing null to the function.
package
{
import flash.display.Sprite;
public class ClassTest extends Sprite
{
function ClassTest()
{
makeObject(Object);
makeObject(Sprite);
makeObject(null);
}
private function makeObject(type:Class):void
{
trace(typeof type);
var obj:* = new type();
trace(typeof obj);
trace("");
}
}
}
This outputs:
object
object
object
object
object
TypeError: Error #1007: Instantiation attempted on a non-constructor.
at ClassTest/makeObject()
at ClassTest()
how are you passing the class to the function ?
the calling line should have trown an error in the first place if the wanted class wasn't available, this is weird.
can you post the real code ?
here a trick for loading a class compiled in an external swf
var clazz:Class = this.yourLoader.contentLoaderInfo.applicationDomain.getDefinition("yourClassName") as Class;
Turns out I hadn't given the font a class name in CS3. So yes,I was passing null.
I got this idea of expanding my trace() messages.
Why
trace() is all over my code, I want to turn them on/off by a simple command and maybe add some sort of priority functionality to the trace(), i.e.
myTrace.TraceMsg("loosehere",debugme, 0);
myTrace.TraceMsg("winhere",debugme, 1);
And when I run, only the one with the higher priority, "1" in this case, shows.
There is a lot more functionality I would like to add as well, like logging messages to file and so on.
Problem
How do trace() work? -Is it possible to overload trace() somehow? -How would I implement the custom TraceMsg(what code here?) method?
Having some serious problems finding info on this subject on our favourite search engine, so any help would be appreciated.
I have come up with a rather efficient, yet tedious way of using my own trace() function in Flash only projects, but calling it simply with
trace("this", "that", "and that too");
I basically implement one trace() method in every class of my project, that calls a public function (so that i can call the real trace() function from there.
here is what I do : in every class I call this
include "trace_implementation.as";
in the .as file comes a simple method implementation (it could be a static method too).
public function trace(... arguments){
for(var i in arguments){
myTrace(arguments[i]);
}
}
and the myTrace function is defined in its own myTrace.as file
package pt.utils{
import flash.external.ExternalInterface
public function myTrace(_s:String):void{
trace(_s);// this will call the original flash trace() function
ExternalInterface.call("console.log", _s);// to get traces outside of flash IDE
/*implement what you want here*/
}
}
so now when I compile with "omit trace actions", my whole debugging is ignored as if I used trace() simply.
the really good part here is that you could implement custom actions depending on instructions you give in the trace, so :
trace(Debug.DEBUG_MESSAGE, "message to output in debug");
trace(Profile.START_PROFILING, this, 'name');
/*do heavy code*/
trace(Profile.STOP_PROFILING, this);
then dispatch it from myTrace, or a Tracer class or anything :)
Hope this helps future tracers.
trace() itself is a top-level function, not a class, so unfortunately we cannot extend it. That being said, we can utilize it in a simple class to do just what it does normally, only in this case the trace is based on conditions (i.e. Boolean - true|false, etc). First we create the Trace class, which we wouldn't instantiate ourselves because we are utilizing a Factory design pattern through the class below, Tracer. Tracer is built around the singleton design pattern, yet utilizes the Factory pattern to instantiate instances of Trace, when the trace method of Tracer is called.
//This class is handled by Tracer, which is right below it.
//You WILL NOT instantiate these, nor hold references.
package
{
public class Trace
{
private function _value:*;
private function _trace:Boolean;
public function Trace(pValue:*, pTrace:Boolean):void
{
_value = pValue;
_trace = pTrace;
}
public function get value():*
{
return _value;
}
public function get trace():Boolean
{
return _trace;
}
}
}
//This is the important class and the only one you will work with.
package
{
/**
*Utilizes Singleton and Factory design patterns.
*/
public class Tracer
{
private var _traceArray:Array;
private static var _instance:Tracer;
public function Tracer(pvt:PrivateClass = null):void
{
if(pvt == null)
{
throw(new Error("You cannot instantiate this class directly, please use the static getInstance method."));
}
_init();
}
public static function getInstance():Tracer
{
if(Tracer._instance == null)
{
Tracer._instance = new Tracer(new PrivateClass());
}
return Tracer._instance;
}
public function trace(pValue:*, pTrace:Boolean):void
{
var trace:Trace = new Trace(pValue, pTrace);
if(trace.pTrace)
{
trace(pValue);
}
}
//Since we have the option for individual traces to be disabled
//I provide this to get access to any and all later.
public function traceAll():void
{
traceStr:String = _traceArray.toString();
}
public function get traceables():Array
{
return _traceArray;
}
//Here we provide a method to trace all, even if set to false in their constructor.
private function _init():void
{
_traceArray = new Array();
}
}
}
//Here we create a class that is OUTSIDE of the package.
//It can only be accessed from within this class file. We use this
//to make sure this class isn't instantiated directly.
class PrivateClass
{
function PrivateClass():void
{
trace('can only be accessed from within this class file');
}
}
//Now for use in doc class
package
{
import flash.display.Sprite;
import flash.events.Event;
//No need to import Tracer and Trace, they are also in the
//unnamed package.
public class DocumentClass extends Sprite
{
private var _tracer:Tracer;
public function DocumentClass():void
{
if(stage) _init();
else addEventListener(Event.ADDED_TO_STAGE, _init);
}
private function _init(e:Event = null):void
{
_tracer = Tracer.getInstance();
_tracer.trace(10*20, false);
_tracer.trace(10*20, 0); //SAME AS ABOVE
_tracer.trace("I love AS3", true); //traces
_tracer.traceAll(); //Would trace: 200, 200, I love AS3
}
}
}
Keep in mind this is off the hip and very well could have a bug or two, but the idea is there; That is to say that this is not tested, it is merely to give you an idea of how you might implement this.
I hope this helps.
Look at the Flex logging API, particularly the section: Implementing a custom logger with the logging API.
Look up the TraceTarget class as well.
You can't override trace itself, but for ease of typing I like to create a global function called 'tr'. It's a little known fact that you can create global functions in AS3, but it's easy.
Create a file called tr.as inside you main source directory (not in a subdirectory or package), with the contents:
package {
public function tr(msg:String, ...):void {
// add custom trace logic here
trace("tr message: "+msg);
}
}
If you need to have a lot of logic or static storage variables etc, it might be better to make a separate static class, and have the global tr function call out to that, such as:
package {
import org.code.MyTracer;
public function tr(msg:String, ...):void {
MyTracer.tr(msg); // all the tracing logic is inside the MyTracer class
}
}
Here is a super simple custom trace function I use. debugFlag can be set to true/false
elsewhere in the package.
public static function myTrace(... vars) :void {
if (debugFlag) {
var output:Array = new Array;
for each (var arg in vars) {
output.push(arg);
}
trace(output);
}
}
In AS2, it was possible to override the global trace function by doing something like this (taken from memory, might be a bit wrong but the gist of it is there):
public static var realTrace:Function = _global["trace"];
// This is put in some init code somewhere
_global["trace"] = myTrace;
public static function myTrace(... args):void
{
// Do whatever you want with args here, build a nice formatted string or whatever
// before passing to realTrace. Using with MTASC one could add line numbers, class
// names and all sorts of nice meta data. Or just return should you want to turn
// tracing off.
realTrace.apply(args);
}
Unfortunately I haven't found a way to do the same in AS3. Yet.
Trace is a top-level function, so you can't override it, and as far as I know, it does not fire any events. Since it's a top-level function (not contained in any named package), you can use it without import statements.
Here is an example of a top-level "Tracer" class that you can use in place of trace without import statements.
Just call "Tracer.write" or "Tracer.writeError" for tracing Error objects.
"Tracer.write" accepts a variable number of arguments, just like the built-in trace function. "Tracer.writeError" is a helper method that allows you to easily trace Error objects.
Features:
Calls built-in trace.
Keeps a log of all your calls to Tracer.write as an array of strings.
The call log is accessible as a string through getText, which joins all elements in the array with a newline character and will optionally tack on line numbers!
Fires events when new lines are added to the log, so if you have some kind of display window for the log, the display window can listen for Tracer events to update the log display in real-time as the events occur. This is great for displaying trace events when running inside a web browser or stand-alone player.
-Tracer class definition
package
{
import flash.events.EventDispatcher;
public class Tracer extends EventDispatcher
{
private static var traced_text:Array = new Array( "--Start of Trace Log--" );
public static var enabled:Boolean = true;
private static var suspended:Boolean = false;
public static var instance:Tracer = new Tracer();
public static const newline:String = "\n"; //workaround for TextField.appendText bug.. use "\n" instead of "\r". See note and link to bug post in getText method
public function Tracer()
{
}
static public function write( ...args ):void
{
if (enabled && !suspended)
{
trace.apply( null, args );
var text:String = args.join( newline );
var next_index:int = traced_text.length;
traced_text.push( text );
suspended = true; //prevent recursive calls from TracerEvent handler
instance.dispatchEvent( new TracerEvent( text, next_index ) );
suspended = false;
}
}
static public function writeError( e:Error ):void
{
write( "errorID: " + e.errorID, "errorName: " + e.name, "errorMessage: " + e.message, "stackTrace: " + e.getStackTrace() );
}
static public function getText( include_line_numbers:Boolean ):String
{
var line_count:int = traced_text.length;
var lines:Array = traced_text; //store pointer to traced_text; pointer may be changed to reference an altered array that includes line numbers
if (include_line_numbers) //create temporary trace log copy with altered lines; allows quick call to join at end
{
var new_lines:Array = new Array();
for (var i:int = 0; i < line_count; i++)
new_lines.push( i.toString() + ": " + lines[i] );
lines = new_lines;
}
return lines.join( newline ); //do not include last newline character (workaround for bug in appendText method (https://bugs.adobe.com/jira/browse/FP-1982); I have to call appendText with newline character first, otherwise it has issues like not acknoledging the newline thats already there at the end).
}
static public function addEventListener(type:String, listener:Function, useCapture:Boolean = false, priority:int = 0, useWeakReference:Boolean = false):void
{
instance.addEventListener(type, listener, useCapture, priority, useWeakReference);
}
static public function removeEventListener(type:String, listener:Function, useCapture:Boolean = false):void
{
instance.removeEventListener(type, listener, useCapture);
}
static public function willTrigger(type:String):Boolean
{
return instance.willTrigger(type);
}
static public function hasEventListener(type:String):Boolean
{
return instance.hasEventListener(type);
}
}
}
-TracerEvent class definition
package
{
import flash.events.Event;
public class TracerEvent extends Event
{
public static const WRITE:String = "te_write";
public var text:String;
public var index:int; //index of newly traced text in the traced_text array (trace log)
public function TracerEvent( text:String, index:int )
{
super( WRITE, false, false );
this.text = text;
this.index = index;
}
override public function clone():Event
{
return new TracerEvent( text, index );
}
}
}
As mentioned below, there is no way to override trace (at least not if you want your traces to reach the output stream), but it's actually very easy to create your own universally accessable logging function. Plus, you can even define a universally accessable boolean to turn logging on or off:
log.as (note that the filename must reflect the name of the function)
package {
function log(... arguments):void {
trace("Custom logging FTW!");
if (logEnabled)
trace(arguments);
}
}
logEnabled.as (note that the filename must reflect the name of the variable)
package {
var logEnabled:Boolean = true;
}
Main.as
package {
import flash.display.MovieClip;
public class Main extends MovieClip {
public function Main() {
log("Testing");
logEnabled = false;
log("Testing2");
}
}
}
Response
Custom logging FTW!
Testing
Custom logging FTW!
you dont need to override it , just create a function in your project and call it trace then any trace call will point to this.trace ;)
function trace(... arguments){
yourfunction(arguments);
}