I have been using Flex / Flash Builder for a number of years. The latest release of Flash Builder (4.7) seems to come with quite a few problems, the biggest of those being:
Does not detect component IDs in MXML. For example, you cannot Find Usages of the ID of a component. Keeping the cursor on the ID of a component does not even mark occurrences of the ID. Instead, it marks occurrences of the actual id words in the MXML.
Extremely slow.
I am seriously evaluating moving over to IntelliJ IDEA 12, especially after reading many experienced Flex devs raving about it and recommending it.
I tried it. It took me a while to get to terms with the new terminologies of the IDE (made easy by this doc and very helpful support personnel at JetBrains).
I was able to setup my (large) projects in IDEA with Adobe Flex 4.6 SDK and got it to compile fine. But I noticed many "errors" highlighted in my AS files which are all actually false alarms.
The ActionScript editor doesn't seem to recognise the objects defined in MXML. Apparently, this is a known bug in IDEA (tracked here). And this bug has existed for more than 2 years!
Quoting the JetBrains support personnel:
I must admit that highlighting of ActionScript files which do not contain classes, but instead included in mxml as <fx:Script source="some_file.as"/> is probably the only weak part of IntelliJ IDEA code highlighting. False error highlighting will go away if you embed AS code inside CDATA of <fx:Script/> instead of referencing as external *.as file. Though I understand that this is not always desired.
I'm afraid the fix won't go into 12 release because the release is very soon and the fix is too risky. Priority of the issue depends on votes and user feedback. So far we have only 2 votes (http://youtrack.jetbrains.com/issue/IDEA-52598) and as the fix is pretty complex we still haven't implemented it thinking that this is a rare use case. I hope to fix it in one of 12.x update releases.
My project is a huge one, with huge MXML files and even more huge AS code for each MXML. So, for organisation purposes, I need to logically split them into smaller files. So, merging the AS code with the MXML is not practical. The false error highlighting just drastically reduces the readability of the code. Also, it does not allow Control / Command-clicking component IDs in AS code to quickly navigate to the definition of the component in MXML (which, incidentally, is now broken in FB 4.7 as well, but worked well in FB 4.6).
This bug in IDEA is unfortunately a deal-breaker for me. But I am wondering how other Flex devs are able to overcome / work around this seemingly critical bug.
It seems unbelievable to me that just 2 people have been affected by this bug, especially with so many Flex devs recommending IDEA. Maybe I am doing something wrong?
All you Flex developers, I would appreciate your thoughts.
UPDATE
This is in response to RIAStar's excellent and detailed answer. But it doesn't quite help me completely. Let me explain why and how I use <fx:Script source>. I am using Flex 4.x, with almost only Spark components.
Suppose a brand new Flex project. The main application is an MXML file.
In this MXML file, suppose I have a signup form.
On editing the form (in each field), suppose I have to run validations and enable the Submit button only if the form is completely valid. This would mean I need to assign change event handlers to the form items. The event handlers is AS code.
Suppose there is a username field which needs an on-type uniqueness check, by calling the server asynchronously. The server communication code is also AS code.
And then of course, there is the Submit button handler, which is also AS code.
I usually put all the AS code in separate .as files and include it in the MXML using <fx:Script source>. This AS code is usually quite heavy, with a lot of functional and behavioural logic. Many times, based on user action, even the components in the MXML and layout of the elements is modified through this AS code.
If I understand you guys right, none of this event handler code should be in these MXML script files. So, where should it be? How do you guys do it? I am not sure how the Spark Skinning architecture has anything to do with this.
Since I can't think of a gentle way of putting this, I'll just be blunt: I'm afraid the reason only two people think this is a critical bug, is that most seasoned Flex developers will agree that using <fx:Script source="some_file.as"/> is bad practice.
You effectively create two files that represent one class. From a readablity POV, which you seem concerned about, that's not a good move. One of these files (the .as file) is just a bunch of functions that cannot exist in their own right: they are tightly coupled to another file/class, but just looking at the .as file there is no way of knowing which class it is coupled to. Of course you can use some kind of naming convention to work around this, but in the end ActionScript/Flex is supposed te be used as a statically typed language, not a scripting language relying on mixins and naming conventions (don't get me wrong: I'm not saying scripting languages are bad practice; it's just not how ActionScript was conceived).
So what are your alternatives?
I suppose the main reason behind this construct is that you wish to separate MXML from ActionScript code, or in more abstract terms: separate the view from the logic. Fortunately this can be achieved in a few other, cleaner ways. Which solutions are available to you depends whether we're talking Flex 3 (or earlier) or Flex 4.
I realise that you may not have time to refactor your code to one of the proposed solutions, but I didn't want to leave you with just a "that's not good practice" answer.
Flex 3 (mx)
Code behind: A lot of developers used the so-called "code behind" pattern to separate their logic from their view. You can find plenty of information on the topic by Googling "flex code behind". I don't need to repeat all that in here. I'm not much of a fan of the concept because it relies heavily on inheritance and the two resulting classes are still pretty tightly coupled, but at least we're talking two classes. If you design your architecture well, you may even be able to reuse some of your base classes.
Compose model en controller: I used to create a separate "presentation model" class and a "controller" class for each MXML view and then use it something like this:
<!--MyView.mxml-->
<mx:VBox>
<m:MyModel id="model"/>
<c:MyController model="{model}" view="{this}"/>
...
</mx:VBox>
MVC purists won't like this, but it worked pretty well for me in thencontext of Flex applications.
Later when Direct Injection supporting frameworks (like Parsley) made their appearance, I could use injection to wire all those classes instead of hard-wiring them like in this example.
MVC frameworks: My knowledge of this topic is sparse (because in my opinion Flex is a very decent MVC framework that requires no third-party additions, but that's another disussion), but in short: they can help you separate logic from view in a clean way.
Flex 4 (Spark)
With Flex 4, the Spark skinning architecture was introduced, which allows for very nicely separated view and logic. You create a so-called 'host component' class in plain ActionScript, which contains all of the behavioural code, and a 'skin' class in MXML which defines the visual representation of the component. This makes designing reusable components very easy.
As per your request, here's a simplified example of how you might use Spark skinning to create your signup form.
Let's start with the skin class since it's easy to understand. It's just a form with some input fields. The HostComponent metadata tells the skin it's supposed to work together with the SignUp host component.
<!--SignUpSkin.mxml: the visual representation-->
<s:Skin xmlns:fx="http://ns.adobe.com/mxml/2009"
xmlns:s="library://ns.adobe.com/flex/spark">
<fx:Metadata>
[HostComponent("net.riastar.view.SignUp")]
</fx:Metadata>
<s:Form>
<s:FormHeading label="Sign up"/>
<s:FormItem label="User name">
<s:TextInput id="userInput"/>
</s:FormItem>
<s:FormItem label="Password">
<s:TextInput id="passwordInput" displayAsPassword="true"/>
</s:FormItem>
<s:Button id="submitButton" label="Submit"
enabled="{hostComponent.canSave}"/>
</s:Form>
</s:Skin>
And now the host component in pure ActionScript. It has to extend SkinnableComponent to be able to use our skin (there's also SkinnableContainerwhich I've just recently explained in this question: Flex mxml custom component - how to add uicomponents?, but we won't be needing that here).
public class SignUp extends SkinnableComponent {
[SkinPart(required="true")]
public var userInput:SkinnableTextBase;
[SkinPart(required="true")]
public var passwordInput:SkinnableTextBase;
[SkinPart(required="true")]
public var submitButton:IEventDispatcher;
[Bindable]
public var canSave:Boolean;
override protected function partAdded(partName:String, instance:Object):void {
super.partAdded(partName, instance);
switch (instance) {
case userInput:
userInput.addEventListener(TextOperationEvent.CHANGE,
handleUserInputChange);
break;
case passwordInput:
passwordInput.addEventListener(TextOperationEvent.CHANGE,
handlePasswordInputChange);
break;
case submitButton:
submitButton.addEventListener(MouseEvent.CLICK,
handleSubmitButtonClick);
}
}
private function handleUserInputChange(event:TextOperationEvent):void {
validateUsername(userInput.text);
}
...
}
What's important here?
The variables marked as SkinPart will automatically be assigned the components with the same id that exist in the Skin you just created. For instance <s:TextInput id="userInput"/> will be injected into public var userInput:SkinnableTextBase;. Note that the type is different: SkinnableTextBase is the base class of TextInput; this allows us to create another skin with e.g. a TextArea instead of a TextInput and it'll work without touching the host component.
partAdded() is called whenever a SkinPart is added to the display list, so that's where we hook up our event listeners. In this example we're validating the username whenever its value changes.
When the validation is done, you can simply set the canSave property to true or false. The binding in the skin on this property will automatically update the Button's enabled property.
And to use both of these classes together:
<v:SignUp skinClass="net.riastar.skin.SignUpSkin"/>
I actually have become quite fond of using RobotLegs.
In my MXML views I try to keep all logic outside of the MXML and simply dispatch events out to the mediator. From there I can put code in the mediator to the heavier AS needed.
Related
we've got two applications (not modules, two independent applications!): A and B. both are Parsley-managed and we'd like to embed B in A using SWFLoader (but, and i stress that, we don't want to "connect" these applications using Parsley, we just want to do normal Flash embedding) .
that's embed code:
<fx:Script>
<![CDATA[
[Bindable]
private var childDomain:ApplicationDomain =
new ApplicationDomain(ApplicationDomain.currentDomain);
]]>
</fx:Script>
<mx:SWFLoader width="100%" height="100%" source="B.swf"
complete="initNestedAppProps(SWFLoader(event.currentTarget).content);"
loaderContext="{new LoaderContext(false, childDomain, SecurityDomain.currentDomain)}"/>
and it works when i embed B in a dummy app without Parsley.
however, when i copy-paste that embed code in live application A, Parsley throws this famous error:
ReferenceError: Specified ApplicationDomain does not contain the class _B_mx_managers_SystemManager
even if the view that contains embedding code is not Parsley-configured (and doesn't have <Configure/> tag).
i can't post this on Parsley forums unfortunately and googling didn't help as it seems people don't do application embedding too often.
so the question is, why does this error happen (Parsley shouldn't care about stuff in embedded application, should it?) and how can tell Parsley to properly use my childDomain?
The problem is that Parsley is bubbling events up the display list so that a context can use them to inject properties etc.
Despite the fact your sub application is in a separate application domain, events can still bubble up from the swf loader's child to the parent and so on.
What is happening is that your sub application is bubbling events that are getting handled by your shells (or wrapper/loader applications) context, however when parsley then tries to reflect on this object it can't because the object doesn't exist in it's application domain.
The solution is to stop these events getting to your shell application's parsley context. You can do this a number of ways, for example you could just add listeners for the events and stop their propagation. However this would mean you would have to add listeners for all Parsley events, which could change in the future. A better solution is to create a new context in your SWFLoader’s parent that has an autowireFilter that returns ViewAutowireMode.NEVER for displayObjects passed to it.
This context will stop them bubbling any further and will stop parsley reflecting on them, and therefore stop the problem with them not being in the application domain.
See: org.spicefactory.parsley.core.view.impl.DefaultViewAutowireFilter
org.spicefactory.parsley.core.builder.impl.DefaultCompositeContextBuilder
http://opensource.powerflasher.com/jira/browse/PSL-587
Hope this helps.
the above answer is correct.
in our case i solved the problem by writing a flex module and using ModuleLoader instead of SWFLoader, which is nicely integrated with Parsley.
By "refresh" I am completely disposing it and then introducing it again in the application (without closing the application itself - that is). Other than than I think the question is self-explanatory.
Example:
Say I have a component named myComponent. I add that component to the application using MXMl in the standard way <components:myComponent id="myID" />. Say that when a user clicks a button (the button may be in another state), the component with id myID should be garbage-collected and a new instance of it added to the application.
How do I go about doing that? If there are multiple solutions which one is the optimal performance-wise?
I am new to Flash and Flex so excuse me if any incorrect terminology were used.
Remove all the event listeners from the old component; whatever they are using the removeEventListener method:
myButton.removeEventListener(someEvent, someEventHandlerMethod);
Then all variables that refer to the component should be set to null. If created in an MXML file, like this:
<s:Button id="myButton" />
Then all you have to do is set that value to null:
myButton = null;
Once there are no references to the component, it can safely become eligible for garbage collection.
If you want to re-created, then just re-created it. You'll have to re-create it in ActionScript, but the code isn't hard. Conceptually something like this:
myButton = new myButton();
myButton.properties = propertyValues;
myButton.addEventListener(someEvent, someEventHandlerMethod);
parentContainer.addChildAt(myButton, whateverPositionYouWantToADdTheComponentAt);
I'm not sure I see the benefit of doing this. I suspect it'll be much more efficient to tweak the existing button instance in the way you need to as opposed to destroying it and trying to replace it with the exact same thing.
There are lots of examples of how to create skinnable components in AS3 using the new Spark component architecture, however I've yet to find any such examples using MXML.
What I'm mainly referring to is defining the skin parts and skin states. It seems as though the SkinPart metadata is supposed to be associated with properties and as such can't be used in MXML, is this correct?
As far as I can tell, this isn't possible, because - as you rightly pointed out, there's no way to assign property-level metadata in MXML.
But declaring skin contracts in MXML seems like it would go against the spirit of the Skinning architecture, which is to separate the view of the component from it's implementation logic.
And, given that in MXML everything declared has a value, it inherently couples the component to a skin.
Ie:
<!-- labelDisplay has a value -->
<s:Label id="labelDisplay" />
vs
[SkinPart]
public var labelDisplay:Label; // labelDisplay == null;
Likewise, most of the component logic itself would end up in a <fx:Script /> block, which would probably be better suited in a seperate class.
Solution:
if you have the same problem, addElement() instead of addChild() is what did it
I'm trying to move away from mxml to actionsctipt. I have a <s:Rect> that I've created and set its properties, but having trouble adding it.
var aRect:Rect = new Rect();
//set properties like aRect.x, aRect.y, aRect.width, aRect.height
//tried adding it various ways
addChild(aRect);
Application.addChild(aRect);
Application.application.addChild(aRect);
stage.addChild(aRect);
But I keep getting the error
1067: Implicit coercion of a value of type spark.primitives:Rect to an unrelated type flash.display:DisplayObject
Originally in the mxml, it was right inside <s:Application> not nested inside anything
<s:Application>
<s:Rect id="aRect" x="10" y="10" width="15%" height="15%">
//then fill code here, removed for readability
</s:Rect>
</s:Application>
What's the deal, I thought actionscript would be nicer than mxml.
tried changing addChild(aRect); to addElement(aRect); and that worked beautifully.
It's because Flex 4 significantly changed the way the display hierarchy works in MXML-based applications. This is a bit confusing since addChild() no longer works as simply as you'd want it to - you have to add elements to a dataprovider, and then the logic of displaying those elements (which ones to add where, how to skin them, etc) is handled elsewhere. It's kind of a useful change, though, because it forces you separate your concerns in a very concrete way. Once you have your elements all added to your dataProvider you can swap out Layout objects at will (even at runtime) to change the way your application looks.
EDIT: Technically it's not the displayList itself that they've changed. It's the fact that the basic unit used by Flex is now the "Group" - even s:Application extends group. You add your content to a a Group (or to the top level Application) and then you assign the group a layout to tell it how to display the items you've added.
Yes, you need a DisplayObject. I'm not familiar with spark.primitives.Rect, but perhaps you could just create a new Sprite and call methods on its Graphics object to draw the rectangle?
According to the live docs, the addChild method of the Application class does require it to be a displayObject.
Annoyingly we will often struggle to add flash assets ( swf swc ) (display objects) using addElement.
I'm working on a way to do this right now :( more hoops and jumping
Also my swc is not viewable in the package explorer (why not ?)
It's possible within a Flex application, to declare elements, for example a HTTPService elements, both in XML and also in code.
That is, either:
...
or in code:
var hs : HTTPService = ...
My question is when should I prefer which alternative? What are the advantages of having stuff in XML vs. plain old vars in code?
MXML is great for doing declarative layouts, much easier to follow than doing everything programmatically in ActionScript. If you are using something like a ServiceLocator to define HTTPService, RemoteObject, etc in your app then declaring them via MXML is also quick and easy. Basically if you want to add anything to an object's displayList quickly and easily, MXML is a great way.
Elements are more concise and compact than AS3 code, tho of course ultimately it all becomes the same thing. You can, for example in 'one line' of mxml declare an HTTPService and set several of its properties. In script you'd need to do this in many lines of init code. If you have a lot of global variables this can start to become unruly.