As the title says what's better for writing views on mobile application : AS or MXML, in performance and visual?
I've tested the two ways and I found out that sometimes the mxml works better than AS
I disagree with #Apocalyptic0n3: HTML and JS are 2 completely different languages, whereas MXML is just a declarative way to write AS. MXML code is compiled to AS first and the generated AS is then compiled to ABC bytecode, just like regular AS.
This means that as long as you use the Flex component hierarchy for writing your components in AS there will be hardly any performance difference. And anything you can write in MXML can be written in AS. Some of the generated code may be useless to your specific use case, so there's some optimization that could be done there, but it would hardly be noteicable in most cases. (If you write pure AS components outside the Flex framework, that's a whole other story of course)
The main difference is the speed at which you write your code, because you're going to put a lot more time into writing a component in AS as opposed to MXML.
In MXML you would also naturally use data bindings, whereas in a component written in AS you would usually set properties by overriding the commitProperties() method. There is a significant performance difference to be made here, but it doesn't mean you can't use bindings in AS, or you can't override commitProperties in an MXML-based component.
MXML is far superior than AS3 in terms of GUI layouts. It's faster to write and far, far more flexible (and you can still use AS3 to boot), and generally faster in overall speed, especially since you do not need to bind resize listeners. MXML and AS3 are meant to be used together, just as HTML and JS are.
Related
they both seem to accomplish the same things with different syntax, whats the point of using two different technologies. Please highlight every possible benefit of using mxml. Also are there scenarios when one is more beneficial than the other and why.
Please clarify this runtime behavior of mxml vs AS3 as discussed in Oreilly Flex 4 Cookbook page 1:
" Many newcomers to Flex wonder how MXML and ActionScript relate to one another.
The MXML compiler (mxmlc), after parsing through the different idioms, translates
them into the same objects, so that this:
<s:Button id="btn" label="My Button" height="100"/>
and this:
var btn:Button = new Button();
btn.label = "My Button";
btn.height = 100;
produce the same object. The major difference is that while creating that object in
ActionScript (the second example) creates the button and nothing else, creating the
object in MXML adds the button to whatever component contains the MXML code.
The Flex Framework handles calling the constructor of the object described in MXML
and either adding it to the parent or setting it as a property of the parent. "
Usually,
MXML is used to build the user interface
ActionScript is used to code the logic of your application
Of course, you can also build the user interface with ActionScript but :
it will take you more time
MXML views are easier to read than AS views
In terms of user interface, ActionScript should be used only to manage view elements (add, remove, ...) at runtime.
Edit
Within a Flex application, one is expected to favor using mxml over AS, provided both uses are available. In case mxml doesn't provide a solution , the user can fallback to AS3. Flex is a framework, therefore you would expect that the emphasis is put towards mxml, rather than AS3. Like most frameworks , the idea is to hide the details in order to facilitate design and supposedly make programming more accessible.
Everything comes at a cost though, you can't really expect mxml to be faster than AS3 since it's built on top of it. A logical conclusion would be to decide on a case basis. What are your app requirements, is there really a performance gain in using this component or not etc... Apart from the fact that Flex as a whole is bound to be slower than AS3, I don't think you can set general rules. One could add that if one is worried about performance, why choose Flex in the first place...
End of Edit
The Flex framework is geared towards building RIAs, therefore it comes with a bunch of components to that effect and mxml which mainly deals with the layout of these components, Actionscript being the language the Flex framework is built upon, will typically be used for the logic of your application.
The question is not really why choose one over the other. It's the same approach you would have if you decide to develop using a language directly or using a framework based on this language. It's like comparing JQuery & Javascript or CodeIgniter & PHP .I don't mean to imply that the reasons to use JQuery are the same regarding using Flex. Whilst JQuery will attempt to hide browser specificities and make JS more accessible , Flex's purpose is entirely different.
If you're developing RIAs, Flex makes sense, if you're developing games, not really. All you can do with Flex can be done with AS3 , only you wouldn't want to spend time developing DataGrids or HBoxes whilst you could spend more time on your app structure & design.
On the other hand, if you're developing a website or a game , you wouldn't want to add the weight of the Flex framework to your swf, simply because you've added a few Image components and a couple of Labels.
Flex is not adapted to every situation, unlike my previous Javascript or Php examples , Flex's purpose is not to make the language it's based upon more accessible or easier to deal with. Flex has a very specific target.
A similar question might be "what is the advantage of using HTML over Javascript for the UI?"
I suppose you are really asking "what is the advantage of using a declarative UI over an imperative one?"
Declarative UIs are more like configuration -- no logic (or in the case of MXML, very little logic). Declarative UIs separate the specification of the view from the behavior and data.
Tooling almost always prefers declarative UIs as well, because declarative specs tend to be extremely well suited to serialization... much easier for a tool to update MXML than ActionScript with properties and such.
As you get more familiar with the concept, declarative UIs start to feel more natural, and most people tend to prefer them.
Productivity, you create view elements with MXML faster than with AS3.
Under the surface MXML get's converted to the same AS3 bytecode as if you'd write in plain AS3 so the question if MXML makes the app slower comes down to how well the Flex compiler does the conversion.
As for the original question I would say that MXML is for visual development but as said before you hit the limits when trying to create custom components.
Personally I write Flex apps in 99.9% plain AS3 because I can't stand MXML. The only MXML file is the required Entry class so the compiler knows it should be a Flex app that it compiles.
Either way your app is going to get converted to an ActionScript object.
The advantage of using MXML is user friendly and more control on the look and feel of application. sometimes its easy to fix issues aligining your components on the right pane or panel with MXML.
The advantage i see using actionScript is command over your application, you have more control. Sometimes your MXML does not get you the itch of what you are looking and actionScript gives you more power.
I would suggest you to play with both to build an application which can give you less headache and better results.
None, Flex is for learning, as you learn more you use more .as and then one day... you use 0 Flex.
For one it's slow.
(edit - answer to comments)
'Why is flex slow' - it creates computer generated .as code. Equivalent functionality in .as is faster and smaller.
'MXML separates UI from code' - I'd say no it does not. First of ... it's XML.
If you are new, check out other frameworks, like Away3d, FDT and MinimalComponents. As you get more advanced, you use less MXML and one day no MXML, you are writing pure .as.
Anything in Flex can be done in HTML5/.js better. Also .as is much better than HTML5/.j
It's like a learning tricycle, and good for corporate developers that do drag and drop, they think of themselves as programmers. Software engineers... don't stay with Flex for long before evolving.
Here: http://www.youtube.com/watch?v=TL2t8eOs1XE
I feel sorry for you if you stay in Flex and don't advance.
I have a project where we are trying to skin Spark components from a third party library that are built up from the drawing API primitives. Our first attempt involved creating a Flex skin (SWC) in Illustrator/Flash and applying that skin to the Spark components via CSS. We found that even with only a single instance of one of the components skinned in this way on stage, the application was brought to it's knees. For example, it failed to respond in repaint scenarios in a timely manner, and exhibited all the symptoms of being hung.
Our next approach will be to recreate the same skin artifacts in MXML classes that are in turn associated with the third-party components. Although the complexity of some of the skins leads me to believe that we might not avoid performance problems with this approach either.
My question is the following: Is there a correct or recommended way to apply either a SWC or based skin to a component that is built up from the Spark drawing API to begin with? Even as I ask the question I recognize that the approach means using the Spark drawing API in two places and will probably have unexpected consequences.
I've had good luck building the skins using MXML skin classes that inherit from s:Skin in terms of rendering performance. The recommended method (from Adobe) is to start by copying an existing skin file from the SDK and modifying the MXML to suit your needs. Regarding drawing, FXG works very well for defining how you want the skin to be drawn. FXG is XML based, so you don't need to write ActionScript to do your drawing, like we used to do for Flex 3 programmatic skins. You can also create skins using Adobe Catalyst. I have not used Catalyst, so I can't speak to its effectiveness, but many people seem to like it. Hope that helps.
This may be more of an opinion poll than a question, but here goes. I have a fairly complex Flex application that uses many different components. I want the application to be able to custom styling and/or skinning. I also want to be sure that the styling approach makes the most efficient use of system resources (memory, CPU, network), and that it does not adversely impact performance.
We currently are using a hybrid approach to styling the application that uses both CSS and a Flash skin created in CS3. It seems odd to me to have a two-pronged approach to styling an application and ideally would like to settle on one approach or the other. Can someone outline the advantages and disadvantages of using either exclusively CSS or a Flash skin? Is there a justification for using both within the same app?
Using CSS exclusively seems advantageous because it is approachable by anyone with a basic understanding of other uses of CSS, especially when using the Flex 2 Style Explorer. On the other hand, I know that the CS3 skin we use declares up/over/down/disabled/pressed skins for each component where these states are needed and Flex manages the transitions to the appropriate skins as each state is (de)activated. However, my impression is that creation of a Flex skin in CS3 requires somewhat specialized skills and tools.
So what would you recommend for styling a Flex application if performance, customization, and efficient resource utilization are priorities?
First, I would suggest using Flex 4 if you haven't started already. Flex 4 is just about to be released.
With Flex 4, they have a whole new skinning architecture. This means you'll never use Programmatic Skins again. Plus they have Advanced CSS Selectors, so you can do way more with CSS in Flex 4. Programmatic Skins are very hard to make dynamic anyways. In the end, you'll end up creating an architecture like Flex 4 Spark Skins if you try to abstract out the logic for Programmatic Skins.
Customization
What we're doing to make Flex 4 Skins and Styling as dynamic as possible is the following:
Style Palletes: Either static classes, or XML files
Custom Skins
CSS
Style Palletes: These are static classes that hold all of the variables you use in either CSS or Skins. For example, you can have a ColorPalette that has all of your colors: public static const ColorPalette.BRIGHT_RED:uint = 0xff0000;, etc. You can then use these in Flex 4 Skins via binding, such as <mx:Button color="{ColorPalette.RED}"/>. The only problem with static classes is they they're not dynamic (you couldn't define an arbitrary number of colors for them). Same with XML: you'd have to hard-code (eventually) a certain XML structure in order to use them in Skins. You could even have the XML define ColorPalette values (so ColorPalette.ACCENT_COLOR could be set to anything from XML (gold, bright blue, or any accent). Fun stuff.
Custom Skins: Flex 4 allows you to easily skin all the Flex 4 and Flex 3 (datagrid, accordion, etc.). So if you don't like how Panels only have 1 drop shadow and no gradients, you can add a background with 4 gradients and 2 dropshadows (for 2 angles of light), and maybe some rollover effect. And then you could apply your ColorPalette.BRIGHT_RED to the shadow (or a glow :)). And you could have XML configure the ColorPalette settings. Here's a cool custom Flex 4 Skin
CSS: Then in CSS, you could write a very simple Palette parser that applied your ColorPalette values (for example) to CSS. Check out the Degrafa CSS Skin for some code examples.
Performance
The only problem with Palettes and CSS is performance. By definition, static class and binding decrease performance. Binding causes lots of event dispatching in the background, so things like <mx:Button color="{ColorPalette.RED}"/> decrease performance (not noticeable of course, since there are much worse applications, such as using built in Adobe Tweens), but it does to some degree.
Same with CSS. In Flex, every CSS Selector (Panel { properties:values... } is converted into an Object. So if you have 100 selectors, you get not only 100 extra object, but 100 extra classes in your application, at least! And that definitely adds to your swf size. And all of their properties are dynamic, and by definition, dynamic classes are slower than non-dynamic classes. Plus, setting a style (with setStyle(property, value) in Flex is very slow because you have to loop through all the children. If you're setting styles via an admin system (or something like Flex Style Explorer like you suggested), then you don't really care. But because they use setStyle(), you can see a decent performance hit when changing styles.
So if performance was a super big issue, the best option would be to hard code every style value, and not change styles at runtime (who needs to do that anyway (unless your in the admin section)?). How do you hard code styles while still remaining agile? Use Palletes (ColorPalette.BRIGHT_RED) and the like, and convert the binding expressions (<mx:Button color="{ColorPalette.RED}"/>) to static declarations via some preprocessor (I use ruby to compile my swfs, so I just match patterns and replace). So it'd convert <mx:Button color="{ColorPalette.RED}"/> to <mx:Button color="0xff0000"/>, and you get a performance boost.
In the end though, it's best to create custom skins and define the styles there, either through:
static declarations <mx:Button color="0xff0000"/>
palette declarations <mx:Button color="{ColorPalette.RED}"/>
xml declarations <mx:Button color="{myxml.#button_color}"/>
Then you can define some really cool and advanced skins and have them wired up to a palette or xml file and have the user edit that. Trying to build a very cool and modular skin/styling system with Flash or Flex using their built in stuff is not easy (nor desirable). So it's best to
create a set of skins (and package them up as a theme)
wire skins to xml
let user modify xml
Avoid CSS, it's too processor intensive, unless you need advanced selectors like you'd use in HTML because it does make life easy. But for modularity and customizability (and performance), create Skins and wire them to XML.
I myself create a basic theme that I can use in all of my client projects, something that defines my style. That consists of 3 packages (myskins, mx, spark). In there, I copy/paste all the default Flex 3/4 skins and start customizing. You can download a Scale 9 Skin Project for free to have most components ready to skin and test. Then I just merge my design (photoshop or whatever) into Flex however you choose. For all the colors I reuse, I add them to a palette and eventually to XML. Then if I need a slightly different skin and I use that multiple times, I use CSS to define a new style for a skin (maybe a list with no background color or border). My CSS is very lean. Then I can just copy/paste my theme/palettes/css to new projects, modify it as needed, allow the user to edit the colors and positions via xml (or an admin interface) and I'm ready to go.
You can't do that much with just CSS.
Hope that helps.
Lance
I've always exclusively used CSS. You stated it best that it is approachable by anyone. We can delegate the design to people that have no idea how to code in actionscript and they have no problem working with the css and the style explorer (updated for flex 3 at http://examples.adobe.com/flex3/consulting/styleexplorer/Flex3StyleExplorer.html, haven't compared too much to find out whats different vs the flex 2 version).
If performance and flexibility are the two most important considerations for your skins, you should consider Programmatic Skins.
See: Programmatic skinning - QuickStart
To quote Adobe:
"One advantage of programmatic skinning is that it provides you with a great deal of control over styles. For example, you cannot control the radius of a Button control's corners with style sheets or graphical skins, but you can by using programmatic skins. You can also develop programmatic skins directly in your Flex authoring environment or any text editor, without using a graphics tool such as Adobe® Flash®. Programmatic skins also tend to use less memory because they contain no external image files."
Fore more info, you can also read "Creating programmatic skins" in the Adobe LiveDocs Help Docs.
Flex newbie question: can MXML be generated on the fly, like HTML is generated by a server?
If yes, is it ok to do so or am I missing an important Flex architectural principle.
Sort of... but it still needs to be compiled. MXML is not rendered directly, it is first compiled into ActionScript 3.0, and then into a typical SWF... so, you cannot serve your users with MXML. However, like almost all programmming languages, you can use automatic code generation to ease development tasks.
There is IIS/Apache component given by adobe that can generate your mxml -> html+swf using on the fly compilation, and it also caches the last compiled file.
However its not recommended for bigger projects as it has certain problems that you cant make libraries of your code and organize code accordingly. Namespace usage is very limited.
MXML is just a way of describing user interface layout and scripts. This information is then "compiled" into ActionScript, then converted to SWF format for use in the Flash player.
So, yes, you can generate MXML on the fly, in that you can create a text file that contains valid MXML syntax, then use mxmlc to compile it, but there's no way (that I'm aware of) to create MXML and "add" it to your current movie such that the information appears as it would were it compiled.
It is not officially supported but there is a few similar projects doing stuff similar to this.
as3Query let you create things with xml.
Using createComponentFromDescriptor() maybe can get what you want.
Should be more as I remember, but can't find it out at this moment... Searching for "MXML runtime dynamic compilation" or something like that should be helpful...
I always thought that separating the UI from the logic is the way to go like the use of class files in Flash CS3/CS4 or MVC in web frameworks but recently there are plenty of examples and posts using ActionScript embedded in MXML.
Is there a benefit to doing this? Am I missing something.
Don't get too caught up on the "rules" of MVC and the like. Ultimately you want your code to be setup in such a way that it will minimize the impact/churn of future changes. It's less important "where" the code is and more important what and how it's interacting with other code. I see far to many people take a simple component like a login screen and make models, view, controllers, view helpers, service facades, etc. It ends up being like 75+ lines of code for something that should have been about 20. It also spreads the code amoungst a bunch of files rather than one. So in the end, you've made maintenace hard, not easy. I'm not arguing for ignoring design patterns, but rather being practical about it. Hope this helps.
In my experience, there's not much of a difference whether you write a component in ActionScript + MXML, or solely in Actionscript. Personally, if I'm writing something that's layout heavy with a little logic, I'll do it in MXML. Conversely, if it's code heavy and has little layout, I'll write it in Actionscript.
I suppose you could always separate the view and model, but then you have more files to maintain. Sometimes it makes sense, but I have yet to see a compelling reason to do it all the time.
I go by a few simple rules:
If a bit of code is just a single short line and it isn't something you can reuse elsewhere -- mere glue code -- put it directly in the MXML tag.
If it's too long to fit comfortably in the MXML tag or it is something multiple components in that MXML file can use, put it in a function in a Script tag within the same file, and call the function from the MXML tag(s).
Everything else -- full classes, large functions, things used by multiple MXML files -- goes in a separate .as file. This includes single-LOC methods that are part of a larger class, as opposed to the standalone functions the other rules cover.
I agree with Erich Douglass, to a point. Generally, if I can do the layout with MXML it's just easier to maintain. I mean, who wants to write all that crap in a createChildren override if you don't have to? Besides, the MXML is easier to read.
Whatever you do, though, remember that if you are merely calling a separate script file for an MXML component (Foo.mxml calls Foo.as for its ActionScript) you are losing the code highlighting feature that helps you work between the MXML and the ActionScript. I like to keep them both in the same place, so when I click on an ID in an MXML element I can see at a glance where it's being referenced in the AS.