To see a video walk through describing the code base, skip to the Code Walkthrough section.
To get the source, head to the Source Code section.
Unlike my previous article on getting a Photoshop interface into Flex, my goals here are to integrate Flash with Flex, not treat it just as a design asset tool, but a contributor to the functionality of the client; using Flash for enhancing the design, and helping reduce transition code which tends to be verbose in Flex. Below I’ve got a sample application that has integration of a complicated Flash design done using many different techniques. I describe the reasons for why you would want to work this way, compare with other techniques, and walk through the code & files in my example.
To be clear, this is a proof of concept, and uses a variety of techniques for the purposes of showing developers a variety of ways to do things. It is not meant to indicate all ways to integrate Flash & Flex, nor necessarily the best ways for all occasions. Yes, I did in fact abuse this design fo schizzle my nizzle.
Early on in Flex 2’s release, many Flash Developers recognized immediately that there were serious interop issues with Flash & Flex. This stems from the Flash Player 9’s new AVM, or ActionScript Virtual Machine that runs the new ActionScript 3 programming language. ActionScript 2, ActionScript 1, and Flash Player 4 scripting on down uses the old AVM engine. For security & various other engineering reasons, the AVM’s cannot talk to each other. This results in a SWF written in the Flash 8 IDE not being able to “talk” to a Flex 2 SWF. The need for doing such is that the Flash IDE allows creation of design content that Flex cannot create on it’s own. Flex 2 is an awesome programming environment compared to Flash, is approachable by developers, and is more geared towards large programming projects vs. multimedia. Therefore, there is much incentive to get them to work together, and not just by using a loadMovie approach (dynamically loading in the Flash created SWF at runtime).
There are a few solutions out there. One is do just that; load in the content dynamically by using a SWFLoader component in a Flex 2 project. In the cases of a ActionScript 3 only project, you merely use the flash.display.Loader class. If the content can operate on it’s own, and/or show design content correctly, this is satisfactory.
Another is to utilize a small set of classes to communicate via LocalConnection. LocalConnetion is a class that allows multiple SWF’s on the same computer to talk to each other. This helps bridge the AVM gap via viable API’s. This is asynchronous, however, and error checking is faith based.
A fourth option is to create the SWF using the Flash 9 Public Alpha up on labs. It has the ability to not only use ActionScript 3 in the Flash IDE, but can also produce Flash Player 9 SWF’s. One technique in doing this is to have have AS3 in the talk up to Flex who is loading it. There are some small gotchas with “digging” your way up to talk to Flex, but it can be done.
What does integration get you?
I’m making it sound like integration is actually not the best solution, but rather, the less of the evils. It is. This is how things are, and even when Flash 9 is released, things wont be much better. Regardless, Flash can produce design content that Flex cannot, and used together the tools are powerful with the right team. Integrating Flash into the developers work flow rather than a design afterthought is the best thing to do on larger programming projects; which is typically what Flex is used for. Either that, or the primary skill set on your team is traditional programming, and thus the natural choice is Flex vs. Flash, in which case, Flash is the odd man out, but still a valid contributor.
As mentioned above, you can use the same code. You can write a Flash component in ActionScript 3, and use that same component in Flex. You can integrate it into the same Flex project and check it into the same version control. Traditional programmers will get that good feeling they typically don’t get with Flash projects with regards to maintenance.
You can compile the same code in both programs. Since you are using ActionScript 3, Flex Builder, mxmlc, and Flash 9 can compile it. Granted, there are caveats with embedding, and it’s a bitch. If I knew RegExp, I’m sure I could get an Ant script to do it. Leif suggested to look into it, but this isn’t my strength. Either way, you can even create Flash based test cases that you only use to test the Flash symbol class using the new Document class feature. Flex, which typically wraps the symbol class created in Flash in a UIComponent class via composition, can have it’s own test case.
You can compile once in Flash. Like the MTASC days of old, mxmlc, the commandline Flex compiler, has the ability to do bytecode injection; aka compiling incrementally. What this means is, you can compile in Flash for the mere sake of “getting the design assets into theSWF”. If you later change the code in Flex, it’ll merely change the code, but leave the original SWF design assets intact. Since you are embedding the symbol class into a main Flex SWF anyway, it’ll compile the assets in the Flash SWF into the main one, and use the updated code you just changed. This is nice because it basically treats the SWF’s as little design asset DLL’s so to speak, and uses the most up to date code. Unlike DLL’s, the SWF’s aren’t needed at runtime . It also means, you don’t necessarily need Flash to compile the Flex project, which scores major points with some traditional programmers. Third party developers and/or contractors, based on requirements, can deliver just a SWF to some clients or a SWF and a class(es) without having to turn over a FLA. I always give my source to clients, but I know some people have legal reasons for not being able to do so.
Finally, you can produce more types of components. Since Flash can do some crazy design ideas that Flex can’t, or can’t quickly, you can integrate a lot of that into components to be used in Flex; 3D exported to FLV integration, various compositing, and other hardcore visualizations.
How does integration work?
For a simple example, you can see here where I use a Flash 9 button in Flex 2.
Integration starts with coding the component in Flash you wish to use in Flex. Everything is a component; you think in that mentality, no matter how small the asset, and you’re good to go. You write an AS3 class that represents the asset in Flash. Like the AS2 days of old, you make the linkage ID of the symbol in Flash point to your AS3 class. Since everything is a class now in Flash Player 9, your Flash design component asset is too. This class, by my convention, ends in “Symbol”. So, if you are creating a glowing button, it’s “GlowButtonSymbol”. The “Symbol” denotes a Flash based asset. That way, when you see the class amongst hundreds of others, you can immediately identify it as Flash specific. Typically, the only thing you would do in Flex Builder to this class is add the Embed tag at the top. Flash IDE will ignore the Embed tag when compiling.
The second step is to create a UIComponent based class in Flex. This class will use your Flash class via composition. You are not required to extend UIComponent; you could extend Container or Canvas for example. A lot of Flash components, however, are design assets using low-level, boilerplate design functionality, so UIComponent usually is sufficient. Everything has to be at least extend, or be wrapped in, a UIComponent. If it’s not, Flex will throw an exception.
You can take this a little farther. I like to create test cases, classes that test to see if the component, and only that component, works. They basically throw it on the stage, and calls their exposed methods, if any. This makes debugging a lot faster and a lot less frustrating. It may seem to take longer at the beginning; it does. You’ll find, though, in the long run you spend less time compiling your whole application only to find one stupid thing in your component made everything else blow up.
In Flash, this is just test_GlowButton.as. This class is a Document class that you input into the Document class field in the Flash 9 IDE. It attaches the component and hits some of it’s methods. I like to use keyboard commands to test methods and such. You can save this in the same Flash folder as well, thus segregating it from the rest of your code base.
In Flex, I basically do the same thing: make an application class, embed the component, and test it’s methods. Flex Builder 2 makes me keep these on the root folder, though, since they are Application files… *cough* *Flex Team* *cough* *fixplz* *cough*.
Pitfalls & Frustrations
“Is it that simple?”
No. There are a plethora of gotchas. I’ll try to list the most glaring here.
The most annoying is embedding of sound and images in Flex. For example, Flex doesn’t have a library. You therefore use embed tags in your Flex code. This associates an external image or sound with a variable, and you can thenceforth treat that variable as a link to the asset. This embeds it in theSWF, so you don’t have to worry about preloading, unlike using loadMovie on external images or Sound.loadSound for streaming MP3’s. Flex wraps these assets in special classes called “Asset” classes. Sort of the same thing when you import a sound into the Flash Library, and export it with a linkage name. When you embed a Flash MovieClip symbol in an eternal SWF , it wraps it in a mx.core.MovieClipAsset class. When you embed an image, it wraps it in a mx.core.BitmapAsset class. This is nice. You get strong-typing with embedded assets.
The bad news is, Flash doesn’t have access to these classes. I reckon you could use intrinsics, a class that implies “it’ll be in the SWF at runtime”, and allows Flash to compile. These have to be installed in the Flash 9 classes directory, though; if you put it in Flex’s project directory, things could get dazed & confused and Flex isn’t from da erf. If you have to recompile in Flash for whatever reason, you have to comment out the sound code, compile, then undo your comments. Running Flash + Flex Builder at the same time with design intensive files open uses a lot of RAM, thus your comp goes slow, further exacerbating the frustration.
Fonts. Are. The. WORST! Flex 2.0.1, which isn’t yet, apparently has fixed some font importing challenges so you can import more than just polite ttf’s. In the past, if you wanted the Flash Player 8 FlashType (aka Saffron) engine to show really nice looking fonts, you had to embed them using Flash 8, and then reference that SWF as the font source in the Flex CSS. Either way, it’s commonplace to utilize fonts willy nilly in Flash without thinking about what you are embedding. Static text fields, for example, embed fonts by default. There of course is the issue between how they are embedded. Bold or not? Animation anti-alias or readability? Did Flex already embed the font? Who killed JR?
For example, I finally get a build working last night. I then implement the preloader via the preload attribute on the Application tag. Suddenly my navigation fonts don’t work. Huh? After 2 hours, I finally got it working again. The Flashpreloader had 2 static text fields that were using the same font that Flex was using. The difference? The preloader wasn’t using the readability version. It’s crap like that that eats away at your productivity and makes you want to start smoking again. Fonts have always been hard in Flash; now you REALLY have to micro-manage that stuff.
Flex Builder caching is another gotcha. When you recompile in Flash, you have to refresh your Flex Builder project. I usually refresh the whole thing; I click on the Flex project base folder, and hit F5. This can take as long as the dreaded “Building 0%” you see on the bottom right of Flex Builder / Eclipse. Basically, it’s just copying the updated SWF to your bin folder and/or re-compiling it into our main Flex SWF . This can drive you nuts if you are compiling, and fail to see your changes. Refreshing isn’t always enough. Sometimes, if you recompiled fonts or bitmap stuff in Flash, an actual Clean project is appropriate via the Project menu. Takes longer, but it works EVERY time. Good sanity check, that is.
Flash doing whack AS3 builds is another problem. It’s probably because I have alpha bits, because sometimes, if your class path changes, it’ll compile theSWF, but not with the real class. I’m not sure if Flash 9 is caching ABC files (AS3) vs. ASO (AS2) or what, but I’ll call methods on the class, and Flash compiles, but at runtime claims the method doesn’t exist. Go strong-typing. Anyway, you’ll go nuts staring at correct code, when in fact, the problem is your class path. I suffered a lot from this since I transfer code back and forth between my Mac and PC as well as re-factoring twice.
Framerate really can suffer in Flex. There are a plethora of reasons for this, but I’m having a hard time finding the time to make reproducible test cases to track down the issues. Here are some theories. First, I think Flash is better at producing SWF’s that “know” how to preload themselves. So, if you have a long PNG sequence, the SWF will playback well because… well… because Flash is good at doing animations. The same animation in Flex, though, is slow. Now, this isn’t as clear cut as that sentence reads. For starters, most of the Flash stuff I use is in states. States in Flex utilize the Flash Player 9 DisplayList. The DisplayList allows you to have a MovieClip exist, but not be drawn. My guess is, this violates the typically, “Hey, there is a long timeline with a lot of assets about to be played, let’spreload it.” Now, it’s “We don’t have to draw anything, so just chill…”. Thus, it doesn’t preload the required assets, and the animation plays slow.
States have the benefit of removing children that are not applicable to a state you are in. So, in the case of a LoginForm class that has the states of “signIn”, “error”, and “register”, the actual registration fields are not shown. They can actually exist to support databinding, but the actual vector graphics that represents them isn’t drawn. This is great.
For Flash animations, this isn’t so great. Especially when they are shown, and it now has to play the animation AND preload the assets, thus reducing your intended frame rate. That’s my theory, anyway.
Worse, states have the nice feature of removing children from the DisplayList that aren’t being drawn in the current state. My guess is, Flash Player uses this as an opportunity to run Garbage Collection on the not-being-drawn-assets. Remember, the unofficial ways the Flash Player 8 garbage collector works is A) every 60 seconds and/or B) when RAM usages increase by 20% or more. In the case of 60 PNG’s , each utilizing a meg of RAM (compressed!), that’s a ton of memory usage. Naturally, as soon as that animation is removed from the DisplayList, he’s a prime candidate for collection. Meaning, the PNG’s are no longer in RAM. When the animation is played again, it’s slow because the frames have to be loaded into memory yet again. That’s my theory too.
For the preload, my fix was pretty simple. You basically make a frame called “asset preload”, and throw EVERYTHING that animation needs on that frame. You can even put it at the front of the animation, in a graphic, alpha’d to 0. This forces Flash Player to throw all the assets in RAM. Since Flash Player is generally good at not running Garbage Collection while animations are going, you can usually be sure your animation will then run well. Did in my tests, anyway. This does NOT work if your computer itself is running low on RAM. For example, I ran a bunch of these examples at once, and my computers GC + Flash’s I guess were running at the same time, cause it’d freeze during the animations. That’s just my CPU being a pansy, though. No soup for you, sucka.
For the latter… not sure what to do. In Flash, this was easy. You just make a MovieClip’s visible property false. It’s still there, just not being shown, but as soon as you make visible true, it draws immediately. It IS being drawn, though, and taking up system resources. The DisplayList does the same thing with remove child. You can actually have an animation playing, but not in the DisplayList , taking up resources. It’s better than visible = false, that’s for sure. Either way, my guess is, visible = false is a better safeguard against Garbage Collection unloading your preloaded animation from memory vs. removing it from the DisplayList altogether. This is easy in Flash, hard in Flex if you are using states.
Sound compression is kind of frustrating too. For Flash, there are a multitude of ways to compress it with more codecs. In Flex, it’s a little harder. If you are sharing a lot of assets, you may run across the same sound sounding different because Flex compressed it differently. Really annoying. Or, you forgot to re-comment-in your code that you commented out to compile an ebmedded SoundAsset in Flash… thus you never compiled your sound into the SWF. How’s your brain’s cookie trail?
Resource usage is the biggest kicker. Eclipse 3.1 on my PC came, get this, standard with a max of 40 megs in the Eclipse ini. That’s right, 40 megs! AH, to be a full-time Java Developer, and need only that much. When dealing multimedia content, us Flashdevs need all the RAM we can get, and then some. Specially since FlexBuilder can actually render some of those SWF’s in real-time in Design View. It actually got so bad at one point, I couldn’t even compile! Java 1.5 would collapse on my Windows XP page file faster than Mike Foley, the guy who lives in a trailer down by the river, does on your coffee table. I closed every program to no avail. After finding this blog post, I upped mine and immediately everything worked fine. For the record, didn’t have this problem on my Intel MacBook using the standalone install of Flex Builder. In fact, I gave up looking for the Eclipse.ini on the Flex Builder install (non-plugin) because the performance was acceptable.
That’s the worst of the minefield.
The rest really have to do with the new ways Mouse, Keyboard, and focus events mesh with the updated and now built-in EventDispatcher and making sure you know exactly how things are working in ActionScript 3. For example, when listening for MouseEvent.CLICK events from your Flash created button, you’ll get 2. This is normal. By default, it sends one, as does your UIComponent. Since MouseEvent click events bubble, they’ll get to whoever’s listening. You should intercept it in your UIComponent wrapper, call stopPropagation on the event, create a new event, and dispatch something more meaningful.
Here is a video tutorial walking through the code. In it, you’ll see animations created in Flash and then used in Flex in a variety of ways. I’ve also implemented someCairngorm 2 action with real dynamic data to showcase this isn’t just a proof of concept with no substance.
Please note: I cannot give you access to FLA’s and fonts because I legally don’t own the design. The video tutorial should show you enough about how they are made. The actual code that made them I did include, as well as theSWF’s. Therefore, you CAN compile this in Flex using the provided SWF’s.