The biggest problem that I’ve come up against is that there is no easy way to set the Category for the Log.  The Flex Logging Framework allows you to set the category of the log when you retrieve the ILogger instance.  This can in turn be used by the Log Filtering system to specify which categories you would like to log.  Ideally you set the category String to be the name of the Class you are currently Logging, however, due to the rather shaky reflection tools avaliable, there’s no easy way to automatically get the name of the current Class.  To compound the problem even further, getQualifiedClassName returns illegal characters (The semi-colon seperating the Package and the Class Name) and Flex Logging Framework will throw and Illegal Arguments error if you try and use it as the category name.  So, in order to have automatic categories generated, we have to do a bit of a curly shuffle with a StackTrace and a Helper Method

This can now be used inside any Class you wish to Log by delcaring a private, static instance of ILogger, for example:

Although this is helpful it still does not address a couple of other fundimental issues:

• The output from the Flex Logging Framework does not include the enclosing method or line number from which the error was triggered (although this information can extracted from a StackTrace and automatically appended to the Logged Message by the Publisher).
• The Flex Logging Framework does not intelligently handle datatypes - unfortunatley we can only passing String’s to Flex Logger - however, we can get around this by making use of 3rd Party Consoles such as Thunderbolt (which also, helpfully includes a LoggingTarget).
• Although the Flex Logging Framework allows you to supply a custom Filters Array (and this can even be updated at runtime), it does not appear to include a simple way to configure this via an external XML configuration file. However, I’m fairly confident that I can role my own (just a shame that I have to!)

I’m going to be working on the three points above over the next few days and see if I can come up with a solution which stays as faithful to the intrinsic Logging framework as possible but allows the flexibility of some of the other Loggers out there (Luminic Box, XRay, etc).

Creating The Next Generation Happy Meal Toy

Julain Dolce

Julian is lead Developer for Fuel Industies, a creative agency that worked on a large scale campaign for the fast food chain McDonalds. The project was a rich multimedia experience centred on magical fairies and dragons. The application was delivered on CD which was bundled with Happy Meals. The campaign proved very successful pushing over 80 million units in 30 countries. He began his presentation with a brief overview of the product (marred somewhat by the fact the projector was not playing ball and he had to show it to the audience from the screen of his 15” Macbook).

Julian’s project had several key requirements which were taken into consideration during the initial planning stage:

• App. must be Cross Platform (Win32 and Mac OS)
• Installed from a CD, with no internet access required
• Handle multiple windows (for performance) and alpha transparency

This ruled out Adobe Air 1.0 as it can’t be installed solely from a CD, it also has issues with full frame transparency. As a result they settled on a mixture of MDM Zinc 3 and mProjector. They also used Install Builder, which is a Cross Platform (Win32, Linux and Mac) installation creator. Their main Flash tools were FDT and Eclipse, they made use of SVN and constructed a couple of custom tools in C# using WPF on Win32

Julian’s team ended up building most of their application in ActionScript 2, Julain cited this was because of a lack of a mature ActionScript 3 development environment when the project was started in the summer of 2007. They used SVN to keep code versioned, Mantis to track bugs and took advantage of an automated build process to get the builds out to testers. They also kept Excel spreadsheets for each library asset which detailed file locations, settings (offsets, alpha, and compression ratios for all Cell animations, sounds and linkageid’s, etc.

The team made use of JSFL, which can be used to create custom scripts inside Adobe CS Applications - one such example was the ability to apply an affect to the selected symbols in the Document’s Library.

Their major testing problem was performance bound - using fullscreen alpha transparency bought things to a crawl and meant click events went unnoticed on slower machines - as a result they split the animations down into separate transparent windows which moved across the stage (the user’s desktop). They used LocalConnection calls to get these panels to talk to each other and sync up.

Although the project appeared to go smoothly and was obviously successful, Julian did admit to one major failing. In the time between sending the CD’s off for pressing in China and the returned product - Mac OS X 10.5 shipped - and unfortunately, their application did not work on it - whoops.

Adobe Town Hall Meeting

Richard Galvan, Mike Downey, Paul Betlem, Lee Brimelow and Matt Millar.

The second talk of the day consisted of 5 members of the Adobe Team - 3 platform evangelists and two Flash Player engineers. The group asked questions asked by the audience. I managed to ask a question in the session:

“The company I work for are currently porting a large scale AS2 project to AS3 and were disheartened by the lack of support and tools made available by Adobe to assist in this. By using the Proxy class we have managed to create shims for most the intrinsic AS2 classes, but I Would like to know why Adobe did not do this and release it to the community.”

Their reply wasn’t what I was hoping for (but was to be expected). I was told that Actionscript 3 was designed to target new customers and that they wanted to add as many features as possible to the language. I’m not quite sure what to infer from this seeing as AS2 obviously had a massive user base. The group did, however, acknowledge that the lack of community support in the transition from AS2 to AS3 was a problem (although, again, made no indication that any tools would be forthcoming) . They rounded off my question by assuring that ASVM1 Content (AS1 and 2 SWFs) would continue to be usable in Flash Player releases in the future.

Another interesting question that was asked was about the support for Dynamic Runtime Languages to script the published flash applications and allow greater control at runtime. The team responded by saying that a new project - FLACC (FLAsh C++) is in the works which will allows C++ code to run natively inside the Flash Player. This could lead to other language interpreters being made available and will prove a very powerful tool for ActionScript developers.

Adobe also confirmed that Flash 10 is being worked on for the iPhone - but Apple has not confirmed it will be approved. Flash Player 10 will include the new unloadAndStop() method to help with Garbage Collection when unloading external content in the Player and that there is no AS4 on the cards - instead they will work on improving AS3 whilst pushing ECMAScript 4 forwards (they hinted that Private Constructors might be coming back)

After the Session I spoke to an Adobe Flash Player engineer, Matt Millar. I told him about our Library Linkage Problem where Child Objects placed on the stage in frame 2 (or greater) are not linked to the Display Chain on the first tick. Whilst I was asking this question, another developer stepped in and claimed it was a major problem for his team. I gave Matt my business card and the told me he would be in touch - I’m hoping that this will be fixed before FP 10 is released, but I am not holding my breath!

Grant Skinner

Things every ActionScript developer should know.

Grant is a well revered programmer in the ActionScript community, his Blog provides an excellent resource and he heads up his own development team. Grant’s talk was aimed at developers of all skill levels, but focused mainly on the grass roots. Grant gave an interesting presentation which, in his words, he wished someone had given to him 5 years ago - I’m not sure I can agree it was as life changing as Grant made out - but that may be in part due to the best practices which Mind Candy’s dev team strives towards ;)

Grant started off by assuring the audience that were no right answers in programming - don’t be driven to fake absolutes, although there are no definitive right answers there are plenty of wrong ones! Code is art, Grant stressed, learn the rules before you can start to break them and don’t get hooked on Design Patterns – they’re not the answer to everything.

Grant then proceeded to lay down some programming fundamentals - Enforce code standards as it makes inheriting files easier. Think of code like Lego, not Playdoh. If you are building a space ship in Playdoh and want to change the engine you have to pull it all apart - with Lego you can take apart and reassemble with ease - your code should be like this too. Code flows down in specialisation - at the very top of the Object stack is the Main Application Controller which includes very specific logic for this one particular app - however as you move outwards, objects should become more generic and reusable. When working with reusable code, remember inside looking out - prefer composition and event dispatching over tight coupling and inheritance.

Grant also recommended the use of Service delegates to act as a layer between Server Side logic and Controller Logic in Flash, that way, the server side logic can change with only the one class needing modification on the Flash side.

Moving towards Flash Player specifics, Grant told the audience to always keep learning - read every API you can - you don’t have to specialise but at least know what FP10 is capable of. Also remember that AS3 needs you to be very particular about Garbage Collection - always create destructors in your AS3 objects to handle this in one place.

Grant bought up the concept of using Timeline based Event dispatching. His example was a simple graphical dialogue box (like we use in Moshi), where the box just dispatches events when it finished tweening in / out - that way Classes using this Timeline code / Library Linage just need to register for its events removing any form of tight coupling. Grant proceeded to talk about he makes use of JSFL to create custom Panels and “Graphical Objects” - a single click from the command line laid out all the frames, labels and AS3 timeline code for creating a Graphical Button (including the Linkage name!)

Finally, Grant reminded the audience that we are blessed to have the Flash IDE - we can create prototype and concepts in seconds - get out there an experiment!

You can view Grant’s slides for his presentation over on his site - note that you will need Flash Player 10 beta installed as he makes use of the new 3D features present

Platform Jujitsu

Lee Brimelow

Lee (who runs theflashblog.com) came back in the afternoon to evangelise about the new CS4 platform, showing us some of the new features available across the entire product line (Flash CS4, Flex Builder 4 and Flash Player 10). He started off his speech by outlining that Flash Developers much be proficient in 5 main areas:

• Motion and Video
• Visual Design (using Photoshop and Illustrator)
• Object Orientated ActionScript
• Flash and Flash IDE’s
• Server Side Integration

With the introduction out of the way, Lee went on to talk about some of the new features in Flash Player 10. First up was native 3D - this provides full antialiased 3D inside with AS3, and, as it’s intrinsic to the player, it’s very fast! Lee stressed that this was meant to compliment, not replace more mature and capable 3D engines like PaperVision (but mentioned that there were plenty of new API’s that those guys would be hooking in to to speed things up). To accompany this, the drawing API has received an overhaul to allow drawing and manipulating programmatically generated 3D Objects

Next up was Adobe’s Advanced Text Engine - this is in the form of an AS3 API which Lee described as bewildering to anyone except “Text Nerds” - he said it provided an incredible amount of control - so much so that they were planning to release a CS4 component shortly after release which would allow “normal people” to harness it’s power - he did however say that this component would not ship with CS4.

Flash Player 10 will incorporate a host of “tweaks” including true Hardware Acceleration which will improve the speed of Video and Bitmap Manipulation (Filters, Pixel Bender operations). Local file access will be included (but will require user prompts), the previously mentioned unloadAndStop() method will be added and the new Vector datatype for strongly typed Arrays.

One of the new features that Lee was getting excited about was Pixel Bender - this allows image processing with shaders and filters and looked very interesting. Lee showed a custom demo where he created his own filter in HLSL (High Level Shader Language - which looked a bit like C#). One of the most exciting things about Pixel Bender is the fact that it is threaded separately from Flash Player - Lee said that complex calculations could be passed off to the Pixel Bener to allow for Mutli Threading in Flash Applications - if this is true, and it’s as easy as Lee hinted then this will be a major break through for Flash Player and will be immensely useful for complex Flash Apps.

Lee proceeded to talk about Adobe Air 1.5 and Flex Builder 4, neither of which sounded very exciting - alough Flex 4 will feature (yet another) XML based language for handling graphics in Flex.

Finally, Lee showed us some of the features of the new Flash CS4 IDE including live previews of 3d Images. The Flash Properties Panel now includes a new Z property alongside the traditional x and y which allows you to move symbols in space (away from the camera) - he whipped up a simple, live demonstration of this which was very impressive, however, he insisted the really cold 3D features would only be available via AS3 scripting. He wrapped up his Flash CS4 presentation by mentioning that the Flash IDE ActionScript Panel had not been touched in this release - in his own words - fail.

The Best 8 to 12 Hours of My Life

Robert Hodgin

The day ended with a visual presentation by Robert Hodgin who specialises in Audio Visualisation code - whilst not ActionScript based, his visuals provided some inspiring ideas and concepts along with a recommendation that the audience branch out and try a hit of ‘cid if they haven’t had the privilege.

Magneto Visulisation of Trentemoller’s Miss You

Usually this is a pretty straight forward afair, but I had a bit of a problem - may laptop would only find the TFTP server if it was directly wired to it via a crossover cable (no doubt my POS Speedlink router was partly to blame).  After a couple of hours of faffing I finally managed to get a succesfull install with the following steps:

1. Download a copy of TFTP32 for Windows
2. Directly wire your laptop to your desktop pc (which is running the TFTP server) with a crossover cable
3. Follow the Ubuntu Windows Server Netboot Wiki Entry to get the TFTP Server configured - make sure you set everything up to point towards your Router.  (ie: My router defaults to 192.168.1.254, so I set that as the gateway and gave the DHCP Server in TFTP32 a sensible range of IP Addresses to serve out.
   
4. When the laptop boots via PXE, follow the guide right up to the Network interface seletion (should be just after the Keyboard layout).
5. At this point, unplug the cross over cable and wire the laptop directly into the router, wait a couple of seconds for it to autosense the connection and then proceed with the Ubuntu install - it should discover your router and recieve a new IP Address via DHCP
6. Sit back and smile as your laptop succesfully downloads the required packages from the ubuntu archives.

Hope this helps someone out there

The listener object (in this case, a Controller), will then have a similar method for handling callbacks, again, without using an enum we would be left to switch on a string, something like this:

Obviously this approach works, but it can lead to some quite obvious, yet tricky to find errors. For example, what if I set the value of eventObject.state to ‘Data_Loaded’, or mashed the keyboard and entered ‘data_laoded’? Of course, both these problems can be spotted (the trace() on default will help you debug typos, and a `event.state.toLowerCase()` would help with the first problem, but what about when other people are working with the code - how will they be able to tell which states are available to them, this is especially true if there are multiple listeners to modelEvents, maybe all the state’s aren’t show in each switch() - now we’re starting to see the bigger issue - one of software design. Wouldn’t it be great if we could bring all the Model State messages into one place?

Say hello to the ModelEvent Enum Class:

This class is used a little differently to most others. As the constructor is set to private, we don’t expect developers to be creating instances of ModelEvent’s in their own code (so in the case of the Model’s __onLoadComplete() function which was shown above, we wouldn’t other developers using the new keyword to create their own instance of ModelEvent). Instead, we expect developers to access the pre-defined ModelStates via the static public properties defined inside the class, let’s show a quick example and revisit our Model’s onLoadComplete method:

That was straight forward - an now we don’t have to worry about typo’s or mixing the case, any mistakes at this point will be caught by the Compiler when you build the swf. Okay, that’s cool, but how do we handle recieving the events in the Controller? Even easier!

Of course, if you prefer you could switch based on event.getLevel() with the case statements performing comparisons on ModelEvent.DATA_LOADED.getLevel();

I will try and give a more detailed write up of some of the component features at a later date, but for now it’s back to work rolling out the rest of the enhancements to the video page!

However, one pitfall to be aware of is trying to use Acutes and other “non-latin” characters in dynamic text fields (é, etc). I was having a problem when these characters were not only not showing up, but all characters after the acute were being truncated! Here’s my configuration XML:

Side Note: Flash Player 9 appears to ignore the XML Encoding value specified in XML files, however it’s good practice to adhere to web standards, so make sure your XML’s encoding is set to UTF-8

The next obvious set is to ensure that you embedding all the required glyphs in your .fla file, you do this by clicking on the Embed… button for your Dynamic Textfield - make sure you embed all the Latin characters you need and then specify the acutes and other UTF-8 characters seperatley in the “include these characters:” textbox:

Now this is where the problem comes in - when I try to compile and run this flash movie, instead of getting the text armén in my textfield, I get arm with both the acute (é) and any trailing characters truncated (in some cases (as in the screen shot below), I even get some of the trailing XML… eek!)

After a bit of head scratching I finally managed to track down the problem: The XML file must be saved as UTF-8 encoding! (d’oh!) However, in tracking down the problem I also discovered another workaround using System.useCodepage. Setting this value to ‘true’ in the first frame of your actionscript will force flash to ignore the character encoding of the XML document.

Hope this saves someone some time.

As always, comments are welcome.

The following picture was taken in the town of Aghios Nikolaos, which features a large central lake / dock which is famed for being bottomless and surrounded by restaurants. This particular flower was found all around Crete and looked like a giant dandelion - the “ball” was comprised of hundreds of blue spikes. I took this particular picture in response to a Chris’ own picture, Drunken Bumble Bee to which mine pales in comparison!

Activating User Accounts Via Email Use Case
Goal: To confirm that users are registering with a valid email address, force them to activate their account before they can log in.

1. User registers for an account, all validations passes and $User->save() has been called
2. At this point we flag that the user’s account is pending activation. An email gets sent to the email address the user registered with. The email contains a unique activation link
3. The user recieves the activation email and clicks the activation link
4. The system (your website) handles the incoming link, checks that the activation link is correct (the hash matches) and marks the user’s account as “active” - the user can now log in!
• Alternative Path: The activation link is rejected by the system (it’s invalid / wasn’t copied correctly) - we present some helpful information to the user.

Time for Some Code!
Okay, let’s start simple - the basic User Table we created in the previous article needs to be expanded to include an “active” flag (boolean) to indicate if the user’s account has been activated yet:

Okay, let’s go ahead and hook this into our Users Controller’s login() action to stop “un-activated” users from loging in (after all, that is the primary goal of performing this work).

With this login check in place, we now need to sort out sending out the email which will actually “activate” the user’s account for them. Before we start with the controller actions, let’s defined some custom logic in the Model. As a quick side note, I work to the principle of skinny controllers, fat models (and so should you). What this means, in a nutshell - is that any logic which relates to a Model (in our case, generating the Confirmation Link) should be done in the Model - so let’s do that now.

So, incase you didn’t gather, we can grab a unique Activation Hash for any given user by calling $User->getActivationHash() from inside the controller. Let’s just break down what we are doing in the getActivationHash funciton and the reason why we’re doing it.

When we send the email to the user, we are going to send them a link which they can click on to activate their account. If we don’t create unique activation links then users would be able to “guess” or craft activation links for other users, for example, if we didn’t use an activation hash our links may look like this: http://mysite.com/user/activate/jreeves/ - Hmm, well I know that my username is jreeves, so I could guess pretty easily that /users/activate/dchang is going to active someone elses’ account… not great.

So, what is getActivationHash doing? Basically, it’s taking the datetime of when the user created their account (this will be unique for each user), adding in the Day-Month-Year value (so that activation links only last for 24 hours) and combining the whole shebang with the Security.salt value from CakePHP’s core.ini and Hashing it (with either MD5 or SHA-1 depending on your Cake’s settings). In case you are wondering, this process is called salting and it makes any unique value (such as a password, or MD5 hash), almost impossible to guess.

Okay, enough talk, let’s hook this into the register action so that this email gets sent out.

Okay, now we’re cooking - time to create the Email “views” which will be sent out with the emails - in case you are not familiar with the EmailComponent then now would be a good time to refer to the CookBook. So, let’s create the plain text email template which will contain the activation link set above:

Phew! The end is in sight, just one more controller action to hook up (and probably the most important one) - /users/activate - I’m sure you can figure out what this is going to do.

And there we have it, now when your users register they have to confirm their user accounts via Email - job done!

The problem is caused by this line (line 69, in CakePHP 1.2.0.6311) in /app/webroot/index.php:

Yep, the simple ini_set command - the clues should start coming in thick and fast that ini_set is causing the problem due to the fact taht open_basedir complains about paths with :’s in their structure. (:’s being the delimiter for automatically included paths). To fix the problem you can either remove the if logic that calls “function_exists(’ini_set’)”, or, if you prefer to fix things the corret way, modify the line to remove the final stub:

Good luck, hope that saves someone a headache! :)