I did some Googling and found various forum and blog posts, but whatever I tried, nothing seemed to work. Mindless Labs suggested using mxmlc to embed the game .swf into an Actionscript preloader, but mxmlc couldn’t digest our game .swf. Game Haxe had another approach, but apparently that one broke when haXe 2.03 got released. Using the MochiAd preloader didn’t work out of the box either, because that one would only be run after the whole .swf was loaded. So I was thinking towards creating a separate preloader .swf that would load the main game .swf – it’s a messy, multi-file solution, but at least it’s a solution.

However, I talked about it to a coworker, asking how this would be done in ActionScript. The standard approach is to move all assets to a later frame, and to put some preloader code and assets in the first frame. I decided to give it a try – mind you, the same thing had failed me 2 years ago – and suprizingly, it worked! Back then, the assets had been inaccessible, but some haXe and swfmill versions later it works like a charm.

Here’s how my swfmill .xml file looks like now:

<?xml version="1.0" encoding="iso-8859-1" ?>
<movie version="9" width="600" height="400" framerate="30">
    <background color="#FFFFFF"/>
    <frame /> <!-- add one empty frame -->
    <frame>
        <library>
            <clip id='image' import='images/image.png' />
            <font name='font' import='fonts/font.ttf' />
        </library>
    </frame>
</movie>

Line 4 is important: it adds an empty frame, into which the haXe code is compiled. The second frame contains all the heavy-duty assets, which will be available once the whole .swf has been downloaded.

My main function now looks like this:

package ;
import flash.events.Event;
import flash.Lib;
import haxe.Log;
import game.Game;

class Main
{
    static function main()
    {
        Log.trace("Starting preloader");
        Lib.current.root.addEventListener(Event.ENTER_FRAME, onEnterFrame);
    }
    
    static function onEnterFrame(event : Event)
    {
        Log.trace("Downloaded " + Lib.current.root.loaderInfo.bytesLoaded + " of " + Lib.current.root.loaderInfo.bytesTotal + " bytes");
        
        if (Lib.current.root.loaderInfo.bytesLoaded >= Lib.current.root.loaderInfo.bytesTotal)
        {
            Lib.current.root.removeEventListener(Event.ENTER_FRAME, onEnterFrame);
            Log.trace("Preloading finished");
            
            // The Game class adds images to the screen, sets up listeners, etc.
            var gameInstance : Game = new Game();
        }
    }
}

I hope this post is useful to those who are looking for a haXe preloader. :)

To give a little more insight, all game assets are packed into a .swf file using swfmill. The process is automated with a Python script, so all images, sounds and fonts are lumped together into archive.swf. When compiling the haXe code, this .swf file is passed along, so the code gets compiled into the .swf, to produce the final game .swf. Hesselboom’s guide suggested embedding the mochi library and the assets .swfs into one archive .swf. We first tried to just embed the mochi library .swf into our archive .swf, but that didn’t work. We then tried embedding both into a new .swf, but that didn’t work either.

My coworker did get MochiAds working in a new project though, just not in the actual game. So I started analyzing the swfmill output and I noticed something interesting. The fonts were giving trouble. Apparently swfmill can’t handle the case where an embedded .swf file contains already embedded fonts. However, embedding the fonts into the final .swf didn’t solve the problem…

In the end, I tried something silly: changing the embedding order of the library and the assets .swf. Guess what? It worked!

So just in case you’re working with haXe and integrating MochiAds fails – swapping two lines in your swfmill xml file might just do the trick. :)

Enjoy! :)

It’s meant for a platformer game, but it’s fairly easy to use the output for a different kind of game. Like, say, a strategy game, or just as a title screen layout file.

On the right side, there’s the bare-bone UI: the layer selection and visibility toggle buttons, the delete-this-layer button, and of course a button for adding layers. The red lines are collision lines – I’ve experimented with that a while ago and it turned out to work pretty well, so I’m migrating parts of my collision-line editor/test program to this new editor. On the right side, you can(‘t) see a hidden layer. The blue lines are the grid – which can be toggled and it’s granularity can be fine-tuned. It’s not used for snapping yet, so it’s mostly there to give me some sense of place and size.

I’m still working on various features, but it’s already a useful tool. I wrote it in Python, using Pygame (for the rendering and input handling), pgu (for the buttons) and pywin32 (for the drag and drop support). It has taken me 4 or 5 days so far, a couple of hours each day. :)

Two weeks ago, I wrote a tool that packs smaller images into larger ones. I first prototyped the packing algorithm in Python and then translated it to C#, which I used for the tool itself.

Last week, within a few days, I was able to process, verify and preview almost 2000 files for one of our games, converting them from XML into a tight binary format, all packed into custom archive files to keep things easy for the file-system. I spent most of my time thinking about the file formats and the required checks. Writing the tools took little time.

Just today I set up a few automation scripts for one of our artists. I introduced him to SVN, but, considering he’s an artist, I decided to make things as easy as possible. I wrote a few batch files that took care of updating and committing the files he needs to work with, and another batch file for converting the art. That batch file calls a Python script that generates some information based on the new art, and another script that checks for new files, which are automatically added to SVN. All he has to do now is make art, click a few batch files and press a few Ok buttons.

I’ve also been using Python to prototype a collision system for a spare-time project and to solve some logic problems, as I recently discovered a site called Project Euler.net. Quite fun! Someone also pointed me at the Google Code Jam, that also contains some interesting problems. How about path-finding with portals?

In other words, I’ve come to like Python more and more. It’s still just one of several languages that I use and it’s not the best for every situation, but I do notice how much more productive I’ve become by embracing it.