Thanks for the comments, everyone
It was a rush the past few days, but I do have a result I'm happy with--which is more than I can say for about half of my LD games.
I felt like writing a proper post-mortem, which is an exercise I don't usually do... but it will respond to some of the points others raised (lag, etc.) I never was completely clear about what resources were allowed, etc., so I figured I should at least be upfront about what tools I used and how I used them.
Here's a little bit about how the game works.
All of the game's 113 rooms are really a single scene. When you first launch the game, a variety of XML files are loaded and parsed, creating a sort of database. This uses the FAST XML parsing extension I published recently. I also wrote a debugger prior to the contest. If you were wondering what the text is that displays when the game first launches, that is it. You can drop back to this command prompt if you press backquote+esc when in a proper game scene (not the title screen). Type "help" to see the commands--but most of them don't work. Type "launch" to run the game again.
Each entry in the database contains the floorplan for each level (four height levels in each room), and a list of scene behaviors and attributes. When you enter a new room, the floorplan tells the game what actors to load, and then the list of behavior,attribute,value triples configures them. This process might cause a bit of lag for the more complicated rooms.
The isometric scene renderer is close to the one that is provided with my isometric tools extension. It implements the painter's algorithm to draw the scenes--which is not very optimal. Unfortunately, Flash just isn't a great target for this game. I tried an HTML5 output, but it didn't compile. By then I was on the last day of the jam, so I decided just to go with what I know works. (Note--I haven't tried the newer release of Stencyl yet. I used build 9180).
The physics routine is pretty simple. Every actor has a z-position coordinate and z-velocity value stored as an actor value. Gravity is implemented by subtracting a bit from the z-velocity. The 2D scene is divided into four regions. When an actor falls or rises from one height level to another, its (x,y) scene coordinates are transformed so that the actor is in the right region. This is the strategy I used in the "Thief of Vallas" game and demo I worked on over the summer.
The platforming in this game could have really benefitted from the lighting routines I wrote in May/June. Shadows would make it easier to see where the various structures are positioned in the room. As it is, there are a few rooms with illusions. In some cases, I could take advantage of this to place secret walkways... but that would need to wait for a revision of the game. I didn't use the shadow routines, though. In part, I didn't think it would be within the rules of the contest to use them. On top of that, the game seems to just barely run on Flash as it is. I might do a post-jam version with the lighting, though... and see if it would run better on HTML5.
I'm happy with most of the graphics. I like the style the 6-color palette gave the game. Given more time, there are several graphics I'd like to touch up. I wanted more variety in the look of the enemies. I think I'll keep the diskette statues. I wanted to put a secret room with a joke about diskettes, but didn't have the time.
The funny thing about the hurricane is that I think it actually helped me get this game done on time. I didn't get very far before I realized this was way too ambitious of a project for a 10-day period. It's very scaled back from what I intended--and I really could have used an additional day or two for quality control (debugging, etc.) I did successfully complete it, though.
I really pushed myself to get the isometric scene renderer working the day the jam started--on the Thursday/Friday. I was looking at a very busy week at work the next week, so I knew getting that milestone beforehand would be important.
Friday night the storm arrived. I got a lot done Friday and Saturday morning. The scene renderer and physics were complete by then. Saturday afternoon the full hurricane hit, and then we lost power.
I had no power from Saturday to Tuesday. The other half of my town got power back Sunday, which included my office at work. So I was able to charge flashlight and laptop batteries over there. I spent Sunday mapping out the entire game, and Sunday night/Monday writing the database for all 113 rooms.
My university closed Monday - Wednesday, so I didn't have much else to occupy myself with other than this game. By Thursday I had the levels and code pretty much done. I just needed to draw graphics for the moving sprites. It didn't take me too long, since I used 3D models scaled down to 24x32 pixels. Between the small size and limited color palette, I got functioning graphics fairly quickly. I think they could use smoothing out, though... so in a revision of this game I'd probably just re-draw them.
Final thoughts-- for a 10-day period with a major storm, I think things went about as well as they could. It was good to have a diversion. For the entire jam period after the storm, my town was a virtual island. Almost every route out of town was either flooded out, or was closed due to a sinkhole. There isn't much to do when there's nowhere to go. On the plus side, I was about to put in about 100 hours or so of work.
