Speaking of
Loonyland X-2: Challenge Of The Tunnels: The Withering Tree: Axial Tilt Omega, Woody is now fully animated in-game and running around! Provided you only use a single axe and nothing in your off-hand. In reference to the art layering system from the last post, I think what I've settled on is simplified, which is always the right answer: a full different animation set for each combination of gear. This is the easiest to implement, the easiest art pipeline, and possibly even uses less data. That's because with layered art, I had to make frames for every possible animation you could do, with every weapon, even if that weapon can't do that animation, just to keep the animations aligned. Obviously those could be blank frames, but that would also be a pain, to have to blank those frames each time. This new method means with each type of gear he only has the actual animations that that gear can do, which cut his total animation frames to about 1/3 of what I had (multiplied by the number of different gear combos, but the other method also needed that, just potentially trimmed down to only the different pixels between items). It has its own challenges, but overall I think it's the simplest.
If all that sounds confusing, imagine trying to figure it out in the first place! But in other confusing complexities, here's a fun screenshot from my current work:
The game currently outputs that text file on startup as a test. This is my
Roidvane Sequencing Tool (I am absolutely trying to make 'fetch' happen, since 2011!). It's kind of like a view one level more abstract than the map generator - it is figuring out what items should go where on each floor of the dungeon, and what other items should block your access to them (i.e. Pvt. Public is the guy who gives you the Pine Key, so to get Skill #156, you need to have rescued him first, because Skill #156 is behind a Pine Door). Then the map generator's job will be to produce a physical layout that implements this plan. Note: there aren't 306 presents or 209 skills - each collectible just has a unique ID.
It was a very fun process figuring this out once I actually sat down to do it. Each item has a preferred depth (so you gain certain vital features early on, and there's a general progression of keys and puzzle types, and there are some endgamey elements that will normally appear way at the bottom), and the deeper an item belongs, the more likely it is to be blocked by a roidvane, and it is more likely to be blocked by a roidvane obtained at an earlier dungeon level than it is on (so you don't usually need an item from the 50th floor to get something on the 1st floor... but sometimes you do!). Of course it's all random, so your dungeon may have some real oddities like not getting Phileas (who is
very important!) until way down deep, but that's part of the variety.
The trickiest bit is that it then has to do a check for dependency loops (need the Pine Key to get the Mitten Key, but the Pine Key is behind a Mitten Door, uh oh!). Fingers crossed that I got that right, as it's kind of hard to prove it's not just been lucky in my tests. We'll find out when testers play it!
The parenthetical numbers are the floor that the roidvane is found on, so as you can see above, you have to reach level 38 to rescue Major Fishbug before you can get the present on Floor 2. But that's because, on the first few floors, there's no room for you to have gotten roidvanes above them. And yes, that is something that is handled to a degree! The first 4 floors of the dungeon have greatly reduced chance of needing any roidvanes at all, everything just sits out in the open. By the bottom of the dungeon, almost everything is behind some kind of barrier, but of course by then you have nearly every roidvane so it's not really an impediment.
How this all translates into an actual physical map you explore is the next big challenge, and I'd say probably 14 to 93 times as difficult as this was. I've swung back and forth between pure generation (traditional maze/dungeon algorithms that lay it out one block at a time) and putting together hand-built map segments. I have settled on the latter, but the hand-built rooms might be connected by algorithmic tunnels (which allows the rooms to be of arbitrary size and positioning, rather than evenly spaced on a grid), we shall see. At this stage of the development, lots of the work is just plain mental, imagining these systems in the abstract so I can break them down into something that can be created. That's how I spend my time going to sleep every night, just arguing with myself in my head. It's weird.
And always looming over it all is the terrible fear of how bad the non-Supreme level editors are... maybe porting that over is worth the effort.
