This past month, I’ve been both working on Taiji and crunching on promotional materials for Manifold Garden. (Which is out now on Apple Arcade and the Epic Games Store, by the way! You should check it out if you like puzzle games. I get no extra money if the game does well, so I am just recommending it personally.)
Seeing off Manifold Garden has been exciting. But turning back around to work on my own thing has been a bit depressing. It still has so much further to go before it will be done! I’ve been trying to keep my head on straight but it’s been a bit of a damper on my spirits.
The Breaking Point
Some technical aspects about the visuals in Taiji started to come unraveled earlier this month. One of the decisions I made early on was how to sort all of the individual graphical elements in the game. Although for 3D games, sorting is just handled as part of the perspective (except for translucent objects), in 2D games you usually set up an explicit sorting order.
In Unity, there are actually two systems you can use to handle sorting, the first is a sorting axis, which is equivalent to the Painter’s Algorithm: objects that are further away from the camera are drawn first and closer ones are drawn last.
The other system is Sorting Layers. These are just buckets you can put different objects into and you can set the order in which the buckets draw. My initial idea was to only use 3 sorting layers for the entire game: a layer below the player, the player layer, and a layer above the player. This seemed like it would work, because you are additionally allowed to specify a numerical sorting order for the objects within each layer.
The primary benefit of this approach is that it is player-centric. This means that I know that all objects in the “Below Player” layer will always be drawn below the player, and vice-versa for the “Above Player” layer.
But what happens if I want to have objects that are above the player at one point, and then below the player at another?
There are two types of scenarios where this problem might happen.
One is “vertical” objects that the player can walk around, such as trees. If we place them below the player, the player will be walking above the branches, and if we place them above, then the trunk will float over the players head. This problem is easy to solve by simply placing those objects on the player layer. In this case, Unity will fall back to the sorting axis and sort by distance. However, we can tell Unity to sort using the Y-axis, instead of the Z-axis. This means that objects that are higher on the screen than the player will draw behind them, and those that are lower, draw in front.
The other slippery sorting situation is when the player is underneath an area which they can climb up into. A basic example of this is a bridge over a canyon. The player might be in the canyon, walking underneath the bridge, but they can also climb out of the canyon and end up above the bridge, walking across it.
This scenario is challenging to achieve under a simple 3 layer (Below Player, Player, Above Player) setup. The only real way to do this is to either shuffle all the objects between the above and below layers, or have copies of the objects on both layers, and only enable whichever is appropriate depending on where the player is.
I was using a mixture of both of these systems up until recently. It worked, although it was quite cumbersome. You’re moving around of dozens of objects from layer to layer all the time, and you can’t even see any of the visual issues until you run around the game. But eventually you run into scenarios where there need to be more than two layers, and it all falls apart.
So I made the difficult decision to change the entire sorting system used by the game. Under the new setup, each area in the game has a sorting layer, and the player is moved from layer to layer as they walk around the world, always staying at order 0 in whichever layer they are in. Objects with negative sort values will be below the player in that sorting layer, and those with positive values will sort above the player.
This setup makes so much more sense. Since only the player ever moves around, I never have to worry about the environment looking any different than it does in the editor.
In fact, I feel like I should have changed things over much sooner than I did.
I think this particular type of mistake was misguided optimization, which is even worse than premature optimization. Instead of optimizing for my sanity, and the simplicity of building the game over the long haul, I tried to optimize for the number of layers without being sure that it would ever be an issue. It wasn’t a performance concern, more just an aesthetic one.
I think it’s important to accept that your game is going to be a big icky mess at some point anyway, so you just should just leave the cleanup until you can actually see what you’re dealing with.
In any case, things haven’t been perfectly rosy with the new setup, but I’ll leave that story for next month perhaps. See you soon.
Proof Of Work
Perhaps you’d like to see the work I did related to Manifold Garden? If so, you can check out the following links:
Manifold Garden Instagram (Daily Videos since July 18)
Edge Detection & Anti-Aliasing Comparison 2015 vs 2019
Now Available Trailer (Although about 80% of this was Derek Lieu, and I just polished it up and replaced a few shots)
I also did a ton of odds and ends stuff that I can’t really take the time to list here, but suffice to say it’s been a bit busy.