: This project made me notice that the ladders leading to the Labyrinth have
: their destination points set wrong. They must get fixed during the
: building of the labyrinth, but the ladder data says you'll wind up near
: the center, instead of in the corners.
They are. The algorithm that generates the Labyrinth maze doesn't guarantee that the "corner" points, at (1,1), (1,30), (30,1), and (30,30) will be open, so the first thing it does after generating the maze is move the ladders to the points west-most and east-most open sectors on rows 1 and 30. (The algorithm does guarantee that there will be at least two open sectors on each edge.)
The violet crystal is placed at (16,16) if you've never picked it up. (The maze algorithm fills the level with stone, then makes (16,16) empty, then starts cutting random paths through the stone from there until it can't cut away any more without creating undesirable local patterns. Thus, (16,16) is always open.)
The next step is to add the Shocking Spheres. It adds exactly 20 of them, trying to spread them randomly in open sectors throughout the maze. Unfortunately, there's a bug in the way they use their random number generator, so that the only candidate sectors they consider are (x,y) where x=0..31 and y=x/2 or y=x/2+16. (That is, (0,0), (1,0),(2,1),(3,1),(4,2),(4,3), ... (30,15),(31,15),(0,16),(1,16), ... (30,31),(31,31)). These sectors lie along two lines, one from near the ladder from HHCC to the middle of the right wall, and one from the middle of the left wall to the ladder down to Lasciate Ogni Speranza. That's why when you enter the Labyrinth from HHCC or Lasciate, you're almost immediately set upon by spheres, but when you enter from Beware of Low Flying Nightmares or from Need a Light you always have a quiet arrival. They're trying to put 20 spheres in only 32 sectors, some of which will be rock, so each empty sector along those lines has an extremely high probability of having a Shocking Sphere in it.
Finally, they sprinkle detritus, putting something inconsequential in approximately 5/8 of the open and empty cells.
Something I hadn't worked out until recently is how the program makes the walls look right, putting end-caps on all the right corners. While you're playing the game, you don't even notice the walls, but it takes a fair chunk of code to do it right, and you would notice if they did it wrong.
