mjrlowe / maze_generator Goto Github PK

Low priority

Some algorithms could have functions that pick neighbor cells to connect to that aren't completely random. Some examples:

• Horizontal and vertical bias (e.g. picking a vertical neighbor is twice as likely as picking a horizontal neighbor)
• Pick neighbors according to a perlin/open-simplex noise function (creates blobby patterns)
• Other functions (see this p5.js sketch I made a while ago: change "select neighbor mode")
• Custom functions? It wood be cool if the user was able to give the algorithm a custom function to pick the neighbors with. (Maybe it takes in the cell's x and y position and it's possible neighbors and it outputs which neighbor/direction to chose?)

This will work best for algorithms like the recursive backtracker, where almost every step we have to make a decision about which neighbor we should connect to.

I haven't really thought this one through much but something which allows for the creation of loops in the mazes would be cool.

This would be a property, probably a number, which would be passed in when calling Maze.create().
E.g.:

let recursiveDivisionMaze = Maze.create({
  algorithm: "recursive division",
  width: 10,
  height: 12,
  seed: 28542
})

This should always produce the same maze, provided the other settings are the same.

I am not entirely sure how to implement this though:

• I think it is probably easiest to rely on a third-party module for random seeds but I'm not sure how/if I will be able to get this to work with both Deno and Node.
• Should I overwrite Math.random with the seeded random function? This will mean that the algorithm code will work by itself without the seed function stuff (although the seeds won't work), but it probably isn't good practice to overwrite a standard function(?)

I have started this but I still need to get it working.

An option to have multiple start points for some algorithms could be nice.

Algorithms that this would work well with:

• Recursive backtracker
• Simplified Prim's
• Aldous Broder

I don't know much about them so this will require a little bit of research, but hopefully it will make development easier in the future.

Currently we have two coloring modes: "normal", which is just one color, and "distance", which colors the cells according to the distance from a certain point.

Here are a few ideas for some other coloring modes that might work well:

• Number of neighbors
• Number of times visited
• Direction (or direction bias if that ever gets added)
• Set (particularly for Kruskal's but could work for other algorithms like Eller's)
• Order added/connected
• Order you would pass them if you went from the start using the left/right wall following method.

E.g. hexagonal tiles, circular mazes, and irregular shapes.

This will probably require a rewrite of much of the code, so it is more of a long-term aim.

At the moment, the properties related to colors in the display function can only be in the form of hex strings.

Eventually I hope to support all strings which are valid CSS colors (and perhaps even some formats that aren't).