For my maze generation I decided to create a classical perfect maze, but on hexagonal grid. For this, three things are needed:
- Rendering
- Neighbouring cells
- Depth-first search
For rendering I chose SVG, since it allows me to easily draw lines. Code for this is not very nice, but it works for this occasion. It could be easily replaced with code for generation of any regular polygon, but unfortunately there would be a problem in other methods with this.
There one more method, putDots which just puts blue dots into the top left and bottom right corner, as to make the start and finish of the maze clear.
When generating maze, we need to know which cells are neighbouring for any other - for example when removing a wall, we need to remove it from current cell and the neighbouring cell. Also, for DFS, we need to know where to go next.
This turned out to be a bit difficult for hexagonal grid. If the grid should change, this method would probably have to be adjusted manually. I used two dimensional array for representation of all cells and I believe the problem cannot be overcome even if only one dimensional array is used. For example when going to right or left, in two-dimensional grid only one parameter needs to be changed, but here also the other one has to be changed. This is the main reason I did not try to make the generation general for any polygon, because this method would be difficult to implement.
Opposite to programming complexity of previous methods, DFS turned out to be easy and quite general for any maze. Two things are fixed for this implementation - the number of neighbours (on line 68) and two dimensional position of cell in the array (parameters x and y). Otherwise, the implementation should work with any grid.
I created two small mazes, which differ in parity of size. Also one large maze is generated.
