A* Path finding for PNG mazes, from Ruby now in Crystal! Using StumpyPNG! Now ×40 faster than its Ruby counterpart: A-Star
Build it from source: Clone and enter repository:
git clone https://github.com/Demonstrandum/Crystal-Maze.git && cd Crystal-Maze
then install dependencies with shards
(StumpyPNG):
shards install
and run the make file:
make
sudo make install clean
#sudo make uninstall # to remove the program
To run the program itself:
maze ~/Pictures/yourImage.png
This will solve the maze and make and image with the suffix "-solved"
e.g. yourImage-solved.png
in the same location as the input image.
You can also add optional arguments of heuristic distance type, verbosity and wheather to draw nodes or not. For heuristic distance just type euclidean
or manhattan
as argument for distance calculation type, these are the only two currently supported. Showing nodes is done with the argument --show-nodes
and for verbosity (prints the nodes of the final path) either -v
, verbose
or --verbose
.
For example:
maze euclidean -v hipMaze.png --show-nodes
Maze being mapped out by flood and then solved with A*
You can animate the solving with the --gif
argument, this looks better with --show-nodes
as you can see the nodes being placed and spreading. e.g.
maze mazes/50-maze.png --show-nodes --gif
The gif will most likely be very slow, so you can drop some frames to speed it up:
maze mazes/normal.png --show-nodes --gif --gif-speed 14
The above means: only keep every 14th frame.
which generates mazes/50-maze-solved.gif
Be careful, when making GIFs with large mazes, the program will use a lot of memory and may freeze your computer!
The start and end points of the maze must be white and be from top or left to bottom or right, if this is not true, then it means the maze has a black border, in which case it will find a green pixel (doesn't have to be perfect green) and sets that as the start point and then finds a red pixel (doesn't have to be perfect red) and sets that as the end point.
P.S. It's highly recommended that the maze's walls and paths are 1 pixel wide as the program draws and reads the paths as one pixel.