Changes by Forked

I just run get_puzzle repeatedly (n=500) in get_puzzle.py and save it with np.save.

Slitherlink Solver

For a while I was addicted to the game Slither Link (https://www.puzzle-loop.com/). In an effort to get over it, I went ahead and wrote a solver that could solve these game in an automated manner. This project is the result of that effort.

The code isn't particularly pretty, but it is definitely able to solve these puzzles. In many cases, one can apply a very simple set of rules over and over again, until the game is solved. In situations where the set of rules is insufficient to solve the puzzle, random moves can be made to see if a solution will be found.

I am mostly sick of this game by now, so it seems to have been a successful approach to getting over it.

Getting Puzzles

In order to get puzzles to solve, I also wrote a program to download the puzzle data from https://www.puzzle-loop.com into a simple file format. The program is called get_puzzle.py, and can be edited to download puzzles of various sizes. Read the comments in this short program to see what values to specify. To store the puzzle data, you can run something like this:

    python get_puzzle.py > 10x10_hard_1.txt

Note: To get this program to run, you must have the Python Requests and Beautiful Soup 4 libraries installed. I usually set up a virtual environment to do this.

    virtualenv venv
    source venv/bin/activate
    pip install requests
    pip install beautifulsoup4

Puzzle Format

Here is an example 15x15 puzzle:

15 15
 122 32 2    13
  22  222    3 
   311  1   122
 3  31  2   1  
202  1 113 2222
   21 3221 1  3
1 2     2 3  2 
3   21 2    23 
  1    213 3  3
 32    12212  2
3  3 122 2 2  3
 1    22 2   3 
3    2 1   2023
22 2  32  2    
3   21    231 3

The first line contains the width and height of the puzzle; the remaining lines contain the numbers that go into each cell of the puzzle, or a space for no number.

The puzzles folder contains many such puzzles. The naming convention tends to be that "normal" puzzles are able to be solved entirely by repeated application of the rules, and "hard" puzzles require guessing at various points because the rules are insufficient.

Running the Solver

The solver is called slsolve.py. Currently it doesn't parse command line arguments, so it is only configurable via editing the code. To run the program on a puzzle, do the following:

    python slsolve.py 10x10_hard_1.txt

The program will commence solving the puzzle, attempting to apply its rules for solving until the board no longer changes, and then switching to a random search mechanism. Eventually, the program will print out the solution to the puzzle once it is found.

Issues

Here are the current known issues with the program.

Infrequently, the program crashes with a failed assertion. It is unclear what is causing this failure, but I will probably eventually track it down and fix it.
As mentioned above, the solver doesn't take any command-line options, and it would be nice to have options like verbose output.
A far more interesting question is, how to generate hard slitherlink puzzles? This is a topic worth exploring in the future!

s-n-1-0 / pinkston3_slitherlink