Skip to content

🧩 A Python Project dedicated to providing functionality for solving puzzles.

License

Notifications You must be signed in to change notification settings

GamesCrafters/GamesmanPuzzles

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GamesmanPuzzles

Build Status codecov

A collection of Puzzles bundled together in a simple yet powerful Python interface. Developed as of part of the UC Berkeley GamesCrafters.

Installation

Clone this repository and install the dependencies (it's recommended to use a virtualenv when installing dependencies of any project):

git clone https://github.com/GamesCrafters/GamesmanPuzzles.git
cd GamesmanPuzzles
pip install -r requirements.txt
pip install -e .

To get optional closed-form solver for Lights Out, run the following commands from the project's root directory:

cd puzzlesolver/extern
python setup.py build_ext --inplace

Run from the base directory of the repository

cd puzzlesolver/players
python tui.py hanoi

to play a puzzle of Towers of Hanoi.

Solving Puzzles

You can solve all the puzzles by running the following in the base project directory:

python -m scripts.solve

Serving Puzzles

Run from the base directory of the respository

python -m scripts.server

to access the webserver locally. The server should be running at http://127.0.0.1:9001/.

Routes

  • /<puzzle_id>/<variant_id>/start/

    • Returns the initial position of the variant of ID variant_id of the puzzle of ID puzzle_id. If the puzzle supports randomized starting positions, the content of the response will correspond to a random initial position.

      The response contains two fields:

      • position: (String) The human-readable string representation of the position.
      • autoguiPosition: (String) The AutoGUI-formatted string corresponding to the position, which tells the frontend application how to render the position.

      Below is an example response for /npuzzle/3/start/ which gives a starting position for the variant with variant_id "3" of the puzzle with puzzle_id "npuzzle" (i.e., the 8-Puzzle variant of the Sliding Number Puzzle).

      {
          "position": "7-2453681",
          "autoguiPosition": "1_7-2453681"
      }

      Since the Sliding Number Puzzle supports randomized starting positions, this response is randomized. Below is another example response for /npuzzle/3/start/.

      {
          "position": "62-581374",
          "autoguiPosition": "1_62-581374"
      }
  • /<puzzle_id>/<variant_id>/positions/?p=<position_string>

    • Returns information about the position specified by position_string of the variant specified by variant_id of the game specified by game_id. This is used, for example, by GamesmanUni when it needs to load a new position every time a user makes a move.

      When using this route to get information about a position, position_string should be the human-readable string representation of the position, and NOT the AutoGUI-formatted position string corresponding to that position.

      The response contains the following fields:

      • position: (String) The human-readable string representation of the puzzle position.
      • autoguiPosition: (String) The AutoGUI-formatted string corresponding to the position, which tells the frontend application how to render the position.
      • positionValue: (String) The value of this position. This field will either be win (the puzzle is solvable from this position) or lose (the puzzle cannot be solved from this position).
      • remoteness: (Number, may be undefined) If positionValue is win, i.e., the puzzle is solvable from this position, then this is a non-negative integer indicating the number of moves needed to solve the puzzle from this position. If positionValue is lose, i.e., the puzzle cannot be solved from this position, then this field is undefined.
      • moves: (Array): A list of move objects, in no particular order. If there are no legal moves from this position, this is an empty array. Each move object contains the following fields:
        • move: (String) The human-readable string representation of the move.
        • autoguiMove: (String) An AutoGUI-formatted move string, which tells the frontend application how to render the button that the user can click to make this move.
        • position, autoguiPosition, positionValue, remoteness of the child position reached after making this move.

      Below is the response for position -87125364 of the 8-Puzzle variant (variant_id of "3") of the Sliding Number Puzzle (puzzle_id of "npuzzle"), i.e., the response for /npuzzle/3/positions/?p=-87125364. The puzzle can be solved in as few as 28 moves and both legal moves make progress toward the solved state.

      {
          "position": "-87125364", 
          "autoguiPosition": "1_-87125364",
          "positionValue": "win",
          "remoteness": 28,
          "moves": [
              {
                  "autoguiMove": "M_1_0_x",
                  "autoguiPosition": "1_8-7125364",
                  "move": "8",
                  "position": "8-7125364",
                  "positionValue": "win",
                  "remoteness": 27
              },
              {
                  "autoguiMove": "M_3_0_x",
                  "autoguiPosition": "1_187-25364",
                  "move": "1",
                  "position": "187-25364",
                  "positionValue": "win",
                  "remoteness": 27
              }
          ]
      }

Testing (Broken)

To run all the tests, run the following command:

pytest --cov puzzlesolver

Exploring GamesmanPuzzles

Tips for exploring this repository:

  1. Follow the guides and learn how to create a puzzle and a solver!
  2. Definitely explore the puzzlesolver in depth.
  3. Understand what a puzzle tree is.

Contributing to GamesmanPuzzles

See contributing

Contributors:

Spring 2020: Anthony Ling, Mark Presten, Arturo Olvera

Fall 2020: Anthony Ling, Mark Presten, Brian Delaney, Yishu Chao, Sophia Xiao

Spring 2021: Anthony Ling, Mark Presten, Mia Campdera-Pulido

Fall 2022: Linh Tran

Spring 2023: Christopher Nammour

Current: Cameron Cheung, Robert Shi

About

🧩 A Python Project dedicated to providing functionality for solving puzzles.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages