chessboard 1.5.3

CLI to solve combinatoric chess puzzles.

Stable release:

Development:

Motivation

This project started its life as a coding challenge I was asked to solve while interviewing in 2015 for a software engineering position at Uber.

After the interview proccess ended, I kept toying with the code, as a playground to test some optimization strategies in Python. It is now a boilerplate that I use to:

• bootstrap CLI-based projects powered with Click,
• keep up with the current state-of-art of Python packaging,
• streamline the integration of a data stack (Numpy, Pandas, Seaborn and Conda),
• automate testing and quality checks (unit-tests, coverage, coding style and packaging),
• provide an auto-generated documentation.

Install

This package is available on PyPi, so you can install the latest stable release and its dependencies with a simple pip call:

$ pip install chessboard

In case it is not available on your system, see pip installation instructions.

Usage

List global options and commands:

$ chessboard --help
Usage: chessboard [OPTIONS] COMMAND [ARGS]...

  CLI to solve combinatoric chess puzzles.

Options:
  --version      Show the version and exit.
  -v, --verbose  Print much more debug statements.
  --help         Show this message and exit.

Commands:
  benchmark  Benchmark the solver.
  graph      Plot solver performances.
  solve      Solve a chess puzzle.

Solver specific options:

$ chessboard solve --help
Usage: chessboard solve [OPTIONS]

  Solve a puzzle constrained by board dimensions and pieces.

Options:
  -l, --length INTEGER  Length of the board.  [required]
  -h, --height INTEGER  Height of the board.  [required]
  -s, --silent          Do not render result boards in ASCII-art.
  -p, --profile         Produce a profiling graph.
  --queen INTEGER       Number of queens.
  --king INTEGER        Number of kings.
  --rook INTEGER        Number of rooks.
  --bishop INTEGER      Number of bishops.
  --knight INTEGER      Number of knights.
  --help                Show this message and exit.

Benchmark specific options:

$ chessboard benchmark --help
Usage: chessboard benchmark [OPTIONS]

  Run a benchmarking suite and measure time taken by the solver.

  Each scenario is run in an isolated process, and results are appended to
  CSV file.

Options:
  --help  Show this message and exit.

Plotting specific options:

$ chessboard plot --help
Usage: chessboard graph [OPTIONS]

  Update all kind of performance graphs from the benchmark data.

  All data come from CSV database.

Options:
  --help  Show this message and exit.

Examples

Simple 3x3 board with 2 kings and a rook:

$ chessboard solve --length=3 --height=3 --king=2 --rook=1
<SolverContext: length=3, height=3, pieces={'rook': 1, 'king': 2, 'queen': 0, 'bishop': 0, 'knight': 0}>
Searching positions...
┌───┬───┬───┐
│ ♚ │   │   │
├───┼───┼───┤
│   │   │ ♜ │
├───┼───┼───┤
│ ♚ │   │   │
└───┴───┴───┘
┌───┬───┬───┐
│   │   │ ♚ │
├───┼───┼───┤
│ ♜ │   │   │
├───┼───┼───┤
│   │   │ ♚ │
└───┴───┴───┘
┌───┬───┬───┐
│ ♚ │   │ ♚ │
├───┼───┼───┤
│   │   │   │
├───┼───┼───┤
│   │ ♜ │   │
└───┴───┴───┘
┌───┬───┬───┐
│   │ ♜ │   │
├───┼───┼───┤
│   │   │   │
├───┼───┼───┤
│ ♚ │   │ ♚ │
└───┴───┴───┘
4 results found in 0.03 seconds.

Famous eight queens puzzle, without printing the solutions to speed things up:

$ chessboard solve --length=8 --height=8 --queen=8 --silent
<SolverContext: length=8, height=8, pieces={'rook': 0, 'king': 0, 'queen': 8, 'bishop': 0, 'knight': 0}>
Searching positions...
92 results found in 119.87 seconds.

Huge combinatoric problem can take some time to solve:

$ chessboard solve --length=7 --height=7 --king=2 --queen=2 --bishop=2 --knight=1 --silent
<SolverContext: length=7, height=7, pieces={'rook': 0, 'king': 2, 'queen': 2, 'bishop': 2, 'knight': 1}>
Searching positions...
3063828 results found in 9328.33 seconds.

The CLI allow the production of a profiling graph, to identify code hot spots and bottleneck:.

$ chessboard solve --length=6 --height=6 --king=2 --queen=2 --bishop=2 --knight=1 --silent --profile
<SolverContext: length=6, height=6, pieces={'rook': 0, 'king': 2, 'queen': 2, 'bishop': 2, 'knight': 1}>
Searching positions...
23752 results found in 207.25 seconds.
Execution profile saved at /home/kevin/chessboard/solver-profile.png

Performances

Results below are given in seconds, and were run with the --silent option.

Pieces	Size	Solutions	MacBook Air [1]	C1 instance [2]
2 kings, 1 rook	3x3	4	0.01	0.04
2 rooks, 4 knights	4x4	8	0.12	0.91
1 queen	1x1	1	0	0
2 queens	2x2	0	0	0
3 queens	3x3	0	0	0.02
4 queens	4x4	2	0.02	0.10
5 queens	5x5	10	0.10	0.80
6 queens	6x6	4	0.90	7.10
7 queens	7x7	40	8.53	65.55
8 queens	8x8	92	85.80	673.28
9 queens	9x9	352	900.20	7 282.56
2 kings, 2 queens, 2 bishops, 1 knight	5x5	8	3.29	23.79
6x6	23 752	187.40	1 483.31
7x7	3 063 828	8 150.86	62 704.99

To run the standard benchmark suite and add results to the database, run the benchmark in a detached background process:

$ nohup chessboard benchmark > /dev/null 2>&1 &

[1]	MacBook Air 5,2; x86 64 bits 2.0 GHz 2-cores i7-3667U CPU; 8 GB 1600 MHz DDR3 RAM; SSD Flash storage; OSX Yosemite 10.10.5; Python 2.7.10.

[2]	Scaleway C1 compute instance; ARMv7 32 bits 4-cores Marvell Cortex A9 Armada 370/XP CPU; 2 GB RAM; SSD Flash storage; Ubuntu Vivid 15.04; Python 2.7.9.

N-queens problem solving time:

Third-party

This project package’s boilerplate is sourced from the code I wrote for Scaleway’s postal-address module, which is published under a GPLv2+ License.

The CLI code is based on the one I wrote for the kdenlive-tools module, published under a BSD license.

Other resources

• Awesome Chess: curated list of assets
• Combinatorial Generation
• Applied Combinatorics
• Extremal Problems

Changes for v1.5.3 (2016-11-30)

• Fix rendering of changelog link in RST.

Full changelog.