9 releases

Uses old Rust 2015

0.3.0 May 9, 2017
0.2.7 Jan 19, 2017

#1850 in Game dev

32 downloads per month

MIT license

400KB
7K SLoC

alcibiades

A framework for writing chess engines in Rust.

Why chess engines?

Simple! Chess is the greatest thing humans have invented. Computers follow closely ;)

Why a framework?

There is lots of knowledge out there about how to write a chess engine, and there is a lot of room for innovation also. Writing a chess engine is fun, but even for the simplest engine there is a lot of complex (and boring) things that have to be implemented first: the UCI protocol communication, the rules, the static exchange evaluator, and many more. Thousands of programmers have been re-implementing those things over and over again.

So, if you want to write your own chess engine, you face an unpleasant choice: You either roll up your sleeves and implement all the hard stuff from scratch, or you take someone else's chess engine and struggle to understand its cryptic, undocumented source code, hoping that it will be general enough to allow modification. This unfortunate situation stifles innovation.

Features

  • Modular design. Users can write their own implementations for every part of the chess engine.

  • Very good default implementations — move generation, quiescence search, static exchange evaluation, time management, iterative deepening, multi-PV, aspiration windows, generic transposition table.

  • Very complete UCI support (including "searchmoves").

Usage

Here is how simple it is to create a chess engine using the framework:

extern crate alcibiades;
use alcibiades::stock::*;
use alcibiades::engine::run_uci;

fn main() {
    type HashTable = StdHashTable<StdHashTableEntry>;
    type SearchNode = StdSearchNode<StdQsearch<StdMoveGenerator<SimpleEvaluator>>>;
    type SearchExecutor = Deepening<SimpleSearch<HashTable, SearchNode>>;
    run_uci::<SearchExecutor, StdTimeManager>("My engine", "John Doe", vec![]);
}

This engine is assembled from the "in stock" implementations of the different framework traits.

In reality, you will probably want to write your own implementations for some of the framework traits. Thanks to Rust's incredible generic programming capabilities, you are not limited to implementing only the methods required by the traits. For example you may write your own static position evaluator which has a consult_endgame_table method. Then you will be able to write a search algorithm that uses this method.

Last but not least, the framework has an extensive documentation.

Dependencies

~4MB
~78K SLoC