fastchess 0.1.1

High-performance Python chess library in C: legal move generation, SAN/UCI parsing, and NumPy tensor export

fastchess

A high-performance chess library written in C with Python bindings. Designed as a fast drop-in replacement for python-chess in data pipelines and machine learning workflows.

Installation

pip install fastchess

Requires a C compiler and NumPy. On most systems pip will handle this automatically.

Usage

import fastchess

board = fastchess.Board()                  # start position
board = fastchess.Board("rnbqkbnr/...")    # from FEN

# Move generation
moves = board.legal_moves_uci()           # ['e2e4', 'd2d4', ...]

# Apply moves
board.push_uci("e2e4")
board.push_san("e5")

# Board state
print(board.fen())
print(board.turn)                         # fastchess.WHITE or fastchess.BLACK
print(board.is_check())

# Tensor for ML (18 x 8 x 8 numpy array)
tensor = board.to_tensor(canonical=True)

# Mirroring
mirrored = board.mirror()

API

fastchess.Board(fen=None)

Method	Description
copy()	Return a copy of the board
mirror()	Return a vertically mirrored copy with swapped colors
push_uci(uci)	Apply a UCI move (e.g. "e2e4")
push_san(san)	Apply a SAN move; returns the UCI string
legal_moves_uci()	List all legal moves as UCI strings
piece_at(sq)	(piece_type, color) at square index, or None
is_check()	True if the side to move is in check
has_kingside_castling_rights(color)	Castling rights query
has_queenside_castling_rights(color)	Castling rights query
has_legal_en_passant()	True if a legal en passant capture exists
to_tensor(canonical=False)	Board as (18, 8, 8) numpy array
fen()	Current FEN string

Constants

WHITE, BLACK, PAWN, KNIGHT, BISHOP, ROOK, QUEEN, KING

License

MIT