jv-rs

A Rust library for linear assignment problems, based on the Jonker–Volgenant algorithm. Supports minimization and maximization, square and rectangular matrices, and integer (i32, i64, isize) or floating-point (f32, f64) costs.

Installation

Requires Rust 1.85 or later. Use the GitHub repository as a Cargo dependency:

[dependencies]
jv = { package = "jv-rs", git = "https://github.com/ReinoutA/jv-rs", branch = "main" }

The Cargo package is named jv-rs; the library is imported as jv.

Usage

use jv::DenseMatrix;

let costs = DenseMatrix::from_rows(&[[4, 1, 3], [2, 0, 5], [3, 2, 2]]).unwrap();
let a = jv::minimize(&costs).unwrap();
assert_eq!(a.cost, 5);

let b = jv::maximize(&costs).unwrap();
assert_eq!(b.cost, 11);

Every successful solve assigns exactly min(rows, cols) pairs. Each row and column participates at most once; surplus rows or columns remain unassigned. Empty matrices are supported. row_to_col and col_to_row contain None for unassigned entries.