ripopt

A memory-safe interior point optimizer written in Rust, inspired by Ipopt.

What is ripopt?

ripopt solves nonlinear programming (NLP) problems of the form:

min  f(x)
s.t. g_l <= g(x) <= g_u
     x_l <= x    <= x_u

It implements a primal-dual interior point method (IPM) with a logarithmic barrier formulation. The solver is written entirely in Rust (~21,700 lines) with no external C/Fortran dependencies.

At a glance

Key features

• Primal-dual IPM with logarithmic barrier and fraction-to-boundary rule
• Mehrotra predictor-corrector with Gondzio centrality corrections (enabled by default)
• Filter line search with switching condition and second-order corrections
• Two-phase restoration: fast Gauss-Newton + full NLP restoration subproblem
• Multi-solver fallback: L-BFGS → Augmented Lagrangian → SQP → slack reformulation
• Dense condensed KKT (Schur complement) for tall-narrow problems: 100-800x speedup on m >> n
• Adaptive and monotone barrier strategies with automatic mode switching
• NE-to-LS reformulation for overdetermined nonlinear equation systems
• Parametric sensitivity analysis (sIPOPT-style)
• Preprocessing: fixed variable elimination, redundant constraint removal, bound tightening
• C API mirroring Ipopt's interface; Pyomo, GAMS, and Julia/JuMP wrappers

Quick example

use ripopt::{NlpProblem, SolveStatus, SolverOptions};

struct Hs071;

impl NlpProblem for Hs071 {
    // ... (see API Guide for full implementation)
}

fn main() {
    let result = ripopt::solve(&Hs071, &SolverOptions::default());
    assert_eq!(result.status, SolveStatus::Optimal);
    println!("f* = {:.6}", result.objective);  // 17.014017
}

See the API Guide for a complete walkthrough.