Linear Programming with `good_lp`: Define a tolerance gap like in PuLP

Question

I have the following example of a small example of a linear program in Rust:

use good_lp::{variables, variable, default_solver, SolverModel, Solution, constraint};

fn main() {
    // Declaring binary variables
    let mut vars = variables!();
    let x = vars.add(variable().binary());
    let y = vars.add(variable().binary());

    // Creating the modell and solving it
    let mut model = vars
        .maximise(2.5 * x - y)
        .using(default_solver);

    model = model.with(constraint!(2 * x + y <= 1)); // Adding Constrain
    model = model.with(constraint!(x + y <= 1)); // Adding Constraint

    let solution = model.solve().unwrap();

    // Printing the result's
    println!("x = {}", solution.value(x));
    println!("y = {}", solution.value(y));
}

I want to allow in the code that the found solution doesn't have to be 100% perfect. For example 99% would be also fine. If you use PuLP_CBC with Python you only need to write a command like "PULP_CBC_CMD(gapRel=0.01)".

I searched in crates.io in the docs of coin_cbc for some similar syntax/notation but I couldn't find anything so far. Did I miss something or much better do you know how to write it?

hallo007 · Accepted Answer · 2025-04-13 19:07:49Z

Ok I finally could find it in good_lp which is a bit hidden: crates.io

I also added it into my code from above. Here is the reuslt. I added a commend "NEW !!!" on some spots which are new in the code:

use good_lp::{variables, variable, default_solver, SolverModel, Solution, constraint};
use good_lp::solvers::WithMipGap; // NEW !!!

fn main() {
    // Declaring binary variables
    let mut vars = variables!();
    let x = vars.add(variable().binary());
    let y = vars.add(variable().binary());

    // Creating the modell and solving it
    let mut model = vars
        .maximise(2.5 * x + y)
        .using(default_solver)
        .with_mip_gap(0.05) // NEW !!!
        .unwrap();  // NEW !!!

    model = model.with(constraint!(2 * x + y <= 1)); // Adding Constrain
    model = model.with(constraint!(x + y <= 1)); // Adding Constrain

    let solution = model.solve().unwrap();

    // Printing the result's
    println!("x = {}", solution.value(x));
    println!("y = {}", solution.value(y));
}

hkBst · Accepted Answer · 2025-04-13 07:02:04Z

0

There is a with_mip_gap method provided by the lp_solvers feature/crate. Perhaps you can just replace default_solver with default_solver.with_mip_gap(0.05).unwrap()?

answered Apr 13 at 7:02

hkBst

3,43017 silver badges37 bronze badges

2 Comments

hallo007 Apr 13 at 17:59

That doesn't work. I use good_lp and not lp_solvers. These are two different packages.

hallo007 Apr 13 at 18:16

Ok, I just translated my program into the syntax of lp_solvers, based on this example docs.rs/crate/lp-solvers/1.0.1 But I noticed that "StrExpression("x - y".to_string())" is different to StrExpression("x-y".to_string()) (gives wrong solution!!) which is in my opinion a very bad and unflexible design especially if you have quite more complex expressions like terms with more than 50 summands. It shouldn't be impoartant if you use Whitespaces or not!

Collectives™ on Stack Overflow

Linear Programming with `good_lp`: Define a tolerance gap like in PuLP

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related