ztlpn / minilp Goto Github PK

Is this project still maintained?

Hi, and thanks for this great crate! I got a panic at minilp-0.2.2/src/solver.rs:1301:10:

self.lu_factors = lu_factorize(
    basic_vars.len(),
    |c| {
        orig_constraints_csc
            .outer_view(basic_vars[c])
            .unwrap()
            .into_raw_storage()
    },
    0.1,
    &mut self.scratch,
)
.unwrap(); // TODO: When is singular basis matrix possible? Report as a proper error.
// ^^^^^^^^^^^^^^^^^^^^^ this line

I have a somehow weird query, which contains 30 variables in format a[i] = a[i-1] * 1000 and two variables x, y with x = a[30] and y = a[30] + 1 and x <= y conditions. I guess singularity occurs due floating point errors, but I know nothing about simplex and your code.

Unrelated to this, how reliable is an Infeasible answer? People can attack my code by creating some constraints that are actually feasible but my code detects it as Infeasible, so even a panic is better than a false Infeasible for me. Would it work correct even in case of floating point errors? How I can validating the result, or restrict the input to prevent that?

Hi!

I noticed there is not Continuous Integration for this one. Let me know if a PR (github actions I suppose) for that would be interesting.

Have you looked at integrating parallelisation via:
https://gitlab.com/lappsufrn/MulticoreParallelSimplex

explained in

Pdf Paper

And of course better pivot selection Langrangian or others?

I'm just starting to use minilp and it panics with an unwrapped NonSortedIndices error from the sprs crate when calling

problem.add_constraint(
                    &[
                        (node_var, 1.0),
                        (other_node_var, -1.0),
                        (additional_var, -1.0),
                    ],
                    ComparisonOp::Le,
                    0.0,
                );

The panic originates from minilp's lib.rs line 279

  CsVec::new(self.obj_coeffs.len(), expr.vars, expr.coeffs),

What am I doing wrong?