Struct CrossoverReport 

pub struct CrossoverReport {
    pub phase: Option<CrossoverPhase>,
    pub declined: Option<CrossoverDecline>,
    pub n_iter: u32,
    pub n_qp_solves: u32,
    pub active_bounds: usize,
    pub active_constraints: usize,
    pub estimated_active: usize,
    pub kkt_before: Number,
    pub kkt_after: Number,
    pub compl_after: Number,
}

What crossover did. Retrieved from crate::application::IpoptApplication::crossover_report.

Fields

phase: Option<CrossoverPhase>

The phase that produced the accepted point; None when crossover declined.

declined: Option<CrossoverDecline>

Why it declined; None when it did not.

n_iter: u32

Outer iterations spent in the step-4 fallback. 0 on the step-3 path.

n_qp_solves: u32

QP subproblems solved across both phases, step 3’s included.

active_bounds: usize

Variable bounds in the identified active set (AtLower, AtUpper or Fixed), read off the returned point where the primal test is exact. See [identify_at].

active_constraints: usize

Constraint rows in the identified active set (equalities included — they are unconditionally active).

estimated_active: usize

Rows and bounds the §7 step-2 tolerance test called active at the interior iterate, before any pivoting.

Compare against active_bounds + active_constraints, which is the same question answered at the crossed-over point. They differ exactly where the interior iterate could not support the inference — the measurement this phase exists to make.

kkt_before: Number

max(stationarity, constraint violation) at the interior iterate.

kkt_after: Number

The same at the returned point. Never worse than kkt_before beyond the tolerances accepts allows.

compl_after: Number

Max-norm complementarity at the returned point, measured against the declared bounds — see [complementarity_at]. NaN when crossover declined.

This exists because the interior method’s own complementarity is measured against the relaxed bounds, and after crossover the two frames disagree by the entire relaxation: an iterate sitting exactly on a declared bound is bound_relax_factor inside the relaxed one, so the relaxed reading is |multiplier| · δ — around 1e-8 — where the truth is zero. Reporting that as the solve’s complementarity printed a converged run as Overall NLP error above tol (#646). The caller substitutes this figure when the point was accepted.

Implementations

impl CrossoverReport

pub fn accepted(&self) -> bool

Did crossover replace the interior iterate?

Trait Implementations

impl Clone for CrossoverReport

fn clone(&self) -> CrossoverReport

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for CrossoverReport

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.