Struct Nlp 

pub struct Nlp<P: Problem> { /* private fields */ }

Builder: Nlp::new(problem) then .var_bounds(..) / .x0(..) (which fix the number of variables) and .solve().

Implementations

impl<P: Problem + 'static> Nlp<P>

pub fn new(problem: P) -> Self

A new builder for problem. The number of variables is inferred from the first of var_bounds / x0 you set (they must agree); the number of constraints comes from Problem::n_constraints. Variable bounds default to ±∞, constraint bounds to 0, and x0 to the origin.

pub fn var_bounds(self, lo: &[f64], hi: &[f64]) -> Self

Variable bounds x_l ≤ x ≤ x_u (use ±2e19 for ∞). Fixes the number of variables.

pub fn constraint_bounds(self, lo: &[f64], hi: &[f64]) -> Self

Constraint bounds g_l ≤ g(x) ≤ g_u (g_l == g_u is an equality).

pub fn x0(self, x0: &[f64]) -> Self

Initial guess. Fixes the number of variables.

pub fn option_num(self, tag: &str, value: f64) -> Self

A numeric solver option (e.g. ("tol", 1e-8)).

pub fn option_int(self, tag: &str, value: i32) -> Self

An integer solver option (e.g. ("max_iter", 500)).

pub fn option_str(self, tag: &str, value: &str) -> Self

A string solver option (e.g. ("mu_strategy", "adaptive")).

pub fn capture_iterations(self) -> Self

Record the per-iteration trajectory: one IterRecord per Newton iteration into Solution::stats.iterations (empty without this call).

Interior-point solves only, since the per-iteration event is emitted by the IPM engine, so on the active-set SQP engine (solver_selection=qp-active-set/algorithm=active-set-sqp) stats.iterations stays empty even though stats.iteration_count still reports the iterations run.

pub fn solve(self) -> Solution

Build the TNLP adapter and run the interior-point solver.

Panics

If the number of variables was never fixed (no var_bounds or x0).