Struct Constraint 

pub struct Constraint {
    pub fun: Box<dyn Fn(&[f64]) -> f64 + Send + Sync>,
    pub jac: Option<Box<dyn Fn(&[f64]) -> Array1<f64> + Send + Sync>>,
    pub kind: ConstraintKind,
    pub lb: Option<f64>,
    pub ub: Option<f64>,
}

Constraint type for constrained optimization.

The constraint callable is stored as a boxed trait object so that a Vec<Constraint> can hold heterogeneous closures (issue #126). Closures that capture outer variables (for example a threshold) are accepted via Constraint::new. An optional analytical Jacobian (gradient of the constraint with respect to each variable) can be attached via Constraint::with_jacobian (issue #127); when present the solvers use it instead of finite differences.

Fields

fun: Box<dyn Fn(&[f64]) -> f64 + Send + Sync>

The constraint function

jac: Option<Box<dyn Fn(&[f64]) -> Array1<f64> + Send + Sync>>

Optional analytical Jacobian (gradient) of the constraint, returning a length-n vector of partial derivatives. None selects finite differences.

kind: ConstraintKind

The type of constraint (equality or inequality)

lb: Option<f64>

Lower bound for a box constraint

ub: Option<f64>

Upper bound for a box constraint

Implementations

impl Constraint

pub const EQUALITY: ConstraintKind = ConstraintKind::Equality

Constant for equality constraint

pub const INEQUALITY: ConstraintKind = ConstraintKind::Inequality

Constant for inequality constraint

pub fn new<F>(fun: F, kind: ConstraintKind) -> Self

where F: Fn(&[f64]) -> f64 + Send + Sync + 'static,

Create a new constraint.

Accepts any callable implementing Fn(&[f64]) -> f64, including closures that capture outer variables and plain fn pointers.

pub fn new_bounds(lb: Option<f64>, ub: Option<f64>) -> Self

Create a new box constraint

pub fn with_jacobian<J>(self, jac: J) -> Self

where J: Fn(&[f64]) -> Array1<f64> + Send + Sync + 'static,

Attach an analytical Jacobian (gradient) for this constraint (issue #127).

The supplied callable returns the length-n gradient of the constraint with respect to each variable, which fills one row of the constraint Jacobian matrix. When present, the solvers use it instead of finite differences.

pub fn is_bounds(&self) -> bool

Check if this is a box constraint

Trait Implementations

impl Debug for Constraint

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

Formats the value using the given formatter.