Enum LdltError 

pub enum LdltError {
    ZeroPivot(usize),
    NearZeroPivot {
        column: usize,
        pivot: f64,
        scale: f64,
        suggested_shift: f64,
    },
    InvalidInput(&'static str),
    SizeMismatch {
        expected: usize,
        got: usize,
    },
}

Failure modes of the factorization and solves.

Eq is deliberately not derived: LdltError::NearZeroPivot carries f64 payloads.

Variants

ZeroPivot(usize)

A zero pivot (D[k] == 0) was hit at this column: the matrix is singular or the un-pivoted factorization broke down there.

NearZeroPivot

A pivot that is not exactly zero but has lost every significant digit to cancellation: |D[k]| < NEAR_ZERO_PIVOT_REL * scale.

This is the honest report of the case that used to be returned silently, and it matters because the sign pattern of D is the matrix inertia (Sylvester’s law). A pivot at this magnitude still has a sign, but that sign is rounding noise, so the inertia read from the factorization would be noise too - and downstream that inertia is a Sturm eigenvalue count, i.e. an eigenvalue or buckling load. Returning it is strictly better than returning a number nobody can tell is wrong.

Recover by moving off the near-singular point: either shift the matrix yourself, or call SparseLdlt::factor_shifted, which does exactly that and reports the shift it used through SparseLdlt::shift.

Fields

column: usize

The column at which the pivot collapsed.

pivot: f64

The computed pivot value. Its sign is not trustworthy at this magnitude.

scale: f64

The largest absolute diagonal entry of the input matrix - the reference the tolerance is relative to.

suggested_shift: f64

A diagonal shift large enough to clear the breakdown: sqrt( NEAR_ZERO_PIVOT_REL ) * scale, i.e. comfortably outside the tolerance band rather than on its edge. Factoring A + suggested_shift * I is an exact factorization of a NEARBY matrix, not of A.

InvalidInput(&'static str)

The CSC arrays were inconsistent (bad length, col_ptr not monotonic, an index out of range, or a non-finite value).

SizeMismatch

A right-hand side (or multi-RHS row) did not match the factored matrix’s order.

Fields

expected: usize

The order of the factored matrix.

got: usize

The length that was supplied.

Trait Implementations

impl Clone for LdltError

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LdltError

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

Formats the value using the given formatter.

impl PartialEq for LdltError

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for LdltError