Struct VerifyConfig 

pub struct VerifyConfig {
    pub weights: Weights,
    pub type3_min_composite: f64,
    pub type2_min_lexical: f64,
    pub high_confidence: f64,
    pub medium_confidence: f64,
    pub exact_epsilon: f64,
    pub alignment_band: usize,
    pub max_alignment_cells: usize,
}

Tuning for verification. Thresholds are provisional and calibrated against the mutation corpus.

Fields

weights: Weights

Composite-score weights.

type3_min_composite: f64

Smallest composite a Type-3 pair must reach to be a clone at all.

Calibrated against the labelled corpora: the pairs deliberately built to share a skeleton while computing different things reach 0.69, and the weakest pair that is a real copy reaches 0.71. The threshold sits in that gap.

The gap is narrow, and which side an unlabelled pair falls on is decided almost entirely by lexical agreement — the lookalikes reach 0.77 there while the weakest real copy reaches 0.91. A composite near this threshold is therefore weak evidence by construction, which is what the low confidence band exists to say.

type2_min_lexical: f64

Smallest lexical agreement required before an otherwise exact structural match can be called a safe Type-2 clone.

Below this floor, matching shape alone is insufficient evidence that the differing code is a consistent rename or literal substitution. Such pairs remain eligible for the ordinary Type-3 decision.

high_confidence: f64

Composite at or above which a Type-3 finding is high confidence.

medium_confidence: f64

Composite at or above which a Type-3 finding is medium confidence.

Kept above Self::type3_min_composite, or the low band could never be reached and a finding sitting just over the acceptance threshold would be reported as confidently as one well clear of it.

exact_epsilon: f64

Tolerance for treating a similarity as exactly 1.0.

alignment_band: usize

How far the statement alignment may stray from the diagonal that joins the two sequences’ ends, in statements.

The band bounds the alignment’s cost at O(min(n, m) * band) instead of O(n * m). Widening it can only raise a pair’s similarity, so the banded result is a lower bound: a real clone, whose alignment hugs that diagonal, is measured exactly, while a pair that would need to wander further is scored no higher than it deserves.

max_alignment_cells: usize

Largest alignment table, in cells.

The band alone bounds the table for units of comparable length; this bounds it for a pair whose lengths also differ widely, by narrowing the band further until the table fits. Narrowing only weakens the lower bound, so nothing is dropped — such a pair cannot align well enough to be a clone in any case.

Trait Implementations

impl Clone for VerifyConfig

fn clone(&self) -> VerifyConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for VerifyConfig

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

Formats the value using the given formatter.

impl Default for VerifyConfig

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for VerifyConfig

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for VerifyConfig