Struct VectorMetricQuery 

pub struct VectorMetricQuery<'a> { /* private fields */ }

Metric scorer bound to a single query vector.

Use this when ranking many candidates against one query. It preserves the same scores and error contract as VectorMetric::distance, but avoids recomputing query-only metric state for every candidate.

Implementations

impl<'a> VectorMetricQuery<'a>

pub const fn metric(&self) -> VectorMetric

Return the metric this scorer uses.

pub const fn query(&self) -> &'a VectorValue

Return the bound query vector.

pub fn distance(&self, candidate: &VectorValue) -> CoreResult<f64>

Compute this bound query’s lower-is-better distance to candidate.

Errors

Returns CoreError::VectorDimensionMismatch if dimensions differ. VectorMetric::Cosine also returns CoreError::VectorZeroNorm when candidate has zero magnitude.

pub fn distance_with_candidate_squared_norm( &self, candidate: &VectorValue, candidate_squared_norm: f64, ) -> CoreResult<f64>

Compute distance using a precomputed candidate squared norm.

When candidate_squared_norm is the candidate’s actual squared norm, this is equivalent to Self::distance. It lets ANN indexes cache centroid norms for cosine scoring while still using the canonical metric kernels and error behavior. Non-cosine metrics ignore candidate_squared_norm.

Errors

Returns CoreError::VectorDimensionMismatch if dimensions differ. VectorMetric::Cosine returns CoreError::VectorZeroNorm when the supplied candidate squared norm is not positive and finite.

Trait Implementations

impl<'a> Clone for VectorMetricQuery<'a>

fn clone(&self) -> VectorMetricQuery<'a>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<'a> Copy for VectorMetricQuery<'a>

impl<'a> Debug for VectorMetricQuery<'a>

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

Formats the value using the given formatter.