Enum DistanceMetric 

#[non_exhaustive]
pub enum DistanceMetric {
    Cosine,
    L2,
    Dot,
    L1,
    Hamming,
    Jaccard,
}

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Cosine

L2

Dot

L1

L1 / Manhattan distance (Σ|xᵢ − yᵢ|). Exact/brute-force only — Lance ANN indexes do not support it, so an L1 column cannot build an ANN index.

Hamming

Hamming distance over binary vectors: the number of differing bits (Σ popcount(aᵢ ⊕ bᵢ) across u8 lanes). Applies only to DataType::BinaryVector columns. Exact/brute-force only — Lance ANN over binary metrics is not wired here.

Jaccard

Jaccard distance over binary vectors: 1 − |A ∩ B| / |A ∪ B| computed bitwise (two all-zero vectors are defined as distance 0). Applies only to DataType::BinaryVector columns. Exact/brute-force only — Lance has no Jaccard ANN.

Implementations

impl DistanceMetric

pub fn compute_distance(&self, a: &[f32], b: &[f32]) -> f32

Computes the distance between two vectors using this metric.

All metrics follow LanceDB conventions so that lower values indicate greater similarity:

• L2: squared Euclidean distance.
• Cosine: 1.0 - cosine_similarity (range [0, 2]).
• Dot: negative dot product.
• L1: Manhattan distance (Σ|xᵢ − yᵢ|).

Panics

Panics if a and b have different lengths.

pub fn is_binary(&self) -> bool

Returns true if this metric operates on binary vectors (&[u8] lanes) rather than float vectors — i.e. DistanceMetric::Hamming or DistanceMetric::Jaccard.

Callers use this to route between DistanceMetric::compute_distance and DistanceMetric::compute_distance_binary, and to decide that a column is scored exact/brute-force (binary metrics have no ANN backend).

pub fn compute_distance_binary(&self, a: &[u8], b: &[u8]) -> f32

Computes the distance between two binary vectors (u8 lanes) using this metric, following the LanceDB “lower is more similar” convention.

• Hamming: number of differing bits, Σ popcount(aᵢ ⊕ bᵢ).
• Jaccard: 1 − |A ∩ B| / |A ∪ B| computed bitwise; two all-zero vectors are defined as distance 0.

Panics

Panics if a and b have different lengths, or if self is a float metric (DistanceMetric::L2, Cosine, Dot, or L1) — those are computed via DistanceMetric::compute_distance.

Trait Implementations

impl Clone for DistanceMetric

fn clone(&self) -> DistanceMetric

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DistanceMetric

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for DistanceMetric

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for DistanceMetric

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for DistanceMetric

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for DistanceMetric