Implementation:Lance format Lance Java DistanceType

Knowledge Sources	Lance
Domains	Java_SDK, Indexing
Last Updated	2026-02-08 19:33 GMT

Overview

Description

DistanceType is a Java enum in the org.lance.index package that defines the supported distance metrics used for vector similarity search in Lance indices. It provides four distance metric variants: L2 (Euclidean distance), Cosine (cosine similarity), Dot (dot product), and Hamming (Hamming distance). This enum is used throughout the Java SDK to configure the distance function when creating vector indices via VectorIndexParams.

Usage

DistanceType is passed as a parameter when constructing vector index configurations. It determines how Lance computes the similarity between query vectors and indexed vectors during nearest-neighbor search. The most common usage is with the VectorIndexParams factory methods or builder, where the caller specifies which distance metric the index should use.

Code Reference

Source Location

java/src/main/java/org/lance/index/DistanceType.java

Signature

public enum DistanceType {
    L2,
    Cosine,
    Dot,
    Hamming;
}

Import

import org.lance.index.DistanceType;

I/O Contract

Enum Values

Value	Description
L2	Euclidean (L2) distance. Measures the straight-line distance between two vectors.
Cosine	Cosine similarity. Measures the angular distance between two vectors, normalized by their magnitudes.
Dot	Dot product distance. Uses the inner product of two vectors as the similarity measure.
Hamming	Hamming distance. Counts the number of positions where corresponding elements differ. Typically used with binary vectors.

Usage Examples

// Use L2 distance for a vector index
VectorIndexParams params = VectorIndexParams.ivfFlat(256, DistanceType.L2);

// Use cosine similarity
VectorIndexParams cosineParams = VectorIndexParams.ivfFlat(128, DistanceType.Cosine);

// Use dot product with IVF-PQ
VectorIndexParams dotParams = VectorIndexParams.ivfPq(
    256, 8, 16, DistanceType.Dot, 50
);

Related Pages

• VectorIndexParams - Uses DistanceType to configure vector index distance metrics
• IvfBuildParams - IVF build parameters used alongside distance type configuration
• IndexType - Enum of all index types including vector index variants