CHANGELOG.md

Changelog

This file documents recent notable changes to this project. The format of this file is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.18.0 - 2025-11-19

Changed

• Updated ndarray to 0.17.

0.13.0 - 2025-05-03

Changed

• Requires Rust 1.81 or newer.

Performance

• Improved performance of BallTree::query (#77).

0.12.0 - 2025-03-05

Added

• Added Cosine Distance Metric Implementation
  • Implemented the Cosine distance metric by defining the Cosine struct and implementing the Metric trait. Cosine distance is defined as 1 - cosine similarity, where [ \text{cosine similarity} = \frac{\mathbf{A} \cdot \mathbf{B}}{|\mathbf{A}| |\mathbf{B}|} ]
  • Ensured thread safety and used generic floating-point types (Float) for flexibility in numerical calculations.

0.11.0 - 2024-08-07

Changed

• Requires Rust 1.74 or newer.
• Updated ndarray to 0.16.

0.10.0 - 2023-12-20

Changed

• Instead of Float trait, FloatCore trait is used as a bound for T in BallTree and VantagePointTree.
• Requires Rust 1.70 or newer.

0.9.0 - 2023-08-07

Added

• Euclidean derives Eq.

Changed

• Switched to edition 2021.
• Requires Rust 1.64 or newer.

0.8.0 - 2022-04-18

Added

• Add pairwise distance computation function.

Changed

• Make fields of Node, BallTree accessible for user.
• MSRV is changed to 1.53.0.

0.7.1 - 2021-07-06

Added

• Implement Clone and Sync for Euclidean struct.

0.7.0 - 2021-06-23

Changed

• Add Metric trait, use it for distance calculation instead of function pointer.
• Add Euclidean struct, which implements Metric trait.

0.6.0 - 2021-03-29

Changed

• Upgrade ndarray to 0.15.0
• Requires Rust 1.49 or later.

0.5.2 - 2021-02-16

Changed

• Fixed build on aarch64.

0.5.1 - 2020-12-30

Changed

• Upgrade ndarray to 0.14.0

0.5.0 - 2020-08-20

Changed

• BallTree and VantagePointTree accept an ndarray as a point.

Fixed

• No longer panics when a coordinate is NaN; the distance from a point with NaN in its coordinate and another point is considered greater than the distance between any two points without NaN in their coordinates.
• BallTree::query returns empty vectors, rather than panics, when k is zero.

0.4.0 - 2020-06-01

Added

• A vantage point tree data structure to find nearest points.
• BallTree accepts not only an f64 array but also an f32one.
• BallTree::euclidean to create a ball tree withoug having to pass a distance metric as an argument.

Changed

• The codinates of each point must be stored in a contiguous area in memory.
• A distance metric is now a function, not a trait.

0.3.0 - 2020-04-17

Changed

• The ownership of the input can be transferred to BallTree, which accepts both an owned array and a view.
• An error is returned, rather than a panic, if an empty array is given to construct a BallTree.
• query_one has been renamed query_nearest.
• query returns indices and distances separately, so that data of the same type are stored together.

0.2.0 - 2020-04-09

Changed

• BallTree takes ArrayBase as its input, instead of ArrayView, to allow more types in ndarray.

0.1.0 - 2019-11-20

Added

• A ball tree data structure to find nearest neighbors.