Theon is a Rust library that abstracts Euclidean spaces and integrates with various crates and types in the Rust ecosystem. Theon can be used by libraries to avoid choosing specific math or linear algebra crates on behalf of their dependents.
Theon provides geometric traits that model Euclidean spaces. These traits are not always mathematically rigorous, but this allows them to be implemented for many types. APIs are designed for computations in lower dimensional Euclidean spaces, but traits and types are generic with respect to dimension.
Theon provides optional implementations for commonly used crates in the Rust
ecosystem, including glam
and ultraviolet
. These implementations can be
enabled using Cargo features.
Feature | Default | Crate | Support |
---|---|---|---|
geometry-cgmath |
No | cgmath |
Complete¹ |
geometry-glam |
No | glam |
Complete¹ |
geometry-mint |
No | mint |
Partial |
geometry-nalgebra |
Yes | nalgebra |
Complete¹ |
geometry-ultraviolet |
No | ultraviolet |
Partial² |
Integrated crates are re-exported in the integration
module. Because a given
version of Theon implements traits for specific versions of integrated crates,
care must be taken to align these versions. Dependent crates can either use the
re-exported crates provided by Theon with no direct dependency or ensure that
the version of a supported crate resolves to the same version for which Theon
implements its traits.
[1]: Because Theon is still in its initial development phase, complete does not necessarily mean that all traits and features are implemented, but instead that all traits and features can be feasibly supported and are implemented for common use cases.
[2]: Importantly, traits and features are not yet implemented for SIMD types
like Wec3
.
Geometric queries can be performed using any types that implement the appropriate geometric traits.
use nalgebra::Point2;
use theon::adjunct::Converged;
use theon::query::{Aabb, Intersection, Ray, Unit};
use theon::space::EuclideanSpace;
type E2 = Point2<f64>;
let aabb = Aabb::<E2> {
origin: EuclideanSpace::origin(),
extent: Converged::converged(1.0),
};
let ray = Ray::<E2> {
origin: EuclideanSpace::from_xy(-1.0, 0.5),
direction: Unit::x(),
};
assert_eq!(Some((1.0, 2.0)), ray.intersection(&aabb));
assert_eq!(None, ray.reverse().intersection(&aabb));
In the above example, it is possible to replace the E2
type definition with
types from cgmath
, glam
, or any other type that implements
EuclideanSpace
, etc.
Some queries require solving linear systems of arbitrary and non-trivial size.
To support these queries, the lapack
feature depends on ndarray
and
LAPACK libraries. For example, Plane::from_points
is enabled by the
lapack
feature and computes a best-fit plane using a singular value
decomposition.
The lapack
feature only supports Linux at this time.