MathAssertions 0.5.0
Prefix Reserveddotnet add package MathAssertions --version 0.5.0
NuGet\Install-Package MathAssertions -Version 0.5.0
<PackageReference Include="MathAssertions" Version="0.5.0" />
<PackageVersion Include="MathAssertions" Version="0.5.0" />
<PackageReference Include="MathAssertions" />
paket add MathAssertions --version 0.5.0
#r "nuget: MathAssertions, 0.5.0"
#:package MathAssertions@0.5.0
#addin nuget:?package=MathAssertions&version=0.5.0
#tool nuget:?package=MathAssertions&version=0.5.0
MathAssertions
Scope: Test projects only. Not intended for production code.
Framework-agnostic core of the MathAssertions package family. Math assertion helpers for .NET tests, spanning tolerance comparisons, sequences, statistics, linear algebra, number theory, and 3D geometry. Test-framework-specific entry points ship in adapter packages (currently MathAssertions.TUnit for TUnit).
Most users want
MathAssertions.TUnit, not this package directly. This package ships the framework-agnostic helper classes (MathTolerance,Sequences,Statistics,LinearAlgebra,NumberTheory, plus theGeometry3Dnamespace); the adapter package adds the fluentAssert.That(value).Method(...)entry points your test framework expects.
What's in this package
Pure, NaN-aware, infinity-aware mathematical-assertion helpers. Six static classes plus the Geometry3D namespace cover ~85 methods:
MathTolerance: scalarIsApproximatelyEqualfordouble/floatplus theSystem.Numericscompounds (Vector2/Vector3/Vector4,QuaternionwithIsRotationallyEquivalentand (v0.2.0+)HasAxisAngleApproximately,Matrix4x4,PlanewithIsGeometricallyEquivalent,Complex), span / tensor overloads (ReadOnlySpan<double>/<float>, genericReadOnlyTensorSpan<T>), ULP-distance equality (IsCloseInUlps), combined relative+absolute tolerance, finiteness/probability/percentage predicates,HasRoundtripIdentityfor invertible-transformation checks, and (v0.4.0+) the pose primitivesRotationAngleDegrees,PositionDistance, andIsPoseApproximatelyEqual.- (v0.2.0+) Adapter packages such as
MathAssertions.TUnituse an internal per-component / per-cell / first-failing-element failure-message renderer so every compoundIsApproximatelyEqualTofailure shows the diverging axis and its delta. The renderer is reachable across the assembly boundary via[InternalsVisibleTo]; consumers do not interact with it directly. Sequences: monotonicity (strict + non-strict), boundedness, arithmetic and geometric progressions,ConvergesTo, single-stepIsCauchyConvergent, generic length predicates overReadOnlySpan<T>.Statistics: Welford'sMeanAndVariance,HasMean/Variance/StdDev/Sum/Median/PercentileApproximately,IsWithinSigmasOfMean,AreAllWithinSigmasOfMean. Median and percentile use overflow-safe forms.LinearAlgebra:Matrix4x4invariants (IsSymmetric,IsOrthogonal,IsRotation,IsIdentity,HasDeterminantApproximately,HasTraceApproximately,IsInvertible) plusVector3pair properties (AreOrthogonal,AreParallel(scale-invariant),AngleBetween,AreLinearlyIndependent).NumberTheory: exact integer predicates overlong:IsDivisibleBy,IsPrime,AreCoprime,GreatestCommonDivisor,LeastCommonMultiple,IsPowerOf,IsPerfectSquare,IsCongruent. Overflow-safe andlong.MinValue-aware.Geometry3D: eight primitiverecord structtypes (Sphere,AxisAlignedBox,OrientedBox,Ray3D,LineSegment3D,Triangle3D,Capsule,Cylinder) and five static classes:Properties(degeneracy, collinearity, coplanarity),Containment(point/box/sphere/OBB/convex hull),Distance(point-to-plane/segment/triangle),Intersection(sphere-sphere, AABB-AABB, ray-plane/sphere/triangle/AABB with citations to Real-Time Rendering §§22.6, 22.7, 22.8),Pointcloud(boundedness, centroid, on-plane / on-sphere). Algorithm citations preserved per Real-Time Collision Detection and Real-Time Rendering.
All tolerance- and sigma-taking methods validate the bound up front; invalid input throws ArgumentOutOfRangeException even on early-return paths.
Test-framework adapters
| Package | Test framework | Status |
|---|---|---|
MathAssertions.TUnit |
TUnit | Available now |
MathAssertions.NUnit |
NUnit | Possible if there is demand |
MathAssertions.xUnit |
xUnit | Possible if there is demand |
MathAssertions.MSTest |
MSTest | Possible if there is demand |
If you'd find a non-TUnit adapter useful, open a feature request. Adapters are not built proactively.
When to install this package directly
Only when authoring a non-TUnit adapter for the assertion family, or when calling MathTolerance from a [GenerateAssertion] extension on a private type. For any TUnit testing use case, install MathAssertions.TUnit.
Installation
dotnet add package MathAssertions.TUnit
MathAssertions comes transitively. You don't need to install it directly unless you're building your own adapter package.
Stability
The public surface above is semver-bound. Breaking changes require a major version bump. The exact text of any future failure-message rendering produced inside MathTolerance is not stable and may gain extra detail or change formatting in any release.
Repository
github.com/JohnVerheij/MathAssertions.TUnit for the full README, design notes, and roadmap.
License
MIT. Copyright (c) 2026 John Verheij.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- System.Numerics.Tensors (>= 10.0.9)
NuGet packages (1)
Showing the top 1 NuGet packages that depend on MathAssertions:
| Package | Downloads |
|---|---|
|
MathAssertions.TUnit
TUnit-native math assertions for .NET tests. Fluent .Method(...) entry points (generated via TUnit's [GenerateAssertion]) covering tolerance comparisons (scalar, vector, quaternion, matrix, plane, complex, double[]/float[]), sequences, statistics, Vector3 linear algebra, long-integer number theory, and the Geometry3D primitive surface (containment, distance, intersection, pointcloud aggregates). Extensible: consumers add tolerance assertions for their own domain types via [GenerateAssertion] extensions calling MathTolerance helpers. AOT-compatible, trimmable, no reflection. |
GitHub repositories
This package is not used by any popular GitHub repositories.
View the rendered release notes: https://github.com/JohnVerheij/MathAssertions.TUnit/releases/tag/v0.5.0
Minor release. Fixes the scale-dependence of the parallel check and adds an angle-between assertion and a proper-rotation matrix assertion.
### Fixed
- **`AreParallel` / `IsParallelTo` are now scale-invariant.** The check compared the raw cross-product magnitude `|u x v|` against an absolute tolerance, so whether two vectors read as parallel depended on their length, not just their direction: the same angular deviation passed at unit scale and failed at large scale. It now compares the sine of the angle, `|u x v| / (|u| |v|)`, against the tolerance (an angular measure), the same normalization the package already applied in `IsCollinear`. A zero or shorter-than-tolerance vector is still treated as parallel to any vector. Tolerances chosen for unit vectors are unchanged in meaning; tolerances applied to non-unit vectors now express an angle rather than a magnitude.
### Added
- **`Assert.That(vector).HasAngleBetweenApproximately(other, expectedRadians, tolerance)`** asserts the unsigned angle between two vectors (on `[0, pi]`) is within tolerance of an expected value. The angle is computed as `atan2(|u x v|, u . v)`, which stays accurate across the whole range where `acos` of the normalized dot product loses precision near `0` and `pi`. `LinearAlgebra.AngleBetween(u, v)` exposes the same computation on the framework-agnostic core.
- **`Assert.That(matrix).IsRotation(tolerance)`** asserts a `Matrix4x4` is a proper rotation: orthogonal (`M * M^T = I`) with determinant `+1`. Reflections (determinant `-1`) and matrices carrying translation or scale are rejected. Exposed on the core as `LinearAlgebra.IsRotation`.
### Changed
- Bumped `PackageValidationBaselineVersion` from `0.4.2` to `0.4.3` on both packages so ApiCompat strict-mode validates `0.5.0` against the most recently published baseline.