Make Test a weak dependency

Project.toml

@@ -1,6 +1,6 @@
 name = "KernelFunctions"
 uuid = "ec8451be-7e33-11e9-00cf-bbf324bd1392"
-version = "0.10.58"
+version = "0.10.59"
 
 [deps]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
@@ -21,6 +21,12 @@ TensorCore = "62fd8b95-f654-4bbd-a8a5-9c27f68ccd50"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 ZygoteRules = "700de1a5-db45-46bc-99cf-38207098b444"
 
+[weakdeps]
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[extensions]
+KernelFunctionsTestExt = "Test"
+
 [compat]
 ChainRulesCore = "1"
 Compat = "3.7, 4"

ext/KernelFunctionsTestExt.jl

@@ -0,0 +1,283 @@
+module KernelFunctionsTestExt
+
+using KernelFunctions
+using KernelFunctions: TestUtils, LinearAlgebra, Random
+using Test
+
+"""
+    test_interface(
+        k::Kernel,
+        x0::AbstractVector,
+        x1::AbstractVector,
+        x2::AbstractVector;
+        rtol=1e-6,
+        atol=rtol,
+    )
+
+Run various consistency checks on `k` at the inputs `x0`, `x1`, and `x2`.
+`x0` and `x1` should be of the same length with different values, while `x0` and `x2` should
+be of different lengths.
+
+These tests are intended to pick up on really substantial issues with a kernel implementation
+(e.g. substantial asymmetry in the kernel matrix, large negative eigenvalues), rather than to
+test the numerics in detail, which can be kernel-specific.
+"""
+function TestUtils.test_interface(
+    k::Kernel,
+    x0::AbstractVector,
+    x1::AbstractVector,
+    x2::AbstractVector;
+    rtol=1e-6,
+    atol=rtol,
+)
+    # Ensure that we have the required inputs.
+    @assert length(x0) == length(x1)
+    @assert length(x0) ≠ length(x2)
+
+    # Check that kernelmatrix_diag basically works.
+    @test kernelmatrix_diag(k, x0, x1) isa AbstractVector
+    @test length(kernelmatrix_diag(k, x0, x1)) == length(x0)
+
+    # Check that pairwise basically works.
+    @test kernelmatrix(k, x0, x2) isa AbstractMatrix
+    @test size(kernelmatrix(k, x0, x2)) == (length(x0), length(x2))
+
+    # Check that elementwise is consistent with pairwise.
+    @test kernelmatrix_diag(k, x0, x1) ≈ LinearAlgebra.diag(kernelmatrix(k, x0, x1)) atol =
+        atol rtol = rtol
+
+    # Check additional binary elementwise properties for kernels.
+    @test kernelmatrix_diag(k, x0, x1) ≈ kernelmatrix_diag(k, x1, x0)
+    @test kernelmatrix(k, x0, x2) ≈ permutedims(kernelmatrix(k, x2, x0)) atol = atol rtol =
+        rtol
+
+    # Check that unary elementwise basically works.
+    @test kernelmatrix_diag(k, x0) isa AbstractVector
+    @test length(kernelmatrix_diag(k, x0)) == length(x0)
+
+    # Check that unary pairwise basically works.
+    @test kernelmatrix(k, x0) isa AbstractMatrix
+    @test size(kernelmatrix(k, x0)) == (length(x0), length(x0))
+    @test kernelmatrix(k, x0) ≈ permutedims(kernelmatrix(k, x0)) atol = atol rtol = rtol
+
+    # Check that unary elementwise is consistent with unary pairwise.
+    @test kernelmatrix_diag(k, x0) ≈ LinearAlgebra.diag(kernelmatrix(k, x0)) atol = atol rtol =
+        rtol
+
+    # Check that unary pairwise produces a positive definite matrix (approximately).
+    @test LinearAlgebra.eigmin(Matrix(kernelmatrix(k, x0))) > -atol
+
+    # Check that unary elementwise / pairwise are consistent with the binary versions.
+    @test kernelmatrix_diag(k, x0) ≈ kernelmatrix_diag(k, x0, x0) atol = atol rtol = rtol
+    @test kernelmatrix(k, x0) ≈ kernelmatrix(k, x0, x0) atol = atol rtol = rtol
+
+    # Check that basic kernel evaluation succeeds and is consistent with `kernelmatrix`.
+    @test k(first(x0), first(x1)) isa Real
+    @test kernelmatrix(k, x0, x2) ≈ [k(xl, xr) for xl in x0, xr in x2]
+
+    tmp = Matrix{Float64}(undef, length(x0), length(x2))
+    @test kernelmatrix!(tmp, k, x0, x2) ≈ kernelmatrix(k, x0, x2)
+
+    tmp_square = Matrix{Float64}(undef, length(x0), length(x0))
+    @test kernelmatrix!(tmp_square, k, x0) ≈ kernelmatrix(k, x0)
+
+    tmp_diag = Vector{Float64}(undef, length(x0))
+    @test kernelmatrix_diag!(tmp_diag, k, x0) ≈ kernelmatrix_diag(k, x0)
+    @test kernelmatrix_diag!(tmp_diag, k, x0, x1) ≈ kernelmatrix_diag(k, x0, x1)
+end
+
+"""
+    test_interface([rng::AbstractRNG], k::Kernel, ::Type{T}=Float64; kwargs...) where {T}
+
+Run the [`test_interface`](@ref) tests for randomly generated inputs of types `Vector{T}`,
+`Vector{Vector{T}}`, `ColVecs{T}`, and `RowVecs{T}`.
+
+For other input types, please provide the data manually.
+
+The keyword arguments are forwarded to the invocations of [`test_interface`](@ref) with the
+randomly generated inputs.
+"""
+function TestUtils.test_interface(k::Kernel, T::Type=Float64; kwargs...)
+    return TestUtils.test_interface(Random.default_rng(), k, T; kwargs...)
+end
+
+function TestUtils.test_interface(
+    rng::Random.AbstractRNG, k::Kernel, T::Type=Float64; kwargs...
+)
+    return TestUtils.test_with_type(TestUtils.test_interface, rng, k, T; kwargs...)
+end
+
+"""
+    test_type_stability(
+        k::Kernel,
+        x0::AbstractVector,
+        x1::AbstractVector,
+        x2::AbstractVector,
+    )
+
+Run type stability checks over `k(x,y)` and the different functions of the API 
+(`kernelmatrix`, `kernelmatrix_diag`). `x0` and `x1` should be of the same 
+length with different values, while `x0` and `x2` should be of different lengths.
+"""
+function TestUtils.test_type_stability(
+    k::Kernel, x0::AbstractVector, x1::AbstractVector, x2::AbstractVector
+)
+    # Ensure that we have the required inputs.
+    @assert length(x0) == length(x1)
+    @assert length(x0) ≠ length(x2)
+    @test @inferred(kernelmatrix(k, x0)) isa AbstractMatrix
+    @test @inferred(kernelmatrix(k, x0, x2)) isa AbstractMatrix
+    @test @inferred(kernelmatrix_diag(k, x0)) isa AbstractVector
+    @test @inferred(kernelmatrix_diag(k, x0, x1)) isa AbstractVector
+end
+
+function TestUtils.test_type_stability(k::Kernel, ::Type{T}=Float64; kwargs...) where {T}
+    return TestUtils.test_type_stability(Random.default_rng(), k, T; kwargs...)
+end
+
+function TestUtils.test_type_stability(
+    rng::Random.AbstractRNG, k::Kernel, ::Type{T}; kwargs...
+) where {T}
+    return TestUtils.test_with_type(TestUtils.test_type_stability, rng, k, T; kwargs...)
+end
+
+"""
+    test_with_type(f, rng::AbstractRNG, k::Kernel, ::Type{T}; kwargs...) where {T}
+
+Run the functions `f`, (for example [`test_interface`](@ref) or
+[`test_type_stable`](@ref)) for randomly generated inputs of types `Vector{T}`,
+`Vector{Vector{T}}`, `ColVecs{T}`, and `RowVecs{T}`.
+
+For other input types, please provide the data manually.
+
+The keyword arguments are forwarded to the invocations of `f` with the
+randomly generated inputs.
+"""
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{T}; kwargs...
+) where {T}
+    @testset "Vector{$T}" begin
+        TestUtils.test_with_type(f, rng, k, Vector{T}; kwargs...)
+    end
+    @testset "ColVecs{$T}" begin
+        TestUtils.test_with_type(f, rng, k, ColVecs{T}; kwargs...)
+    end
+    @testset "RowVecs{$T}" begin
+        TestUtils.test_with_type(f, rng, k, RowVecs{T}; kwargs...)
+    end
+    @testset "Vector{Vector{$T}}" begin
+        TestUtils.test_with_type(f, rng, k, Vector{Vector{T}}; kwargs...)
+    end
+end
+
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{Vector{T}}; kwargs...
+) where {T<:Real}
+    return f(k, randn(rng, T, 11), randn(rng, T, 11), randn(rng, T, 13); kwargs...)
+end
+
+function TestUtils.test_with_type(
+    f,
+    rng::Random.AbstractRNG,
+    k::MOKernel,
+    ::Type{Vector{Tuple{T,Int}}};
+    dim_out=3,
+    kwargs...,
+) where {T<:Real}
+    return f(
+        k,
+        [(randn(rng, T), rand(rng, 1:dim_out)) for i in 1:11],
+        [(randn(rng, T), rand(rng, 1:dim_out)) for i in 1:11],
+        [(randn(rng, T), rand(rng, 1:dim_out)) for i in 1:13];
+        kwargs...,
+    )
+end
+
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{<:ColVecs{T}}; dim_in=2, kwargs...
+) where {T<:Real}
+    return f(
+        k,
+        ColVecs(randn(rng, T, dim_in, 11)),
+        ColVecs(randn(rng, T, dim_in, 11)),
+        ColVecs(randn(rng, T, dim_in, 13));
+        kwargs...,
+    )
+end
+
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{<:RowVecs{T}}; dim_in=2, kwargs...
+) where {T<:Real}
+    return f(
+        k,
+        RowVecs(randn(rng, T, 11, dim_in)),
+        RowVecs(randn(rng, T, 11, dim_in)),
+        RowVecs(randn(rng, T, 13, dim_in));
+        kwargs...,
+    )
+end
+
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{<:Vector{Vector{T}}}; dim_in=2, kwargs...
+) where {T<:Real}
+    return f(
+        k,
+        [randn(rng, T, dim_in) for _ in 1:11],
+        [randn(rng, T, dim_in) for _ in 1:11],
+        [randn(rng, T, dim_in) for _ in 1:13];
+        kwargs...,
+    )
+end
+
+function TestUtils.test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{Vector{String}}; kwargs...
+)
+    return f(
+        k,
+        [Random.randstring(rng) for _ in 1:3],
+        [Random.randstring(rng) for _ in 1:3],
+        [Random.randstring(rng) for _ in 1:4];
+        kwargs...,
+    )
+end
+
+function test_with_type(
+    f, rng::Random.AbstractRNG, k::Kernel, ::Type{ColVecs{String}}; dim_in=2, kwargs...
+)
+    return f(
+        k,
+        ColVecs([Random.randstring(rng) for _ in 1:dim_in, _ in 1:3]),
+        ColVecs([Random.randstring(rng) for _ in 1:dim_in, _ in 1:3]),
+        ColVecs([Random.randstring(rng) for _ in 1:dim_in, _ in 1:4]);
+        kwargs...,
+    )
+end
+
+function TestUtils.test_with_type(f, k::Kernel, T::Type{<:Real}; kwargs...)
+    return TestUtils.test_with_type(f, Random.default_rng(), k, T; kwargs...)
+end
+
+"""
+    example_inputs(rng::AbstractRNG, type)
+
+Return a tuple of 4 inputs of type `type`. See `methods(example_inputs)` for information
+around supported types. It is recommended that you utilise `StableRNGs.jl` for `rng` here
+to ensure consistency across Julia versions.
+"""
+function TestUtils.example_inputs(rng::Random.AbstractRNG, ::Type{Vector{Float64}})
+    return map(n -> randn(rng, Float64, n), (1, 2, 3, 4))
+end
+
+function TestUtils.example_inputs(
+    rng::Random.AbstractRNG, ::Type{ColVecs{Float64,Matrix{Float64}}}; dim::Int=2
+)
+    return map(n -> ColVecs(randn(rng, dim, n)), (1, 2, 3, 4))
+end
+
+function TestUtils.example_inputs(
+    rng::Random.AbstractRNG, ::Type{RowVecs{Float64,Matrix{Float64}}}; dim::Int=2
+)
+    return map(n -> RowVecs(randn(rng, n, dim)), (1, 2, 3, 4))
+end
+
+end # module

src/KernelFunctions.jl

@@ -53,6 +53,7 @@ using CompositionsBase
 using Distances
 using FillArrays
 using Functors
+using Random
 using LinearAlgebra
 using Requires
 using SpecialFunctions: loggamma, besselk, polygamma
@@ -125,6 +126,9 @@ include("chainrules.jl")
 include("zygoterules.jl")
 
 include("TestUtils.jl")
+if !isdefined(Base, :get_extension)
+    include("../ext/KernelFunctionsTestExt.jl")
+end
 
 function __init__()
     @require Kronecker = "2c470bb0-bcc8-11e8-3dad-c9649493f05e" begin