Improve code coverage

examples/basics.jl

@@ -257,6 +257,8 @@ hcat(initialization(hmm_est_concat), initialization(hmm))
 
 # ## Tests  #src
 
+@test startswith(string(hmm), "Hidden")  #src
+@test length.(values(rand(hmm, T))) == (T, T);  #src
 control_seq = fill(nothing, last(seq_ends));  #src
 test_coherent_algorithms(rng, hmm, control_seq; seq_ends, hmm_guess)  #src
 test_type_stability(rng, hmm, control_seq; seq_ends, hmm_guess)  #src

examples/temporal.jl

@@ -184,5 +184,6 @@ map(mean, hcat(obs_distributions(hmm_est, 2), obs_distributions(hmm, 2)))
 # ## Tests  #src
 
 @test mean(obs_seqs[1][1:2:end]) < 0 < mean(obs_seqs[1][2:2:end])  #src
+@test length.(values(rand(hmm, control_seq))) == (10, 10);  #src
 test_coherent_algorithms(rng, hmm, control_seq; seq_ends, hmm_guess, init=false)  #src
 test_type_stability(rng, hmm, control_seq; seq_ends, hmm_guess)  #src

ext/HiddenMarkovModelsDistributionsExt.jl

@@ -27,6 +27,10 @@ function HiddenMarkovModels.fit_in_sequence!(
     return dists[i] = fit(typeof(dists[i]), reduce(hcat, x_vecs), w)
 end
 
+#=
+
+# Matrix distribution fitting not supported by Distributions.jl at the moment
+
 function HiddenMarkovModels.fit_in_sequence!(
     dists::AbstractVector{<:MatrixDistribution},
     i::Integer,
@@ -37,5 +41,6 @@ function HiddenMarkovModels.fit_in_sequence!(
 end
 
 dcat(M1, M2) = cat(M1, M2; dims=3)
+=#
 
 end

src/inference/forward.jl

@@ -39,7 +39,6 @@ struct ForwardBackwardStorage{R,M<:AbstractMatrix{R}}
     Bβ::Matrix{R}
 end
 
-Base.eltype(::ForwardStorage{R}) where {R} = R
 Base.eltype(::ForwardBackwardStorage{R}) where {R} = R
 
 const ForwardOrForwardBackwardStorage{R} = Union{

src/inference/viterbi.jl

@@ -17,8 +17,6 @@ struct ViterbiStorage{R}
     ψ::Matrix{Int}
 end
 
-Base.eltype(::ViterbiStorage{R}) where {R} = R
-
 """
 $(SIGNATURES)
 """

src/types/hmm.jl

@@ -38,10 +38,6 @@ struct HMM{
     end
 end
 
-function Base.copy(hmm::HMM)
-    return HMM(copy(hmm.init), copy(hmm.trans), copy(hmm.dists))
-end
-
 function Base.show(io::IO, hmm::HMM)
     return print(
         io,

test/distributions.jl

@@ -1,5 +1,6 @@
 using Distributions
-using HiddenMarkovModels: LightCategorical, LightDiagNormal, logdensityof, rand_prob_vec
+using HiddenMarkovModels:
+    LightCategorical, LightDiagNormal, logdensityof, rand_prob_vec, rand_trans_mat
 using LinearAlgebra
 using Statistics
 using StatsAPI: fit!
@@ -8,6 +9,29 @@ using Test
 
 rng = StableRNG(63)
 
+function test_randprobvec(p)
+    @test all(>=(0), p)
+    @test sum(p) ≈ 1
+end
+
+function test_randtransmat(A)
+    foreach(eachrow(A)) do p
+        test_randprobvec(p)
+    end
+end
+
+@testset "Rand prob" begin
+    n = 10
+    test_randprobvec(rand_prob_vec(n))
+    test_randprobvec(rand_prob_vec(rng, n))
+    test_randprobvec(rand_prob_vec(Float32, n))
+    test_randprobvec(rand_prob_vec(rng, Float32, n))
+    test_randtransmat(rand_trans_mat(n))
+    test_randtransmat(rand_trans_mat(rng, n))
+    test_randtransmat(rand_trans_mat(Float32, n))
+    test_randtransmat(rand_trans_mat(rng, Float32, n))
+end
+
 function test_fit_allocs(dist, x, w)
     dist_copy = deepcopy(dist)
     allocs = @allocated fit!(dist_copy, x, w)
@@ -17,6 +41,7 @@ end
 @testset "LightCategorical" begin
     p = rand_prob_vec(rng, 10)
     dist = LightCategorical(p)
+    @test startswith(string(dist), "LightCategorical")
     x = [(@inferred rand(rng, dist)) for _ in 1:100_000]
     # Simulation
     val_count = zeros(Int, length(p))
@@ -38,6 +63,7 @@ end
     μ = randn(rng, 10)
     σ = rand(rng, 10)
     dist = LightDiagNormal(μ, σ)
+    @test startswith(string(dist), "LightDiagNormal")
     x = [(@inferred rand(rng, dist)) for _ in 1:100_000]
     # Simulation
     @test mean(x) ≈ μ atol = 2e-2