bugfix logpdf broadcasting (#364)

* removed broadcasting bug for logpdfs

* rm prints

* sep Par. and Vec.Of.Par. tests

* add univariate eval case

* format

* codecov

src/ParameterDistributions.jl

@@ -649,19 +649,33 @@ sample(d::VectorOfParameterized) = sample(Random.GLOBAL_RNG, d, 1)
 Obtains the independent logpdfs of the parameter distributions at `xarray`
 (non-Samples Distributions only), and returns their sum.
 """
-logpdf(d::Parameterized, xarray::AbstractVector{FT}) where {FT <: Real} = logpdf.(d.distribution, xarray)
+logpdf(d::Parameterized, x::FT) where {FT <: Real} = logpdf(d, [x]) # make into 1D array
 
-function logpdf(d::VectorOfParameterized, xarray::AbstractVector{FT}) where {FT <: Real}
+function logpdf(d::Parameterized, xarray::VV) where {VV <: AbstractVector}
+    dimension = ndims(d)
+    if dimension != length(xarray)
+        throw(
+            DimensionMismatch("cannot evaluate logpdf with distribution $dimension on array length $(length(xarray))"),
+        )
+    end
+    if dimension == 1 # if univariate, requires scalar evaluation
+        return logpdf(d.distribution, xarray[1])
+    else
+        return logpdf(d.distribution, xarray)
+    end
+end
+
+function logpdf(d::VectorOfParameterized, xarray::VV) where {VV <: AbstractVector}
     # get the index of xarray chunks to give to the different distributions.
     batches = batch(d)
     dimensions = get_dimensions(d)
     lpdfsum = 0.0
     # perform the logpdf of each of the distributions, and returns their sum    
-    for (i, dd) in enumerate(d.distribution)
-        if dimensions[i] == 1
-            lpdfsum += logpdf.(dd, xarray[batches[i]])[1]
+    for (i, dd, dimen, batch) in zip(1:length(d.distribution), d.distribution, dimensions, batches)
+        if dimen == 1
+            lpdfsum += logpdf(dd, xarray[batch][1]) # needs to be eval on a scalar
         else
-            lpdfsum += logpdf(dd, xarray[batches[i]])
+            lpdfsum += logpdf(dd, xarray[batch])
         end
     end
     return lpdfsum
@@ -672,7 +686,7 @@ function logpdf(pd::ParameterDistribution, xarray::AbstractVector{FT}) where {FT
     if any(isa.(pd.distribution, Samples))
         throw(
             ErrorException(
-                "Cannot compute logpdf of Samples distributions. Consider using a Parameterized type for your prior.",
+                "No implementation of logpdf of Samples distributions. Consider using a Parameterized type for your prior.",
             ),
         )
     end
@@ -692,6 +706,8 @@ function logpdf(pd::ParameterDistribution, xarray::AbstractVector{FT}) where {FT
     return sum(sum(logpdf(d, xarray[batches[i]])) for (i, d) in enumerate(pd.distribution))
 end
 
+logpdf(pd::ParameterDistribution, x::FT) where {FT <: Real} = logpdf(pd, [x])
+
 #extending StatsBase cov,var
 """
     var(pd::ParameterDistribution)

test/ParameterDistributions/runtests.jl

@@ -538,11 +538,26 @@ using EnsembleKalmanProcesses.ParameterDistributions
         @test_throws ErrorException logpdf(u, zeros(ndims(u)))
         x_in_bd = [0.5, 0.5, 0.5]
         Random.seed!(seed)
-        lpdf3 = sum([logpdf(Beta(2, 2), x_in_bd[1])[1], logpdf(MvNormal(zeros(2), 0.1 * I), x_in_bd[2:3])[1]]) #throws deprecated warning without "."
-
+        # for VectorOfParameterized
+        lpdf3 = sum([logpdf(Beta(2, 2), x_in_bd[1])[1], logpdf(MvNormal(zeros(2), 0.1 * I), x_in_bd[2:3])[1]]) #throws deprecated warning without "."     
         Random.seed!(seed)
         @test isapprox(logpdf(u3, x_in_bd) - lpdf3, 0.0; atol = 1e-6)
         @test_throws DimensionMismatch logpdf(u3, [0.5, 0.5])
+        # for Parameterized Multivar
+        x_in_bd = [0.0, 0.0, 0.0, 0.0]
+        @test isapprox(logpdf(u1, x_in_bd) - logpdf(MvNormal(zeros(4), 0.1 * I), x_in_bd)[1], 0.0, atol = 1e-6)
+        @test_throws DimensionMismatch logpdf(u1, [1])
+        # for Parameterized Univar
+        u5 = constrained_gaussian("u5", 3.0, 1.0, -Inf, Inf)
+        x_in_bd = 0.0
+        @test isapprox(logpdf(u5, x_in_bd) - logpdf(Normal(3.0, 1.0), x_in_bd)[1], 0.0, atol = 1e-6)
+        @test_throws DimensionMismatch logpdf(u1, [1, 1])
+        @test isapprox(
+            logpdf(Parameterized(Normal(3.0, 1.0)), x_in_bd) - logpdf(Normal(3.0, 1.0), x_in_bd)[1],
+            0.0,
+            atol = 1e-6,
+        )
+        @test_throws DimensionMismatch logpdf(Parameterized(Normal(3.0, 1.0)), [1, 1])
 
         #Test for cov, var        
         block_cov = cat([cov(d1), var(d2), cov(d3), cov(d4)]..., dims = (1, 2))