Merge branch 'JuliaStats:master' into doctests

Author: abhro

Project.toml

@@ -1,7 +1,7 @@
 name = "Distributions"
 uuid = "31c24e10-a181-5473-b8eb-7969acd0382f"
 authors = ["JuliaStats"]
-version = "0.25.119"
+version = "0.25.120"
 
 [deps]
 AliasTables = "66dad0bd-aa9a-41b7-9441-69ab47430ed8"
@@ -44,7 +44,7 @@ ForwardDiff = "0.10, 1"
 JSON = "0.21"
 LinearAlgebra = "<0.0.1, 1"
 OffsetArrays = "1"
-PDMats = "0.10, 0.11"
+PDMats = "0.11.35"
 Printf = "<0.0.1, 1"
 QuadGK = "2"
 Random = "<0.0.1, 1"

docs/src/univariate.md

@@ -72,6 +72,7 @@ pdfsquaredL2norm
 insupport(::UnivariateDistribution, x::Any)
 pdf(::UnivariateDistribution, ::Real)
 logpdf(::UnivariateDistribution, ::Real)
+gradlogpdf(::Distribution, ::Any)
 loglikelihood(::UnivariateDistribution, ::AbstractArray)
 cdf(::UnivariateDistribution, ::Real)
 logcdf(::UnivariateDistribution, ::Real)

src/common.jl

@@ -256,7 +256,7 @@ be disabled by using `@inbounds`.
 Instead of `logpdf` one should implement `_logpdf(d, x)` which does not have to check the
 size of `x`.
 
-See also: [`pdf`](@ref).
+See also: [`pdf`](@ref), [`gradlogpdf`](@ref).
 """
 @inline function logpdf(
     d::Distribution{ArrayLikeVariate{N}}, x::AbstractArray{<:Real,M}
@@ -280,6 +280,17 @@ See also: [`pdf`](@ref).
     end
 end
 
+"""
+    gradlogpdf(d::Distribution, x)
+
+Evaluate the gradient of the logarithm of the probability density function of `d` at `x`.
+
+For univariate distributions, return the derivative.
+
+See also: [`logpdf`](@ref).
+"""
+function gradlogpdf end
+
 # `_logpdf` should be implemented and has no default definition
 # _logpdf(d::Distribution{ArrayLikeVariate{N}}, x::AbstractArray{<:Real,N}) where {N}
 

src/multivariate/mvnormalcanon.jl

@@ -170,20 +170,13 @@ sqmahal!(r::AbstractVector, d::MvNormalCanon, x::AbstractMatrix) = quad!(r, d.J,
 
 # Sampling (for GenericMvNormal)
 
-unwhiten_winv!(J::AbstractPDMat, x::AbstractVecOrMat) = unwhiten!(inv(J), x)
-unwhiten_winv!(J::PDiagMat, x::AbstractVecOrMat) = whiten!(J, x)
-unwhiten_winv!(J::ScalMat, x::AbstractVecOrMat) = whiten!(J, x)
-if isdefined(PDMats, :PDSparseMat)
-    unwhiten_winv!(J::PDSparseMat, x::AbstractVecOrMat) = x[:] = J.chol.PtL' \ x
-end
-
 function _rand!(rng::AbstractRNG, d::MvNormalCanon, x::AbstractVector)
-    unwhiten_winv!(d.J, randn!(rng, x))
+    invunwhiten!(d.J, randn!(rng, x))
     x .+= d.μ
     return x
 end
 function _rand!(rng::AbstractRNG, d::MvNormalCanon, x::AbstractMatrix)
-    unwhiten_winv!(d.J, randn!(rng, x))
+    invunwhiten!(d.J, randn!(rng, x))
     x .+= d.μ
     return x
 end