Fix / on Julia >= 1.9 and broken tests

Project.toml

@@ -1,6 +1,6 @@
 name = "PDMats"
 uuid = "90014a1f-27ba-587c-ab20-58faa44d9150"
-version = "0.11.17"
+version = "0.11.18"
 
 [deps]
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"

src/pdiagmat.jl

@@ -58,8 +58,12 @@ function \(a::PDiagMat, x::AbstractVecOrMat)
 end
 function /(x::AbstractVecOrMat, a::PDiagMat)
     @check_argdims a.dim == size(x, 2)
-    # return matrix for 1-element vectors `x`, consistent with LinearAlgebra
-    return reshape(x, Val(2)) ./ permutedims(a.diag) # = (a' \ x')'
+    if VERSION < v"1.9-"
+        # return matrix for 1-element vectors `x`, consistent with LinearAlgebra < 1.9
+        return reshape(x, Val(2)) ./ permutedims(a.diag) # = (a' \ x')'
+    else
+        return x ./ (x isa AbstractVector ? a.diag : a.diag')
+    end
 end
 Base.kron(A::PDiagMat, B::PDiagMat) = PDiagMat(vec(permutedims(A.diag) .* B.diag))
 

src/pdmat.jl

@@ -49,8 +49,19 @@ end
 *(a::PDMat, x::AbstractVector) = a.mat * x
 *(a::PDMat, x::AbstractMatrix) = a.mat * x
 \(a::PDMat, x::AbstractVecOrMat) = a.chol \ x
-# return matrix for 1-element vectors `x`, consistent with LinearAlgebra
-/(x::AbstractVecOrMat, a::PDMat) = reshape(x, Val(2)) / a.chol
+function /(x::AbstractVecOrMat, a::PDMat)
+    # /(::AbstractVector, ::Cholesky) is not defined
+    if VERSION < v"1.9-"
+        # return matrix for 1-element vectors `x`, consistent with LinearAlgebra
+        return reshape(x, Val(2)) / a.chol
+    else
+        if x isa AbstractVector
+            return vec(reshape(x, Val(2)) / a.chol)
+        else
+            return x / a.chol
+        end
+    end
+end
 
 ### Algebra
 

src/scalmat.jl

@@ -53,8 +53,12 @@ function \(a::ScalMat, x::AbstractVecOrMat)
 end
 function /(x::AbstractVecOrMat, a::ScalMat)
     @check_argdims a.dim == size(x, 2)
-    # return matrix for 1-element vectors `x`, consistent with LinearAlgebra
-    return reshape(x, Val(2)) / a.value
+    if VERSION < v"1.9-"
+        # return matrix for 1-element vectors `x`, consistent with LinearAlgebra < 1.9
+        return reshape(x, Val(2)) / a.value
+    else
+        return x / a.value
+    end
 end
 Base.kron(A::ScalMat, B::ScalMat) = ScalMat(A.dim * B.dim, A.value * B.value )
 

test/pdmtypes.jl

@@ -88,25 +88,29 @@ using Test
         A = rand(1, 1)
         x = randn(1)
         y = x / A
-        @assert x / A isa Matrix{Float64}
-        @assert size(y) == (1, 1)
 
         for M in (PDiagMat(vec(A)), ScalMat(1, first(A)))
-            @test x / M isa Matrix{Float64}
-            @test x / M ≈ y
+            z = x / M
+            @test typeof(z) === typeof(y)
+            @test size(z) == size(y)
+            @test z ≈ y
         end
 
         # requires https://github.com/JuliaLang/julia/pull/32594
         if VERSION >= v"1.3.0-DEV.562"
-            @test x / PDMat(A) isa Matrix{Float64}
-            @test x / PDMat(A) ≈ y
+            z = x / PDMat(A)
+            @test typeof(z) === typeof(y)
+            @test size(z) == size(y)
+            @test z ≈ y
         end
 
         # right division not defined for CHOLMOD:
         # `rdiv!(::Matrix{Float64}, ::SuiteSparse.CHOLMOD.Factor{Float64})` not defined
         if !HAVE_CHOLMOD
-            @test x / PDSparseMat(sparse(first(A), 1, 1)) isa Matrix{Float64}
-            @test x / PDSparseMat(sparse(first(A), 1, 1)) ≈ y
+            z = x / PDSparseMat(sparse(first(A), 1, 1)) 
+            @test typeof(z) === typeof(y)
+            @test size(z) == size(y)
+            @test z ≈ y
         end
     end
 

test/specialarrays.jl

@@ -63,17 +63,17 @@ using StaticArrays
         x = rand(5)
         X = rand(2, 5)
         Y = rand(5, 2)
-        @test P * x ≈ A * x
-        @test P * Y ≈ A * Y
+        @test P * x ≈ x
+        @test P * Y ≈ Y
         # Right division with Cholesky requires https://github.com/JuliaLang/julia/pull/32594
         if VERSION >= v"1.3.0-DEV.562"
-            @test X / P ≈ X / A
+            @test X / P ≈ X
         end
-        @test P \ x ≈ A \ x
-        @test P \ Y ≈ A \ Y
-        @test X_A_Xt(P, X) ≈ X * A * X'
-        @test X_invA_Xt(P, X) ≈ X * (A \ X')
-        @test Xt_A_X(P, Y) ≈ Y' * A * Y
-        @test Xt_invA_X(P, Y) ≈ Y' * (A \ Y)
+        @test P \ x ≈ x
+        @test P \ Y ≈ Y
+        @test X_A_Xt(P, X) ≈ X * X'
+        @test X_invA_Xt(P, X) ≈ X * X'
+        @test Xt_A_X(P, Y) ≈ Y' * Y
+        @test Xt_invA_X(P, Y) ≈ Y' * Y
     end
 end