Support pointer(A::PermDimsArray) and strides(A) if parent(A) support…

…s them

(cherry picked from commit 2bc83ca)
ref #20385

Qualify PermutedDimsArray as it isn't exported on release-0.5

base/permuteddimsarray.jl

@@ -26,6 +26,20 @@ Base.parent(A::PermutedDimsArray) = A.parent
 Base.size{T,N,perm}(A::PermutedDimsArray{T,N,perm})    = genperm(size(parent(A)),    perm)
 Base.indices{T,N,perm}(A::PermutedDimsArray{T,N,perm}) = genperm(indices(parent(A)), perm)
 
+Base.unsafe_convert{T}(::Type{Ptr{T}}, A::PermutedDimsArray{T}) = Base.unsafe_convert(Ptr{T}, parent(A))
+
+# It's OK to return a pointer to the first element, and indeed quite
+# useful for wrapping C routines that require a different storage
+# order than used by Julia. But for an array with unconventional
+# storage order, a linear offset is ambiguous---is it a memory offset
+# or a linear index?
+Base.pointer{T}(A::PermutedDimsArray{T}, i::Integer) = throw(ArgumentError("pointer(A, i) is deliberately unsupported for PermutedDimsArray"))
+
+function Base.strides{T,N,perm}(A::PermutedDimsArray{T,N,perm})
+    s = strides(parent(A))
+    ntuple(d->s[perm[d]], Val{N})
+end
+
 @inline function Base.getindex{T,N,perm,iperm}(A::PermutedDimsArray{T,N,perm,iperm}, I::Vararg{Int,N})
     @boundscheck checkbounds(A, I...)
     @inbounds val = getindex(A.parent, genperm(I, iperm)...)

test/arrayops.jl

@@ -478,6 +478,12 @@ end
 @test_throws ArgumentError permutedims(s, (1,1,1))
 @test_throws ArgumentError Base.PermutedDimsArrays.PermutedDimsArray(a, (1,1,1))
 @test_throws ArgumentError Base.PermutedDimsArrays.PermutedDimsArray(s, (1,1,1))
+cp = Base.PermutedDimsArrays.PermutedDimsArray(c, (3,2,1))
+@test pointer(cp) == pointer(c)
+@test_throws ArgumentError pointer(cp, 2)
+@test strides(cp) == (9,3,1)
+ap = Base.PermutedDimsArrays.PermutedDimsArray(collect(a), (2,1,3))
+@test strides(ap) == (3,1,12)
 
 ## ipermutedims ##