Maybe we should flip nested wrapper constructors

[oxinabox]

#120 makes me think that
Maybe we should revisit the decision not to do overloads of
Symmetric(NamedDimsArray(x)) --> NamedDimsArray(Symmetric(x))
and Diagonal(NamedDimsArray(x)) --> NamedDimsArray(Diagonal(x))
etc.
cc @nickrobinson251 @mcabbott

[glennmoy]

Is/was there already an issue for this?

[nickrobinson251]

we discussed this before somewhere, right? might be good to go look back at what the considerations were.
I suspect it's that Symmetric(x) isa Symmetric should always be true (or an error)?

[Tokazama]

I'd just like to note that I'm getting close to reaching feature parity with ImageCore.jl by combining NamedDims.jl and AxisIndices.jl. This one has been a bit of a headache. I know it's easier in the short term to have NamedDimsArray the top most layer, but it ends up changing the behavior of stuff that expects a certain array type following methods like permutedims.

[mcabbott]

Not sure there was an issue, but at some point I had a PR to make PermutedDimsArray(nda, ...) return a NamedDimsArray, and there was some discussion there, in which some felt strongly that T(x) isa T should always hold. (I have forgotten who though.)

I guess I still think that, in the absence of a lower-case constructor like transpose, the type-name Symmetric plays two roles, and so it's not so weird to overload it.

[rafaqz]

I remember talking about it here somewhere ages ago. DimensionalArray in DimensionalData.jl is always the outer wrapper and I've had no issues with it so far.