preservedims in tables

[rafaqz]

Forgot to PR this a while ago

Closes #896

@asinghvi17 want to review?

[asinghvi17]

Will test it out, thanks! @alex-s-gardner this is thing number 1 for saving glacier height histogram as GeoJSON

[asinghvi17]

This works on this branch but fails when I merge it with master, just FYI

[asinghvi17]

Also this kind of works the other way around than I had expected, I thought that preservedims would make the dimarray of that dimension, but it makes a dimarray of the other dimensions. See this example:

julia> da = @d sin.(hypot.(X(LinRange(-9,9,100)), Y(LinRange(0, 18, 100))) .+ Ti(LinRange(0, 6pi, 100)))
┌ 100×100×100 DimArray{Float64, 3} ┐
├──────────────────────────────────┴─────────────────────────────────────────────────────────────────────────────── dims ┐
  ↓ X  Sampled{Float64} LinRange{Float64}(-9.0, 9.0, 100) ForwardOrdered Regular Points,
  → Y  Sampled{Float64} LinRange{Float64}(0.0, 18.0, 100) ForwardOrdered Regular Points,
  ↗ Ti Sampled{Float64} LinRange{Float64}(0.0, 18.84955592153876, 100) ForwardOrdered Regular Points
└────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
[:, :, 1]
 ⋮      ⋱  

julia> DimTable(da; preservedims=(Ti,))
DimTable with 100 rows, 2 columns, and schema:
 :Ti      …  Float64
 :layer1     DimMatrix{Float64, Tuple{X{Sampled{Float64, LinRange{Float64, Int64}, ForwardOrdered, Regular{Float64}, Points, NoMetadata}}, Y{Sampled{Float64, LinRange{Float64, Int64}, ForwardOrdered, Regular{Float64}, Points, NoMetadata}}}, Tuple{Ti{Sampled{Float64, LinRange{Float64, Int64}, ForwardOrdered, Regular{Float64}, Points, NoMetadata}}}, SubArray{Float64, 2, Array{Float64, 3}, Tuple{Slice{OneTo{Int64}}, Slice{OneTo{Int64}}, Int64}, true}, Symbol, NoMetadata}

where I would have expected the dimtable to have columns (X, Y, layer1) where layer1 was a vector of DimVectors in Ti.

[rafaqz]

Thats what I thought it did too 😆