Allow pairs in constructors

[ParadaCarleton]

Some constructors that feel like they should work, but don't ATM:

julia> z = randn(3, 3) |> x->x' * x |> x->DimArray(x, (ax1=(:x, :y, :z), ax2=(:x, :y, :z)))
ERROR: MethodError: no method matching format(::Tuple{Symbol, Symbol, Symbol}, ::Type{Dim{:ax1}}, ::Base.OneTo{Int64})

Closest candidates are:
  format(::DimensionalData.Dimensions.LookupArrays.LookupArray, ::Type, ::AbstractRange)
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:39
  format(::DimensionalData.Dimensions.LookupArrays.NoLookup, ::Type, ::Any, ::AbstractRange)
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:43
  format(::DimensionalData.Dimensions.LookupArrays.Categorical, ::Type, ::Any, ::AbstractRange)
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:60
  ...

Stacktrace:
   [1-5] ⋮ internal
       @ DimensionalData.Dimensions, DimensionalData
     [6] DimArray
       @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/array/array.jl:326 [inlined]
     [7] #36
       @ Main ./REPL[47]:1 [inlined]
     [8] |>
       @ Base ./operators.jl:915 [inlined]
     [9] #35
       @ Main ./REPL[47]:1 [inlined]
    [10] |>(x::Matrix{Float64}, f::var"#35#37")
       @ Base ./operators.jl:915
Use `err` to retrieve the full stack trace.

Can't use pairs either:

julia> z = randn(3, 3) |> x->x' * x |> x->DimArray(x, (:axis1=>[:x, :y, :z], :axis2=>[:x, :y, :z]))
ERROR: MethodError: no method matching _format(::Pair{Symbol, Vector{Symbol}}, ::Base.OneTo{Int64})

Closest candidates are:
  _format(::Symbol, ::AbstractRange)
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:29
  _format(::DimensionalData.Dimensions.LookupArrays.Irregular{DimensionalData.Dimensions.LookupArrays.AutoBounds}, ::Any, ::Any)
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:99
  _format(::Type{D}, ::AbstractRange) where D<:DimensionalData.Dimensions.Dimension
   @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/Dimensions/format.jl:30
  ...

Stacktrace:
   [1-4] ⋮ internal
       @ DimensionalData.Dimensions, DimensionalData
     [5] DimArray(data::Matrix{Float64}, dims::Tuple{Pair{Symbol, Vector{Symbol}}, Pair{Symbol, Vector{Symbol}}})
       @ DimensionalData ~/.julia/packages/DimensionalData/4TpBG/src/array/array.jl:326
     [6] (::var"#58#60")(x::Matrix{Float64})
       @ Main ./REPL[72]:1
     [7] |>(x::Matrix{Float64}, f::var"#58#60")
       @ Base ./operators.jl:915
     [8] (::var"#57#59")(x::Matrix{Float64})
       @ Main ./REPL[72]:1
     [9] |>(x::Matrix{Float64}, f::var"#57#59")
       @ Base ./operators.jl:915
Use `err` to retrieve the full stack trace.

[rafaqz]

The lookups very much need to be AbstractArray:

(ax1=(:x, :y, :z), ax2=(:x, :y, :z))

So we would have to collect them in format. (probably should have a better error otherwise). Any reason not to just make them arrays from the start?

This looks like it should work though... I remember even making this work at one point writing it and deciding not to just to keep it simpler:

(:axis1=>[:x, :y, :z], :axis2=>[:x, :y, :z])

Not sure how to weight that up

[ParadaCarleton]

So we would have to collect them in format. Any reason not to just make them arrays from the start?

No reason for using a tuple, I just keep forgetting whether it's a tuple or an array :sweat_smile: but I assume if this is a problem for me, I'm probably not the only one.

[rafaqz]

Probably needs a better error catching Tuple and whatever else then.

[ParadaCarleton]

That would work, but is there a problem with just collecting the Tuple?

[rafaqz]

The problem for me is that if it works, people will use it, and its confusing to see a Tuple working as an AbstractArray