@cast works for hcat but not for [ ]

[prittjam]

This works,

@cast P[i,j,k] := hcat(R[:,:,k],X₁[:,k])[i,j]

but this does not

@cast P[i,j,k] := [R[:,:,k] X₁[:,k]][i,j]

It gives the error,

DimensionMismatch: stack expects uniform slices, got axes(x) == (1:3,) while first had (1:3, 1:4)

Should it? Thanks!

[mcabbott]

Here's what it's doing:

julia> @pretty @cast P[i,j,k] := [R[:,:,k] X₁[:,k]][i,j]
begin
    @boundscheck ndims(R) == 3 || throw(ArgumentError("expected a 3-tensor R[:, :, k]"))
    @boundscheck axes(R, 3) == axes(X₁, 2) || throw(DimensionMismatch("range of index k must agree"))
    @boundscheck ndims(X₁) == 2 || throw(ArgumentError("expected a 2-tensor X₁[:, k]"))
    local vicuña = sliceview(R, (:, :, *))
    local alligator = sliceview(X₁, (:, *))
    local dolphin = @__dot__([vicuña alligator])  # <-- this broadcast does nothing
    local redpanda = lazystack(dolphin)
    P = redpanda
end

julia> :([vicuña alligator]) |> dump
Expr
  head: Symbol hcat
  args: Array{Any}((2,))
    1: Symbol vicuña
    2: Symbol alligator

julia> @. hcat(1:2, 3:4)
2-element Vector{Matrix{Int64}}:
 [1 3]
 [2 4]

julia> @. [1:2 3:4]
2×2 Matrix{Int64}:
 1  3
 2  4

It wouldn't be impossible to make it recognise Expr(:hcat, ...) at some early stage, and convert it to a function call. Likewise vcat, vect... from :([1 2 3; 4 5]) |> dump also row not hvcat? And Float32[1:2 3:4] is typed_hcat. But I haven't done so. Probably writing out hcat is best for now.

What will never work is Vector[1:2, 3:4], as the macro cannot distinguish this from indexing.