Ternary operator not understood about tensorcast.jl HOT 4 CLOSED

Right now this is equivalent to q = @. (x> 0.5) ? z : y, which gives the same error. What will work is ifelse:

julia> @cast q[i, j] := ifelse(x[i, j] > 0.5, z[i, j],  y[i, j])
5×5 Matrix{Float64}:
 0.0  1.0  0.0  1.0  1.0
 1.0  0.0  1.0  0.0  0.0
 1.0  0.0  0.0  0.0  0.0
 0.0  1.0  1.0  0.0  1.0
 0.0  1.0  1.0  1.0  1.0

But perhaps your real use has more complicated functions on the right? It might not be hard to detect this, and re-write it as q = @. ((x,z,y) -> (x> 0.5) ? z : y)(x, z, y)

from tensorcast.jl.

But perhaps your real use has more complicated functions on the right? It might not be hard to detect this, and re-write it as q = @. ((x,z,y) -> (x> 0.5) ? z : y)(x, z, y)

Yeah that's my use-case. I use ifelse right now, but I think the ternary version would avoid some computation. It's not really a big deal for me, but I figured I'd report it anyways.

from tensorcast.jl.

Sure. So the transformation needed is something like this:

@cast q[i, j] := (abs(x[i, j]) > 0.5) ? exp(y[i]) : atan(y[i], z[j]+t[j,i])
# q = @. ((x,y,z,t) -> abs(x)>0.5 ? exp(y) : atan(y,z+t))(x, y, z', t')

Perhaps:

julia> @pretty @cast q[i, j] := (abs(x[i, j]) > 0.5) ? exp(y[i]) : atan(y[i], z[j]+t[j,i])
begin
    local quelea = transmute(z, (nothing, 1))
    local shrew = transmute(t, (2, 1))
    local waterbuffalo(x, y, quelea, shrew) = begin
                if abs(x) > 0.5
                    exp(y)
                else
                    atan(y, quelea + shrew)
                end
            end
    q = @__dot__(waterbuffalo(x, y, quelea, shrew))
end

from tensorcast.jl.

This should be fixed on master. Evil test cases would be welcome, the ones I tested are quite simple:
https://github.com/mcabbott/TensorCast.jl/blob/master/test/four.jl#L44

from tensorcast.jl.