In-place map of vector functions across slices of multidimensional arrays

[stevengj]

@tianrluo

I guess this is going to limit the capability for applications like sequentially broadcasting multiple vector functions along different dimensions, is that right?

If you are doing more general operations along different dimensions of multidimensional arrays, I think looping is a lot more powerful than trying to shoehorn it into broadcasting — especially using Julia’s CartesianIndex machinery to easily express loops over N-dimensional arrays in code that is generic for any N. For example, see this code for wave propagation in N dimensions.