[julia-users] Loops versus maps, 1-D vs. 2-D access

[Michael Grant]

Consider the current implementation of a particular instance of vcat in 
abstractarray.jl, line 518:

function vcat(X::Number...)
    T = None
    for x in X
        T = promote_type(T,typeof(x))
    end
    hvcat_fill(Array(T,length(X)), X)
end

Is there a specific reason why an implementation like this was not chosen?

function vcat(X::Number...)
    T = promote_type(map(typeof,X))
    hvcat_fill(Array(T,length(X)), X)
end

In addition, consider the implementation of hvcat_fill, same file, line 827:

function hvcat_fill(a, xs)
    k = 1
    nr, nc = size(a,1), size(a,2)
    for i=1:nr
        @inbounds for j=1:nc
            a[i,j] = xs[k]
            k += 1
        end
    end
    a
end

For the *specific *usage above, is there any reason not to prefer this:

function hvcat_fill(a, xs)
    @inbounds for j=1:length(a)
        a[j] = xs[j]
    end
    a
end

This function is used in three different places; in each case, the array is 
indeed contiguous.