I’ve ported a bit of code from R/Matlab/C++ to Julia a couple of times over the past few years and can get things to work, but I’m still not familiar with the best-practice, particularly around type declarations. My goal is improved efficiency – I already have working code in the 3 languages mention…

[image] baptiste: I’m inferring that Vector is an alias for 1D array, and probably Matrix an alias for 2D arrays. This leaves me with the AbstractArray concept – probably overly general for a regular kind of array? Array is the built-in multidimensional array type, and in particular it is a f…

I think you probably want something like: struct Material{T<:Real} wavelength::Vector{T} alpha::Vector{Complex{T}} medium::T end That way, you aren’t defining a single type, you are defining a family of types Material{T} for any real type that you might want, e.g. Material{Float64} for…

That’s very helpful, thanks! The different kinds of arrays are stlll a bit mysterious to me, between Vector, Matrix, Array, AbstractArray, RowVector(which I saw mentioned but doesn’t seem to exist). I’m inferring that Vector is an alias for 1D array, and probably Matrix an alias for 2D arrays. Thi…

[image] baptiste: And then to initialise or define types of their entries I’m not sure if I should go for Complex{Float64}, or Complex{AbstractFloat}, or Complex{Real}. Here, go with Complex{Float64}. But instead of initializing as Ajk = Array{Complex{Float64}}(undef,(3, 3)) it is much eas…

[image] baptiste: function propagator_freespace_labframe(R::Matrix{Float64}, kn::Float64, AlphaBlocks::Array{Complex{Float64}}) N = size(R,2) A = Matrix{Complex{Float64}}(I, 3*N, 3*N) I3 = Matrix{Complex{Float64}}(I, 3, 3) For function arguments, you almost always want to be more generic. In…

Thanks so much for all the helpful directions. The code is indeed a direct port from Matlab (and R, and C++ Armadillo, all 3 with the same basic syntax), and while I was pretty sure this wouldn’t produce good Julia code, it’s hard for a novice to know the “rights” and “wrongs” without some sort of …

[image] stevengj: On a separate note, note that you are allocating tons of copies of your arrays with slicing expressions like AlphaBlocks[:,:,jj] , and your vectorized expressions allocate zillions of temporary arrays. You need to be cautious about writing Matlab-style code in Julia, which does…

[image] baptiste: My biggest doubt about this approach though, is that ultimately A appears in a 3Nx3N linear system A x = b (coupled-dipole equations), and if A is a NxN matrix of SMatrices I don’t really see how it’s going to work (ultimately some Lapack routine will need the 3Nx3N matrix, won…

Some doubts about types

New to Julia

Topic		Replies	Views
Multiplication times for complex matrices: julia vs Matlab New to Julia question	18	3023	March 21, 2020
Function argument Array{Array{ComplexF64,2},1} seems to be ignored New to Julia	19	732	March 21, 2020
Initialise array without specific types Performance question	25	2074	April 30, 2021
Drastic performance hit matrix multiply different types. Internal cast julia vs numpy? Numerics	15	2590	November 4, 2018
StaticArrays, parametric types, type stability General Usage parametric-types , type-stability , staticarrays	7	646	April 13, 2021

Some doubts about types

Related topics