I'm writing to get some general advice from more advanced Julia users than me (which is probably most users). Coming from Matlab, the concepts of pass-by-reference and array-views initially scared me a bit (in terms of generating hard to find bugs). Pass-by-reference is a bit less scary when using the syntactic convention foo!(x,y) but I'm still a bit wary of the array-view concept. This came up when I recently realized that the output of vec(x) and reshape(A, dims) can share memory with it's arguments. Wow, I had no idea. Indeed, the only way I would know that vec shared memory is if I happened to test it:
a = rand(2,2) b = vec(a) b[1] = 0.0 a[1] == 0.0 #<- true The documentation doesn't give a hint: help> vec INFO: Loading help data... Base.vec(Array) -> Vector Vectorize an array using column-major convention. Is there a general way I can extrapolate when a function's return value will share memory with inputs? For example, I could imagine diag, diagm, transpose, reshape, hcat, vcat could all possibly sharing memory with inputs but I would, a-priori have no idea which? To me, this situation warrants another syntactic convention like foo%, foo> or foo*. Besides wanting to know this for myself, I'm considering using Julia in my intro stats classes. Having to explain shared memory stuff is too complicated for beginners, in my opinion, and having a syntactic convention would allow me to say..."stay away from those unless you know what your going". Thanks!
