Creating iterators for tuples of arbitrary dimension

General Usage
1

Suppose I wish to take as input some integer d and iterate over all tuples (i_1, ... i_d), with each entry lying in 1 to 10. For example, I may take as input a vector v of length d and compute the sum v_1 i_1 + ... + v_d i_d for every such tuple of dimension d. How can I achieve this ? CartesianIndices does not seem to solve this.

2

Maybe I don’t understand the question, but some variant of

d = 2
[sum(Tuple(ι)) for ι in CartesianIndices(ntuple(_ -> 1:10, d))]

should work (make d a Val type for type stability and/or wrap in a vec or flatten if needed, etc).

2 Likes
3

I’m not certain what you are trying to do, but see if this Iteration utilities · The Julia Language is useful.

4

Thanks ! CartesianIndices(ntuple(_ -> 1:10, d)) seems like a great way to generate the indices. I seems to be a collection through which I can iterate, and each iteration is of type CartesianIndex{d}.

  1. Is it possible to convert an object of type CartesianIndex{d} into a vector. You showed how to convert it into a Tuple?
  2. If not, can we take a linear combinations, such as v_1 i_1 + ... + v_d i_d ?
5

Yes, eg

ι = CartesianIndex(1, 2, 3)
collect(Tuple(ι))

but I would recommend SVector(Tuple(ι)...) for efficiency.

1 Like
6

SVector(Tuple(I)) works.

2 Likes
7

Thank you, it worked !

8

Thank you, it worked !