What about an `undefs` function in `Base`?

General Usage
61

Perhaps there should be a @CuArray macro if there is not one already. StaticArrays.jl has a @SArray macro.

julia> @SArray [x for x in 1:5]
5-element SVector{5, Int64} with indices SOneTo(5):
 1
 2
 3
 4
 5
62

Funny that SA doesn’t work with comprehensions.

julia> SA[1, 2, 3]
3-element SVector{3, Int64} with indices SOneTo(3):
 1
 2
 3

julia> SA[x for x=1:3]
ERROR: MethodError: Cannot `convert` an object of type
  Int64 to an object of type
  SA

So for comprehensions, you have to use the macro? Sidenote, the space between the macrocall and the brackets is optional:

julia> SArray[@SArray[1,2,3] for _=1:4]
4-element Vector{SArray}:
 [1, 2, 3]
 [1, 2, 3]
 [1, 2, 3]
 [1, 2, 3]

julia> @SArray[@SArray[1,2,3] for _=1:4]
4-element SVector{4, SVector{3, Int64}} with indices SOneTo(4):
 [1, 2, 3]
 [1, 2, 3]
 [1, 2, 3]
 [1, 2, 3]
63

I think custom array types can define constructors that accept generators, e.g.

CuArray(Int, (i*j for i=1:2, j=1:2))
2 Likes
64

Should that be part of AbstractArray interface?

CuArray{Int}(i*j for i=1:2, j=1:2)
1 Like
65

Undefs.jl is now registered:

2 Likes