Convert array into matrix in place

Is there any way to convert an array into a matrix in place?
Thank you.

A matrix is a type of array. Do you mean convert a 1d array (a “vector”) into a 2d array (a “matrix”)? Use reshape:

julia> v = [0:24;]
25-element Vector{Int64}:
  0
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24

julia> reshape(v, 5,5)
5×5 Matrix{Int64}:
 0  5  10  15  20
 1  6  11  16  21
 2  7  12  17  22
 3  8  13  18  23
 4  9  14  19  24
3 Likes

To complete the answer of stevengj, a Matrix{T} is nothing else than an alias for Array{T,2}, as Vector{T} is nothing else than an alias for Array{T,1}

2 Likes

Thank you, but I noticed that using reshape function in a multi threading for loop eats too much memory.
Are there in place alternatives?

reshape is in-place:

julia> v = rand(10^6);

julia> using BenchmarkTools

julia> @btime reshape($v, 1000,1000);
  42.719 ns (2 allocations: 96 bytes)

Notice that the allocation here is quite small and is essentially just a wrapper object.

You can actually use an explicit wrapper object of type Base.ReshapedArray with some undocumented internals:

julia> reshape2(v, dims) = Base.__reshape((v,IndexLinear()), (1000,1000))
reshape2 (generic function with 1 method)

julia> v = rand(10^6);

julia> @btime reshape2($v, (1000,1000));
  2.034 ns (0 allocations: 0 bytes)

(Beware that the above reshape2 function doesn’t do any argument checking, e.g. it doesn’t check whether the sizes are compatible.)

See also the discussion in reshape(::Array, ...) is slow and allocates · Issue #36313 · JuliaLang/julia · GitHub

2 Likes

A safer version (with basically the same performance, but doing argument checking) should be:

reshape2(a, dims) = invoke(Base._reshape, Tuple{AbstractArray,typeof(dims)}, a, dims)
2 Likes

Does this work?

reshape(view(v, :), 1000, 1000)

Inserting the “view” first seems to avoid all heap allocations.