# Convert 2d matrix to 3d

**URL:** <https://discourse.julialang.org/t/convert-2d-matrix-to-3d/55752>\
**Category:** General Usage\
**Created:** [February 22, 2021, 12:15am UTC](https://discourse.julialang.org/t/convert-2d-matrix-to-3d/55752 "2021-02-22T00:15:34Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![stephenll](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/stephenll/32/28751_2.png) [@stephenll](https://discourse.julialang.org/u/stephenll)\
**Post date:** [February 22, 2021, 12:15am UTC](https://discourse.julialang.org/t/convert-2d-matrix-to-3d/55752/1 "2021-02-22T00:15:34Z")

</div>

I have a large matrix, `n x m` in size.

I like to convert it to a 3d matrix, taking `k` rows at a time. `n` is always divisible by `k`.

newmatrix’s dimensions should be ‘k x M x n/k’

newmatrix[:,:,1] .= bigmat[1:k,:]  
newmatrix[:,:,2] .= bigmat[k+1:2k,:]  
newmatrix[:,:,3] .= bigmat[2k+1:3k,:]

I’ve tried `reshape` and `permutedims` so far with no luck.

I guess I could create the new matrix and write a loop to do the indexing and assigning. I figured there would be an easier way.

Thanks for any help.

---

<div class="post-metadata">

**Author:** ![mcabbott](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mcabbott/32/6603_2.png) [@mcabbott](https://discourse.julialang.org/u/mcabbott)\
**Post date:** [February 22, 2021, 12:32am UTC](https://discourse.julialang.org/t/convert-2d-matrix-to-3d/55752/2 "2021-02-22T00:32:46Z")

</div>

I made a package for doing such things, when I got tired of getting the permutations wrong by hand:

```julia
julia> using TensorCast

julia> N,M,K = 6,7,2; bigmat = rand(Int8,N,M)
6×7 Matrix{Int8}:
 -115 -123 56 64 -60 -99 102
   16 7 5 -97 -56 -101 -19
   79 119 -88 -110 -51 113 62
   21 -125 32 81 -125 105 -50
  -57 15 -34 3 85 -67 8
 -116 17 -11 67 -109 -104 25

julia> @cast out[i,m,k] := bigmat[(i,k),m] k:K
3×7×2 PermutedDimsArray(::Array{Int8, 3}, (1, 3, 2)) with eltype Int8:
[:, :, 1] =
 -115 -123 56 64 -60 -99 102
   16 7 5 -97 -56 -101 -19
   79 119 -88 -110 -51 113 62

[:, :, 2] =
   21 -125 32 81 -125 105 -50
  -57 15 -34 3 85 -67 8
 -116 17 -11 67 -109 -104 25

julia> @pretty @cast out[i,m,k] := bigmat[(i,k),m] k:K
begin
    @assert_ ndims(bigmat) == 2 "expected a 2-tensor bigmat[(i, k), m]"
    local (sz_i, sz_k, sz_m) = (:, K, size(bigmat, 2))
    local quetzal = PermutedDimsArray(reshape(bigmat, (sz_i, sz_k, sz_m)), (1, 3, 2))
    out = quetzal
end

julia> out == permutedims(reshape(bigmat, :,K,M),(1,3,2))
true

```

Here `k:K` means `for k in 1:K`, and `(i,k)` is a combined index, with `i` changing fastest.