I have noticed a difference in behavior between python's numpy.permute_dims and julia's Base.permutedims. On an input 3x3x3 matrix containing elements 0:26, inclusive in both languages, they agree for the axes argument (1,2,0) but disagree for (0,2,1). As far as I can tell from the docs, these functions should be equivalent. There's a note about permutedims being non-recursive, but I don't see why that should have behavioral consequences. Julia is also column-major order, but again I don't see why that should effect the overall behavior. Python code: arr = np.array([[[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, 25, 26]]]) arr_perm = np.permute_dims(arr, axes=[0,2,1]) Output: array([[[ 0, 3, 6], [ 1, 4, 7], [ 2, 5, 8]], [[ 9, 12, 15], [10, 13, 16], [11, 14, 17]], [[18, 21, 24], [19, 22, 25], [20, 23, 26]]]) Julia code: arr = [ 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 25 26 ] arr_perm = permutedims(arr, [1,3,2]) Output: 3×3×3 Array{Int64, 3}: [:, :, 1] = 0 9 18 3 12 21 6 15 24 [:, :, 2] = 1 10 19 4 13 22 7 16 25 [:, :, 3] = 2 11 20 5 14 23 8 17 26 Continue reading...
