From R array to Numpy array

Question

Lets say, I have a following R array

a <- array(1:18, dim = c(3, 3, 2))

r$> a
, , 1

     [,1] [,2] [,3]
[1,]    1    4    7
[2,]    2    5    8
[3,]    3    6    9

, , 2

     [,1] [,2] [,3]
[1,]   10   13   16
[2,]   11   14   17
[3,]   12   15   18

and now I want to have the same array in Python numpy. I use

a = np.arange(1, 19).reshape((3, 3, 2))

array([[[ 1,  2],
        [ 3,  4],
        [ 5,  6]],

       [[ 7,  8],
        [ 9, 10],
        [11, 12]],

       [[13, 14],
        [15, 16],
        [17, 18]]])

But somehow, those two do not look like the same. how can one replicate the same array in Python?

I also tried

a = np.arange(1, 19).reshape((2, 3, 3))

array([[[ 1,  2,  3],
        [ 4,  5,  6],
        [ 7,  8,  9]],

       [[10, 11, 12],
        [13, 14, 15],
        [16, 17, 18]]])

which is also not identical.

score 5 · Accepted Answer · 2023-10-11 09:37:49Z

Basically the difference is in the ordering of the array.

Arrays can be row-major or column-major. See here about some info.

R by default constructs column-major arrays, and Python constructs row-major arrays. That’s why the default constructors you use will not give the same output. To fix that you can tell Python to construct the array in column-major by saying it should be in Fortran contiguous (column-major):

np.reshape(np.arange(1,19), (3,3,2), "F")

array([[[ 1, 10],
        [ 4, 13],
        [ 7, 16]],

       [[ 2, 11],
        [ 5, 14],
        [ 8, 17]],

       [[ 3, 12],
        [ 6, 15],
        [ 9, 18]]])

It looks different, but the underlying data is the same as in R.

If you do indexing you can see that it performs exactly the same:

R:

a[1,1,]
[1]  1 10

Python:

a[0,0,]
>>> array([ 1, 10])

Comparing that to the formerly top-rated answer, which will not give you the same values. It may visually look the same, but it does not function the same.

b = np.arange(1, 19).reshape((2, 3, 3)).transpose(0,2,1)
b[0,0,]
>>> array([1, 4, 7])

See also this great vignette by the reticulate team about the differences: https://cran.r-project.org/web/packages/reticulate/vignettes/arrays.html

It does, it just doesn't look the same, because Python doesn't provide structured output like R — 14 hours ago
That is what confused me, I also tried the same, and because the printed versions did not look the same, I assumed it produced something else. — 14 hours ago

Nayem Jaman TusherNayem Jaman Tusher 3843 silver badges14 bronze badges · Accepted Answer · 2023-10-11 09:27:43Z

The order of the items in NumPy must match the column-major order in order to reproduce the same array as in R.

This may be accomplished by reshaping first, then transposing the axes.

Here is how I go about it:

import numpy as np

# Create the array in NumPy with the desired shape and order
a = np.arange(1, 19).reshape((3, 3, 2), order='F')

print(a)

This will give you an array in NumPy that matches the order of the R array:

array([[[ 1, 10],
        [ 2, 11],
        [ 3, 12]],

       [[ 4, 13],
        [ 5, 14],
        [ 6, 15]],

       [[ 7, 16],
        [ 8, 17],
        [ 9, 18]]])

SotosSotos 51.2k6 gold badges33 silver badges67 bronze badges · Accepted Answer · 2023-10-11 09:40:42Z

In addition to np.arrange you can also use the more flexible np.linspace, i.e.

np.linspace(1, 18, 18).reshape((3, 3, 2), order='F')

array([[[ 1., 10.],
        [ 4., 13.],
        [ 7., 16.]],

       [[ 2., 11.],
        [ 5., 14.],
        [ 8., 17.]],

       [[ 3., 12.],
        [ 6., 15.],
        [ 9., 18.]]])

From R array to Numpy array