What is the Julian equivalent of Python's `np.eye(5, 5, 1)`

[originalsouth]

What is the Julian equivalent of Python’s np.eye(5, 5, 1)

% python -q
>>> import numpy as np
>>> np.eye(5, 5, 1)
array([[0., 1., 0., 0., 0.],
       [0., 0., 1., 0., 0.],
       [0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 1.],
       [0., 0., 0., 0., 0.]])

[nilshg]

It’s I in the LinearAlgebra standard library:

In [1]: using LinearAlgebra

In [2]: I(5)
5×5 Diagonal{Bool, Vector{Bool}}:
 1  ⋅  ⋅  ⋅  ⋅
 ⋅  1  ⋅  ⋅  ⋅
 ⋅  ⋅  1  ⋅  ⋅
 ⋅  ⋅  ⋅  1  ⋅
 ⋅  ⋅  ⋅  ⋅  1

[RoyiAvital]

Probably using diagm().

In your case:

diagm(1 => [1, 1, 1, 1]);

If you want efficient data structure you may use BandedMatrices.jl.

I am not aware of a data structure specialized for the case the diagonals have the values 0 or 1 (Namely no need for multiplication for Matrix Multiplication).

[originalsouth]

Note that the third index shifts the diagonal up creating an upper shift matrix instead of an identity.

[originalsouth]

Almost diagm(1 => [1, 1, 1, 1]) because of 5x5. Thanks!

[nilshg]

Ah, apologies - an unusual use of eye then, given that the name as I understand it comes from Matlab and is a pun on the Identity matrix.