How do I create a 1,2,3 3x1 Matrix?

While introducing Arrays, I want to explain the difference in Julia between a normal column vector and a matrix with a single column… Is there a way to create a one-column matrix like b = collect([1 2 3]') in a simpler way (and without using rand, zeros, the Array constructor with undef/nothing…) ?


Or even

reshape([1:3;], :, 1)

if you do not want to care about number of elements.

1 Like

There is no way to create it with just using the brackets syntax without introducing other functions, including reshape ?

[ i for i=1:3, j=1:1] but it is not compact.

FYI, probably still not what you looking for though.

[1:3 1:3][:, [1]]

Matrix{Float64}(undef, 3, 1)
Does that work for you?

normally I write hcat(1:3)


Or maybe fill(0,(3,1)).

yeah, thanks everyone, I think the answer is that it is not possible to use some simple a = [1 ; 2 ; 3 ;] like syntax to create a one-column matrix, as it is reinterpreted as a column vector… I’ll use one of the methods you posted, it is just that they all use already some other concepts…

a=[1 2 3]'

[1, 2, 3][:,:]
3×1 Matrix{Int64}:

And if @DNF was online, he would likely suggest:

permutedims([1 2 3])

The limitation is that in bracket notation a space is used to indicate hcat and it is ignored if you have only one argument passed.

1 Like

One wouldn’t recommend this, but it is interesting:

using LinearAlgebra
julia> [1, 2, 3]*I
3×1 Matrix{Int64}:

Note that there is a PR, which, if I understand it correctly, means that a = [1 ; 2 ; 3 ;;] may be a future syntax for this. Specifically see the proposed NEWS entry.


I thought a Vector was a synonym for a 1 Column Matrix? Why would there be a need to differentiate?

1 Like

There are many reasons, e.g.:

  • they have different ndims
  • you can resize Vector, but you cannot resize Matrix
  • there are functions that accept only Vector or only Matrix (especially in LinearAlgebra)
  • Tables.jl tables usually require AbstractVector, but disallow 1-column AbstractMatrix as columns
1 Like

Mathematically they are the same, but they are different objects in Julia. [1,2,3] is a uni-dimensional array, collect([1 2 3]') is a two- dimensional array where “it happens” that the second dimension has size 1, i.e. there is only one column. So, if a function requires AbstractArray{T,1} the first will fit, but the second will not.
Note that this possible confusion doesn’t arise with row vectors, as all row vectors in Julia have 2 dimensions.

You can also use


if that’s clearer than collect

@tomerarnon, you probably meant:

Matrix([1 2 3]')
3×1 Matrix{Int64}:
1 Like