The `sort` only change order of the first column in matrix?

General Usage
1

Hi friends,

I just found a strange behave of the method of sort

a1 = [ 
    669.57  52.3
    657.52  56.18
    638.38  65.09
    634.17  72.41
    632.08  80.61]
sort(a1,dims=1)

will give me below result

5×2 Matrix{Float64}:
 632.08  52.3
 634.17  56.18
 638.38  65.09
 657.52  72.41
 669.57  80.61

why the order of 2nd column donot change?

my Julia version is Version 1.9.3 (2023-08-24)

2

Your second column is already sorted. What change do you expect?

2 Likes
3

I think you’re looking for sortslices

5 Likes
4

Oh, I see. I just misunderstand the sort is sorting each row.

5

You explictly told Julia to sort columns with dims=1. Julia is column-major so the first dimension goes down the column. If you want to sort within the rows, use dims=2.

1 Like
6

That is generally the case also for row-major convention. Column/row-majority refers to memory layout, not indexing order.

1 Like
7

What’s wrong with using sort with dims keyword?

8

I think OP implicitly assumed a structure to his matrix where elements are grouped row-wise so sorting any column would mean all columns are sorted by the sorting order of that column (like when sorting a DataFrame by the values of a single column). E.g.

julia> x = [2 "b"; 1 "c"; 3 "a"]
3×2 Matrix{Any}:
 2  "b"
 1  "c"
 3  "a"

gives

julia> sort(x, dims = 1)
3×2 Matrix{Any}:
 1  "a"
 2  "b"
 3  "c"

with sort but

julia> sortslices(x, dims = 1)
3×2 Matrix{Any}:
 1  "c"
 2  "b"
 3  "a"

with sortslices (note the pairwise correspondence 1 => c, 2 => b, 3 => a is retained here).

1 Like
9

Ah, perhaps. I say, that’s well reasoned, I would never have guessed.