Just a thought: for 1D KeyedArray sort would seem to be a valid operation. Currently t

Sorting about axiskeys.jl HOT 3 CLOSED

mcabbott commented on June 12, 2024

Sorting

from axiskeys.jl.

Comments (3)

mcabbott commented on June 12, 2024

Do you have an example of this failing? It ought to work, but perhaps you found a bug:

julia> sort(wrapdims(rand(Int8, 5), 'a':'e'))
1-dimensional KeyedArray(...) with keys:
↓   5-element Vector{Char}
And data, 5-element Array{Int8,1}:
 ('e')  -38
 ('a')   -5
 ('d')    3
 ('c')   59
 ('b')   65

from axiskeys.jl.

kcajf commented on June 12, 2024

Ah great. I just realised I didn't have a 1D array, but a 1xN matrix:

julia> sort(wrapdims(rand(Int8, 1, 5), 1:1, 'a':'e'), dims=2)
2-dimensional KeyedArray(...) with keys:
↓   1-element UnitRange{Int64}
→   5-element OneTo{Int}
And data, 1×5 Array{Int8,2}:
       (1)  (2)  (3)  (4)  (5)
 (1)  -127    4   65   80   96

The behaviour therefore makes sense, given how multidimensional sort works. I would need to drop the first dimension to get the behaviour I wanted.

from axiskeys.jl.

mcabbott commented on June 12, 2024

Cool. There is also sortslices for higher-dim arrays:

julia> wrapdims(rand(Int8, 2, 5), 11:12, 'a':'e')
2-dimensional KeyedArray(...) with keys:
↓   2-element UnitRange{Int64}
→   5-element StepRange{Char,...}
And data, 2×5 Array{Int8,2}:
        ('a')  ('b')  ('c')  ('d')   ('e')
 (11)  122     83     81     53     -28
 (12)  -62     39     61      0     -88

julia> sortslices(ans, dims=2)
2-dimensional KeyedArray(...) with keys:
↓   2-element UnitRange{Int64}
→   5-element Vector{Char}
And data, 2×5 Array{Int8,2}:
        ('e')  ('d')  ('c')  ('b')   ('a')
 (11)  -28     53     81     83     122
 (12)  -88      0     61     39     -62

from axiskeys.jl.

Recommend Projects