reshape with zero dimensions

[cossio]

I think this should be able to return without error.

julia> reshape(randn(0,4),0,:)
ERROR: DivideError: integer division error
Stacktrace:
 [1] div
   @ ./int.jl:284 [inlined]
 [2] divrem
   @ ./div.jl:162 [inlined]
 [3] divrem
   @ ./div.jl:158 [inlined]
 [4] _reshape_uncolon
   @ ./reshapedarray.jl:127 [inlined]
 [5] reshape(parent::Matrix{Float64}, dims::Tuple{Int64, Colon})
   @ Base ./reshapedarray.jl:118
 [6] reshape(::Matrix{Float64}, ::Int64, ::Colon)
   @ Base ./reshapedarray.jl:117
 [7] top-level scope
   @ REPL[1]:1

[giordano]

What shape do you expect it to return?

[cossio]

@giordano (0, 4) ?

[giordano]

Why not (0, 6829)?

[cossio]

@giordano `(0,4) is the original size. I'd say it should prefer that?

[giordano]

Then my next question would be why you're reshaping it if you want to keep the same shape 🙂

[cossio]

Okay. How about reshape(zeros(0,4,3),0,:)? I think it could return zeros(0,12).

[jishnub]

What shape do you expect it to return?

I think it's reasonable to expect it to be consistent with this:

julia> reshape(zeros(0,4), Val(2))
0×4 Matrix{Float64}

[JeffBezanson]

Heh, never realized this works, though of course it makes sense:

julia> reshape(zeros(0,4),0,5)
0×5 Matrix{Float64}

Still it seems to me the example in the OP might as well work.

[nlw0]

Why not (0, 6829)?

Anything goes, so it should be able to return without an error...

[frankwswang]

This discussion in the Julia Discourse might be relevant.

I think if the returned array's shape is the same as the input array's shape and the input array is a container of plain data (e.g. Float64) at the same time, reshape should just return the input array regardless of its dimensions, like the identity.

[nsajko]

In the meantime I opened an issue that turns out to partially duplicate this one: #54245

Then I fixed it in #54261 (not in v1.11).

Perhaps the sizing for empty arrays should be considered again? The linked PR implements this behavior:

julia> a = randn(0,4)
0×4 Matrix{Float64}

julia> size(reshape(a, 0, :))
(0, 0)

julia> ans == size(reshape(a, Val(2)))
false