Zero-dimensional views of AbstractVectors assume that the parent is 1-indexed

[jishnub]

This is the same issue reported in OffsetArrays, but I'm posting it here as I'm not sure if the fix should go into Base. Please close this if it's more appropriate for the fix to be in OffsetArrays.

For a zero-dimensional array a we usually expect a[] == a[1]. However this does not work with zero-D views of OffsetVectors

julia> using OffsetArrays

julia> a = OffsetArray(1:3, 0:2);

julia> b = @view a[0]
0-dimensional view(OffsetArray(::UnitRange{Int64}, 0:2), 0) with eltype Int64:
1

julia> b[]
1

julia> b[1]
2

[stevengj]

I've traced it down to this line, which seems like it incorrectly assumes that linear indices start at 1.

@timholy, it seems that you wrote that code back in #16260, can you comment?

[timholy]

@timholy, it seems that you wrote that code back in #16260, can you comment?

I often struggle to remember last Tuesday, so I don't recall anything useful, but that looks like it predates offset axes (going by the blog post date). Thanks for catching this use case, @jishnub.

@stevengj, that line enacts the "linear indexing starts at 1" policy with the exemption for 1d arrays on

julia/base/subarray.jl

Lines 377 to 378 in 41e603e

	compute_offset1(parent, stride1::Integer, dims::Tuple{Int}, inds::Tuple{Slice}, I::Tuple) =
	(@_inline_meta; compute_linindex(parent, I) - stride1*first(axes(parent, dims[1]))) # index-preserving case

I'm guessing we need a 0-dimensional exemption as well?

[stevengj]

I'm guessing we need a 0-dimensional exemption as well?

The underlying array is 1d in this case, and since this is computing an offset index into the original array shouldn't the 1d exemption apply?

[fghorow]

A question: would this bug affect the ordering of an OffsetArray (I_Ωpp) being written out via a Numpy call of the form np.save("./I_Omega_pp_R_interpolated.npy",I_Ωpp)?
If so, I've been chasing problems caused by this for weeks! :-(

Ooops. Edited to add: using PyCall; np = pyimport("numpy") prior to referencing np...