use the first linear index instead of 1 for zero-argument getindex (redo #195)

[johnnychen94]

JuliaLang/julia#39379 (comment)

Here's the root problem:

https://github.com/JuliaLang/julia/blob/69d24532980bda6db510b6d4b71f66e7ffff2e81/base/abstractarray.jl#L1196

Edit: This should probably be first(eachindex(A)) or somesuch.

Changes:

# master
OffsetArray([6], 2:2)[] # BoundsError
# this PR
OffsetArray([6], 2:2)[] == 6

Forked from JuliaLang/julia#39393, I'll wait until that gets merged.

Technically this is a breaking change, but since v1.5.2 was released yesterday, I'd see this as a bug fix and bump a patch version for it.

cc: @jishnub @mbauman

[codecov]

Codecov Report

Merging #196 (043b26d) into master (1263513) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #196      +/-   ##
==========================================
- Coverage   98.93%   98.93%   -0.01%     
==========================================
  Files           5        5              
  Lines         283      281       -2     
==========================================
- Hits          280      278       -2     
  Misses          3        3              

Impacted Files	Coverage Δ
src/axes.jl	100.00% <ø> (ø)
src/OffsetArrays.jl	98.93% <100.00%> (+<0.01%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1263513...2b82093. Read the comment docs.

[mbauman]

Typically we don’t consider making errors functional breaking. This went in the perfect order to be nonbreaking: bug -> error -> functional. :)

[jishnub]

Nice! I was not sure if A[] should behave like only(A) treating A as an iterator, or just deal with the indices as they're provided, in which case it had to work with A[1]. I chose the more conservative approach because this may be changed later in a non-breaking way, as in this PR.

This is the correct solution as it resolves the iterator/array debate. This works identically for both.

[johnnychen94]

@jishnub I reverted the version bump here, hopefully, we can include #197 in OffsetArrays v1.5.3, too. I'll leave it to you to release the new version.