Fix #133: compute_offset1 for an IdOffsetRange axis

[jishnub]

Now

julia> a12 = zeros(3:8, 3:4);

julia> r = OffsetArrays.IdOffsetRange(Base.OneTo(3), 5)
6:8

julia> a12[r, 4] .= 3
3-element view(OffsetArray(::Array{Float64,2}, 3:8, 3:4), 6:8, 4) with eltype Float64 with indices 6:8:
 3.0
 3.0
 3.0

julia> a12
6×2 OffsetArray(::Array{Float64,2}, 3:8, 3:4) with eltype Float64 with indices 3:8×3:4:
 0.0  0.0
 0.0  0.0
 0.0  0.0
 0.0  3.0
 0.0  3.0
 0.0  3.0

[codecov]

Codecov Report

Merging #134 into master will increase coverage by 2.03%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #134      +/-   ##
==========================================
+ Coverage   84.26%   86.29%   +2.03%     
==========================================
  Files           2        2              
  Lines         197      197              
==========================================
+ Hits          166      170       +4     
+ Misses         31       27       -4     

Impacted Files	Coverage Δ
src/axes.jl	85.36% <ø> (+9.75%)	⬆️

---

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 c6fae67...cb4a803. Read the comment docs.

[jishnub]

Fixes #133

[johnnychen94]

Apparently you've figured out what it should be! I myself is still a little lost here, approve because the test cases are good. It would be better if @timholy can take a look before I merge this.

[jishnub]

This is similar to an issue in Base that I had described here, the fix is this:

compute_offset1(parent, stride1::Integer, dims, inds::Tuple{AbstractRange}, I::Tuple) =
    (@_inline_meta; compute_linindex(parent, I) - stride1*first(axes1(inds[1]))) # potentially index-offsetting case

This is equivalent to our fix here, as Base.axes1(r) == r for an IdOffsetRange.

[mbauman]

I'd hope that this method can be deleted entirely after JuliaLang/julia#37204 is merged/backported — those base behaviors should be correct (but this fix makes sense to help ensure correctness on older versions).