eltype and ndims of an OffsetArray matches that of the parent

[jishnub]

Fixes #162

Now

julia> OffsetMatrix{Float64, Matrix{ComplexF64}}
ERROR: TypeError: in Type, in AA, expected AA<:AbstractArray{Float64,2}, got Type{Array{Complex{Float64},2}}

also,

julia> OffsetMatrix{Float64, Vector{Float64}}
ERROR: TypeError: in Type, in AA, expected AA<:AbstractArray{Float64,2}, got Type{Array{Float64,1}}

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@87666ee). Click here to learn what that means.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##             master     #163   +/-   ##
=========================================
  Coverage          ?   99.21%           
=========================================
  Files             ?        4           
  Lines             ?      256           
  Branches          ?        0           
=========================================
  Hits              ?      254           
  Misses            ?        2           
  Partials          ?        0           

Impacted Files	Coverage Δ
src/OffsetArrays.jl	99.44% <100.00%> (ø)

---

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 87666ee...f02d12a. Read the comment docs.

[jishnub]

Leaving open as I'm not entirely sure if there are cases where the ndims might not be identical

[johnnychen94]

I don't think we need that much flexibility here.

There are many ways to insert the missing dimension without memory allocation, e.g.,

Base.ReshapedArray(A, (1, 4, 4), ())

[timholy]

Better to call reshape, though, rather than the constructor itself. Normal users don't want to have to deal with Base.MultiplicativeInverses.

[timholy]

Hmm, this turned out to be pretty breaking for a patch release (at least for ImageCore). In retrospect we should have bumped to 1.4. I would still call this a bug fix, but it's a pretty big (& important) one.

[jishnub]

I'm not familiar with ImageCore, could you point me towards an example that breaks as a consequence of this change?

[johnnychen94]

JuliaImages/ImageCore.jl#149, no worries, that breakage is "expected" and is fixable. Just that this patch release breaks something...

[timholy]

Yep, @johnnychen94 is right. Really, it was misguided of OffsetArrays to implement those methods in the first place.

I've released my share of patch releases that broke something. Just thought I'd mention it since these questions of what to call releases come up all the time (typically, without a crystal-clear answer).