Add conversion from AbstractArray to mxarray

[tqml]

Adds a conversion method to convert Julia abstract array to mxarray to avoid having them treated as structs in MATLAB.

Example:

a = transpose(rand(10)) # transpose
x = mxarray(a) # x is an actual matrix in matlab, was a struct before
y = jvalue(x) # the return value is now a matrix instead of a dictionary

Related issue #196

[tqml]

I removed the definitions for Ranges and BitArrays since they are only be a special case of AbstractArray, would unnecessary allocate data and get copied to MATLAB anyway

Old defs:

mxarray(a::BitArray) = mxarray(convert(Array{Bool}, a))
mxarray(a::AbstractRange) = mxarray([a;])

New definition was tested with additional special matrix types such as Diagonal, transpose and LowerTriangular.

function mxarray(a::AbstractArray{T}) where {T<:MxRealNum}
    mx = mxarray(T, size(a))
    ptr = data_ptr(mx)
    na = length(a)
    dat = unsafe_wrap(Array{T}, ptr, na)
    for (i, ix) in enumerate(eachindex(a))
        dat[i] = a[ix]
    end
    return mx
end

[ViralBShah]

@tqml Would it be helpful to give you commit access to be able to merge your PRs and perform maintenance on the package?

[tqml]

Hey @ViralBShah
Are there other active maintainers as well? Having commit access to merge this PR would be fine, however I can not actively maintain it besides the bare minimum, because I do not really use this package anymore and my MATLAB student license will expire at some point (which will make testing difficult)

[ViralBShah]

I don't know of other active maintainers. It is always welcome for someone to do the minimum to maintain a package, and make releases etc. It is not necessarily a commitment to take on major development work.

If you're ok to get this one merged and release, that would already be quite welcome.

[musm]

Yeah my apologies. I'm the last active maintainer of this package. Haven't been keeping up to speed with all the pull requests.

[musm]

In any case this looks good to me

[musm]

This and the function below still work?

[ViralBShah]

Thanks @musm.

[tqml]

Thanks @ViralBShah for the invite and thanks @musm for merging this PR :)

[ViralBShah]

Is it worth tagging a new release? Seems like a few things have accumulated since the last release.

[tqml]

Good idea, the last release also was also some time ago. Can we justify v0.9?
The different handling of matrix types is kinda a breaking change, but i can hardly imagine that people really use matlab structs for their non-standard matrices.

I'm also not familiar with the Julia release process. Do you just create a new Git Tag and some automation does the rest?

[tqml]

If its fine for you, I would include #231 in the release as well, it adds some new conversations for tuples

[ViralBShah]

Yes, please do include. The process of releasing is fairly straightforward. The way I do it is:

1. Bump the version number in Project.toml in line with the semver spec. I think 0.9 is justified.
2. Click the Register button on https://juliahub.com/ui/Packages (and login with your github account).

[tqml]

@ViralBShah @musm are you familiar why the package needs a build.jl, hardcodes the matlab paths into a deps.jl file and subsequently stores Refs and Ptrs to functions that could be a ccall?

Were there issues to get the file path at "runtime" (or when imported)?
Or was this a limitation of earlier versions of Julia (considering this package has been around a while)

[ViralBShah]

This definitely sounds like how things were done a few years ago. Much of this went away with BinaryBuilder and artifacts but a lot of that won’t help here. Still I wouldn’t be surprised if folks can do better.