Implement of the RClass idea

[randy3k]

close #138

An example of usage can be found in datatime.jl.

@richardreeve, I guess you may be interested in it.

[richardreeve]

Great, I have a brutalist version of my code (EcoJulia/Phylo.jl#3) that uses this and it seems to work using @rget, though I'm having to create RObjects again in a definitely-wrong sort of way. Are there plans for exporting RClass and rcall_p, which I currently seem to need, or should I just use call using on them specifically?

[randy3k]

RClass and rcall_p are not meant to be used by normal users. I guess we should keep them unexported.

[richardreeve]

Would it be crazy to create a submodule like RCall.Extend for things like protect, unprotect, rcall_p, RClass, isObject, isS4 to make it easy to extend the conversion functionality without having loads of using statements?

[randy3k]

There are too many functions falling into this category. I am not sure if it becomes necessary if most of the internal functions are put in say, RCall.API

[richardreeve]

Yes, I assumed there would be lots of them tbh, which was why I thought something like RCall.API would be useful. Obviously there's also a second much smaller group of functions that need importing too - rcopy, rcopytype, sexp...

[richardreeve]

So, I'm completely happy with this, and I don't want to open a can of worms, but when multiple people start providing converters for the same R type, is there going to be a way forward? As things stand (if I understand Julia multiple dispatch correctly, which I probably don't!) are they going to overwrite one another, with the last man standing? That may be fine of course, but I thought I ought to raise it...

[randy3k]

I believe only the last overload counts (untested). I guess there is no easy fix for this. Since there should be one and only one output, it should not be very bad to use the last known type.

[richardreeve]

Yes, that's what I thought. I guess it's okay - it would be a strange situation using two different Julia types in a single piece of code that match the same R type (and wanting to do conversions!).