Replace `SEXPREC` with something truly opaque

[CGMossa]

Replacement for #205, with a pattern from Rust for Rustacean book, see https://rust-for-rustaceans.com/#errata

This is a far more complete approach

• No longer deriving Copy or Clone for something we cannot construct, so how would we copy/clone it?
• Using c_void to indicate that we don't know anything at all about it, but still have a typed SEXP.
• Even if we have access to SEXPREC, in the defining module, we cannot construct it in this definition, while the zero-sized arrays could potentially be constructed.

@yutannihilation What do you think?

Depends on

• [ ] #208

[yutannihilation]

My comment stays the same as https://github.com/extendr/libR-sys/pull/205#issuecomment-1817496032; I believe there's no need to expose SEXPREC at all, but I'm approving.

[CGMossa]

I see. To be honest with you, I felt very proud of this PR, because I thought I basically did exactly what you asked.

I can now see the subtle difference in what you said and what I did here.

I honestly like this. I followed what was presented in the issue I created over at bindgen repo.

[yutannihilation]

I get you. But, my understanding is that, the comments on https://github.com/rust-lang/rust-bindgen/issues/2685 is probably under the assumption that other APIs use SEXPREC* in the signatures because it's a common case. Actually, this example in Rustonomicon is the same:

struct Foo; /* Foo is a structure, but its contents are not part of the public interface */
struct Bar;
void foo(struct Foo *arg);
void bar(struct Bar *arg);

In this case, the Rust bindings should provide Foo and Bar. Apparently, the things are different in the R API. All the signatures use a type alias SEXP instead of SEXPREC*. So, I don't think we need SEXPREC in theory. But, I might be wrong, and I don't have a strong preference on this.

[Lokathor]

The winapi crate is an example of a major crate offering the pointer alias as well as the pointed-to types as well.

Generally, I would suggest defining every single part as close as possible to the C version, including making SEXP just be a type alias of a pointer.

[yutannihilation]

Just curious. Why do you think it's better? I agree "defining every single part as close as possible to the C version" sounds good as a general rule, but I don't see any actual benefits in this case.

[Lokathor]

Users needing example code, and/or needing to talk to others about the lib (usage, bug reports, etc), will generally need to be able to talk about it in terms of C code. Then they'll have to convert that discussion about C mentally back into Rust code before they can actually write Rust. Having the most minimal difference possible, ideally none at all, reduces the mental load of the whole cycle.

[yutannihilation]

I see. Then, I personally think this crate is not the case as I haven't seen anyone talk about SEXPREC whether it's C code or Rust code. But, it might be just my preference. Thanks for the explanation.