Consider changing this Altrep-interface

[CGMossa]

Currently, the bindings generate

pub type R_altrep_Coerce_method_t =
    ::std::option::Option<unsafe extern "C" fn(arg1: SEXP, arg2: ::std::os::raw::c_int) -> SEXP>;

in particular, this c_int. There is little documentation on it, but this function prototype is satisfied by a function
that looks like this

static SEXP altrep_Coerce_default(SEXP x, int type) { return NULL; }

Thus the second argument is supposed to be a type, presumably SEXPTYPE, which we have now made into an enum.

[CGMossa]

There are no meaningful tests in extendr for R_altrep_Coerce_method_t. And I can't find even a coerce method available to use with extendr made altrep objects. We should device a test and see if this abstraction causes an issue.

[yutannihilation]

Thus the second argument is supposed to be a type, presumably SEXPTYPE, which we have now made into an enum.

I think you are correct. It seems the Coerce method is what is called here.

https://github.com/wch/r-source/blob/ab9c4114cd774d359f6d4c3b4a69e7ba9d2864c8/src/main/coerce.c#L1182-L1191

SEXP coerceVector(SEXP v, SEXPTYPE type)
{
    if (TYPEOF(v) == type)
	return v;


    SEXP ans = R_NilValue;	/* -Wall */
    if (ALTREP(v)) {
	PROTECT(v); /* the methods should protect, but ... */
	            /* also "v" is protected by caller */
	ans = ALTREP_COERCE(v, type);

[CGMossa]

Now I get it. So for extendr, one has to call the altrep as an Robj to invoke it. I needed this information, many times thanks!