Hey,
I was messing around with typemaps and I came up with the following
code, which seems to work with Inline 0.43:
-----------------------------------------------
TYPEMAP
NN * WRAPPED_C_OBJECT
INPUT
WRAPPED_C_OBJECT
$var = (($type)SvIV(SvRV($arg)))
OUTPUT
WRAPPED_C_OBJECT
/* Use a block so I can declare variables here */
{
SV *container, *obj;
container = newSViv(0);
obj = newSVrv(container, class);
sv_setiv(obj, (IV)$var);
SvREADONLY_on(obj);
$arg = container;
}
-----------------------------------------------
This code is taken from the Inline::C-Cookbook, with one change -
there's a 'class' variable here that's completely undeclared. The only
reason it works is that (so far) each function that returns a NN* is a
constructor function, called like "AI::NodeLib"->new(), and I have
declared that function in C with something like this:
NN *new (char *class) {
...
}
So if you consider the wrapper function Inline will make
(XS_AI__NodeLib_new), you'll note that the typemap code will have access
to the 'class' variable.
I guess my question is, can I count on this behavior, or is there a
better way to accomplish what I'm trying to do? This way results in
some pretty slick Inlined code with a minimum of XS junk, but I feel
like I'm abusing the concept of the typemap a little bit - it's not just
performing a translation of one thing to another, instead it's using
some external data that just happens to be there.
-Ken
