in reply to Re: Scope, Reference Counting, Mortality... a question about Object Caching
in thread Scope, Reference Counting, Mortality... a question about Object Caching

Here's how the SVs are created: (slimmed down version)

SV * newSVObj(void *object, char * classname) { SV * result; if (!object) return newSVsv(&PL_sv_undef); HV * h = newHV(); hv_store(h, "ETK", 3, newSViv((long)object), 0); result = newRV((SV*)h); sv_bless(result, gv_stashpv(classname, FALSE)); SvREFCNT_dec(h); return result; }

As you can see, the SV is a hashref storing only a pointer to the original object, and through the class I keep track of the type of the object, so the typemap calls functions like:

SV * newSVEtkButtonPtr(Etk_Button *o) { return newSVObj(o, "Etk::Button"); }

And when these SVs are passed around as data, the get wrapped into a new SV that is made mortal, for example, the code passing data to the callback function: (cbd being a certain struct that hold relevant information to the callback)

PUSHMARK(SP) ; XPUSHs(sv_2mortal(newSVsv(cbd->perl_object))); // self XPUSHs(sv_2mortal(newSVsv(cbd->perl_data))); // data PUTBACK ; call_sv(cbd->perl_callback, G_DISCARD);
--
Leviathan.