Front page | perl.perl5.porters |
Postings from May 2003
John Peacock <jpeacock@rowman.com> writes: >Dan Sugalski wrote: >> What I'm in favor of is getting overloading and tying working together. >> I still think the two should remain separate, with assignment living in >> the tie part. I don't have the tuits to throw at it, unfortunately. > >I'm not sure how that would work. Would you tie the blessed object or bless the >tie? No ;-) - you tie the variable and bless the value. What it means in practice (probably) is SV gets both kinds of magic. >The OO interface (overloading) has to be dominant, so that the correct >method gets called with the correct object[s]. They are both OO interfaces. > >FETCH could naturally be mapped to numify()/stringify() based on context, But there is no context to a FETCH. >I >suppose, but the overloading already does that now. STORE is the only piece >missing from the possible overload methods. No - they should remain distinct - they might be different packages: tie $foo,'SomeThing' $foo = OverLoaded->new; What has to happen in the general case is that you need call mg_get() for the tie which will call FETCH and _then_ do the overload thing on the SV's magic at that point. Finally do mg_set() to do STORE. IIRC it nearly works. The FETCH happens - but the overloaded-ness of the FETCHed value is not honoured. > >Would you be more favorably inclined towards an overload::Tied module which >added a Tie-like interface to overloaded objects? Not really. > >John -- Nick Ing-Simmons http://www.ni-s.u-net.com/Thread Previous | Thread Next