Re: Featurewish: Context aware operator overloading

On Tue, May 25, 2010 at 04:44:32PM +0200, Rafael Garcia-Suarez wrote:
> On 25 May 2010 15:36, Steffen Schwigon <[email protected]> wrote:
> > Hi!
> >
> > Forgive me this passive way of just wishing a feature. I can't provide
> > more than a user's perspective. Anyway:
> >
> > Would it make sense to make operator overloading context aware?
> >
> > Currently an overloaded operator only runs in scalar
> > context. Therefore I do nasty dereferencing like
> >
> >  my @list = @{ $data ~~ dpath '//affe/zomtec' };
> >
> > just because the overloaded '~~' cannot return an array.
> >
> > What I would want is
> >
> >  my @list = $data ~~ dpath '//affe/zomtec';
> >  my $head = $data ~~ dpath '//affe/zomtec';
> >
> > with the sub behind the overloaded '~~' doing the usual wantarray
> > drill.
> >
> > I have, however, no more knowledge about this, except that someone
> > once said the original author of overloading hadn't found it
> > neccessary, IIRC.
> >
> > And yes, I know that this would probably introduce trouble on
> > backwards compat. Maybe it can be protected by some “use feature”
> > magic.
> 
> A feature pragma wouldn't be needed if we pass the context as a new
> parameter to the overloading method, with the default (undef)
> indicating scalar context.


That would break code, wouldn't it? 

If

    my @foo = X ~~ Y

now calls the overload sub in scalar context, changing it to run it in
list context will cause code breakage, regardless how you pass the
information around.

Or am I misunderstanding things?


Not that I would care too much about this breakage - I'd call it a bug fix ;-)


Abigail