Erm, no. I'm just talking about making the method called by the normal overload call examine the instance in question to decide what to do. I don't see how that would change anything about the automagically generated boolification or numification logic.

#!/usr/bin/perl
use strict;
use warnings;

package Ovl;

use overload '""' => \&mystring;

sub new {
  my $ret = bless {}, shift();
  
  $ret->{_name}     = shift;
  $ret->{_str_code} = \&_str_default;
  
  return $ret;
}
      
sub _str_default { my $self = shift; "!" . $self->{_name} . "!" }

sub mystring {
  my $self = shift;
  
  my $str_code = $self->{_str_code} || \&_str_default;
  
  $self->$str_code( @_ );
}

package main;

my $o = Ovl->new( "default" );

print "$o\n";

my $o_undef = Ovl->new( undef );
$o_undef->{_str_code} = sub { "I R UNDEFINED" };

print "$o_undef\n" if $o_undef;

my $o_numeric = Ovl->new( 15 );
$o_numeric->{_str_code} = sub { sprintf( "%0.5f", shift->{_name} ) };

print "$o_numeric\n" if int( $o_numeric ) == 15;

exit 0;

__END__
[download]

Update: AAAAAHHH. Just reread the OP's code and I see what you're getting at. In order to implement arbitrary overload operations on instances (i.e. what operations are overloaded on the instance, not just what each operation does on each instance) you would need to have defined handlers for all possible operations for the class itself and then yes you would need duplicate dispatching logic. I was thinking of the case where each instance shared the same set of overloaded operations, whereas his allows instance level selection of what's overloaded. I just glanced over his code and missed that he was trying to get different sets of overload behavior per instance. Bah, it's quitting time on Friday.