Consider this:

use strict;
use Scalar::Util qw(weaken isweak);

my $a = 1;

my $r1 = *main::a{SCALAR};
my $r2 = *main::a{SCALAR};

weaken $r1;

print "r1 is weak\n" if isweak $r1;
print "r2 is weak\n" if isweak $r2;
[download]

Here I get two copies of the typeglob. Both are references to $a but they are different one as proved by weakening one and observing that the other one has not been weakened.

So I guess you don't need the special snytax for array-refs, hash-refs etc, you just generically get a new copy of whatever is in the corresponding slot of the typeglob.

That's a very good idea, and one I hadn't thought of.

But it's just half the problem. The solution requires that I know the typeglob slot. All I have is a reference and I don't know a general way of going from the reference to the typeglob slot.

In fact, I can't see from the documentation any situation in which a format reference can be used for any purpose whatsover. (Unless you count taking its address and using that as a very low quality pseudo-random number generator as a "use".)

All I have is a reference and I don't know a general way of going from the reference to the typeglob slot.

What? I thought that the typeglob slots were chosen to match what ref returns (except you'd want Scalar::Util::reftype because of the unfortunate decision to overload ref to returning class names and you have to deal with just a few special cases like the unfortunate and silly decision to have ref(\\$x) return "REF" instead of "SCALAR" and the more reasonable features of returning "LVALUE" or "VSTRING" instead of "SCALAR").

I don't see how having to deal with 3 special cases prevents this from being a "general" solution.

Updated with minor wording changes.

- tye        

I think it is even simpler:

Consider:

use strict;

use Scalar::Util qw(weaken isweak);
my $a = [];

my $r1 = $a;
my $r2 = $a;

weaken $r1;

print "a is weak\n" if isweak $a;
print "r1 is weak\n" if isweak $r1;
print "r2 is weak\n" if isweak $r2;
[download]

So here we have a array_ref $a of which we make two further copies for which we prove again that they are different.

So to answer your initial question:

Given a reference you simple assign it to a scalar to get another reference to the same referent - no need for special syntax or type-glob fiddling...