Hi nikmit,

I expected the below code to create a fresh copy of @arr every time gimme() is executed.

In the code you showed, @arr is only initialized once, so the anonymous hashes are only created once. The first and only element of @arr is a reference to an anonymous hash. When you say my @result = @arr;, the elements of @arr are copied into @result, so that reference is copied exactly, so it still points to the same anonymous hash. This is known as a "shallow copy" and is how arrays and hashes are copied in Perl. (Note: Making @arr Readonly won't change that.)

To get a "deep copy" or "clone", choroba already pointed you to Storable and Clone, but there's another solution:

sub gimme {
    my @result = ( {
        box => { attr => 'foo' },
        } );
    return \@result;
}
[download]

This causes the array and the anonymous hashes to be re-created on every call to gimme.

Hope this helps,
-- Hauke D

Update: Minor clarifications to wording.

G'day nikmit,

I probably would have written that more like this:

#!/usr/bin/env perl -l

use strict;
use warnings;

use Data::Dump;

for (1..3) {
    my $ref1 = gimme();
    $ref1->[0]{box}{y}{z} = $_;
    $ref1->[0]{box}{w} = 'x' if $_ == 2;
    $ref1->[0]{box}{a}{d} = 4 if $_ == 3;
    trace($ref1);

    my $ref2 = gimme();
    trace($ref2);
}

sub gimme {
    [
        {   
            box => {
                a => { b => 2, c => 3 },
                y => {}
            }   
        }   
    ]
}

sub trace {
    my ($ref) = @_;

    print 'Tracing: ';
    print join ' -> ',
        $ref, $ref->[0], $ref->[0]{box}, $ref->[0]{box}{y};
    dd $ref;

    return;
}
[download]

Notes:

• The BEGIN block, as well as @arr, have gone.
• &gimme uses no variables: it just returns an anonymous data structure that's created afresh every time it's called.
• There are no multiple instances of \@result hanging around. That equates to no instances of @results in existence whose reference counts are greater than zero: they can't be garbage collected until their reference count reaches zero.
• I threw in some additional modifications on the 2nd and 3rd iterations: no changes to $ref1 were propagated to $ref2.
• Note that you don't have to use the '->' between every element of the data structure: just use it once, e.g. $ref1->[0]{box}{a}{d}.

Here's the output from a sample run:

Tracing: 
ARRAY(0x7fdc5307b7c8) -> HASH(0x7fdc5282e280) -> HASH(0x7fdc5282de48) 
+-> HASH(0x7fdc5282dd88)
[{ box => { a => { b => 2, c => 3 }, y => { z => 1 } } }]

Tracing: 
ARRAY(0x7fdc530b1f18) -> HASH(0x7fdc530b4bc0) -> HASH(0x7fdc530251a0) 
+-> HASH(0x7fdc53025170)
[{ box => { a => { b => 2, c => 3 }, y => {} } }]

Tracing: 
ARRAY(0x7fdc530b4bc0) -> HASH(0x7fdc52846220) -> HASH(0x7fdc5282def0) 
+-> HASH(0x7fdc5282dd70)
[
  { box => { a => { b => 2, c => 3 }, w => "x", y => { z => 2 } } },
]

Tracing: 
ARRAY(0x7fdc530b8a48) -> HASH(0x7fdc53025188) -> HASH(0x7fdc530b1ee8) 
+-> HASH(0x7fdc530b7968)
[{ box => { a => { b => 2, c => 3 }, y => {} } }]

Tracing: 
ARRAY(0x7fdc530b8a48) -> HASH(0x7fdc5282dda0) -> HASH(0x7fdc53094b50) 
+-> HASH(0x7fdc5282de48)
[
  { box => { a => { b => 2, c => 3, d => 4 }, y => { z => 3 } } },
]

Tracing: 
ARRAY(0x7fdc530b1f48) -> HASH(0x7fdc530b7a88) -> HASH(0x7fdc530aabb0) 
+-> HASH(0x7fdc530b1e58)
[{ box => { a => { b => 2, c => 3 }, y => {} } }]
[download]