(\ my %all_keys)->@{keys %A, keys %Z} = ();
[download]

'It seems to work without the "\".'

That seemed odd to me so I checked it. I modified the output slightly for ease of comparison.

With \:

$ perl -Mstrict -Mwarnings -E '
    my %A = (a => 1, b => 2, D => 3);
    my %Z = (z => 9, y => 8, D => 7);
    (\ my %all_keys)->@{keys %A, keys %Z} = ();
    say for "@{[sort keys %all_keys]}";
'
D a b y z
[download]

Without \:

$ perl -Mstrict -Mwarnings -E '
    my %A = (a => 1, b => 2, D => 3);
    my %Z = (z => 9, y => 8, D => 7);
    (my %all_keys)->@{keys %A, keys %Z} = ();
    say for "@{[sort keys %all_keys]}";
'
D a b y z
[download]

So I can confirm that behaviour. I then ran both of those through B::Deparse.

With \:

$ perl -Mstrict -Mwarnings -MO=Deparse -E '
    my %A = (a => 1, b => 2, D => 3);
    my %Z = (z => 9, y => 8, D => 7);
    (\ my %all_keys)->@{keys %A, keys %Z} = ();
    say for "@{[sort keys %all_keys]}";
'
use warnings;
use strict;
use feature 'current_sub', 'bitwise', 'evalbytes', 'fc', 'isa', 'modul
+e_true', 'postderef_qq', 'say', 'signatures', 'state', 'unicode_strin
+gs', 'unicode_eval';
my(%A) = ('a', 1, 'b', 2, 'D', 3);
my(%Z) = ('z', 9, 'y', 8, 'D', 7);
@(\my %all_keys){keys %A, keys %Z} = ();
say $_ foreach (join $", @{[sort(keys %all_keys)];});
-e syntax OK
[download]

No huge surprises there.

Without \:

$ perl -Mstrict -Mwarnings -MO=Deparse -E '
    my %A = (a => 1, b => 2, D => 3);
    my %Z = (z => 9, y => 8, D => 7);
    (my %all_keys)->@{keys %A, keys %Z} = ();
    say for "@{[sort keys %all_keys]}";
'
use warnings;
use strict;
use feature 'current_sub', 'bitwise', 'evalbytes', 'fc', 'isa', 'modul
+e_true', 'postderef_qq', 'say', 'signatures', 'state', 'unicode_strin
+gs', 'unicode_eval';
my(%A) = ('a', 1, 'b', 2, 'D', 3);
my(%Z) = ('z', 9, 'y', 8, 'D', 7);
@all_keys{keys %A, keys %Z} = ();
say $_ foreach (join $", @{[sort(keys %all_keys)];});
-e syntax OK
[download]

The lexical variable my %all_keys appears to have been changed in favour of the package variable %all_keys (no my). A search through perlref provided no answers; although, it's not impossible that I missed something. I don't have time to investigate further; perhaps another monk can shed some light on this.

All of the above used:

$ perl -v | head -2 | tail -1
This is perl 5, version 39, subversion 3 (v5.39.3) built for cygwin-th
+read-multi
[download]

I repeated everything with the earliest Perl version I have available:

$ perl -v | head -2 | tail -1
This is perl 5, version 30, subversion 0 (v5.30.0) built for cygwin-th
+read-multi
[download]

The results were the same except the "use feature" list was slightly different: signatures out; switch in.

use feature 'current_sub', 'bitwise', 'evalbytes', 'fc', 'postderef_qq
+', 'say', 'state', 'switch', 'unicode_strings', 'unicode_eval';
[download]

I don't believe that difference has any relevance to what's being tested.

— Ken

This works also.

my %all_keys;
%all_keys->@{keys %A, keys %Z} = ();
[download]

deparser says

my %all_keys;
@all_keys{keys %A, keys %Z} = ();
[download]

So it doesn't seem to be requiring a hash reference in that context.

\ returns the reference

Postfix Reference Slicing is only defined for refs, no idea why it works without.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
see Wikisyntax for the Monastery}

The OP wasn't really a SSCCE tho. Duplicated keys would have shown the problem. :)

Personally I would go for a clear 2 line approach, all this dereferencing doesn't make it more readable.

Or to repeat my old mantra:

• Perl would be better of with autobox-methods¹ {%A, %Z}->keys

Hmm...People will misunderstand this.

What I mean is that many things are elegantly solved if you don't need 'my'. A language design problem.

¹) compare autobox

LOL, I cannot condone that foolery.