Unfortunately, without the wrapper (push $future_arrayref, @values) that will fail with an error. Shouldn't that autovivify $future_arrayref?

I don't think push should autovivify because that would be a surprise (no built-ins autovivify) and it would hide delay discovering errors (pushing onto the wrong variable).

I feel the current documentation for push is wrong, as push doesn't really dereference anything. The new experimental feature merely saves you from manually having to dereferencing to satisfy prototypes -- push always took an array ref as an argument but required you to type  push @array instead of  push \@array and now it drops the requirement so you don't have to write  push @{$array}

Only dereferencing in lvalue context autovivifies (more on this in Re: undefined value as an ARRAY reference sometimes ok ) and plenty of newbies complain about this already :)( see When DOESN'T "Use of uninitialized value" show up? )

So why is this new experimental feature more than mere keystroke saving sugar? Because it doesn't autovivify. Consider this

#!/usr/bin/perl --
use strict; use warnings;
use Data::Dump;

my @foo;
my $bar;
silentWin( \@foo );
silentFail( $bar );
dd \@foo;
dd $bar;

aliasUglyWin( $bar );
dd $bar;

print "bye\n";
silentWin( undef );
print "bye\n";


sub silentWin {
   my( $arrayref ) = @_;
   push $arrayref, "win";
}
sub silentFail {
   my( $arrayref ) = @_;
   push @{$arrayref}, "fail";
}
sub aliasUglyWin {
    push @{ $_[0] } , "ugly win";
}
__END__
["win"]
undef
["ugly win"]
bye
Not an ARRAY reference ...
[download]

aliasUglyWin() works but its ugly :)  $_[0] is aliased to $bar, so the correct variable ($bar) is autovivified

silentFail() fails because when you copy  $_[0] to $arrayref, $arrayref is not aliased to $bar , so the wrong variable is autovivified ( not $bar)

silentWin() saves you three keystrokes @{} and it doesn't autovivify the wrong variable but dies instead . It also changes the diagnostics message from Type of arg 1 to push must be array ... to Not an ARRAY reference

So I don't think push should autovivify, I feel it defeats what little advantage this new experimental feature provides to the uninitiated