I agree with simply operating on the refs in-place. On the other hand, functional style is is good, and if you don't want to modify the input hashes, you can still have an elegant call/return by using refs throughout the code:

my $x = {"a"  => "red"};
my $y = {"b"  => "green"};
my $z = {"c"  => "black"};

my ($new_x, $new_y, $new_z) = &modfifyHash($x, $y, $z);

# original $x is unchanged, $new_x has new value
print "\$x->{a} = $x->{a}\t\$new_x->{a} = $new_x->{a}\n";

sub modfifyHash {
    my ($x, $y, $z) = @_;
    # If we don't want to modify the originals,
    # then make a shallow copy for each
    $_ = { %$_ } for $x,$y,$z;
    
    $x->{ "a" } = "circle";
    $y->{ "b" } = "square";
    $z->{ "c" } = "rectangle";
    return($x, $y, $z)
}
[download]