• All package variables (well, technically, package variables are the only variables with a "namespace").
• All lexical variables that share the same scope as your subroutine.

Abigail

That is, lexical variables that don't share a scope are unreachable.

Not strictly true since you can modify lexical variables with the help of PadWalker. Here's some code from the docs

sub increment_my_x {
    my $h = peek_my (1);
    ${$h->{'$x'}}++;
}

my $x=5;
increment_my_x;
print $x;  # prints 6
[download]

Although you'd have to be quite a disturbed individual to use this sort of thing in production any code ;-)
HTH

_________
broquaint

I could see one or two valid uses. The one place I've used it for production was a tied hash package. We wanted to be able to alias package vars in the hash values with vars from the caller. Retrofitting this to older code was a pain as they HAD to be package variables so we changed the alias sub to look at caller's lexicals and then the package vars.

-Lee

"To be civilized is to deny one's nature."

---

Update
While on the subject, I should mention a bug/feature of PadWalker. It will NOT see all lexicals visable where the subroutine is called if you have a bare block

#!/usr/bin/perl -w
use strict;

use PadWalker qw(peek_my);


my ($oa,$ob) = ("Outer One", "Outer Two");

{
    my $inb = "Bare Block"; # Will not be seen by peek_my
    peeper();

}

sub peeper {
   my $c = peek_my(1);
   print  "Lexicals:\n", (map { "Found $_ with value $c->{$_}\n" } sor
+t keys %$c ),"\n";

}
[download]