1. Is it possible to inject new code into a subroutine that will execute within the lexical scope of the subroutine body? That is, can I take

   sub {
       BLAH
   }
   [download]

   and convert it (programmatically, not by grovelling around in the original definition) into

   sub  {
       BLAH
       YOUR CODE HERE
   }
   [download]

   Specifically, what I have in mind is taking a subroutine that modifies @_—possibly without modifying its original arguments (say, by splicing entries into the middle)—and applying a user-specified function to the resulting mutated array. (A natural thought is some wrapper like

   sub inject {
       my ( $outer, $inner ) = @_;
       return sub {
           $inner->(@_);
           goto &$outer;
       }
   }
   [download]

   but that doesn't work, since $outer is executing in a different lexical scope.)

   I know that B::Deparse will handle the simple case of functions that don't close over external variables, and Data::Dump::Streamer will handle more cases, but I'm looking for something really robust.

2. Is there a subroutine uneval such that eval uneval($a) always returns $a? Even if it's possible to do this just when $a is a simple scalar (i.e., not a reference), then I'm happy—but I'd like to preserve any ‘magic’ associated to $a (for example, not lose its numeric value by just stringifying). If one could control the lexical scope in which eval'd strings executed, then this would be easy:

   sub uneval {
       my ( $a ) = @_;
       return ( in_current_scope => '$a' );
   }
   [download]

   but I'm not sure how to do it otherwise.

UPDATE: I forgot to mention that I wanted to do the unevaling with lexicals, not globals. It's easy with globals:

our @cache;
my $id = 0;
sub uneval {
    $cache[$id++] = $_[0];
    return __PACKAGE__ . '::$cache[' . $id . ']';
}
[download]