... and probably XS defined subroutines and AUTOLOADed methods and so and so forth. If I had listed all the caveats it would have taken up more space than the code ;-) This was posted not so much as a solution to the OPs question, rather as a riposte to the bald statement that it couldn't be done.

/J\

It may not be big but it certainly is clever!

To those kind people who pointed out that:
(a)I could execute the sub without knowing its name - I knew that
(b)That it might not have a name - in my case I knew it had a name but not what the name was.

My problem was that the coderef was being passed in from somewhere else but I needed to get at the subroutine name, not to execute it but to print for debug, logging, etc.

As usual, perlmonks has excelled itself for speed and quality of response ...

Thanks

I can tell you're already happy with the solution you've got, but I'd like to recommend a few alternates anyway. :-)

I think you're losing a lot of flexibility by requiring the passed in coderef to be a named sub in the same namespace. Any time you're passing around coderefs, anonymous subs become more and more useful. To this end, I can think of a few ways to allow them:

• Pass in a label along with the coderef. This would allow the caller to use anonymous coderefs as well as refs to named subs, and would avoid the "in another namespace" problem.
• Pass in a label, but only optionally. If there is no label, the symbol table is searched. If there is one, it's used instead.
• Pass in an object. At first this sounds a bit ugly, but it could be something as simple as:

      foobar(Code::Wrapper->new( code => \&foo ));
      foobar(Code::Wrapper->new( code => sub { "baz" }, label => "baz" )
  +)
  [download]

  Then this object could encapsulate the behavior mentioned earlier, by either searching the symbol table or using the specified label.

Which one you choose depends on how much flexibility you want in the future, and how much you want to insulate certain parts of your code from others. I have a feeling you will just go with the symbol table scan, but hopefully you'll at least give my ideas some consideration.

To make that even simpler, you could write a subroutine called Sub that takes a coderef argument and records its file and line number and returns an object that when used as a coderef invokes the coderef, but when used as a string returns the filename/linenumber of creation. How's that for slick? I may even have to code that up for a snippet. Usage would be something like:

my $sub = Sub { my $x = shift; print $x };
$sub->("foo"); # prints foo
print "that was from $sub\n";
[download]

Pretty simple to do.

-- Randal L. Schwartz, Perl hacker
Be sure to read my standard disclaimer if this is a reply.

---

update: see Track the filename/line number of an anonymous coderef.

Would caller help? It's used to examine the call stack. For example, the Carp module (included with perl) uses it to warn/die with stack traces or from the perspecive of the caller.