Alternatively, if you don't like the ref solutions, you can use one of perl's Obscure Open Tricks

   sub your_function {
     my ( $string, $in_fh, $out ) = @_;
     local *OUTFH;

     # opens $out if it's a filename
     # dupes $out if it's a filehandle
     open OUTFH, ">$out" or die "$out: $!\n";

     # do whatever with $in_fh and $string
     while( <$in_fh> ) {
       /$string/o and print OUTFH "$string: $_"
     }

     # close OUTFH: either we opened it or we duped it
     # Either way we don't need it any more.
     close OUTFH;
   }
[download]

Then your_function() just relies on its caller to do the right thing. If you want to pass your_function() a filename to take as output, then you'd call it as you'd expect:

    your_function( $string, \*INPUT, '/tmp/somefile' );
[download]

On the other hand, if you want to pass your_function() an already opened filehandle, and use that for output, then you'd call it as:

    # print to STDOUT instead of to disk
    your_function( $string, \*INPUT, "&STDOUT" );

    # print to already-opened filehandle
    open OUTPUT, '> /tmp/somefile'
        or die "/tmp/somefile: $!\n";
    your_function( $string, \*INPUT, "&OUTPUT" );
[download]

For more info on this, check out Obscure Open Tricks in perlopentut. There's a recommendation in there that you qualify the package name where OUTPUT is located, which I've ommitted in the code above.

One thing worth noting is that the line

     open OUTFH, ">$out" or die "$out: $!\n";
[download]

Must appear as written (i.e. with no spaces between the > and $out). Avoid these following variations, since they don't do the same thing as the line above

     # wrong: given, eg, &STDOUT, creates a file named '&STDOUT'
     open OUTFH, "> $out" or die "$out: $!\n";

     # wrong: also creates, eg, '&STDOUT'
     open OUTFH, '>', $out or die "$out: $!\n";
[download]