7.9.1. Problem

7.9.2. Solution

open(FH, "<&=", $FDNUM)       # open FH to the descriptor itself
open(FH, "<&",  $FDNUM);      # open FH to a copy of the descriptor

use IO::Handle;
$fh = IO::Handle->new_from_fd($FDNUM, "r");

7.9.3. Discussion

Occasionally you have a file descriptor but no filehandle. Perl's I/O system uses filehandles instead of file descriptors, so you have to make a new filehandle from an already open file descriptor. The "<&", ">&", and "+<&" access modes to open do this for reading, writing, and updating, respectively. Adding an equals sign to these—making them "<&=", ">&=", and "+<&="—is more parsimonious of file descriptors and nearly always what you want. That's because the underlying implementation of Perl's open statement uses only a C-level fdopen(3) function from the C library, not a dup2(2) syscall that calls the kernel.

The new_from_fd IO::Handle object method is equivalent to:

use IO::Handle;

$fh = new IO::Handle;
$fh->fdopen($FDNUM, "r");            # open fd 3 for reading

$fd = $ENV{MHCONTEXTFD};
open(MHCONTEXT, "<&=", $fd)   or die "couldn't fdopen $fd: $!";
# after processing
close(MHCONTEXT)              or die "couldn't close context file: $!";

use POSIX;

POSIX::close(3);              # close fd 3