Filtering Your Own Output (Perl Cookbook, 2nd Edition)

16.5.1. Problem

You want to postprocess your program's output without writing a separate program to do so.

16.5.2. Solution

Use the forking form of open to attach a filter to yourself. For example, this will restrict your program to a hundred lines of output:

head(100);
while (<>) {
    print;
} 

sub head {
    my $lines = shift || 20;
    return if $pid = open STDOUT, "|-";
    die "cannot fork: $!" unless defined $pid;
    while (<STDIN>) {
        print;
        last unless --$lines;
    } 
    exit;
}

16.5.3. Discussion

It's easy to add an output filter. Just use the forking open on your own STDOUT, and let the child filter STDIN to STDOUT, performing whatever alterations you care about. Notice that we install the output filter before we generate the output. This makes sense—you can't filter your output if it has already left your program. Any such filters should be applied in LIFO order—the last one inserted is the first one run.

Here's an example that uses two output filters. One numbers lines; the other quotes the lines like a mail reply. When run on /etc/motd, you get something like:

1:  > Welcome to Linux, version 2.0.33 on a i686
2:  >
3:  >    "The software required `Windows 95 or better',
4:  >     so I installed Linux."

If you reversed the order of the two filters, you'd get:

> 1: Welcome to Linux, Kernel version 2.0.33 on a i686
> 2: 
> 3:     "The software required `Windows 95 or better',
> 4:      so I installed Linux."

The program is in Example 16-1.

Example 16-1. qnumcat

  #!/usr/bin/perl
  # qnumcat - demo additive output filters
  
  number( );                   # push number filter on STDOUT
  quote( );                    # push quote filter on STDOUT
  
  while (<>) {                # act like /bin/cat
      print;
  } 
  close STDOUT;               # tell kids we're done--politely
  exit;
  
  sub number {
      my $pid;
      return if $pid = open STDOUT, "|-";
      die "cannot fork: $!" unless defined $pid;
      while (<STDIN>) { printf "%d: %s", $., $_ } 
      exit;
  } 
  
  sub quote {
      my $pid;
      return if $pid = open STDOUT, "|-";
      die "cannot fork: $!" unless defined $pid;
      while (<STDIN>) { print "> $_" } 
      exit;
  }

As with all process forks, doing this a zillion times has some cost, but it's fine for a couple of processes, or even a couple dozen. If the system was actually designed to be multitasking right from the start, as Unix was, this is far cheaper than you might imagine. Virtual memory and copy-on-write makes this efficient. Forking is an elegant and inexpensive solution to many, if not most, multitasking needs.

16.5.4. See Also

The open function in Chapter 29 of Programming Perl and in perlfunc(1); Recipe 16.4


16.4. Reading or Writing to Another Program		16.6. Preprocessing Input

16.5. Filtering Your Own Output

16.5.1. Problem

16.5.2. Solution

16.5.3. Discussion

Example 16-1. qnumcat

16.5.4. See Also