At work I have a collection of long-running Perl programs exec-ed by a rc script that have their noise redirected to a single intermingled log. The scripts themselves and the modules they use are littered with carps, warns and print STDERRs depending upon when and by whom they were written. I can differentiate their output (more or less) but I’d like it timestamped for forensic purposes.

My first thought, heaven help me, involving tying filehandles is exampled by this hack:

!/usr/local/bin/perl -w 

use Tie::Handle;
use Carp;
use strict; 

package TimeStamper; 
@ISA = qw(Tie::Handle); 

sub wrap { 
    my $fh = shift; 
    my $fh_name = $fh; 
    $fh_name =~ s/^\*//;
    local *MYFH; 
    open(MYFH, ">&$fh_name") or die("Failed to dupe [$fh_name]: $!"); 
    tie($fh,'TimeStamper',*MYFH); 
} 

sub TIEHANDLE { 
     my ($class,$fh) = @_; 
     my $obj = bless [$fh],$class; 
     return $obj; 
} 

sub PRINT { 
    my $self = shift; 
    my $fh   = $self->[0]; 
    my @t = localtime(); 
    print sprintf("[%4d/%02d/%02d %02d:%02d:%02d] ", 
                  $t[5]+1900,@t[4,3,2,1,0]),@_; 
} 

package main; 

TimeStamper::wrap(*STDERR);
print STDERR "This is timestamped.\n";
warn('This is timestamped.'); 
carp('This is timestamped.');
print "This is not timestamped.\n";

Neat. Even half clever. But I’m not sure it’s a good idea.