Object-Oriented Perl

I've recently started learning to play the game of Go. Go and Perl have many things in common -- the basic stuff of which they are made, the rules of the game, are relatively simple, and hide an amazing complexity of possibilities beneath the surface. But I think the most interesting thing I've found that Go and Perl have in common is that there are various different stages in your development as you learn either one. It's almost as if there are several different plateaus of experience, and you have to climb up a huge hill before getting onto the next plateau.

For instance, a Go player can play very simply and acquit himself quite decently, but to stop being a beginner and really get into the game, he has to learn how to attack and defend economically. Then, to move on to the next stage, he has to master fighting a repetitive sequence called a "ko." As I progress, I expect there to be other difficult strategies I need to master before I can become a better player.

Perl, too, is not without its plateaus of knowledge, and in my experience, the one that really separates the beginner from the intermediate programmer is an understanding of object-oriented (OO) programming. Once you've understood how to use OO Perl, the door is opened to a huge range of interesting and useful CPAN modules, new programming techniques, and mastery of the upper plateaus of Perl programming.

So what is it?

Object-oriented programming is one of those buzzwordy manager-speak phrases, but unlike most of them, it actually means something. Let's take a look at some perfectly ordinary procedural Perl code, bread and butter programming to most beginning programmers:

my $request = accept_request($client);
my $answer = process_request($request);
answer_request($client, $answer);
$new_request = redirect_request($client, $request, $new_url);

The example here is of something like a Web server: we receive a request from a client, process it in some way to obtain an answer, and send the answer to the client. Additionally, we can also redirect the request to a different URL.

The same code, written in an object-oriented style, would look a little different:

my $request = $client->accept();
$new_request = $request->redirect($new_url);

What's going on here? What are these funny arrows? The thing to remember about object-oriented programming is that we're no longer passing the data around to subroutines, to have subroutines do things for us -- now, we're telling the data to do things for itself. You can think of the arrows, (->, formally the "method call operator") as instructions to the data. In the first line, we're telling the data that represents the client to accept a request and pass us something back.

What is this "data that represents the client," and what does it pass back? Well, if this is object-oriented programming, we can probably guess the answer: they're both objects. They look like ordinary Perl scalars, right? Well, that's just because objects really are like ordinary Perl scalars.

The only difference between $client and $request in each example is that in the object-oriented version, the scalars happen to know where to find some subroutines that they can call. (In OO speak, we call them "methods" instead of "subroutines.")

This is why we don't have to say process_request in the OO case: if we're calling the process method on something that knows it's a request, it knows that it's processing a request. Simple, eh? In OO speak, we say that the $request object is in the Request "class" -- a class is the "type of thing" that the object is, and classes are how objects locate their methods. Hence, $request->redirect and $mail->redirect will call completely different methods if $request and $mail are in different classes; what it means to redirect a Request object is very different to redirecting a Mail object.

You might wonder what's actually going on when we call a method. Since we know that methods are just the OO form of subroutines, you shouldn't be surprised to find that methods in Perl really are just subroutines. What about classes? Well, the purpose of a class is to distinguish one set of methods from another. And what's a natural way to distinguish one set of subroutines from another in Perl? You guessed it -- in Perl, classes are just packages. So if we've got an object called $request in the Request class and we call the redirect method, this is what actually happens:

# $request->redirect($new_url)

Request::redirect($request, $new_url)

That's right -- we just call the redirect subroutine in the appropriate package, and pass in the object along with any other parameters. Why do we pass in the object? So that redirect knows what object it's working on.

At a very basic level, this is all OO Perl is -- it's another syntax for writing subroutine calls so that it looks like you're performing actions on some data. At that, for most users of OO Perl modules, is as much as you need to know.

Why is it a win?

So if that's all it is, why does everyone think that OO Perl is the best thing since sliced bread? You'll certainly find that a whole host of interesting and useful modules out there depend on OO techniques. To understand what everyone sees in it, let's go back to procedural code for a moment. Here's something that extracts the sender and subject of a mail message:

sub mail_subject {
    my $mail = shift;
    my @lines = split /\n/, $mail;
    for (@lines) {
        return $1 if /^Subject: (.*)/;
        return if /^$/; # Blank line ends headers
sub mail_sender {
    my $mail = shift;
    my @lines = split /\n/, $mail;
    for (@lines) {
        return $1 if /^From: (.*)/;
        return if /^$/;

my $subject = mail_subject($mail);
my $from    = mail_sender($mail);

All well and good, but notice that we have to run through the whole mail each time we want to get new information about it. Now, it's true we could replace the body of these two subroutines with quite a complicated regular expression, but that's not the point: we're still doing more work than we ought to.

For our equivalent OO example, let's use the CPAN module Mail::Header. This takes a reference to an array of lines, and spits out a mail header object to which we can then do things.

my @lines = split /\n/, $mail;
my $header = Mail::Header->new(\@lines);

my $subject = $header->get("subject");
my $from    = $header->get("from");

Not only are we now looking at the problem from a perspective of "doing things to the header", we're also giving the module an opportunity to make this more efficient. How come?

One of the main benefits of CPAN modules is that they give us a set of functions we can call, and we don't have to care how they're implemented. OO programming calls this "abstraction" - the implementation is abstracted from the user's perspective. Similarly, we don't have to care what $mail_obj really is. It could just be our reference to an array of lines, but on the other hand, Mail::Header can do clever things with it.

In reality, $header is a hash reference under the hood. Again, we don't need to care whether or not it's a hash reference or an array reference or something altogether different, but as it's a hash reference, this allows the constructor, new (a constructor is just a method that creates a new object) to do all the pre-processing on our array of lines once and for all, and then store the subject, sender, and all sorts of other fields into some hash keys. All that get does, essentially, is retrieve the appropriate value from the hash. This is obviously vastly more efficient than running through the whole message each time.

That's what an object really is: it's something that the module can rearrange and use any representation of your data that it likes so that it's most efficient to operate on in the future. You, as an end user, get the benefits of a smart implementation (assuming, of course, that the person who wrote the module is smart...) and you don't need to care about, or even actually see, see what's going on underneath.

Using it

We've seen a simple use of OO techniques by using Mail::Header. Let's now look at a slightly more involved program, to solidify our knowledge. This is a very simple system information server for a Unix machine. (Don't be put off -- these programs will work on non-Unix systems as well.) Unix has a client/server protocol called "finger," by which you can contact a server and ask for information about its users. I run "finger" on my username at a local machine, and get:

% finger simon
Login name: simon       (messages off)  In real life: Simon Cozens
Office: Computing S
Directory: /v0/xzdg/simon               Shell: /usr/local/bin/bash
On since Nov  6 10:03:46                5 minutes 38 seconds Idle Time
   on pts/166 from riot-act.somewhere
On since Nov  6 12:28:08
   on pts/197 from riot-act.somewhere
Project: Hacking Perl for Sugalski

Insert amusing anecdote here.

What we're going to do is write our own finger client and a server which dishes out information about the current system, and we're going to do this using the object-oriented IO::Socket module. Of course, we could do this completely procedurally, using Socket.pm, but it's actually comparatively much easier to do it this way.

First, the client. The finger protocol, as much as we need to care about it, is really simple. The client connects and sends a line of text -- generally, a username. The server sends back some text, and then closes the connection.

By using IO::Socket to manage the connection, we can come up with something like this:

use IO::Socket::INET;

my ($host, $username) = @ARGV;

my $socket = IO::Socket::INET->new(
                        PeerAddr => $host,
                        PeerPort => "finger"
                      ) or die $!;


while ($_ = $socket->getline) {

This is pretty straightforward: the IO::Socket::INET constructor new gives us an object representing the connection to peer address $host on port finger. We can then call the print and getline methods to send and receive data from the connection. Compare this with the non-OO equivalent, and you may realize why people prefer dealing with objects:

#!/usr/bin/perl -w
use strict;
use Socket;
my ($remote,$port, $iaddr, $paddr, $proto, $user);

($remote, $user) = @ARGV; 

$port    = getservbyname('finger', 'tcp')   || die "no port";
$iaddr   = inet_aton($remote)               || die "no host: $remote";
$paddr   = sockaddr_in($port, $iaddr);

$proto   = getprotobyname('tcp');
socket(SOCK, PF_INET, SOCK_STREAM, $proto)  || die "socket: $!";
connect(SOCK, $paddr)                       || die "connect: $!";
print SOCK "$user\n";
while (<SOCK>)) {

close (SOCK)            || die "close: $!";

Now, to turn to the server. We'll also use another OO module, Net::hostent, which allows us to treat the result of gethostbyaddr as an object, rather than as a list of values. This means we don't have to worry about remembering which element of the list means what we want.

#!/usr/bin/perl -w
use IO::Socket;
use Net::hostent;

my $server = IO::Socket::INET->new( Proto     => 'tcp',
                                    LocalPort => 'finger',
                                    Listen    => SOMAXCONN,
                                    Reuse     => 1);
die "can't setup server" unless $server;

while ($client = $server->accept()) {
  $hostinfo = gethostbyaddr($client->peeraddr);
  printf "[Connect from %s]\n", $hostinfo->name || $client->peerhost;
  my $command = client->getline();
  if    ($command =~ /^uptime/) { $client->print(`uptime`); }
  elsif ($command =~ /^date/)   { $client->print(scalar localtime, "\n"); }
  else  { $client->print("Unknown command\n");

This is chock-full of OO Perl goodness -- a method call on nearly every line. We start in a very similar way to how we wrote the client: using IO::Socket::INET->new as a constructor. Did you notice anything strange about this? IO::Socket::INET is a package name, which means it must be a class, rather than an object. But we can still call methods on classes (they're generally called "class methods," for obvious reasons) and this is how most objects actually get instantiated: the class provides a method called new that produces an object for us to manipulate.

The big while loop calls the accept method that waits until a client connects and, when one does, returns another IO::Socket::INET object representing the connection to the client. We can call the client's autoflush method, which is the equivalent to setting $| for its handle; the peeraddr method returns the address of the client, which we can feed to gethostbyaddr.

As we mentioned earlier, this isn't the usual Perl gethostbyadd, but one provided by Net::Hostent, and it returns yet another object! We use the name method from this object, which represents information about a given host, to find the host's name.

The rest isn't anything new. If you think back to our client, it sent a line and awaited a response -- so our server has to read a line, and send a response. You get bonus points for adding more possible responses to our server.


So there we are. We've seen a couple of examples of using object-oriented modules. It wasn't that bad, was it? Hopefully now you'll be well-enough equipped to be able to start using some of the many OO modules on CPAN for yourself.

If, on the other hand, you feel you need a little more in-depth coverage of OO Perl, you could take a look at the "perlboot," "perltoot," and "perltootc" pages in the Perl documentation. The Perl Cookbook, an invaluable book for any serious Perl programmer, has a very comprehensive and easy to follow treatment of OO techniques. Finally, the most in-depth treatment of all can be found in Damian Conway's "Object-Oriented Perl", which will see you through from a complete beginner way up to Perl 4 or 5 dan...

Visit the home of the Perl programming language: Perl.org

Sponsored by

Monthly Archives

Powered by Movable Type 5.13-en