Let's Send Some Mail!

  use strict;
  use Email::Sender::Simple qw(sendmail);
  use Email::Simple;
  use Email::Simple::Creator;

  my $email = Email::Simple->create(
    header => [
      To      => '"Xavier Q. Ample" <x.ample@example.com>',
      From    => '"Bob Fishman" <orz@example.mil>',
      Subject => "don't forget to *enjoy the sauce*",
    ],
    body => "This message is short, but at least it's cheap.\n",
  );

  sendmail($email);

That's it. Your message goes out into the internet and tries to get delivered to "x.ample@example.com".

In the example above, $email could be an Email::Simple object, a MIME::Entity, a string containing an email message, or one of several other types of input. If "Email::Abstract" can understand a value, it can be passed to Email::Sender::Simple. Email::Sender::Simple tries to make a good guess about how to send the message. It will usually try to use the sendmail program on unix-like systems and to use SMTP on Windows. You can specify a transport, if you need to, but normally that shouldn't be an issue. (See "Picking a Transport", though, for more information.)

Also note that we imported and used a "sendmail" routine in the example above. This is exactly the same as saying:

  Email::Sender::Simple->send($email);

...but it's a lot easier to type. You can use either one.

envelope information

We didn't have to tell Email::Sender::Simple where to send the message. If you don't specify recipients, it will use all the email addresses it can find in the To and Cc headers by default. It will use Email::Address to parse those fields. Similarly, if no sender is specified, it will use the first address found in the From header.

In most email transmission systems, though, the headers are not by necessity tied to the addresses used as the sender and recipients. For example, your message header might say "From: mailing-list@example.com" while your SMTP client says "MAIL FROM:<verp-1234@lists.example.com>". This is a powerful feature, and is necessary for many email application. Being able to set those distinctly is important, and Email::Sender::Simple lets you do this:

  sendmail($email, { to => [ $to_1, $to_2 ], from => $sender });

in case of error

When the message is sent successfully (at least on to its next hop), "sendmail" will return a true value -- specifically, an Email::Sender::Success object. This object only rarely has much use. What's more useful is what happens if the message can't be sent.

If there is an error sending the message, an exception will be thrown. It will be an object belonging to the class Email::Sender::Failure. This object will have a "message" attribute describing the nature of the failure. There are several specialized forms of failure, like Email::Sender::Failure::Multi, which is thrown when more than one error is encountered when trying to send. You don't need to know about these to use Email::Sender::Simple, though. All you need to know is that "sendmail" returns true on success and dies on failure.

If you'd rather not have to catch exceptions for failure to send mail, you can use the "try_to_send" method, which can be imported as "try_to_sendmail". This method will return just false on failure to send mail.

For example:

  Email::Sender::Simple->try_to_send($email, { ... });

  use Email::Sender::Simple qw(try_to_sendmail);
  try_to_sendmail($email, { ... });

Some Email::Sender transports can signal success if some, but not all, recipients could be reached. Email::Sender::Simple does its best to ensure that this never happens. When you are using Email::Sender::Simple, mail should either be sent or not. Partial success should never occur.

Picking a Transport

If Email::Sender::Simple doesn't pick the transport you want, or if you have more specific needs, you can specify a transport in several ways. The simplest is to build a transport object and pass it in. You can read more about transports elsewhere. For now, we'll just assume that you need to send mail via SMTP on an unusual port. You can send mail like this:

  my $transport = Email::Sender::Transport::SMTP->new({
    host => 'smtp.example.com',
    port => 2525,
  });

  sendmail($email, { transport => $transport });

Now, instead of guessing at what transport to use, Email::Sender::Simple will use the one you provided. This transport will have to be specified for each call to "sendmail", so you might want to look at other options, which follow.

specifying transport in the environment

If you have a program that makes several calls to Email::Sender::Simple, and you need to run this program using a different mailserver, you can set environment variables to change the default. For example:

  $ export EMAIL_SENDER_TRANSPORT=SMTP
  $ export EMAIL_SENDER_TRANSPORT_host=smtp.example.com
  $ export EMAIL_SENDER_TRANSPORT_port=2525

  $ perl your-program

It is important to note that if you have set the default transport by using the environment, no subsequent "transport" args to "sendmail" will be respected. If you set the default transport via the environment, that's it. Everything will use that transport. (Also, note that while we gave the host and port arguments above in lower case, the casing of arguments in the environment is flattened to support systems where environment variables are of a fixed case. So, "EMAIL_SENDER_TRANSPORT_PORT" would also work.

This is extremely valuable behavior, as it allows you to audit every message that would be sent by a program by running something like this:

  $ export EMAIL_SENDER_TRANSPORT=Maildir
  $ perl your-program

In that example, any message sent via Email::Sender::Simple would be delivered to a maildir in the current directory.

subclassing to change the default transport

If you want to use a library that will behave like Email::Sender::Simple but with a different default transport, you can subclass Email::Sender::Simple and replace the "build_default_transport" method.

Testing

  use Test::More;
  BEGIN { $ENV{EMAIL_SENDER_TRANSPORT} = 'Test' }
  use YourCode;

  YourCode->run;

  my @deliveries = Email::Sender::Simple->default_transport->deliveries;

Now you've got an array containing every delivery performed through Email::Sender::Simple, in order. Because you set the transport via the environment, no other code will be able to force a different transport.

When testing code that forks, Email::Sender::Transport::SQLite can be used to allow every child process to deliver to a single, easy to inspect destination database.

Hey, where's my Bcc support?

For example, here is a set of headers:

  From: sender@example.com
  To:   to_rcpt@example.com
  Cc:   cc_rcpt@example.com
  Bcc:  the_boss@example.com

In this case, we'd expect the message to be delivered to three people: to_rcpt, cc_rcpt, and the_boss. This is why it's often suggested that the Bcc header should be a source for envelope recipients. In fact, though, a message with a Bcc header should probably be delivered only to the Bcc recipients. The "B" in Bcc means "blind." The other recipients should not see who has been Bcc'd. This means you want to send two messages: one to to_rcpt and cc_rcpt, with no Bcc header present; and another to the_boss only, with the Bcc header. If you just pick up Bcc addresses as recipients, everyone will see who was Bcc'd.

Email::Sender::Simple promises to send messages atomically. That is: it won't deliver to only some of the recipients, and not to others. That means it can't automatically detect the Bcc header and make two deliveries. There would be a possibility for the second to fail after the first succeeded, which would break the promise of a pure failure or success.

The other strategy for dealing with Bcc is to remove the Bcc header from the message and then inject the message with an envelope including the Bcc addresses. The envelope information will not be visible to the final recipients, so this is safe. Unfortunately, this requires modifying the message, and Email::Sender::Simple should not be altering the mutable email object passed to it. There is no "clone" method on Email::Abstract, so it cannot just build a clone and modify that, either. When such a method exists, Bcc handling may be possible.

Example Bcc Handling

If you want to support the Bcc header now, it is up to you to deal with how you want to munge the mail and inject the (possibly) munged copies into your outbound mailflow. It is not reasonable to suggest that Email::Sender::Simple do this job.

Example 1: Explicitly set the envelope recipients for Bcc recipients

Create the email without a Bcc header, send it to the Bcc users explicitly and then send it to the To/Cc users implicitly.

  my $message = create_email_mime_msg;  # <- whatever you do to get the message

  $message->header_set('bcc');          # delete the Bcc header before sending
  sendmail($message, { to => $rcpt });  # send to explicit Bcc address
  sendmail($message);                   # and then send as normal

Example 2: Explicitly set the envelope recipients for all recipients

You can make a single call to "sendmail" by pulling all the recipient addresses from the headers yourself and specifying all the envelope recipients once. Again, delete the Bcc header before the message is sent.

This is awesome! Where can I learn more?