NAME
Email::Store - Framework for database-backed email storage
SYNOPSIS
use Email::Store 'dbi:mysql:mailstore';
Email::Store->setup; # Do this once
Email::Store::Mail->store( $rfc822 );
Email::Store::Mail->retrieve( $msgid );
...
DESCRIPTION
"Email::Store" is the ideal basis for any application which needs to
deal with databases of email: archiving, searching, or even storing mail
for implementing IMAP or POP3 servers.
"Email::Store" itself is a very lightweight framework, meaning it does
not provide very much functionality itself; in effect, it is merely a
Class::DBI interface to a database schema which is designed for storing
email. Incidentally, if you don't know much about "Class::DBI", you're
going to need to in order to get much out of this.
Despite its minimalist nature, "Email::Store" is incredibly powerful.
Its power comes from its extensibility, through plugin modules and hooks
which allow you to add new database tables and concepts to the system,
and so access the mail store from a "different direction". In a sense,
"Email::Store" is a blank canvas, onto which you can pick and choose (or
even write!) the plugins which you want for your application.
For instance, the core "Email::Store::Entity" plugin module addresses
the idea of "people" in the email universe, allowing you to search for
mails to or from particular people; (despite their changing names or
email addresses) "Email::Store::Thread" interfaces "Email::Store" to
"Mail::Thread" allowing you to navigate mails by their position in a
mail thread; the planned non-core "Email::Store::Plucene" module plugs
into the indexing process and stores information about emails in a
Plucene search index for quick retrieval later, and so on.
Core "Email::Store" modules
To get you started with a useful database, "Email::Store" provides a few
core plugin modules which comprise the basics of a mailstore. Each of
the modules provides one or more database tables, representing important
concepts in the email world, and one or more relationships between these
concepts and the other tables in the system. It's a little less
complicated than that, as we'll see when we go through each module in
turn. Here is a quick summary of what the core modules do:
"Email::Store::Mail"
This is, in a sense, the plugin of plugins. "Email::Store::Mail"
encapsulates individual email messages. Its "store" method is the
means in which emails are indexed and enter the mailstore. How this
storing is done, however, is influenced by all the other plugins.
"Email::Store::List"
"List" is one of the easiest plugins to understand. To our concept of
the mail, it adds the concept of a mailing list.
"Email::Store::List" hooks into the indexing process and examines a
mail to see if it came via a mailing list; if so, it associates the
mail with one or more lists. This means you can ask a mail object for
its "lists", and a list object for its "posts". Because of this,
instead of looking at messages by their message ID, you can start by
looking for a mailing list you're interested in and then navigate to
the messages you want.
"Email::Store::Entity"
"Entity" is the most fundamental of the plugins but (or perhaps,
"thus") the most complex. This module adds the concept of an
addressing, which abstracts out the From, To, Cc and Bcc headers of
an email. A "To" header, for instance, says that the mail is
addressed to a particular name and address, but
"Email::Store::Entity" also provides the potential for associating
different names and addresses with the concept of an entity, a unique
individual. That is, not all mails addressed to the name "Simon
Cozens" are to me (due to the existence of multiple Simon Cozenses in
the world) but all mails to ".*@simon-cozens.org" are, despite their
being multiple email addresses which match that pattern.
If that has you confused, (and believe me, it has me confused) ignore
the "entity" bit and know that you can navigate from names, addresses
and the intersection of the two, to emails involving them. More
details in Email::Store::Entity as you'd expect.
"Email::Store::Attachment"
As you might be able to guess, this adds the concept of an
attachment. It also ambushes the indexing process, and strips all the
MIME attachments off an email, placing them in the attachments table.
It then quietly slips the de-MIMEd email back into the mail table,
and now you can ask a mail for its "attachments".
All these modules have some degree of POD, so you should consult them
for more details on the interface that they provide. Over time, there
will be additional modules that you can install from CPAN.
USAGE
When you use "Email::Store", you should pass a DBI connection string to
its "use" statement:
use Email::Store 'dbi:SQLite:dbname=mailstore.db';
In order to create the tables used by the plugin modules, you should
then say
Email::Store->setup;
You should do this on the initial set-up of your database, and then
again on installing any additional plugin modules, to create the new
tables they want to use. Note that this does not retroactively index
existing mail with the new functions provided by the modules you've just
installed! - a "reindex" method is planned, but is not there yet.
This is all the functionality that "Email::Store" itself provides. See
the documentation to the various plugins for their public interface,
chiefly Email::Store::Mail.
THE PLUGIN SYSTEM
If you want to write your own plugins, you will need to know how the
plugin system works.
The first thing to note is that when "Email::Store" indexes a mail,
whether for the first time or when it re-indexes a mail it's seen
before, it loads up all the modules it can find under the
"Email::Store::*" hierarchy. Additionally, when "Email::Store->setup" is
called, all the "Email::Store::*" modules are required. So, to register
your new plugin, all you need to do is call it
"Email::Store::"*something* and put it in Perl's include path in the
usual way.
Each plugin module should be a self-contained description of some
concepts, the database schema that encapsulates them, their relationship
to the rest of the system, and any hooks or additional functionality
provided.
Let us write a very simple plugin as a first example. This will
introduce the concept of a mail annotation, an open-ended space where we
can store "sticky notes" which relate to a particular email. We'll call
the plugin "Email::Store::Annotation", and we start by putting the
following in Email/Store/Annotation.pm:
package Email::Store::Annotation;
use base 'Email::Store::DBI';
This makes us a "Class::DBI"-based package. Next we need to do the usual
"Class::DBI" thing and ddeclare our table and columns:
Email::Store::Annotation->table("mail_annotation");
Email::Store::Annotation->columns(All => qw/id mail content/);
Next we declare how this fits into the rest of the world: an
"Email::Store::Mail" has many "annotations":
Email::Store::Mail->has_many(annotations => "Email::Store::Annotation");
Annotations are something that the utility which uses "Email::Store" is
going to create, modify and delete manually; we can hardly auto-generate
a user-defined annotation when a mail is indexed, so we don't need to
define any hooks into the indexing process. In fact, this is all the
code we need to write, so we end the package in the usual way:
1;
If we did need to hook into a different part of "Email::Store", we'd
have to use Module::Pluggable::Ordered's plugin mechanism. See
Email::Store::Mail for the hooks provided and how to hook into them.
But where does this "mail_annotation" table come from? How does
"Email::Store" know how to create it? The answer comes when we put the
schema into the "__DATA__" section: "Email::Store->setup" reads all the
"DATA" sections for the plugins that it finds, and executes them as SQL
in the database. As pretty much every database's SQL is subtly
different, the schema should be written in MySQL's SQL and
"Email::Store" will magically translate it for the database in use:
__DATA__
CREATE TABLE IF NOT EXISTS mail_annotation (
id INTEGER auto_increment NOT NULL PRIMARY KEY,
mail INTEGER,
content TEXT
);
With this module complete and installed, an "Email::Store" user can now
say:
my $mail = Email::Store::Mail->retrieve( $msg_id );
$mail->add_to_annotations({ content => "I like this mail" });
print "Things I know about this mail:\n";
print $_->content, "\n" for $mail->annotations;
The really big advantage of this architecture is that everything about a
concept and its relationship to the mailstore is encapsulated in a
single file and can be dropped in and out at will, without disturbing
the rest of the code. This is fantastic extensibility. "Email::Store"
does not need to define a schema of every single table you might
possibly need up front, but everything is modularised.
The really big disadvantage is that the interface of one part of the
system, such as "Email::Store::Mail" isn't collected in one place, but
gets added to by pretty much every other plugin that gets loaded up. If
you look in the "Email::Store::Mail" POD you'll see nothing about the
"add_to_annotations" method that we've just called.
However, since every plugin should document its interface thoroughly and
its relationship to other parts of the system, this should not really be
a problem for end-users.
SEE ALSO
Understanding Class::DBI is fundamental to using "Email::Store".
The core modules: Email::Store::Mail, Email::Store::List,
Email::Store::Entity, Email::Store::Thread, Email::Store::Attachment.
Please do read through their documentation to see the whole of the
"Email::Store" API.
Any other "Email::Store::*" modules you find on CPAN.
Module::Pluggable::Ordered is the pluggable hooks system used throughout
"Email::Store". Those developing additional modules might want to look
at its documentation to understand how to hook into the indexing,
reindexing and other processes.
AUTHOR
Simon Cozens, <simon@cpan.org>
COPYRIGHT AND LICENSE
Copyright 2004 by Simon Cozens
This library is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.