Mail::MboxParser - read-only access to UNIX-mailboxes

    use Mail::MboxParser;

    my $parseropts = {
        enable_cache    => 1,
        enable_grep     => 1,
        cache_file_name => 'mail/cache-file',
    };
    my $mb = Mail::MboxParser->new('some_mailbox', 
                                    decode     => 'ALL',
                                    parseropts => $parseropts);

    # -----------

    # slurping
    for my $msg ($mb->get_messages) {
        print $msg->header->{subject}, "\n";
        $msg->store_all_attachments(path => '/tmp');
    }

    # iterating
    while (my $msg = $mb->next_message) {
        print $msg->header->{subject}, "\n";
        # ...
    }

    # we forgot to do something with the messages
    $mb->rewind;
    while (my $msg = $mb->next_message) {
        # iterate again
        # ...
    }

    # subscripting one message after the other
    for my $idx (0 .. $mb->nmsgs - 1) {
        my $msg = $mb->get_message($idx);
    }

This module attempts to provide a simplified access to standard UNIX-mailboxes. It offers only a subset of methods to get 'straight to the point'. More sophisticated things can still be done by invoking any method from MIME::Tools on the appropriate return values.

Mail::MboxParser has not been derived from Mail::Box and thus isn't acquainted with it in any way. It, however, incorporates some invaluable hints by the author of Mail::Box, Mark Overmeer.

See also the section ERROR-HANDLING much further below.

More to that, see the relevant manpages of Mail::MboxParser::Mail, Mail::MboxParser::Mail::Body and Mail::MboxParser::Mail::Convertable for a description of the methods for these objects.

new(mailbox, options)

new(scalar-ref, options)

new(array-ref, options)

new(filehandle, options)

This creates a new MboxParser-object opening the specified 'mailbox' with either absolute or relative path. new() can also take a reference to a variable containing the mailbox either as one string (reference to a scalar) or linewise (reference to an array), or a filehandle from which to read the mailbox. The following option(s) may be useful. The value in brackets below the key is the default if none given.

    key:      | value:     | description:
    ==========|============|===============================
    decode    | 'NEVER'    | never decode transfer-encoded
    (NEVER)   |            | data
              |------------|-------------------------------
              | 'BODY'     | will decode body into a human-
              |            | readable format
              |------------|-------------------------------
              | 'HEADER'   | will decode header fields if
              |            | any is encoded
              |------------|-------------------------------
              | 'ALL'      | decode any data
    ==========|============|===============================
    uudecode  | 1          | enable extraction of uuencoded
    (0)       |            | attachments in MIME::Parser
              |------------|-------------------------------
              | 0          | uuencoded attachments are
              |            | treated as plain body text
    ==========|============|===============================
    newline   | 'UNIX'     | UNIXish line-endings 
    (AUTO)    |            | ("\n" aka \012)
              |------------|-------------------------------
              | 'WIN'      | Win32 line-endings
              |            | ("\n\r" aka \012\015)
              |------------|-------------------------------
              | 'AUTO'     | try to do autodetection
              |------------|-------------------------------
              | custom     | a user-given value for totally
              |            | borked mailboxes
    ==========|============|===============================
    oldparser | 1          | uses the old (and slower) 
    (0)       |            | parser (but guaranteed to show
              |            | the old behaviour)
              |------------|-------------------------------
              | 0          | uses Mail::Mbox::MessageParser
    ==========|============|===============================
    parseropts|            | see "Specifying parser opts"
              |            | below
    ==========|============|===============================

The newline option comes in handy if you have a mbox-file that happens to not conform to the rules of your operating-system's character semantics one way or another. One such scenario: You are using the module under Win but deliberately have mailboxes with UNIX-newlines (or the other way round). If you do not give this option, 'AUTO' is assumed and some basic tests on the mailbox are performed. This autoedection is of course not capable of detecting cases where you use something like '#DELIMITER' as line-ending. It can as to yet only distinguish between UNIX and Win32ish newlines. You may be lucky and it even works for Macintoshs. If you have more extravagant wishes, pass a costum value:

    my $mb = new Mail::MboxParser ("mbox", newline => '#DELIMITER');

You can't use regexes here since internally this relies on the $/ var ($INPUT_RECORD_SEPERATOR, that is). When passing either a scalar-, array-ref or \*STDIN as first-argument, an anonymous tmp-file is created to hold the data. This procedure is hidden away from the user so there is no need to worry about it. Since a tmp-file acts just like an ordinary mailbox-file you don't need to be concerned about loss of data or so once you have been walking through the mailbox-data. No data will be lost and it'll all be fine and smooth.

When available, the module will use CWMail::Mbox::MessageParser to do the parsing. To get the most speed out of it, you can tweak some of its options. Arguably, you even have to do that in order to make it use caching. Options for the parser are given via the parseropts switch that expects a reference to a hash as values. The values you can specify are:

enable_cache

When set to a true value, caching is used but only if you gave cache_file_name. There is no default value here!

cache_file_name

The file used for caching. This option is mandatory if enable_cache is true.

enable_grep

When set to a true value (which is the default), the extern grep(1) is used to speed up parsing. If your system does not provide a usable grep implementation, it silently falls back to the pure Perl parser.

When the module was unable to create a CWMail::Mbox::MessageParser object, it will fall back to the old parser in the hope that the construction of the object then succeeds.

open(source, options)

Takes exactly the same arguments as new() does just that it can be used to change the characteristics of a mailbox on the fly.

get_messages

Returns an array containing all messages in the mailbox respresented as Mail::MboxParser::Mail objects. This method is _minimally_ quicker than iterating over the mailbox using CWnext_message but eats much more memory. Memory-usage will grow linearly for each new message detected since this method creates a huge array containing all messages. After creating this array, it will be returned.

get_message(n)

Returns the n-th message (first message has index 0) in a mailbox. Examine CW$mb->error which contains an error-string if the message does not exist. In this case, CWget_message returns undef.

next_message

This lets you iterate over a mailbox one mail after another. The great advantage over CWget_messages is the very low memory-comsumption. It will be at a constant level throughout the execution of your script. Secondly, it almost instantly begins spitting out Mail::MboxParser::Mail-objects since it doesn't have to slurp in all mails before returing them.

set_pos(n)

rewind

current_pos

These three methods deal with the position of the internal filehandle backening the mailbox. Once you have iterated over the whole mailbox using CWnext_message MboxParser has reached the end of the mailbox and you have to do repositioning if you want to iterate again. You could do this with either CWset_pos or CWrewind.

    $mb->rewind;  # equivalent to
    $mb->set_pos(0);

CWcurrent_pos reveals the current position in the mailbox and can be used to later return to this position if you want to do tricky things. Mark that CWcurrent_pos does *not* return the current line but rather the current character as returned by Perl's tell() function.

    my $last_pos;
    while (my $msg = $mb->next_message) {
        # ...
        if ($msg->header->{subject} eq 'I was looking for this') {
            $last_pos = $mb->current_pos;
            last; # bail out here and do something else
        }
    }

    # ...
    # ...

    # now continue where we stopped:
    $mb->set_pos($last_pos)
    while (my $msg = $mb->next_message) {
        # ...
    }

WARNING: Be very careful with these methods when using the parser of CWMail::Mbox::MessageParser. This parser maintains its own state and you shouldn't expect it to always be in sync with the state of CWMail::MboxParser. If you need some finer control over the parsing, better consider to use the public interface as described in the manpage of Mail::Mbox::MessageParser. Use CWparser() to get the underlying parser object. This however may expose you to the same problems turned around: CWMail::MboxParser may loose its sync with its parser when you do that. Therefore: Just avoid any of the above for now and wait till CWMail::Mbox::MessageParser has a stable interface.

make_index

You can force the creation of a message-index with this method. The message-index is a mapping between the index-number of a message (0 .. CW$mb->nmsgs - 1) and the byte-position of the filehandle. This is usually done automatically for you once you call CWget_message hence the first call for a particular message will be a little slower since the message-index first has to be built. This is, however, done rather quickly. You can have a peek at the index if you are interested. The following produces a nicely padded table (suitable for mailboxes up to 9.9999...GB ;-).

    $mb->make_index;
    for (0 .. $mb->nmsgs - 1) {
        printf "%5.5d => %10.10d\n", 
                $_, $mb->get_pos($_);
    }

get_pos(n)

This method takes the index-number of a certain message within the mailbox and returns the corresponding position of the filehandle that represents that start of the file. It is mainly used by CWget_message() and you wouldn't really have to bother using it yourself except for statistical purpose as demonstrated above along with make_index.

nmsgs

Returns the number of messages in a mailbox. You could naturally also call get_messages in scalar-context, but this one wont create new objects. It just counts them and thus it is much quicker and wont eat a lot of memory.

parser

Returns the bare CWMail::Mbox::MessageParser object. If no such object exists returns CWundef. You can use this method to check whether the module actually uses the old or new parser. If CWparser returns a false value, it is using the old parsing routines.

error

Call this immediately after one of the methods above that mention a possible error-message.

log

Sort of internal weirdnesses are recorded here. Again only the last event is saved.

Mail::MboxParser provides a mechanism for you to figure out why some methods did not function as you expected. There are four classes of unexpected behavior:

(1) bad arguments

In this case you called a method with arguments that did not make sense, hence you confused Mail::MboxParser. Example:

  $mail->store_entity_body;           # wrong, needs two arguments
  $mail->store_entity_body(0);        # wrong, still needs one more

In any of the above two cases, you'll get an error message and your script will exit. The message will, however, tell you in which line of your script this error occured.

(2) correct arguments but...

Consider this line:

  $mail->store_entity_body(50, \*FH); # could be wrong

Obviously you did call store_entity_body with the correct number of arguments. That's good because now your script wont just exit. Unfortunately, your program can't know in advance whether the particular mail ($mail) has a 51st entity. So, what to do? Just be brave: Write the above line and do the error-checking afterwards by calling CW$mail->error immediately after store_entity_body:

        $mail->store_entity_body(50, *\FH);
        if ($mail->error) {
                print "Oups, something wrong:", $mail->error;
        }

In the description of the available methods above, you always find a remark when you could use CW$mail->error. It always returns a string that you can print out and investigate any further.

(3) errors, that never get visible

Well, they exist. When you handle MIME-stuff a lot such as attachments etc., Mail::MboxParser internally calls a lot of methods provided by the MIME::Tools package. These work splendidly in most cases, but the MIME::Tools may fail to produce something sensible if you have a very queer or even screwed up mailbox. If this happens you might find information on that when calling CW$mail->log. This will give you the more or less unfiltered error-messages produced by MIME::Tools. My advice: Ignore them! If there really is something in CW$mail->log it is either because you're mails are totally weird (there is nothing you can do about that then) or these errors are smoothly catched inside Mail::MboxParser in which case all should be fine for you.

(4) the apocalyps

If nothing seems to work the way it should and CW$mail->error is empty, then the worst case has set in: Mail::MboxParser has a bug. Needless to say that there is any way to get around of this. In this case you should contact and I'll examine that.

I have been working hard on making Mail::MboxParser eat less memory and as quick as possible. Due to that, two time and memory consuming matters are now called on demand. That is, parsing out the MIME-parts and turning the raw header into a hash have become closures.

The drawback of that is that it may get inefficient if you often call

 $mail->header->{field}

In this case you should probably save the return value of CW$mail->header (a hashref) into a variable since each time you call it the raw header is parsed.

On the other hand, if you have a mailbox of, say, 25MB, and hold each header of each message in memory, you'll quickly run out of that. So, you can now choose between more performance and more memory.

This all does not happen if you just parse a mailbox to extract one header-field (eg. subject), work with that and exit. In this case it will need both less memory and is still considerably quicker. :-)

Some mailers have a fancy idea of how a To: - or Cc: -line should look. I have seen things like:

        To: "\"John Doe"\" <john.doe@example.com>

The splitting into name and email, however, does still work here, but you have to remove these silly double-quotes and backslashes yourself.

The way of counting the messages and detecting them now complies to RFC 822. This is, however, no guarentee that it all works seamlessly. There are just so many mailboxes that get screwed up by mal-formated mails.

Apart from new bugs that almost certainly have been introduced with this release, following things still need to be done:

Transfer-Encoding

Still, only quoted-printable encoding is correctly handled.

Tests

Clean-up of the test-scripts is desperately needed. Now they represent rather an arbitrary selection of tested functions. Some are tested several times while others don't show up at all in the suits.

Thanks to a number of people who gave me invaluable hints that helped me with Mail::Box, notably Mark Overmeer for his hints on more object-orientedness.

Kenn Frankel (kenn AT kenn DOT cc) kindly patched the broken split-header routine and added get_field().

David Coppit for making me aware of CWMail::Mbox::MessageParser and designing it the way I needed to make it work for my module.

This is version 0.54.

Tassilo von Parseval <tassilo.von.parseval@rwth-aachen.de>

Copyright (c) 2001-2005 Tassilo von Parseval. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

MIME::Entity

Mail::MboxParser::Mail, Mail::MboxParser::Mail::Body, Mail::MboxParser::Mail::Convertable

Mail::Mbox::MessageParser