=encoding utf8 =head1 NAME Mail::Box - manage a mailbox, a folder with messages =head1 INHERITANCE Mail::Box is a Mail::Reporter Mail::Box is extended by Mail::Box::Dir Mail::Box::File Mail::Box::Net =head1 SYNOPSIS use Mail::Box::Manager; my $mgr = Mail::Box::Manager->new; my $folder = $mgr->open(folder => $ENV{MAIL}, ...); print $folder->name; # Get the first message. print $folder->message(0); # Delete the third message $folder->message(3)->delete; # Get the number of messages in scalar context. my $emails = $folder->messages; # Iterate over the messages. foreach ($folder->messages) {...} # all messages foreach (@$folder) {...} # all messages $folder->addMessage(Mail::Box::Message->new(...)); Tied-interface: tie my(@inbox), 'Mail::Box::Tie::ARRAY', $inbox; # Four times the same: $inbox[3]->print; # tied $folder->[3]->print; # overloaded folder $folder->message(3)->print; # usual print $folder->[3]; # overloaded message tie my(%inbox), 'Mail::Box::Tie::HASH', $inbox; # Twice times the same $inbox{$msgid}->print; # tied $folder->messageId($msgid)->print;# usual =head1 DESCRIPTION A L creates C objects. But you already knew, because you started with the L manual page. That page is obligatory reading, sorry! C is the base class for accessing various types of mailboxes (folders) in a uniform manner. The various folder types vary on how they store their messages, but when some effort those differences could be hidden behind a general API. For example, some folders store many messages in one single file, where other store each message in a separate file within the same directory. No object in your program will be of type C: it is only used as base class for the real folder types. C is extended by Extends L<"DESCRIPTION" in Mail::Reporter|Mail::Reporter/"DESCRIPTION">. =head1 OVERLOADED =over 4 =item overload: B<""> (stringification) The folder objects stringify to their name. This simplifies especially print statements and sorting a lot. example: use overloaded folder as string # Three lines with overloading: resp. cmp, @{}, and "" foreach my $folder (sort @folders) { my $msgcount = @$folder; print "$folder contains $msgcount messages\n"; } =item overload: B<@{}> When the folder is used as if it is a reference to an array, it will show the messages, like L and L would do. example: use overloaded folder as array my $msg = $folder->[3]; my $msg = $folder->message(3); # same foreach my $msg (@$folder) ... foreach my $msg ($folder->messages) ... # same =item overload: B (string comparison) folders are compared based on their name. The sort rules are those of the build-in C. =back =head1 METHODS Extends L<"METHODS" in Mail::Reporter|Mail::Reporter/"METHODS">. =head2 Constructors Extends L<"Constructors" in Mail::Reporter|Mail::Reporter/"Constructors">. =over 4 =item Mail::Box-EB(%options) Open a new folder. A list of labeled %options for the mailbox can be supplied. Some options pertain to Mail::Box, and others are added by sub-classes. To control delay-loading of messages, as well the headers as the bodies, a set of C<*_type> options are available. C determines whether we want delay-loading. -Option --Defined in --Default access 'r' body_delayed_type Mail::Message::Body::Delayed body_type coerce_options [] create extract 10240 field_type undef fix_headers folder $ENV{MAIL} folderdir undef head_delayed_type Mail::Message::Head::Delayed head_type Mail::Message::Head::Complete keep_dups lock_file undef lock_timeout 1 hour lock_type Mail::Box::Locker::DotLock lock_wait 10 seconds locker undef log Mail::Reporter 'WARNINGS' manager undef message_type ::Message multipart_type Mail::Message::Body::Multipart remove_when_empty save_on_exit trace Mail::Reporter 'WARNINGS' trusted =over 2 =item access => MODE Access-rights to the folder. Folders are opened for read-only (which means write-protected) by default! MODE can be =over 4 =item C<'r'>: read-only (default) =item C<'a'>: append =item C<'rw'>: read-write =item C<'d'>: delete =back These MODE has no relation to the modes actually used to open the folder files within this module. For instance, if you specify C<"rw">, and open the folder, only read permission on the folder-file is required. Be warned: writing a MBOX folder may create a new file to replace the old folder. The permissions and owner of the file may get changed by this. =item body_delayed_type => CLASS The bodies which are delayed: which will be read from file when it is needed, but not before. =item body_type => CLASS|CODE When messages are read from a folder-file, the headers will be stored in a C object. For the body, however, there is a range of choices about type, which are all described in L. Specify a CODE-reference which produces the body-type to be created, or a CLASS of the body which is used when the body is not a multipart or nested. In case of a code reference, the header structure is passed as first argument to the routine. Do I return a delayed body-type (like C<::Delayed>), because that is determined by the C option while the folder is opened. Even delayed message will require some real body type when they get parsed eventually. Multiparts and nested messages are also outside your control. For instance: $mgr->open('InBox', body_type => \&which_body); sub which_body($) { my $head = shift; my $size = $head->guessBodySize || 0; my $type = $size > 100000 ? 'File' : 'Lines'; "Mail::Message::Body::$type"; } The default depends on the mail-folder type, although the general default is L. Please check the applicable manual pages. =item coerce_options => ARRAY Keep configuration information for messages which are coerced into the specified folder type, starting with a different folder type (or even no folder at all). Messages which are coerced are always fully read, so this kind of information does not need to be kept here. =item create => BOOLEAN Automatically create the folder when it does not exist yet. This will only work when access is granted for writing or appending to the folder. Be careful: you may create a different folder type than you expect unless you explicitly specify L. =item extract => INTEGER | CODE | METHOD | 'LAZY'|'ALWAYS' Defines when to parse (process) the content of the message. When the header of a message is read, you may want to postpone the reading of the body: header information is more often needed than the body data, so why parse it always together? The cost of delaying is not too high, and with some luck you may never need parsing the body. If you supply an INTEGER to this option, bodies of those messages with a total size less than that number will be extracted from the folder only when necessary. Messages where the size (in the C field) is not included in the header, like often the case for multiparts and nested messages, will not be extracted by default. If you supply a CODE reference, that subroutine is called every time that the extraction mechanism wants to determine whether to parse the body or not. The subroutine is called with the following arguments: CODE->(FOLDER, HEAD) where FOLDER is a reference to the folder we are reading. HEAD refers to the L head of the message at hand. The routine must return a C value (extract now) or a C value (be lazy, do not parse yet). Think about using the L and L on the header to determine your choice. The third possibility is to specify the NAME of a method. In that case, for each message is called: FOLDER->NAME(HEAD) Where each component has the same meaning as described above. The fourth way to use this option involves constants: with C all messages will be delayed. With C you enforce unconditional parsing, no delaying will take place. The latter is usuful when you are sure you always need all the messages in the folder. $folder->new(extract => 'LAZY'); # Very lazy $folder->new(extract => 10000); # Less than 10kB # same, but implemented yourself $folder->new(extract => &large); sub large($) { my ($f, $head) = @_; my $size = $head->guessBodySize; defined $size ? $size < 10000 : 1 }; # method call by name, useful for Mail::Box # extensions. The example selects all messages # sent by you to be loaded without delay. # Other messages will be delayed. $folder->new(extract => 'sent_by_me'); sub Mail::Box::send_by_me($) { my ($self, $header) = @_; $header->get('from') =~ m/\bmy\@example.com\b/i; } =item field_type => CLASS The type of the fields to be used in a header. Must extend L. =item fix_headers => BOOLEAN Broken MIME headers usually stop the parser: all lines not parsed are added to the body of the message. With this flag set, the erroneous line is added to the previous header field and parsing is continued. See L. =item folder => FOLDERNAME Which folder to open (for reading or writing). When used for reading (the C option set to C<"r"> or C<"a">) the mailbox should already exist and must be readable. The file or directory of the mailbox need not exist if it is opened for reading and writing (C<"rw">). Write-permission is checked when opening an existing mailbox. The folder name can be preceded by a C<"=">, to indicate that it is named relative to the directory specified in new(folderdir). Otherwise, it is taken as relative or absolute path. =item folderdir => DIRECTORY Where are folders to be found by default? A folder-name may be preceded by a equals-sign (C<=>, a C convension) to explicitly state that the folder is located below the default directory. For example: in case C '/tmp'> and C '=abc'>, the name of the folder-file is C<'/tmp/abc'>. Each folder type has already some default set. =item head_delayed_type => CLASS The headers which are delayed: which will be read from file when it is needed, but not before. =item head_type => CLASS The type of header which contains all header information. Must extend L. =item keep_dups => BOOLEAN Indicates whether or not duplicate messages within the folder should be retained. A message is considered to be a duplicate if its message-id is the same as a previously parsed message within the same folder. If this option is false (the default) such messages are automatically deleted, because it is considered useless to store the same message twice. =item lock_file => FILENAME The name of the file which is used to lock. This must be specified when locking is to be used. =item lock_timeout => SECONDS When the lock file is older than the specified number of SECONDS, it is considered a mistake. The original lock is released, and accepted for this folder. =item lock_type => CLASS|STRING|ARRAY The type of the locker object. This may be the full name of a CLASS which extends Mail::Box::Locker, or one of the known locker types C, C, C, C, C, C, or C. If an ARRAY is specified, then a Multi locker is built which uses the specified list. =item lock_wait => SECONDS SECONDS to wait before failing on opening this folder. =item locker => OBJECT An OBJECT which extends L, and will handle folder locking replacing the default lock behavior. =item log => LEVEL =item manager => MANAGER A reference to the object which manages this folder -- typically an L instance. =item message_type => CLASS What kind of message objects are stored in this type of folder. The default is constructed from the folder class followed by C<::Message>. For instance, the message type for C is C =item multipart_type => CLASS The default type of objects which are to be created for multipart message bodies. =item remove_when_empty => BOOLEAN Determines whether to remove the folder file or directory automatically when the write would result in a folder without messages nor sub-folders. =item save_on_exit => BOOLEAN Sets the policy for saving the folder when it is closed. A folder can be closed manually (see L) or in a number of implicit ways, including on the moment the program is terminated. =item trace => LEVEL =item trusted => BOOLEAN Flags whether to trust the data in the folder or not. Folders which reside in your C will be trusted by default (even when the names if not specified staring with C<=>). Folders which are outside the folderdir or read from STDIN (L) are not trused by default, and require some extra checking. If you do not check encodings of received messages, you may print binary data to the screen, which is a security risk. =back =back =head2 The folder =over 4 =item $obj-EB($message, %options) Add a message to the folder. A message is usually a L object or a sub-class thereof. The message shall not be in an other folder, when you use this method. In case it is, use L or L via the manager. Messages with id's which already exist in this folder are not added. BE WARNED that message labels may get lost when a message is moved from one folder type to an other. An attempt is made to translate labels, but there are many differences in interpretation by applications. -Option--Default share =over 2 =item share => BOOLEAN Try to share the physical resource of the current message with the indicated message. It is sometimes possible to share messages between different folder types. When the sharing is not possible, than this option is simply ignored. Sharing the resource is quite dangerous, and only available for a limited number of folder types, at the moment only some L folders; these file-based messages can be hardlinked (on platforms that support it). The link may get broken when one message is modified in one of the folders.... but maybe not, depending on the folder types involved. =back example: $folder->addMessage($msg); $folder->addMessages($msg1, $msg2, ...); =item $obj-EB(@messages) Adds a set of message objects to the open folder at once. For some folder types this may be faster than adding them one at a time. example: $folder->addMessages($msg1, $msg2, ...); =item Mail::Box-EB(%options) Append one or more messages to an unopened folder. Usually, this method is called by the L, in which case the correctness of the folder type is checked. For some folder types it is required to open the folder before it can be used for appending. This can be fast, but this can also be very slow (depends on the implementation). All %options passed will also be used to open the folder, if needed. -Option --Default folder message undef messages undef share =over 2 =item folder => FOLDERNAME The name of the folder to which the messages are to be appended. The folder implementation will avoid opening the folder when possible, because this is resource consuming. =item message => MESSAGE =item messages => ARRAY-OF-MESSAGES One reference to a MESSAGE or a reference to an ARRAY of MESSAGEs, which may be of any type. The messages will be first coerced into the correct message type to fit in the folder, and then will be added to it. =item share => BOOLEAN Try to share physical storage of the message. Only available for a limited number of folder types, otherwise no-op. =back example: my $message = Mail::Message->new(...); Mail::Box::Mbox->appendMessages ( folder => '=xyz' , message => $message , folderdir => $ENV{FOLDERS} ); better: my Mail::Box::Manager $mgr; $mgr->appendMessages($message, folder => '=xyz'); =item $obj-EB(%options) Close the folder, which usually implies writing the changes. This will return C when writing is required but fails. Please do check this result. WARNING: When moving messages from one folder to another, be sure to write the destination folder before writing and closing the source folder. Otherwise you may lose data if the system crashes or if there are software problems. -Option --Default force save_deleted false write MODIFIED =over 2 =item force => BOOLEAN Override the L setting which was specified when the folder was opened. This option only has an effect if its value is TRUE. NOTE: Writing to the folder may not be permitted by the operating system, in which case even C will not help. =item save_deleted => BOOLEAN Do also write messages which where flagged to be deleted to their folder. The flag for deletion is conserved (when possible), which means that a re-open of the folder may remove the messages for real. See L. =item write => 'ALWAYS'|'NEVER'|'MODIFIED' Specifies whether the folder should be written. As could be expected, C means always (even if there are no changes), C means that changes to the folder will be lost, and C only saves the folder if there are any changes. =back example: my $f = $mgr->open('spam', access => 'rw') or die "Cannot open spam: $!\n"; $f->message(0)->delete if $f->messages; $f->close or die "Couldn't write $f: $!\n"; =item $obj-EB($folder, %options) Copy the folder's messages to a new folder. The new folder may be of a different type. -Option --Default delete_copied select 'ACTIVE' share subfolders =over 2 =item delete_copied => BOOLEAN Flag the messages from the source folder to be deleted, just after it was copied. The deletion will only take effect when the originating folder is closed. =item select => 'ACTIVE'|'DELETED'|'ALL'|LABEL|!LABEL|FILTER Which messages are to be copied. See the description of L about how this works. =item share => BOOLEAN Try to share the message between the folders. Some L folder types do support it by creating a hardlink (on UNIX/Linux). =item subfolders => BOOLEAN|'FLATTEN'|'RECURSE' How to handle sub-folders. When false (C<0> or C), sub-folders are simply ignored. With C, messages from sub-folders are included in the main copy. C recursively copies the sub-folders as well. By default, when the destination folder supports sub-folders C is used, otherwise C. A value of true will select the default. =back example: my $mgr = Mail::Box::Manager->new; my $imap = $mgr->open(type => 'imap', host => ...); my $mh = $mgr->open(type => 'mh', folder => '/tmp/mh', create => 1, access => 'w'); $imap->copyTo($mh, delete_copied => 1); $mh->close; $imap->close; =item $obj-EB(%options) Remove the specified folder file or folder directory (depending on the type of folder) from disk. Of course, THIS IS DANGEROUS: you "may" lose data. Returns a C value on success. WARNING: When moving messages from one folder to another, be sure to write the destination folder before deleting the source folder. Otherwise you may lose data if the system crashes or if there are software problems. -Option --Default recursive 1 =over 2 =item recursive => BOOLEAN =back example: removing an open folder my $folder = Mail::Box::Mbox->new(folder => 'InBox', access => 'rw'); ... some other code ... $folder->delete; example: removing an closed folder my $folder = Mail::Box::Mbox->new(folder => 'INBOX', access => 'd'); $folder->delete; =item $obj-EB( [$directory] ) Get or set the $directory which is used to store mail-folders by default. example: print $folder->folderdir; $folder->folderdir("$ENV{HOME}/nsmail"); =item $obj-EB() Returns the name of the folder. What the name represents depends on the actual type of mailbox used. example: print $folder->name; print "$folder"; # overloaded stringification =item $obj-EB() Returns how the folder is organized: as one C with many messages, a C with one message per file, or by a C server. =item $obj-EB() Returns the size of the folder in bytes, not counting in the deleted messages. The error in the presented result may be as large as 10%, because the in-memory representation of messages is not always the same as the size when they are written. =item $obj-EB() Returns a name for the type of mail box. This can be C, C, C, or C. =item $obj-EB(%options) Read new messages from the folder, which where received after opening it. This is quite dangerous and shouldn't be possible: folders which are open are locked. However, some applications do not use locks or the wrong kind of locks. This method reads the changes (not always failsafe) and incorporates them in the open folder administration. The %options are extra values which are passed to the L method which is doing the actual work here. =item $obj-EB() Represent the folder as a URL (Universal Resource Locator) string. You may pass such a URL as folder name to L. example: print $folder->url; # may result in # mbox:/var/mail/markov or # pop3://user:password@pop.aol.com:101 =back =head2 Folder flags =over 4 =item $obj-EB() Returns the access mode of the folder, as set by L =item $obj-EB() Checks if the folder, as stored in memory, is modified. A true value is returned when any of the messages is to be deleted, has changed, or messages were added after the folder was read from file. WARNING: this flag is not related to an external change to the folder structure on disk. Have a look at L for that. =item $obj-EB( [BOOLEAN] ) Sets whether the folder is modified or not. =item $obj-EB() Checks whether the current folder is writable. example: $folder->addMessage($msg) if $folder->writable; =back =head2 The messages =over 4 =item $obj-EB( [$number|$message|$message_id] ) Some mail-readers keep the I message, which represents the last used message. This method returns [after setting] the current message. You may specify a $number, to specify that that message number is to be selected as current, or a $message/$message_id (as long as you are sure that the header is already loaded, otherwise they are not recognized). example: $folder->current(0); $folder->current($message); =item $obj-EB($message_id) Like L, this method searches for a message with the $message_id, returning the corresponding message object. However, C will cause unparsed message in the folder to be parsed until the message-id is found. The folder will be scanned back to front. =item $obj-EB( $label, [BOOLEAN, [$msgs]] ) Find the first message which has this $label with the correct setting. The BOOLEAN indicates whether any true value or any false value is to be found in the ARRAY of $msgs. By default, a true value is searched for. When a message does not have the requested label, it is taken as false. example: looking for a labeled message my $current = $folder->findFirstLabeled('current'); my $first = $folder->findFirstLabeled(seen => 0); my $last = $folder->findFirstLabeled(seen => 0, [ reverse $self->messages('ACTIVE') ] ) =item $obj-EB( $index, [$message] ) Get or set a message with on a certain index. Messages which are flagged for deletion are counted. Negative indexes start at the end of the folder. example: my $msg = $folder->message(3); $folder->message(3)->delete; # status changes to `deleted' $folder->message(3, $msg); print $folder->message(-1); # last message. =item $obj-EB( $message_id, [$message] ) With one argument, returns the message in the folder with the specified $message_id. If a reference to a message object is passed as the optional second argument, the message is first stored in the folder, replacing any existing message whose message ID is $message_id. (The message ID of $message need not match $message_id.) !!WARNING!!: when the message headers are delay-parsed, the message might be in the folder but not yet parsed into memory. In this case, use L instead of C if you really need a thorough search. This is especially the case for directory organized folders without special indexi, like L. The $message_id may still be in angles, which will be stripped. In that case blanks (which origin from header line folding) are removed too. Other info around the angles will be removed too. example: my $msg = $folder->messageId(''); $folder->messageId("", $msg); my $msg = $folder->messageId('complex-message.id'); my $msg = $folder->messageId('garbage trash'); =item $obj-EB() Returns a list of I message-ids in the folder, including those of messages which are to be deleted. For some folder-types (like MH), this method may cause all message-files to be read. See their respective manual pages. example: foreach my $id ($folder->messageIds) { $folder->messageId($id)->print; } =item $obj-EB( <'ALL'|$range|'ACTIVE'|'DELETED'|$label| !$label|$filter> ) Returns multiple messages from the folder. The default is C which will return (as expected maybe) all the messages in the folder. The C flag will return the messages not flagged for deletion. This is the opposite of C, which returns all messages from the folder which will be deleted when the folder is closed. You may also specify a $range: two numbers specifying begin and end index in the array of messages. Negative indexes count from the end of the folder. When an index is out-of-range, the returned list will be shorter without complaints. Everything else than the predefined names is seen as labels. The messages which have that label set will be returned. When the sequence starts with an exclamation mark (!), the search result is reversed. For more complex searches, you can specify a $filter, which is simply a code reference. The message is passed as only argument. example: foreach my $message ($folder->messages) {...} foreach my $message (@$folder) {...} # twice the same my @messages = $folder->messages; my @messages = $folder->messages('ALL'); # Selection based on a range (begin, end) my $subset = $folder->messages(10,-8); # twice the same: my @not_deleted= grep {not $_->isDeleted} $folder->messages; my @not_deleted= $folder->messages('ACTIVE'); # scalar context the number of messages my $nr_of_msgs = $folder->messages; # third message, via overloading $folder->[2]; # Selection based on labels $mgr->moveMessages($spam, $inbox->message('spam')); $mgr->moveMessages($archive, $inbox->message('seen')); =item $obj-EB(%options) Simply calls L in scalar context to return a count instead of the messages itself. Some people seem to understand this better. Note that nrMessages() will default to returning a count of C messages in the folder, including both C and C. The %options are passed to (and explained in) L. =item $obj-EB($message, $message_ids, $timespan, $window) You start with a $message, and are looking for a set of messages which are related to it. For instance, messages which appear in the 'In-Reply-To' and 'Reference' header fields of that message. These messages are known by their $message_ids and you want to find them in the folder. When all message-ids are known, then looking-up messages is simple: they are found in a plain hash using L. But Mail::Box is lazy where it can, so many messages may not have been read from file yet, and that's the preferred situation, because that saves time and memory. It is not smart to search for the messages from front to back in the folder: the chances are much higher that related message reside closely to each other. Therefore, this method starts scanning the folder from the specified $message, back to the front of the folder. The $timespan can be used to terminate the search based on the time enclosed in the message. When the constant string C is used as $timespan, then the search is not limited by that. When an integer is specified, it will be used as absolute time in time-ticks as provided by your platform dependent C