source: npl/mailserver/cyrus-dspamlearn/cyrus_dspamlearn.pl @ b4abfab

#!/usr/bin/perl -w
#Syn-3 spam learning cronjob - (C) 2005 DatuX 
#Edwin Eefting

use strict;
my $maildir="/home/system/cyrus-imap/maildir/user";
my $mailcachefile="/home/system/cyrus-dspam/mails.cache";
my $learnedfile="/home/system/cyrus-dspam/learn.cache";
my $test=0;

if ($ARGV[0])
{
    $test=1;
    print "TEST MODE - Not changing anything!\n";
}


#read mailcache, so that we know what's processed
my %mailcache;
open (MAILCACHE,$mailcachefile);
while (<MAILCACHE>)
{
        chomp();        
        #1=processed before, not checked this time after processing
        #2=processed before, and still there
        $mailcache{$_}=1;
}
close (MAILCACHE);

#read learning list, so that we don't learn something twice and
#can unlearn stuff as well
my %learned;
open (LEARNED,$learnedfile);
while (<LEARNED>)
{
        chomp();
        my ($signature,$time)=split(/=/,$_);    
        $learned{$signature}=$time;
        #$learned{$signature}{'mode'}=$mode;
}
close (LEARNED);


#learn by correcting a message with dspam
sub Learn
{
        my ($class,$user,$signature)=@_;
        $user=~ s/\'/_/g;
        $signature=~ s/\'/_/g;
        $class=~ s/\'/_/g;
        if (!exists($learned{$signature}))
        {
                system("dspam --mode=teft --source=error --user '$user' --signature='$signature' --class='$class'");
        }
        else
        {
                print "(Skipping, already learned)\n";
        }
        $learned{$signature}=time();
};


open (USERS,"find '$maildir' -maxdepth 1 -printf '%f\n'|") or die ;
my $user;
while ($user=<USERS>)
{
        chomp ($user);
        if (-e "$maildir/$user/Spam")
        {
                print "Processing user $user\n"; 
        }
        else
        {
                print "Skipping user $user (no Spam folder)\n";
                next;
        }
        #find all mails, but skip trash and send mails
        open (MAILS,"find '$maildir/$user' -type f -name '*.'|grep -v '/Trash/'|");
        my $mail;
        my $reconstruct=0;
        while ($mail=<MAILS>)
        {
                chomp($mail);
                #not yet processed?
                #TODO: only process if the mail is in the folder for more than 24 hours?
                #(in case the user made a mistake and can move it back within 24 hours)
                if (! $mailcache{$mail})
                {
                        #read first 5k of mail
                        open (MAILDATA,$mail);
                        my $maildata;
                        read(MAILDATA,$maildata,5000);
                        my ($headercheck,$spamresult)=($maildata=~ /(.*)\nX-DSPAM-Result: ([a-zA-Z]*)\r/s);
                        my ($signature)=($maildata=~ /.*\nX-DSPAM-Signature: ([0-9a-zA-Z]*)\r/s);
                        if ($spamresult && !($headercheck=~/\r\n\r\n/))
                        {
                                if ($test)
                                {
                                        print "Test: Found $mail ($signature) result $spamresult\n";
                                }
                                #is it NOT spam, in the spamfolder?
                                if ($spamresult ne 'Spam' && $mail=~/Spam/)
                                {
                                        if ($test)
                                        {
                                                print "Test: Would have corrected to SPAM: $mail ($signature)\n";
                                        }
                                        else
                                        {
                                                #learn as spam 
                                                print "Correcting to SPAM: $mail ($signature)\n";
                                                Learn("spam",$user,$signature);
                                                #delete it 
                                                unlink("$mail");
                                                #mailbox needs reconstruction because of our action
                                                $reconstruct=1;
                                        }
                                }
                                #is it spam, but NOT in spamfolder?
                                elsif ($spamresult eq 'Spam' && ! ($mail=~/Spam/))
                                {
                                        if ($test)
                                        {
                                                print "Test: Would have corrected to INNOCENT: $mail ($signature)\n";
                                        }
                                        else
                                        {
                                                #learn as innocent
                                                print "Correcting to INNOCENT: $mail ($signature)\n";
                                                Learn("innocent",$user,$signature);
                                        }
                                }
                                #other combinations
                                else
                                {
                                        #TODO: program could be extended to UNLEARN mails that are
                                }
                        }
                        else
                        {       
                                #unkown, just ignore it to be safe
                                if ($test)
                                {
                                        print "Test: Cant find headers in $mail?\n";
                                }
                        }
                        
                }
                #mark as found and processed
                $mailcache{$mail}=2;
        }
        close(MAILS);
        
        if (!$test && $reconstruct)
        {
                print "Reconstructing spamfolder of $user...\n";
                system ("su - cyrus -c '/usr/cyrus/bin/reconstruct user.$user.Spam'");
        }

}


if ($test)
{
        print "Test mode, not storing caches\n";
        exit;
}

#store new mailcache
open (MAILCACHE,">$mailcachefile");
foreach my $mail (keys(%mailcache))
{
        #2=mail processed this time
        if ($mailcache{$mail}==2)
        {
                print MAILCACHE "$mail\n";
        }       
}
close (MAILCACHE);


#store new learned list
open (LEARNED,">$learnedfile");
foreach my $signature (keys(%learned))
{
        #TODO:remove old entries
        print LEARNED "$signature=$learned{$signature}\n";      
}
close (LEARNED);