On Sat, Jun 07, 2008 at 01:33:39PM +0200, Olav Vitters wrote: > Note: I plan to integrate the discarded / accepted messages with > Spamassassin (bayes stuff). This has been done. See the attachments for the cron script + mailman patch. The ~/.procmailrc is trivial, not attaching that. -- Regards, Olav
diff -ru q/mailman-2.1.10/Mailman/Cgi/admindb.py mailman-2.1.10/Mailman/Cgi/admindb.py
--- q/mailman-2.1.10/Mailman/Cgi/admindb.py 2008-04-21 20:28:22.000000000 +0200
+++ mailman-2.1.10/Mailman/Cgi/admindb.py 2008-06-07 22:32:45.000000000 +0200
@@ -197,6 +197,11 @@
' ' +
_('Discard all messages marked <em>Defer</em>')
))
+ form.AddItem(Center(
+ CheckBox('trainbayesp', 0).Format() +
+ ' ' +
+ _('Use messages marked as <em>Discard</em> and <em>Accept</em> to train the Bayesian classifier')
+ ))
# Add a link back to the overview, if we're not viewing the overview!
adminurl = mlist.GetScriptURL('admin', absolute=1)
d = {'listname' : mlist.real_name,
@@ -245,6 +250,11 @@
' ' +
_('Discard all messages marked <em>Defer</em>')
))
+ form.AddItem(Center(
+ CheckBox('trainbayesp', 0).Format() +
+ ' ' +
+ _('Use messages marked as <em>Defer</em> and <em>Accept</em> to train the Bayesian classifier')
+ ))
form.AddItem(Center(SubmitButton('submit', _('Submit All Data'))))
doc.AddItem(mlist.GetMailmanFooter())
print doc.Format()
@@ -695,6 +705,10 @@
discardalldefersp = cgidata.getvalue('discardalldefersp', 0)
except ValueError:
discardalldefersp = 0
+ try:
+ trainbayesp = cgidata.getvalue('trainbayesp', 0)
+ except ValueError:
+ trainbayesp = 0
for sender in senderactions.keys():
actions = senderactions[sender]
# Handle what to do about all this sender's held messages
@@ -709,6 +723,13 @@
preserve = actions.get('senderpreserve', 0)
forward = actions.get('senderforward', 0)
forwardaddr = actions.get('senderforwardto', '')
+ if trainbayesp and not forward:
+ if action == mm_cfg.DISCARD:
+ forward = 1
+ forwardaddr = 'XXX gnome.org'
+ elif action == mm_cfg.APPROVE:
+ forward = 1
+ forwardaddr = 'XXX gnome.org'
comment = _('No reason given')
bysender = helds_by_sender(mlist)
for id in bysender.get(sender, []):
Attachment:
bayes.png
Description: PNG image
#!/bin/sh # # Mailman has been patched to forward messages marked as Dischard + Accept to # the amavis user. # # It uses plussed addresses for that. # # There is a procmail rule which sends the body of that email to two mailboxes. # The Mailman email has content type message/rfc822. Meaning: the original # message is preserved by only sending the body to the mbox. # # This scripts runs in cron and uses the mbox to train the bayes classifier. # cd $HOME type="spam" if [ -e bayes-$type/new ]; then rm -rf bayes-$type/train mv -f bayes-$type/new bayes-$type/train sa-learn --$type bayes-$type/train fi type="ham" if [ -e bayes-$type/new ]; then rm -rf bayes-$type/train mv -f bayes-$type/new bayes-$type/train sa-learn --$type bayes-$type/train fi
