sb_evoscore.py

用python实现的邮件过滤器

#! /usr/bin/env python

# Copyright (C) 2003 Python Software Foundation; All Rights Reserved
#
# Licensed under the Python Software Foundation License, which you should have
# received as part of the Spambayes distribution.
#
# Author: Barry Warsaw <[e-mail]>

"""A shim for integrating Spambayes and Ximian Evolution.

Evolution is a free email client for Linux and Solaris, developed by Ximian.
See www.ximian.com for details.  Evolution is sometimes called 'Evo' for
short.

Actually Evo is more than that -- you can think of it as an Outlook-alike for
Linux -- but all we care about here is the mail reader.  Evo can connect to a
POP or IMAP server, so you can of course use Spambayes' normal POP or IMAP
filters.

For a variety of reasons, I don't like hooking things up this way.  In Evo,
you can specify filters which run on folders whenever a message shows up in
that folder.  The filter can have any number of criteria, and if the criteria
match, Evo will execute some number of actions.  One of the things a filter
can do is pipe the message to a program's standard in, and then check the exit
code of that program.  That's how we'll hook things together.

You'll need to start sb_xmlrpcserver.py, as this script is a client of that
server.  You'll use sb_imapfilter.py to train a database on the machine local
to your Evo client (yes, if you use many different workstations, you'll need
your database on all of them).

sb_evoscore.py takes the message from standard in, sends it to the xmlrpc
server and receives the float spam score for the message.  Then it compares
this to your ham_cutoff and spam_cutoff options.  It exits with a return code
that your Evo filter will check.  The return codes are:

   -1 - Error
    0 - Ham
    1 - Unsure
    2 - Spam

So, to hook things up do the following:

- In Evo, go to Tools -> Filters... to bring up the filter rules

- Select 'incoming' as the filter rule direction

- Add a new filter rule called 'Spambayes Spam'

- Select 'Pipe Message to Shell Command' as the first and only criteria.
  Point the command at this script, e.g. /usr/local/bin/sb_evoscore.py

- Select the match criteria to be 'returns 2'

- In the 'Then' section, select what you want to have happen for Spam.
  Personally, I have a folder called SBInbox, and inside that folder I have
  four subfolders: HamTrain, SpamTrain, Spam, and Unsure.  My action for the
  'Spambayes Spam' filter is then 'Move to Folder SBInbox/Spam'.

- Now do the same thing with a second filter rule called 'Spambayes Unsure',
  except this time, match a return value of 1, and move these messages to
  SBInbox/Unsure.

- Finally, make sure Evo will run filters automatically when your inbox
  receives new messages.

Now, what I do is throw a bunch of known ham in SBInbox/HamTrain and a bunch
of known spam in SBInbox/SpamTrain.  I use 'sb_imapfilter -t -v -p' to train a
database on my local machine.  Then I start up sb_xmlrpcserver.py.

NOTE: you must edit the variable RPCURL below to match how you invoke
sb_xmlrpcserver.py.

For a while, I watch the Unsure folder, moving mistakes to SpamTrain and
HamTrain respectively.  Every once in a while, I retrain my database, and copy
my database to all my other desktops.  One caveat: I've found that if I kill
the xmlrpc server while Evo is still running, it can cause Evo to hang, choke,
or start spewing endless error messages.  It's best to exit Evo before killing
sb_xmlrpcserver.py.
"""

import sys
import xmlrpclib
from spambayes.Options import options

RPCURL = 'http://localhost:8881'


def main():
    msg = sys.stdin.read()
    try:
        server = xmlrpclib.ServerProxy(RPCURL)
        score = server.score(xmlrpclib.Binary(msg))
    except:
        import traceback
        traceback.print_exc()
        return -1
    else:
        if score < options['Categorization', 'ham_cutoff']:
            return 0
        elif score < options['Categorization', 'spam_cutoff']:
            return 1
        return 2


status = main()
sys.exit(status)