On Fri, Feb 27, 2004 at 10:49:03AM +0000, Chris Withers <[EMAIL PROTECTED]> wrote: > Matthew Hunter wrote: > >The maildir format stores messages in ./new and ./cur within the > >maildir. So, simply pass in both of those directories to > >sa-learn, rather than the top-level directory. This is how I > >have been doing it, and it seems to be working fine. > Hmm, I noticed it "learned" from two messages, even for empty maildirs, so > I suspect here may be a couple of non-message files lying in there :-S
The maildir spec allows metafiles to be stored in the top level directory. Many IMAP servers store message data indexes there, for example. > >Or perhaps even > >recursively, because I have a maildir heirarchy, and learning > >from everything in that requires a bit of shell magic. > Hear hear hear hear :-) FWIW, here's how I learn currently (learning from all read messages, except known spam, and mailing lists dealing with spam): find /home/matthew/Maildir/ -type d -name "cur" | grep -v -i spam | xargs --max-args=1 -t sa-learn --no-rebuild --ham sa-learn --showdots --no-rebuild --spam /home/matthew/Maildir/.Personal/.Spam/cur sa-learn --rebuild It would probably be more efficient to use a temporary store and delete the message after learning, rather than counting on SA to skip messages it has already seen, but I haven't thought that through yet. -- Matthew Hunter ([EMAIL PROTECTED]) Public Key: http://matthew.infodancer.org/public_key.txt Homepage: http://matthew.infodancer.org/index.jsp Politics: http://www.triggerfinger.org/index.jsp
