From: Paul Lesniewski <paul@sq...> - 2007-02-14 05:33:35
On 2/13/07, Tom Williams <tomdkat@...> wrote:
> Paul Lesniewski wrote:
> > On 2/13/07, Tom Williams <tomdkat@...> wrote:
> >> Once I got all that setup, spam_buttons was able to run sa-learn on a
> >> per-user basis. I have more testing to do to see if it's actually
> >> working or not but this will be at the Spamassassin level.
> > You can check the user totals for spam and ham after you click to
> > report one or the other. On the command line, run sa-learn with the
> > --dump option, although it is a long report usually, so pipe it
> > through head....
> > sa-learn --username=jose@... --dump | head
> Now, I don't have spam_buttons configured to use the --username option
> but to use the --dbpath option, to make sure it will use the bayes
> database stored in each user's mail directory. Since I'm using
> Qmail/Vpopmail, the user's mail is stored in:
> /home/vpopmail/domains/[whatever domain]/[whatever user]
> I pass "--dbpath /home/vpopmail/domains/###DOMAIN###/###USERNAME###" to
> sa-learn in spam_button's config.php and sa-learn does find the bayes
> database ok. I'm not sure how sa-learn will know where to look when
> using the --username option, but I'll give it a whirl and see what happens.
The --username option in my example was just that - an exmple. If you
don't have a per-user bayes store, you just need to give sa-learn what
command-line switches it needs to give you the information from the
target bayes store, so don't use --username if you don't need it. The
command will probably look much the same as your learning command but
with --dump and without input for any message to be learned.