A couple people liked the graph I produced that other day and asked
for info, so here it is.

In .procmailrc *after SA has touched the mail* I added:
--
SPAM_STATUS=`grep X-Spam-Status $i |cut -d' ' -f3|cut -d'=' -f2 |sort -n`
MESSAGE_ID=`formail -xMessage-ID:: \
| sed -e 's/[;\`\\]/ /g' \
| expand | sed -e 's/^[ ]*//g' -e 's/[ ]*$//g'`

DATE=`date +"%m/%d/%y %H:%M"`
nada=`echo $DATE $SPAM_STATUS $MESSAGE_ID >> $HOME/email-scores`
--

I couldn't execute a command in procmail except by using the kludge
'nada' as an assignment.  If there is a better way, let me know.

I use gnuplot for the plotting.  The command file looks like this:
--
set timefmt "%m/%d/%y %H:%M"
set format x "%m/%d"
set title "Distribution of spam values"
set xlabel "Date mail arrived"
set ylabel "Hit value"
set nolabel
set xdata time
set terminal png color
set yrange [0:50]
plot '/home/wellner/email-scores' using 1:3 with points 20, 5 with lines 3, 7 with 
lines 7, 10 with li\nes 5
--

The result is at: http://wellner.org/~wellner/email-scores.png

I only just started keeping scores for the non-spam messages a few
minutes ago so there isn't much data below threshold 5 yet.  There is
currently no facility for marking false negs/pos's.

rw2


_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to