[Spambayes] Outlook addin using bsddb
Tim Stone - Four Stones Expressions
tim at fourstonesExpressions.com
Fri Feb 7 07:20:24 EST 2003
2/6/2003 11:46:30 PM, "Mark Hammond" <mhammond at skippinet.com.au> wrote:
>[Resending - appears to have hit the bit-bucket!]
>I have just checked in a change to the Outlook plugin that will use a bsddb
>style database if a reasonable implementation can be found. Currently, a
>"reasonable" implementation means:
>* A bsddb3 module can be imported.
>* If Python is 2.3 or greater, bsddb can be imported (as this is the new
>So everyone using Python 2.2, and without bsddb3, should continue to use
>pickles and see no change. Anyone lucky enough to have a working bsddb will
>see a *huge* performance win when starting and stopping outlook. NOTE: I
>have only tested with python 2.3 + bsddb. I have not tried bsddb3.
>If you system picks up the bsddb implementation, then you will need to
>perform a complete retrain. There is no "migration" code in place.
You might well be able to use dbExpImp.py to export your database before your
upgrade, then import it afterwards. All will be well then, and it'll save you
a retrain. - TimS
>Regarding performance: A "full re-train" is about 10% slower using a DB.
>After a full retrain, the entire database needs to be flushed, so saving the
>initial bsddb file takes twice as long as saving the pickle.
>Subsequent startup times are then radically reduced. On my machine with
>~6000 messages, the pickles load in approx 2000ms, whereas the bsddb files
>load in approx 2ms. A huge win by any means. Subsequent shutdown times
>will depend on how much training has been done (ie, how many words need to
>be flushed to the DB), but is generally many many times faster.
>Please let me know if there are any problems, or any suggestions.
>Spambayes mailing list
>Spambayes at python.org
c'est moi - TimS
More information about the Spambayes