[cvs] bogofilter/src score.c,1.44,1.45
Matthias Andree
matthias.andree at gmx.de
Mon Jan 17 09:59:26 CET 2005
- Previous message (by thread): [cvs] bogofilter/src score.c,1.44,1.45
- Next message (by thread): [cvs] bogofilter/src common.h, 1.51, 1.52 datastore_db.c, 1.158, 1.159 datastore_olddb.c, 1.17, 1.18 datastore_tdb.c, 1.42, 1.43 db_lock.c, 1.7, 1.8
- Messages sorted by:
[ date ]
[ thread ]
[ subject ]
[ author ]
David Relson <relson at osagesoftware.com> writes:
> Tokens in message count files have the proper counts. When using a
> message count file, it's a mistake to look for the token in the
> database.
I have committed a cleanup that removes the "== 0" checks. We don't want
to _guess_ while we actually know. I still don't like the "message count
format input" flag checks down deep. This is actually a higher level
decision to make.
IMO the lookup() needs to move out of a function the name of which
claims to calculate probability from a cnt structure. The actual problem
I'm having with that is that we cannot say if this low-level function
uses "cnt" as input or output.
OK, actually we can but what does a calculate_probability function care
where the input comes from? The cash register doesn't care either if the
price comes from a red "SALE" price label on the package or from a
barcode lookup. More cleanups coming up.
--
Matthias Andree
- Previous message (by thread): [cvs] bogofilter/src score.c,1.44,1.45
- Next message (by thread): [cvs] bogofilter/src common.h, 1.51, 1.52 datastore_db.c, 1.158, 1.159 datastore_olddb.c, 1.17, 1.18 datastore_tdb.c, 1.42, 1.43 db_lock.c, 1.7, 1.8
- Messages sorted by:
[ date ]
[ thread ]
[ subject ]
[ author ]
More information about the bogofilter-dev
mailing list