def unmark_stopword(self, id, word):
if not c.user:
return redirect(url(controller='login', action='signin', id=None, return_to=url.current()))
w = meta.Session.query(model.Stopword).filter_by(user_id=c.user.id, feed_id=id, word=word).first()
if not w:
h.flash("can't remove '%s', because it is not on the list." % word)
log.debug("w: %s" % w)
meta.Session.delete(w)
from sqlalchemy.exceptions import IntegrityError
meta.Session.commit()
h.flash("removed '%s' from stopwords list." % word)
return self.redo(id)
def change_intervall(self, id):
word = request.params.get('word')
settings = meta.Session\
.query(model.BayesFeedSetting)\
.filter_by(user_id = c.user.id, feed_id=id).first()
if settings:
meta.Session.add(settings)
else:
settings = model.BayesFeedSetting()
settings.user_id = c.user.id
settings.feed_id = id
meta.Session.add(settings)
settings.summarize_at = word
meta.Session.commit()
h.flash('changed intervall to %s' % word)
return h.go_back()
def __mark_as__(self, entry, pool, guesser, force=False):
""" when forced the entry is updated even if the db says it is already """
log.debug("entry.id: %s" % entry.id)
classy = meta.Session\
.query(model.Classification)\
.filter_by(user_id = c.user.id, entry_id=entry.id).first()
if not classy:
classy = model.Classification()
classy.user_id = c.user.id
classy.entry_id = entry.id
classy.pool = pool
meta.Session.add(classy)
untrain_id = None
else:
if classy.pool == pool and not force:
h.flash("entry was already classified as %s" % pool)
return h.go_back(h.url_for(controller='feed', action='show_feed', id=entry.feed_id))
classy.pool = pool
meta.Session.add(classy)
untrain_id = entry.id
meta.Session.commit()
guesser.trainer.train(pool, __relevant__(entry), entry.id)
if pool == 'spam':
other_pool = 'ham'
elif pool == 'ham':
other_pool = 'spam'
else:
raise "bad pool"
# if untraind_id:
# guesser.trainer.untrain(other_pool, __relevant__(entry), untrain_id)
guesser.save()
if not force:
h.flash("now known as %s: %s" % (pool, entry.id))
return h.go_back(h.url_for(controller='feed', action='show_feed', id=entry.feed_id))
def mark_stopword(self, id, word):
if not c.user:
return redirect(url(controller='login', action='signin', id=None, return_to=url.current()))
w = model.Stopword()
w.user_id = c.user.id
w.feed_id = id
w.word = word
meta.Session.add(w)
from sqlalchemy.exceptions import IntegrityError
try:
meta.Session.commit()
h.flash("%s? Never heard about it." % word)
except IntegrityError:
h.flash("i already know that '%s' is a stop word." % word)
meta.Session.rollback()
return self.redo(id)
def redo(self, id):
if not c.user:
return redirect(url(controller='login', action='signin', id=None, return_to=url.current()))
c.feed = meta.find(model.Feed, id)
query = meta.Session\
.query(model.Classification)\
.join(model.FeedEntry)\
.filter_by(feed_id=id)
guesser = Guesser(c.feed, c.user, config)
guesser.clear()
cnt = 0
needles_cnt = 0
for entry in query:
# h.flash("%s :%s" % (entry.pool, __relevant__(entry.entry)))
# guesser.trainer.train(entry.pool, __relevant__(entry.entry))
if guesser.is_spam(entry.entry, use_classified=False) and (entry.pool == 'spam'):
needles_cnt += 1
elif not guesser.is_spam(entry.entry, use_classified=False) and (entry.pool == 'ham'):
needles_cnt += 1
self.__mark_as__(entry.entry, entry.pool, guesser, True)
cnt+=1
guesser.save()
log.debug("FOOOOOO")
if needles_cnt > 0:
h.flash("%d entries were needlessly trained (total: %s)" % (needles_cnt, cnt))
else:
h.flash("learned %s entries" % cnt)
return h.go_back()
def update(self, id):
#~ feed = meta.Session.query(model.Feed).get(id)
feed = meta.find(model.Feed, id)
cnt_added = feed.fetch()
h.flash("added %s entries" % cnt_added)
return h.go_back()