view website/issues/extensions/spambayes.py @ 5222:9bf221cebef3

Make properties method return only properties the user can search. See: https://sourceforge.net/p/roundup/mailman/roundup-devel/thread/20170405002844.2004B80690%40vm71.cs.umb.edu/#msg35769250 [Roundup-devel] Bug in context/properties, lists properties user can't search. The HTMLClass::properties() method returns a list of all properties. This is used when creating sort on/group by filters on index pages. However somewhere in the code, a user needs search permission on the property in order for it to be used for grouping or sorting. This means the user can choose to sort/group an index page by a property that they have no search permission for. As a result the sort/group is ignored. This is confusing. I have changed the properties method to only return properties the user has View/Search permissions on. I also added a new cansearch argument set by default to True. If set to False, all properties regardless of Search permission are returned. Doc updated to include the new default operation and mention the use of cansearch argument.
author John Rouillard <rouilj@ieee.org>
date Wed, 05 Apr 2017 21:38:32 -0400
parents ca692423e401
children 198b6e810c67
line wrap: on
line source

import re, math
from roundup.cgi.actions import Action
from roundup.cgi.exceptions import *

import xmlrpclib, socket

REVPAT = re.compile(r'(r[0-9]+\b|rev(ision)? [0-9]+\b)')

def extract_classinfo(db, classname, nodeid):
    node = db.getnode(classname, nodeid)

    authorage = node['creation'].timestamp() - \
                db.getnode('user', node.get('author', node.get('creator')))['creation'].timestamp()

    authorid = node.get('author', node.get('creator'))

    content = db.getclass(classname).get(nodeid, 'content')

    tokens = ["klass:%s" % classname,
              "author:%s" % authorid,
              "authorage:%d" % int(math.log(authorage)),
              "hasrev:%s" % (REVPAT.search(content) is not None)]

    return (content, tokens)

def train_spambayes(db, content, tokens, is_spam):
    spambayes_uri = db.config.detectors['SPAMBAYES_URI']

    server = xmlrpclib.ServerProxy(spambayes_uri, verbose=False)
    try:
        server.train({'content':content}, tokens, {}, is_spam)
        return (True, None)
    except (socket.error, xmlrpclib.Error), e:
        return (False, str(e))


class SpambayesClassify(Action):
    permissionType = 'SB: May Classify'
    
    def handle(self):
        (content, tokens) = extract_classinfo(self.db,
                                              self.classname, self.nodeid)

        if self.form.has_key("trainspam"):
            is_spam = True
        elif self.form.has_key("trainham"):
            is_spam = False

        (status, errmsg) = train_spambayes(self.db, content, tokens,
                                           is_spam)

        node = self.db.getnode(self.classname, self.nodeid)
        props = {}

        if status:
            if node.get('spambayes_misclassified', False):
                props['spambayes_misclassified'] = True

            props['spambayes_score'] = 1.0
            
            s = " SPAM"
            if not is_spam:
                props['spambayes_score'] = 0.0
                s = " HAM"
            self.client.add_ok_message(self._('Message classified as') + s)
        else:
            self.client.add_error_message(self._('Unable to classify message, got error:') + errmsg)

        klass = self.db.getclass(self.classname)
        klass.set(self.nodeid, **props)
        self.db.commit()

def sb_is_spam(obj):
    cutoff_score = float(obj._db.config.detectors['SPAMBAYES_SPAM_CUTOFF'])
    try:
        score = obj['spambayes_score']
    except KeyError:
        return False
    return score >= cutoff_score

def init(instance):
    instance.registerAction("spambayes_classify", SpambayesClassify)
    instance.registerUtil('sb_is_spam', sb_is_spam)
    

Roundup Issue Tracker: http://roundup-tracker.org/