227 lines
6.4 KiB
Python
227 lines
6.4 KiB
Python
# Copyright (C) 2006 Jean-Marie Traissard <jim AT lapin.org>
|
|
# Tomasz Melcer <liori AT exroot.org>
|
|
# Copyright (C) 2006-2014 Yann Leboulanger <asterix AT lagaule.org>
|
|
#
|
|
# This file is part of Gajim.
|
|
#
|
|
# Gajim is free software; you can redistribute it and/or modify
|
|
# it under the terms of the GNU General Public License as published
|
|
# by the Free Software Foundation; version 3 only.
|
|
#
|
|
# Gajim is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
# GNU General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU General Public License
|
|
# along with Gajim. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
"""
|
|
Atom (rfc 4287) feed parser, used to read data from atom-over-pubsub transports
|
|
and services. Very simple. Actually implements only atom:entry.
|
|
Implement more features if you need
|
|
"""
|
|
|
|
# XEP-0277: Microblogging over XMPP
|
|
|
|
# Module is disabled for now because Gajim lacks a good UI for that
|
|
# register the module in connection.py with register_module() to activate again
|
|
|
|
import logging
|
|
import time
|
|
|
|
import nbxmpp
|
|
|
|
from gajim.common.const import PEPEventType
|
|
from gajim.common.exceptions import StanzaMalformed
|
|
from gajim.common.modules.pep import AbstractPEPModule, AbstractPEPData
|
|
|
|
log = logging.getLogger('gajim.c.m.atom')
|
|
|
|
|
|
class AtomData(AbstractPEPData):
|
|
|
|
type_ = PEPEventType.ATOM
|
|
|
|
def __init__(self, atom):
|
|
self.data = atom
|
|
|
|
def get_entry(self):
|
|
return self.data
|
|
|
|
|
|
class Atom(AbstractPEPModule):
|
|
|
|
name = 'atom'
|
|
namespace = 'urn:xmpp:microblog:0'
|
|
pep_class = AtomData
|
|
store_publish = False
|
|
_log = log
|
|
|
|
def _extract_info(self, item):
|
|
entry = item.getTag('entry', namespace=nbxmpp.NS_ATOM)
|
|
if entry is None:
|
|
StanzaMalformed('no entry node')
|
|
|
|
return OldEntry(node=entry) or None
|
|
|
|
def _build_node(self, data):
|
|
raise NotImplementedError
|
|
|
|
|
|
class PersonConstruct(nbxmpp.Node):
|
|
"""
|
|
Not used for now, as we don't need authors/contributors
|
|
in pubsub.com feeds. They rarely exist there
|
|
"""
|
|
|
|
def __init__(self, node):
|
|
''' Create person construct from node. '''
|
|
nbxmpp.Node.__init__(self, node=node)
|
|
|
|
def get_name(self):
|
|
return self.getTagData('name')
|
|
|
|
name = property(
|
|
get_name, None, None,
|
|
'''Conveys a human-readable name for the person. Should not be None,
|
|
although some badly generated atom feeds don't put anything here
|
|
(this is non-standard behavior, still pubsub.com sometimes
|
|
does that.)''')
|
|
|
|
def get_uri(self):
|
|
return self.getTagData('uri')
|
|
|
|
uri = property(
|
|
get_uri, None, None,
|
|
'''Conveys an IRI associated with the person.
|
|
Might be None when not set.''')
|
|
|
|
def get_email(self):
|
|
return self.getTagData('email')
|
|
|
|
email = property(
|
|
get_email, None, None,
|
|
'''Conveys an e-mail address associated with the person.
|
|
Might be None when not set.''')
|
|
|
|
|
|
class Entry(nbxmpp.Node):
|
|
def __init__(self, node=None):
|
|
nbxmpp.Node.__init__(self, 'entry', node=node)
|
|
|
|
def __repr__(self):
|
|
return '<Atom:Entry object of id="%r">' % self.getAttr('id')
|
|
|
|
|
|
class OldEntry(nbxmpp.Node):
|
|
"""
|
|
Parser for feeds from pubsub.com. They use old Atom 0.3 format with their
|
|
extensions
|
|
"""
|
|
|
|
def __init__(self, node=None):
|
|
''' Create new Atom 0.3 entry object. '''
|
|
nbxmpp.Node.__init__(self, 'entry', node=node)
|
|
|
|
def __repr__(self):
|
|
return '<Atom0.3:Entry object of id="%r">' % self.getAttr('id')
|
|
|
|
def get_feed_title(self):
|
|
"""
|
|
Return title of feed, where the entry was created.
|
|
The result is the feed name concatenated with source-feed title
|
|
"""
|
|
if self.parent is not None:
|
|
main_feed = self.parent.getTagData('title')
|
|
else:
|
|
main_feed = None
|
|
|
|
if self.getTag('feed') is not None:
|
|
source_feed = self.getTag('feed').getTagData('title')
|
|
else:
|
|
source_feed = None
|
|
|
|
if main_feed is not None and source_feed is not None:
|
|
return '%s: %s' % (main_feed, source_feed)
|
|
if main_feed is not None:
|
|
return main_feed
|
|
if source_feed is not None:
|
|
return source_feed
|
|
return ''
|
|
|
|
feed_title = property(
|
|
get_feed_title, None, None,
|
|
''' Title of feed. It is built from entry''s original feed title
|
|
and title of feed which delivered this entry. ''')
|
|
|
|
def get_feed_link(self):
|
|
"""
|
|
Get source link
|
|
"""
|
|
try:
|
|
link = self.getTag('feed').getTags('link', {'rel': 'alternate'})
|
|
return link[1].getData()
|
|
except Exception:
|
|
return None
|
|
|
|
feed_link = property(
|
|
get_feed_link, None, None,
|
|
''' Link to main webpage of the feed. ''')
|
|
|
|
def get_title(self):
|
|
"""
|
|
Get an entry's title
|
|
"""
|
|
return self.getTagData('title')
|
|
|
|
title = property(
|
|
get_title, None, None,
|
|
''' Entry's title. ''')
|
|
|
|
def get_uri(self):
|
|
"""
|
|
Get the uri the entry points to (entry's first link element with
|
|
rel='alternate' or without rel attribute)
|
|
"""
|
|
for element in self.getTags('link'):
|
|
if 'rel' in element.attrs and element.attrs['rel'] != 'alternate':
|
|
continue
|
|
try:
|
|
return element.attrs['href']
|
|
except AttributeError:
|
|
pass
|
|
return None
|
|
|
|
uri = property(
|
|
get_uri, None, None,
|
|
''' URI that is pointed by the entry. ''')
|
|
|
|
def get_updated(self):
|
|
"""
|
|
Get the time the entry was updated last time
|
|
|
|
This should be standarized, but pubsub.com sends it in human-readable
|
|
format. We won't try to parse it.
|
|
(Atom 0.3 uses the word «modified» for that).
|
|
|
|
If there's no time given in the entry, we try with <published>
|
|
and <issued> elements.
|
|
"""
|
|
for name in ('updated', 'modified', 'published', 'issued'):
|
|
date = self.getTagData(name)
|
|
if date is not None:
|
|
break
|
|
|
|
if date is None:
|
|
# it is not in the standard format
|
|
return time.asctime()
|
|
|
|
return date
|
|
|
|
updated = property(
|
|
get_updated, None, None,
|
|
''' Last significant modification time. ''')
|
|
|
|
feed_tagline = ''
|