Mercurial > libervia-backend
view src/plugins/plugin_xep_0277.py @ 1055:abcac1ac27a7
plugin otr: first draft
author | Goffi <goffi@goffi.org> |
---|---|
date | Sat, 07 Jun 2014 16:39:08 +0200 |
parents | 301b342c697a |
children | 0b87d029f0a3 |
line wrap: on
line source
#!/usr/bin/python # -*- coding: utf-8 -*- # SAT plugin for microblogging over XMPP (xep-0277) # Copyright (C) 2009, 2010, 2011, 2012, 2013, 2014 Jérôme Poisson (goffi@goffi.org) # This program is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # This program 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 Affero General Public License for more details. # You should have received a copy of the GNU Affero General Public License # along with this program. If not, see <http://www.gnu.org/licenses/>. from sat.core.i18n import _ from sat.core.constants import Const as C from sat.core.log import getLogger log = getLogger(__name__) from twisted.words.protocols.jabber import jid from twisted.internet import defer from sat.core import exceptions from sat.tools.xml_tools import ElementParser from wokkel import pubsub from feed import atom, date from lxml import etree import uuid from time import time import urlparse from cgi import escape NS_MICROBLOG = 'urn:xmpp:microblog:0' NS_XHTML = 'http://www.w3.org/1999/xhtml' NS_PUBSUB = 'http://jabber.org/protocol/pubsub' PLUGIN_INFO = { "name": "Microblogging over XMPP Plugin", "import_name": "XEP-0277", "type": "XEP", "protocols": [], "dependencies": ["XEP-0163", "XEP-0060", "TEXT-SYNTAXES"], "main": "XEP_0277", "handler": "no", "description": _("""Implementation of microblogging Protocol""") } class NodeAccessChangeException(Exception): pass class XEP_0277(object): def __init__(self, host): log.info(_("Microblogging plugin initialization")) self.host = host self.host.plugins["XEP-0163"].addPEPEvent("MICROBLOG", NS_MICROBLOG, self.microblogCB, self.sendMicroblog) host.bridge.addMethod("getLastMicroblogs", ".plugin", in_sign='sis', out_sign='aa{ss}', method=self.getLastMicroblogs, async=True, doc={'summary': 'retrieve items', 'param_0': 'jid: publisher of wanted microblog', 'param_1': 'max_items: see XEP-0060 #6.5.7', 'param_2': '%(doc_profile)s', 'return': 'list of microblog data (dict)'}) host.bridge.addMethod("setMicroblogAccess", ".plugin", in_sign='ss', out_sign='', method=self.setMicroblogAccess, async=True, doc={}) def parseCommentUrl(self, node_url): """Determine the fields comments_service and comments_node of a microblog data from the href attribute of an entry's link element. For example this input: xmpp:sat-pubsub.libervia.org?node=urn%3Axmpp%3Acomments%3A_c5c4a142-2279-4b2a-ba4c-1bc33aa87634__urn%3Axmpp%3Agroupblog%3Asouliane%libervia.org will return (JID(u'sat-pubsub.libervia.org'), 'urn:xmpp:comments:_c5c4a142-2279-4b2a-ba4c-1bc33aa87634__urn:xmpp:groupblog:souliane%libervia.org') @return: a tuple (JID, str) """ parsed_url = urlparse.urlparse(node_url, 'xmpp') service = jid.JID(parsed_url.path) queries = parsed_url.query.split(';') parsed_queries = dict() for query in queries: parsed_queries.update(urlparse.parse_qs(query)) node = parsed_queries.get('node', [''])[0] if not node: raise exceptions.DataError('Invalid comments link') return (service, node) def __removeXHTMLMarkups(self, xhtml): """ Remove XHTML markups from the given string. @param xhtml: the XHTML string to be cleaned @return: a Deferred instance for the cleaned string """ return self.host.plugins["TEXT-SYNTAXES"].convert(xhtml, self.host.plugins["TEXT-SYNTAXES"].SYNTAX_XHTML, self.host.plugins["TEXT-SYNTAXES"].SYNTAX_TEXT, False) @defer.inlineCallbacks def item2mbdata(self, item): """Convert an XML Item to microblog data used in bridge API @param item: domish.Element of microblog item @return: microblog data (dictionary)""" def xpath(elt, path): """Return the XPATH result of an entry element or its descendance, works with both: - no namespace, that means it is inherited from the parent item node --> NS_PUBSUB - empty namespace XXX: check why the received entries have no namespace when they are retrieved from self.host.plugins["XEP-0060"].getItems and they have an empty namespace when they are received with an event. """ result = elt.xpath(path) if len(result) > 0: return result return elt.xpath('/'.join(['ns:%s' % tag for tag in path.split('/')]), namespaces={'ns': NS_PUBSUB}) # convert a date string to float without dealing with the date format date2float = lambda elt, path: unicode(date.rfc3339.tf_from_timestamp(xpath(elt, path)[0].text)) item_elt = etree.fromstring(item.toXml().encode('utf-8')) try: entry_elt = xpath(item_elt, 'entry')[0] except IndexError: raise exceptions.DataError(_('No entry found in the pubsub item %s') % item_elt.get('id', '')) microblog_data = {} for key in ['title', 'content']: # process the textual elements for attr_elt in xpath(entry_elt, key): attr_content = self.__getLXMLInnerContent(attr_elt) if not attr_content.strip(): continue # element with empty value content_type = attr_elt.get('type', 'text').lower() if content_type == 'xhtml': text = self.__decapsulateExtraNS(attr_content) microblog_data['%s_xhtml' % key] = yield self.host.plugins["TEXT-SYNTAXES"].clean_xhtml(text) else: microblog_data[key] = attr_content if key not in microblog_data and ('%s_xhtml' % key) in microblog_data: microblog_data[key] = yield self.__removeXHTMLMarkups(microblog_data['%s_xhtml' % key]) try: # check for mandatory elements microblog_data['id'] = xpath(entry_elt, 'id')[0].text microblog_data['updated'] = date2float(entry_elt, 'updated') assert('title' in microblog_data) # has been processed already except IndexError: log.error(_("Atom entry %s misses a required element") % item_elt.get('id', '')) raise exceptions.DataError if 'content' not in microblog_data: # use the atom title data as the microblog body content microblog_data['content'] = microblog_data['title'] del microblog_data['title'] if 'title_xhtml' in microblog_data: microblog_data['content_xhtml'] = microblog_data['title_xhtml'] del microblog_data['title_xhtml'] # recommended and optional elements with a fallback value try: microblog_data['published'] = date2float(entry_elt, 'published') except IndexError: microblog_data['published'] = microblog_data['updated'] # other recommended and optional elements try: link_elt = xpath(entry_elt, "link")[0] try: assert(link_elt.attrib['title'] == "comments") microblog_data['comments'] = link_elt.attrib['href'] service, node = self.parseCommentUrl(microblog_data["comments"]) microblog_data['comments_service'] = service.full() microblog_data['comments_node'] = node except (exceptions.DataError, RuntimeError, KeyError): log.warning(_("Can't parse the link element of pubsub entry %s") % microblog_data['id']) except: pass try: microblog_data['author'] = xpath(entry_elt, 'author/name')[0].text except IndexError: try: # XXX: workaround for Jappix behaviour microblog_data['author'] = xpath(entry_elt, 'author/nick')[0].text except IndexError: log.warning(_("Can't find author element in pubsub entry %s") % microblog_data['id']) defer.returnValue(microblog_data) def __getLXMLInnerContent(self, elt): """Return the inner content of a lxml.etree.Element. It is not trivial because the lxml tostring method would return the full content including elt's tag and attributes, and elt.getchildren() would skip a text value which is not within an element...""" return self.__getDomishInnerContent(ElementParser()(etree.tostring(elt))) def __getDomishInnerContent(self, elt): """Return the inner content of a domish.Element.""" result = '' for child in elt.children: try: result += child.toXml() # child id a domish.Element except AttributeError: result += child # child is unicode return result def __decapsulateExtraNS(self, text): """Check for XHTML namespace and decapsulate the content so the user who wants to modify an entry will see the text that he entered. Also this avoids successive encapsulation with a new <div>...</div> at each modification (encapsulation is done in self.data2entry)""" elt = ElementParser()(text) if elt.uri != NS_XHTML: raise exceptions.DataError(_('Content of type XHTML must declare its namespace!')) return self.__getDomishInnerContent(elt) def microblogCB(self, itemsEvent, profile): d = defer.Deferred() def manageItem(microblog_data): self.host.bridge.personalEvent(itemsEvent.sender.full(), "MICROBLOG", microblog_data, profile) for item in itemsEvent.items: d.addCallback(lambda ignore: self.item2mbdata(item)) d.addCallback(manageItem) d.callback(None) return d @defer.inlineCallbacks def data2entry(self, data, profile): """Convert a data dict to en entry usable to create an item @param data: data dict as given by bridge method. @return: deferred which fire domish.Element""" _uuid = unicode(uuid.uuid1()) _entry = atom.Entry() _entry.title = '' # reset the default value which is not empty elems = {'title': atom.Title, 'content': atom.Content} synt = self.host.plugins["TEXT-SYNTAXES"] # loop on ('title', 'title_rich', 'title_xhtml', 'content', 'content_rich', 'content_xhtml') for key in elems.keys(): for type_ in ['', 'rich', 'xhtml']: attr = "%s_%s" % (key, type_) if type_ else key if attr in data: if type_: if type_ == 'rich': # convert input from current syntax to XHTML converted = yield synt.convert(data[attr], synt.getCurrentSyntax(profile), "XHTML") else: # clean the XHTML input converted = yield synt.clean_xhtml(data[attr]) elem = elems[key]((u'<div xmlns="%s">%s</div>' % (NS_XHTML, converted)).encode('utf-8')) elem.attrs['type'] = 'xhtml' if hasattr(_entry, '%s_xhtml' % key): raise exceptions.DataError(_("Can't have xhtml and rich content at the same time")) setattr(_entry, '%s_xhtml' % key, elem) else: # raw text only needs to be escaped to get HTML-safe sequence elem = elems[key](escape(data[attr]).encode('utf-8')) elem.attrs['type'] = 'text' setattr(_entry, key, elem) if not getattr(_entry, key).text: if hasattr(_entry, '%s_xhtml' % key): text = yield self.__removeXHTMLMarkups(getattr(_entry, '%s_xhtml' % key).text) setattr(_entry, key, text) if not _entry.title.text: # eventually move the data from content to title _entry.title = _entry.content.text _entry.title.attrs['type'] = _entry.content.attrs['type'] _entry.content.text = '' _entry.content.attrs['type'] = '' if hasattr(_entry, 'content_xhtml'): _entry.title_xhtml = atom.Title(_entry.content_xhtml.text) _entry.title_xhtml.attrs['type'] = _entry.content_xhtml.attrs['type'] _entry.content_xhtml.text = '' _entry.content_xhtml.attrs['type'] = '' _entry.author = atom.Author() _entry.author.name = data.get('author', self.host.getJidNStream(profile)[0].userhost()).encode('utf-8') _entry.updated = float(data.get('updated', time())) _entry.published = float(data.get('published', time())) entry_id = data.get('id', unicode(_uuid)) _entry.id = entry_id.encode('utf-8') if 'comments' in data: link = atom.Link() link.attrs['href'] = data['comments'] link.attrs['rel'] = 'replies' link.attrs['title'] = 'comments' _entry.links.append(link) _entry_elt = ElementParser()(str(_entry).decode('utf-8')) item = pubsub.Item(id=entry_id, payload=_entry_elt) defer.returnValue(item) @defer.inlineCallbacks def sendMicroblog(self, data, profile): """Send XEP-0277's microblog data @param data: must include content @param profile: profile which send the mood""" if 'content' not in data: log.error(_("Microblog data must contain at least 'content' key")) raise exceptions.DataError('no "content" key found') content = data['content'] if not content: log.error(_("Microblog data's content value must not be empty")) raise exceptions.DataError('empty content') item = yield self.data2entry(data, profile) ret = yield self.host.plugins["XEP-0060"].publish(None, NS_MICROBLOG, [item], profile_key=profile) defer.returnValue(ret) def getLastMicroblogs(self, pub_jid, max_items=10, profile_key=C.PROF_KEY_NONE): """Get the last published microblogs @param pub_jid: jid of the publisher @param max_items: how many microblogs we want to get @param profile_key: profile key """ def resultToArray(result): ret = [] for (success, value) in result: if success: ret.append(value) else: log.error('Error while getting last microblog') return ret d = self.host.plugins["XEP-0060"].getItems(jid.JID(pub_jid), NS_MICROBLOG, max_items=max_items, profile_key=profile_key) d.addCallback(lambda items: defer.DeferredList(map(self.item2mbdata, items))) d.addCallback(resultToArray) return d def setMicroblogAccess(self, access="presence", profile_key=C.PROF_KEY_NONE): """Create a microblog node on PEP with given access If the node already exists, it change options @param access: Node access model, according to xep-0060 #4.5 @param profile_key: profile key""" _jid, xmlstream = self.host.getJidNStream(profile_key) if not _jid: log.error(_("Can't find profile's jid")) return C = self.host.plugins["XEP-0060"] _options = {C.OPT_ACCESS_MODEL: access, C.OPT_PERSIST_ITEMS: 1, C.OPT_MAX_ITEMS: -1, C.OPT_DELIVER_PAYLOADS: 1, C.OPT_SEND_ITEM_SUBSCRIBE: 1} def cb(result): #Node is created with right permission log.debug(_("Microblog node has now access %s") % access) def fatal_err(s_error): #Something went wrong log.error(_("Can't set microblog access")) raise NodeAccessChangeException() def err_cb(s_error): #If the node already exists, the condition is "conflict", #else we have an unmanaged error if s_error.value.condition == 'conflict': #d = self.host.plugins["XEP-0060"].deleteNode(_jid.userhostJID(), NS_MICROBLOG, profile_key=profile_key) #d.addCallback(lambda x: create_node().addCallback(cb).addErrback(fatal_err)) change_node_options().addCallback(cb).addErrback(fatal_err) else: fatal_err(s_error) def create_node(): return self.host.plugins["XEP-0060"].createNode(_jid.userhostJID(), NS_MICROBLOG, _options, profile_key=profile_key) def change_node_options(): return self.host.plugins["XEP-0060"].setOptions(_jid.userhostJID(), NS_MICROBLOG, _jid.userhostJID(), _options, profile_key=profile_key) create_node().addCallback(cb).addErrback(err_cb)