def run(self): """ Drops all indices """ global_coll = solr.global_collection() for doc in solr.get_all_docs_by_query(global_coll, '*:*', fields='id'): solr.collection_for_wiki(doc['id']).drop() global_coll.drop() solr.all_pages_collection().drop() solr.all_topics_collection().drop() solr.all_user_pages_collection().drop() solr.wiki_user_collection().drop() solr.user_collection().drop()
def get_all_authors(self): """ Optimized to get all authors :return: an OrderedDict with author dicts :rtype: collections.OrderedDict """ return solr.get_all_docs_by_query(solr.wiki_user_collection(), 'wiki_id_i:%s' % self.wiki_id);
def get_all_users(self): """ Optimized to get all users :return: an OrderedDict with user dicts :rtype: collections.OrderedDict """ return solr.get_all_docs_by_query(solr.wiki_user_collection(), 'wiki_id_i:%s' % self.wiki_id, fields=','.join(UserModel.fields))
def run(self): """ Drops all indices """ global_coll = solr.global_collection() print 'global' global_coll.optimize() for doc in solr.get_all_docs_by_query(global_coll, '*:*', fields='id'): print doc['id'] solr.collection_for_wiki(doc['id']).optimize() print 'all pages' solr.all_pages_collection().optimize() print 'all topics' solr.all_topics_collection().optimize() print 'all user pages' solr.all_user_pages_collection().optimize() print 'wiki user' solr.wiki_user_collection().optimize() print 'user' solr.user_collection().optimize()
def get_topics_for_wiki(self, wiki_id, limit=10, offset=0, **kwargs): """ Gets most important topics for this user on this wiki :param limit: the wiki id :type limit: str :param limit: limit :type limit: int :param offset: offset :type offset: int :return: ordered dict of topic name to auth or a list of dicts for api :rtype: collections.OrderedDict|list """ for doc in solr.get_all_docs_by_query(solr.wiki_user_collection(), 'user_id_i:%d_%d' % (wiki_id, self.user_id), fields=','.join(TopicModel.fields+['attr_entities'])): return doc['attr_entities']
def get_users(self, limit=10, offset=None, **kwargs): """ Provides the top users for a wiki :param limit: number of users you want :type limit: int :param offset: offset :type offset: int :return: list of user dicts :rtype: list """ return solr.get_docs_by_query_with_limit(solr.wiki_user_collection(), 'wiki_id_i:%s' % self.wiki_id, limit=limit, offset=offset, sort='total_page_authority_f desc', fields=','.join(UserModel.fields))
def get_topics_for_wiki(self, wiki_id, limit=10, offset=0, for_api=False): """ Gets most important topics for this user on this wiki :param limit: the wiki id :type limit: str :param limit: limit :type limit: int :param offset: offset :type offset: int :param for_api: if it's for the api, we fix the naming :type for_api: bool :return: ordered dict of topic name to auth or a list of dicts for api :rtype: collections.OrderedDict|list """ for doc in solr.get_all_docs_by_query(solr.wiki_user_collection(), 'name_txt_en:"%s"' % self.user_name): return doc['attr_entities']
def get_wikis(self, limit=10, offset=0, **kwargs): """ Most important wikis for this user Calculated by sum of contribs times global authority :param limit: the limit :type limit: int :param offset: offset :type offset: int :return: an ordereddict of wiki ids to wiki dicts, or a list, for API :rtype: collections.OrderedDict|list """ return solr.get_docs_by_query_with_limit(solr.wiki_user_collection(), 'user_id_i:%d' % self.user_id, limit=limit, offset=offset, boost='scaled_contribs_authority_f', fields='wiki_id_i', **sans_q(kwargs))
def get_wikis(self, limit=10, offset=0, for_api=False): """ Most important wikis for this user Calculated by sum of contribs times global authority :param limit: the limit :type limit: int :param offset: offset :type offset: int :param for_api: if it's for the api, we add less :type for_api: bool :return: an ordereddict of wiki ids to wiki dicts, or a list, for API :rtype: collections.OrderedDict|list """ return solr.get_docs_by_query_with_limit(solr.wiki_user_collection(), 'name_txt_en:%s' % self.user_name, limit=limit, offset=offset, boost='scaled_contribs_authority_f')
def get_authors(self, limit=10, offset=None, for_api=False): """ Provides the top authors for a wiki :param limit: number of authors you want :type limit: int :param offset: offset :type offset: int :param for_api: if it's for the api, we add less :type for_api: bool :return: list of author dicts :rtype: list """ solr.get_docs_by_query_with_limit(solr.wiki_user_collection(), 'wiki_id_i:%s' % self.wiki_id, limit=limit, offset=offset, sort='total_page_authority_f desc')
def analyze_users_globally(): print "Analyzing Users..." user_collection = solr.existing_collection(solr.user_collection()) wiki_user_collection = solr.wiki_user_collection() id_to_docs = dict() for user_doc in solr.get_all_docs_by_query(wiki_user_collection, '*:*'): # these are gonna be wiki-id_user-id doc_id = user_doc['id'].split('_').pop() if doc_id not in id_to_docs: id_to_docs[doc_id] = dict(id=doc_id, attr_entities={'set': []}, name_s={'set': user_doc['name_s']}, name_txt_en={'set': user_doc['name_txt_en']}, wikis_is={'set': []}, attr_wikis={'set': []}, authorities_fs={'set': []}, total_authority_f={'set': 0}, scaled_authority_f={'set': 0}) try: map(id_to_docs[doc_id]['attr_entities']['set'].append, user_doc['attr_entities']) id_to_docs[doc_id]['wikis_is']['set'].append(user_doc['wiki_id_i']) id_to_docs[doc_id]['attr_wikis']['set'].append(user_doc['wiki_name_txt']) id_to_docs[doc_id]['authorities_fs']['set'].append(user_doc['total_page_authority_f']) except KeyError: pass # zero f***s id_to_total_authorities = dict([(uid, sum(doc['authorities_fs']['set'])) for uid, doc in id_to_docs.items()]) user_scaler = MinMaxScaler(id_to_total_authorities.values()) for uid, total_authority in id_to_total_authorities.items(): id_to_docs[uid]['total_authority_f']['set'] = total_authority id_to_docs[uid]['scaled_authority_f']['set'] = user_scaler.scale(total_authority) user_collection.add(id_to_docs.values()) user_collection.commit()
def ingest_data(wiki_id): """ Create Solr documents for a given wiki ID :param wiki_id: the ID of the wiki (int or str) :type wiki_id: int :return: """ # make sure all pages and all user pages exists solr.existing_collection(solr.all_pages_collection()) solr.existing_collection(solr.all_user_pages_collection()) resp = requests.get(u'http://www.wikia.com/api/v1/Wikis/Details', params={u'ids': wiki_id}) items = resp.json()['items'] if wiki_id not in items: print u"Wiki doesn't exist?" return api_data = items[wiki_id] wiki_data = { 'id': api_data['id'], 'wam_f': {'set': api_data['wam_score']}, 'title_s': {'set': api_data['title']}, 'attr_title': {'set': api_data['title']}, 'attr_desc': {'set': api_data['desc']} } for key in api_data['stats'].keys(): wiki_data['%s_i' % key] = {'set': api_data['stats'][key]} wiki_api_data = requests.get(u'%swikia.php' % (api_data[u'url']), params={u'method': u'getForWiki', u'service': u'CrossWikiCore', u'controller': u'WikiaSearchIndexerController'}).json()[u'contents'] wiki_data[u'hub_s'] = wiki_api_data[u'hub_s'] # easier api_data[u'hub_s'] = wiki_api_data[u'hub_s'] collection = solr.existing_collection(solr.collection_for_wiki(wiki_id)) use_caching(is_read_only=True, shouldnt_compute=True) wpe = WikiPageToEntitiesService().get_value(wiki_id) if not wpe: print u"NO WIKI PAGE TO ENTITIES SERVICE FOR", wiki_id return False documents = [] grouped_futures = [] pages_to_authority = WikiAuthorityService().get_value(str(wiki_data['id'])) for counter, (doc_id, entity_data) in enumerate(wpe.items()): documents.append({ 'id': doc_id, 'attr_entities': {'set': list(set(entity_data.get(u'redirects', {}).values() + entity_data.get(u'titles')))}, 'type_s': {'set': 'Page'}, 'authority_f': {'set': pages_to_authority.get(doc_id, 0)}, 'hub_s': wiki_api_data['hub_s'] }) if counter != 0 and counter % 1500 == 0: grouped_futures.append( group(add_with_metadata.s(api_data, grouping) for grouping in iter_grouper(15, documents))() ) documents = [] grouped_futures.append( group(add_with_metadata.s(api_data, grouping) for grouping in iter_grouper(15, documents))() ) # block on completion of all grouped futures completed = 0 total = 0 while len(filter(lambda x: not x.ready(), grouped_futures)) > 1: new_completed = 0 new_total = 0 for future in grouped_futures: new_completed += future.completed_count() new_total += len(future.results) if completed != new_completed or total != new_total: completed = new_completed total = new_total print "Grouped Tasks: (%d/%d)" % (completed, total) sleep(2) all_user_tuples = [] for future in grouped_futures: result = get_with_backoff(future, []) map(all_user_tuples.extend, result) all_user_tuples = list(set(all_user_tuples)) if not all_user_tuples: print "Empty user tuples, bailing" return # assign the unique user ids to the first variable, and the unique usernames to the second all_user_ids, all_users = zip(*all_user_tuples) collection.commit() solr.all_pages_collection().commit() solr.all_user_pages_collection().commit() wiki_data['attr_entities'] = {'set': []} for count, entities in WikiEntitiesService().get_value(str(wiki_id)).items(): for entity in entities: map(wiki_data['attr_entities']['set'].append, [entity] * int(count)) # goddamnit count isn't int wiki_data['user_ids_is'] = {'set': all_user_ids} wiki_data['attr_users'] = {'set': all_users} wiki_data['total_authority_f'] = {'set': sum(pages_to_authority.values())} wiki_data['authorities_fs'] = {'set': pages_to_authority.values()} wiki_collection = solr.existing_collection(solr.global_collection()) wiki_collection.add([wiki_data]) wiki_collection.commit() print "Committed wiki data" print "Retrieving user docs..." futures = group(build_wiki_user_doc.s(api_data, user_tuple) for user_tuple in all_user_tuples)() future_result_len = len(futures.results) while not futures.ready(): print "Progress: (%d/%d)" % (futures.completed_count(), future_result_len) sleep(2) user_docs = get_with_backoff(futures, []) if not user_docs: print "User docs was empty. Possibly connection problems." return authority_scaler = MinMaxScaler([doc['total_page_authority_f']['set'] for doc in user_docs]) contribs_scaler = MinMaxScaler([doc['total_contribs_f']['set'] for doc in user_docs]) for doc in user_docs: scaled_authority = authority_scaler.scale(doc['total_page_authority_f']['set']) scaled_contribs = contribs_scaler.scale(doc['total_contribs_f']['set']) doc['scaled_authority_f'] = {'set': scaled_authority} doc['scaled_contribs_f'] = {'set': scaled_contribs} doc['scaled_contribs_authority_f'] = {'set': scaled_authority * scaled_contribs} wiki_user_collection = solr.existing_collection(solr.wiki_user_collection()) wiki_user_collection.add(user_docs) wiki_user_collection.commit() print "Analyzing topics" futures = group(get_wiki_topic_doc.s(wiki_data['id'], topic) for topic in list(set(wiki_data['attr_entities']['set'])))() future_result_len = len(futures.results) counter = 0 while not futures.ready(): if counter % 5 == 0: print "Progress: (%d/%d)" % (futures.completed_count(), future_result_len) sleep(2) counter += 1 topic_docs = get_with_backoff(futures, []) if not topic_docs: print "No topics, probably a connection error" return collection.add(topic_docs) collection.commit() topic_collection = solr.existing_collection(solr.all_topics_collection()) topic_collection.add(topic_docs) topic_collection.commit()