def start(self):
while True:
try:
# dequeue a processed resource
message = self._mq_client.get_message(self._processed_resource_queue)
resource = self._mq_codec.decode(message.body, Resource)
except EmptyQueueException:
self._logger.info('Empty queue. Sleeping...')
time.sleep(self.EMPTY_QUEUE_TIMEOUT)
continue
resource._id = makeIdFromURI(resource.uri)
story = self._story_provider.provide_for_resource(resource)
resource.story_id = story._id
try:
# save and index the resource
self._resource_collection.insert_model(resource)
self._index_helper.index_resource(resource, self._index)
self._logger.debug('Resource %s: Inserted and indexed.' % resource._id)
# queue update
update = StoryUpdate()
update.story_id = story._id
update.type = STORY_UPDATE_RESOURCE_INSERTED
update.value = resource._id
self._update_collection.insert_model(update)
self._su_client.update(story._id)
except DuplicateKeyError:
self._logger.warning('Resource %s: Already inserted.' % resource._id)
# delete the queue message
self._mq_client.delete_message(self._processed_resource_queue, message.id)
def _enqueue(self, resource):
# Discovered URI - capped collection --db.discovered_uri
mongoConnection = Connection()
db = mongoConnection.feed_reading_discovery
# settings
max_content_length = config['discovery']['resource_enqueueing']['max_content_length']
# message queue
mq_client = message_queue_client_from_config(config['message_queue']['client'])
mq_client.connect()
mq_codec = JSONCodec()
queue = 'discovered_resources'
try:
#insert into capped
db.discovered_uri.insert({'_id':makeIdFromURI(resource.uri), 'uri':resource.uri})
#insert into queue
if len(resource.content) > max_content_length:
self._logger.warning('Skipped %s: Content length is %s.' %
(resource, len(resource.content)))
msg_body = mq_codec.encode(resource)
mq_client.put_message(queue, msg_body)
self._logger.debug("Enqueued: %s" % resource._id)
except DuplicateKeyError:
pass
# remove resource from collection
self._resources_collection.remove_model(resource)
def discovered_uri_process(self, entry):
# Resource fields
kwargs = {}
kwargs['_id'] = makeIdFromURI(entry.link)
kwargs['uri'] = entry.link
return DiscoveredURI(**kwargs)
def resources_process(self, source, entry, config):
self._config = config
# configure logging
logging.config.dictConfig(config['logging'])
self._logger = logging.getLogger()
# Resource fields
kwargs = {}
kwargs['_id'] = makeIdFromURI(entry.link)
kwargs['uri'] = entry.link
kwargs['type'] = 'article'
kwargs['discovered'] = int(time.time())
kwargs['title'] = entry.title
if 'updated_parsed' in entry:
kwargs['published'] = calendar.timegm(entry.updated_parsed)
elif 'date_parsed' in entry:
kwargs['published'] = calendar.timegm(entry.date_parsed)
else:
kwargs['published'] = None
if 'author' in entry:
kwargs['author'] = entry.author
# content
content = ''
if 'summary' in entry:
content = entry.summary
if 'content' in entry:
c = entry.content[0].value
if len(c) > len (content):
content = c
kwargs['feed_content'] = content
kwargs['content'] = ''
kwargs['source'] = source
kwargs['content_extracted'] = 1
return DiscoveredResource(**kwargs)
}
mq_client = message_queue_client_from_config(mq_config)
mq_codec = JSONCodec()
processed_resource_queue = 'processed_resources'
# ElasticSearch
es = ES('localhost:9200', timeout=60)
es_index = 'topic_tracking'
# dequeue one resource
mq_client.connect()
message = mq_client.get_message(processed_resource_queue)
resource = mq_codec.decode(message.body, Resource)
mq_client.delete_message(processed_resource_queue, message.id)
mq_client.disconnect()
# save the resource to mongo
resource._id = makeIdFromURI(resource.uri)
resource_collection.insert_model(resource)
# index the resource
for boost in [1, 1000]:
es_doc = {}
es_doc['content'] = resource.content
es_doc['title'] = resource.title
es_doc['entities'] = build_payload_string(resource.entities, boost)
es_doc['terms'] = build_payload_string(resource.terms, boost)
id = '%s_%d' % (resource._id, boost)
r = es.index(es_doc, es_index, 'resource', id)
pprint(r)
