mygpo/search/index.py

   1
   2 """ Contains code for indexing other objects """
   3
   4 from pyes import ES, QueryStringQuery, FunctionScoreQuery
   5 from pyes.exceptions import IndexAlreadyExistsException
   6
   7 from django.conf import settings
   8
   9 from mygpo.search.json import podcast_to_json
  10 from mygpo.search.models import PodcastResult
  11
  12 import logging
  13 logger = logging.getLogger(__name__)
  14
  15
  16 def get_connection():
  17     """ Create a connection from Django settings """
  18     conn = ES(settings.ELASTICSEARCH_SERVER,
  19               timeout=settings.ELASTICSEARCH_TIMEOUT)
  20     return conn
  21
  22
  23 def index_podcast(sender, **kwargs):
  24     """ Indexes a podcast """
  25
  26     conn = get_connection()
  27     podcast = kwargs['instance']
  28
  29     document = podcast_to_json(podcast)
  30     conn.index(document, settings.ELASTICSEARCH_INDEX,
  31                'podcast', podcast.id.hex)
  32
  33
  34 def create_index():
  35     """ Creates the Elasticsearch index """
  36     conn = get_connection()
  37
  38     logger.info('Creating index %s' % settings.ELASTICSEARCH_INDEX)
  39     try:
  40         conn.indices.create_index(settings.ELASTICSEARCH_INDEX)
  41
  42     except IndexAlreadyExistsException as ex:
  43         logger.info(str(ex))
  44
  45
  46 def search_podcasts(query):
  47     """ Search for podcasts according to 'query' """
  48     conn = ES(settings.ELASTICSEARCH_SERVER)
  49
  50     # we have some "optimal" number of subscribers (eg the max)
  51     # the farther we get from there, the lower the score
  52     decay = FunctionScoreQuery.DecayFunction(
  53             decay_function='gauss',
  54             field='subscribers',
  55             origin=2000,
  56             scale=1000,
  57             decay=.3,
  58         )
  59     # workaround for https://github.com/aparo/pyes/pull/418
  60     decay._internal_name = 'gauss'
  61
  62     q = FunctionScoreQuery(
  63         query=QueryStringQuery(query),
  64         functions=[decay],
  65         boost_mode=FunctionScoreQuery.BoostModes.MULTIPLY,
  66     )
  67     results = conn.search(query=q, indices=settings.ELASTICSEARCH_INDEX,
  68                           doc_types='podcast',
  69                           model=lambda conn, doc: PodcastResult.from_doc(doc))
  70     return results