mongo_client= MongoClient()
mongo_client.drop_database(db_out)
mongo_client.close()
print 'database cleared'
col_tenant_id = 1
col_user_id = 2
col_item_id = 3
num_to_recomm_per_user = 10
num_to_recomm_per_item = 10
conf = SparkConf().setAppName("PysparkCollaborativeFiltering").set("spark.cassandra.connection.host", spark_cassandra_connection_host)
sc = CassandraSparkContext(conf=conf)
sc.setCheckpointDir('checkpoint/')
data = sc.cassandraTable("mykeyspace", "transactions",row_format=1).collect() # row_format: tuple
# (id, tenant_id, user_id, item_id)
tenant_ids = set(list(map(lambda x:x[col_tenant_id],data)))
data_rdd = sc.parallelize(data)
# data_rdd = sc.parallelize(data).map(list)
all_results_per_user = sc.emptyRDD()
all_results_per_item = sc.emptyRDD()
for t_id in tenant_ids:
print("\nComputing recommendation for tenant {}...\n".format(t_id))
per_tenant_rdd = data_rdd.filter(
lambda x: x[col_tenant_id] == t_id).map(
lambda l: ((l[col_user_id],l[col_item_id]),1.0)).reduceByKey(
lambda x,y: x + y).map(
