def get_entities(offset=0, limit=100):
sql = '''
SELECT * FROM entities
OFFSET %s LIMIT %s'''
if limit > 10000: # TODO avoid magic number
raise ValueError('query limit {} too high'.format(limit))
cursor = execute_sql(sql, (offset, limit), **EDB)
rows = cursor.fetchall()
return rows_to_dicts(rows, cursor)
def get_document_relations(document_id):
"""Return relation annotations for document."""
sql = '''
SELECT * FROM relations
WHERE document_id = %s'''
cursor = execute_sql(sql, (document_id, ), **MDB)
rows = cursor.fetchall()
# TODO: KeyError instead of empty list return when doc not in DB?
relations = rows_to_dicts(rows, cursor)
return [_db_relation_to_jsonld(r) for r in relations]
def get_entity_mentions(entity_id, offset, limit):
"""Return mention annotations grounded to entity."""
sql = '''
SELECT * FROM mentions
WHERE data @> %s
OFFSET %s
LIMIT %s'''
e_str = json.dumps({"body": {"id": entity_id}})
cursor = execute_sql(sql, (e_str, offset, limit), **MDB)
rows = cursor.fetchall()
mentions = rows_to_dicts(rows, cursor)
return [_db_mention_to_jsonld(m) for m in mentions]
def _get_relation_instances_denormalized(entity1_id, entity2_id, offset,
limit):
# Implementation using copies of mention data in relation
if entity1_id > entity2_id:
# Assuming symmetry, relations are stored so that the e1id < e2id
entity1_id, entity2_id = entity2_id, entity1_id
sql = '''
SELECT *
FROM relations
WHERE __from_data_body_id = %s
AND __to_data_body_id = %s
OFFSET %s
LIMIT %s'''
cursor = execute_sql(sql, (entity1_id, entity2_id, offset, limit), **MDB)
rows = cursor.fetchall()
relations = rows_to_dicts(rows, cursor)
return [_db_relation_to_jsonld(r) for r in relations]
def _get_relation_instances_join(entity1_id, entity2_id, offset, limit):
# Implementation with double join (slow for large DBs)
sql = '''
SELECT r.*
FROM relations AS r,
mentions AS m1,
mentions AS m2
WHERE m1.data @> %s
AND m2.data @> %s
AND r.from_id = m1.id
AND r.to_id = m2.id
OFFSET %s
LIMIT %s'''
e1_str = json.dumps({'body': {'id': entity1_id}})
e2_str = json.dumps({'body': {'id': entity2_id}})
cursor = execute_sql(sql, (e1_str, e2_str, offset, limit), **MDB)
rows = cursor.fetchall()
relations = rows_to_dicts(rows, cursor)
return [_db_relation_to_jsonld(r) for r in relations]
def get_entity(id_):
"""Get entity with given id."""
sql = '''
SELECT * FROM
(
SELECT DISTINCT ON (LOWER(t2.name)) t2.id AS id, t2.label AS label, t2.name AS name, t2.score AS score FROM
(
SELECT t.oid AS id, c.label AS label, t.text AS name, t.score AS score FROM node_text AS t, node_category AS c WHERE t.oid = c.oid AND t.oid = %s ORDER BY t.text
)
AS t2
)
AS t3
ORDER BY score DESC
LIMIT 1'''
cursor = execute_sql(sql, (id_, ), **MDB)
rows = cursor.fetchall()
if not rows:
raise KeyError
assert len(rows) == 1 # id is primary key
return rows_to_dicts(rows, cursor)[0]