-# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-License-Identifier: GPL-3.0-or-later
#
# This file is part of Nominatim. (https://nominatim.org)
#
-# Copyright (C) 2022 by the Nominatim developer community.
+# Copyright (C) 2024 by the Nominatim developer community.
# For a full list of authors see the git log.
import logging
from itertools import chain
-import psycopg2.extras
+import psycopg
+from psycopg import sql as pysql
from place_inserter import PlaceColumn
from table_compare import NominatimID, DBRow
-from nominatim.indexer import indexer
-from nominatim.tokenizer import factory as tokenizer_factory
+from nominatim_db.indexer import indexer
+from nominatim_db.tokenizer import factory as tokenizer_factory
def check_database_integrity(context):
""" Check some generic constraints on the tables.
"""
- with context.db.cursor() as cur:
+ with context.db.cursor(row_factory=psycopg.rows.tuple_row) as cur:
# place_addressline should not have duplicate (place_id, address_place_id)
cur.execute("""SELECT count(*) FROM
(SELECT place_id, address_place_id, count(*) as c
with context.db.cursor() as cur:
cur.execute("SELECT value FROM osm2pgsql_properties WHERE property = 'db_format'")
row = cur.fetchone()
- if row is None or row[0] == '1':
+ if row is None or row['value'] == '1':
for r in context.table:
last_node = 0
last_way = 0
cur.execute("""INSERT INTO planet_osm_rels (id, tags, members)
VALUES (%s, %s, %s)""",
- (r['id'], psycopg2.extras.Json(tags),
- psycopg2.extras.Json(members)))
+ (r['id'], psycopg.types.json.Json(tags),
+ psycopg.types.json.Json(members)))
@given("the ways")
def add_data_to_planet_ways(context):
with context.db.cursor() as cur:
cur.execute("SELECT value FROM osm2pgsql_properties WHERE property = 'db_format'")
row = cur.fetchone()
- json_tags = row is not None and row[0] != '1'
+ json_tags = row is not None and row['value'] != '1'
for r in context.table:
if json_tags:
- tags = psycopg2.extras.Json({h[5:]: r[h] for h in r.headings if h.startswith("tags+")})
+ tags = psycopg.types.json.Json({h[5:]: r[h] for h in r.headings if h.startswith("tags+")})
else:
tags = list(chain.from_iterable([(h[5:], r[h])
for h in r.headings if h.startswith("tags+")]))
expected rows are expected to be present with at least one database row.
When 'exactly' is given, there must not be additional rows in the database.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
expected_content = set()
for row in context.table:
nid = NominatimID(row['object'])
DBRow(nid, res, context).assert_row(row, ['object'])
if exact:
- cur.execute('SELECT osm_type, osm_id, class from {}'.format(table))
- actual = set([(r[0], r[1], r[2]) for r in cur])
+ cur.execute(pysql.SQL('SELECT osm_type, osm_id, class from')
+ + pysql.Identifier(table))
+ actual = set([(r['osm_type'], r['osm_id'], r['class']) for r in cur])
assert expected_content == actual, \
f"Missing entries: {expected_content - actual}\n" \
f"Not expected in table: {actual - expected_content}"
""" Ensure that no database row for the given object exists. The ID
must be of the form '<NRW><osm id>[:<class>]'.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
NominatimID(oid).query_osm_id(cur, "SELECT * FROM %s where {}" % table)
assert cur.rowcount == 0, \
"Found {} entries for ID {}".format(cur.rowcount, oid)
tokenizer = tokenizer_factory.get_tokenizer_for_db(context.nominatim.get_test_config())
with tokenizer.name_analyzer() as analyzer:
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
for row in context.table:
nid = NominatimID(row['object'])
nid.row_by_place_id(cur, 'search_name',
""" Check that there is noentry in the search_name table for the given
objects. IDs are in format '<NRW><osm id>[:<class>]'.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
NominatimID(oid).row_by_place_id(cur, 'search_name')
assert cur.rowcount == 0, \
All rows must be present as excepted and there must not be additional
rows.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
cur.execute("SELECT *, ST_AsText(geometry) as geomtxt FROM location_postcode")
assert cur.rowcount == len(list(context.table)), \
"Postcode table has {} rows, expected {}.".format(cur.rowcount, len(list(context.table)))
plist.sort()
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
if nctx.tokenizer != 'legacy':
cur.execute("SELECT word FROM word WHERE type = 'P' and word = any(%s)",
(plist,))
and class = 'place' and type = 'postcode'""",
(plist,))
- found = [row[0] for row in cur]
+ found = [row['word'] for row in cur]
assert len(found) == len(set(found)), f"Duplicate rows for postcodes: {found}"
if exclude:
representing the addressee and the 'address' column, representing the
address item.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
for row in context.table:
nid = NominatimID(row['object'])
pid = nid.get_place_id(cur)
""" Check that the place_addressline doesn't contain any entries for the
given addressee/address item pairs.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
for row in context.table:
pid = NominatimID(row['object']).get_place_id(cur)
apid = NominatimID(row['address']).get_place_id(cur, allow_empty=True)
def check_location_property_osmline(context, oid, neg):
""" Check that the given way is present in the interpolation table.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
cur.execute("""SELECT *, ST_AsText(linegeo) as geomtxt
FROM location_property_osmline
WHERE osm_id = %s AND startnumber IS NOT NULL""",
expected rows are expected to be present with at least one database row.
When 'exactly' is given, there must not be additional rows in the database.
"""
- with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:
+ with context.db.cursor() as cur:
expected_content = set()
for row in context.table:
if ':' in row['object']:
if exact:
cur.execute('SELECT osm_id, startnumber from location_property_osmline')
- actual = set([(r[0], r[1]) for r in cur])
+ actual = set([(r['osm_id'], r['startnumber']) for r in cur])
assert expected_content == actual, \
f"Missing entries: {expected_content - actual}\n" \
f"Not expected in table: {actual - expected_content}"