1 # SPDX-License-Identifier: GPL-2.0-only
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2022 by the Nominatim developer community.
6 # For a full list of authors see the git log.
8 Functions for setting up and importing a new Nominatim database.
10 from typing import Tuple, Optional, Union, Sequence, MutableMapping, Any
15 from pathlib import Path
18 from psycopg2 import sql as pysql
20 from nominatim.config import Configuration
21 from nominatim.db.connection import connect, get_pg_env, Connection
22 from nominatim.db.async_connection import DBConnection
23 from nominatim.db.sql_preprocessor import SQLPreprocessor
24 from nominatim.tools.exec_utils import run_osm2pgsql
25 from nominatim.errors import UsageError
26 from nominatim.version import POSTGRESQL_REQUIRED_VERSION, \
27 POSTGIS_REQUIRED_VERSION
29 LOG = logging.getLogger()
31 def _require_version(module: str, actual: Tuple[int, int], expected: Tuple[int, int]) -> None:
32 """ Compares the version for the given module and raises an exception
33 if the actual version is too old.
36 LOG.fatal('Minimum supported version of %s is %d.%d. '
37 'Found version %d.%d.',
38 module, expected[0], expected[1], actual[0], actual[1])
39 raise UsageError(f'{module} is too old.')
42 def _require_loaded(extension_name: str, conn: Connection) -> None:
43 """ Check that the given extension is loaded. """
44 if not conn.extension_loaded(extension_name):
45 LOG.fatal('Required module %s is not loaded.', extension_name)
46 raise UsageError(f'{extension_name} is not loaded.')
49 def check_existing_database_plugins(dsn: str) -> None:
50 """ Check that the database has the required plugins installed."""
51 with connect(dsn) as conn:
52 _require_version('PostgreSQL server',
53 conn.server_version_tuple(),
54 POSTGRESQL_REQUIRED_VERSION)
55 _require_version('PostGIS',
56 conn.postgis_version_tuple(),
57 POSTGIS_REQUIRED_VERSION)
58 _require_loaded('hstore', conn)
61 def setup_database_skeleton(dsn: str, rouser: Optional[str] = None) -> None:
62 """ Create a new database for Nominatim and populate it with the
65 The function fails when the database already exists or Postgresql or
66 PostGIS versions are too old.
68 Uses `createdb` to create the database.
70 If 'rouser' is given, then the function also checks that the user
71 with that given name exists.
73 Requires superuser rights by the caller.
75 proc = subprocess.run(['createdb'], env=get_pg_env(dsn), check=False)
77 if proc.returncode != 0:
78 raise UsageError('Creating new database failed.')
80 with connect(dsn) as conn:
81 _require_version('PostgreSQL server',
82 conn.server_version_tuple(),
83 POSTGRESQL_REQUIRED_VERSION)
85 if rouser is not None:
86 with conn.cursor() as cur:
87 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
90 LOG.fatal("Web user '%s' does not exist. Create it with:\n"
91 "\n createuser %s", rouser, rouser)
92 raise UsageError('Missing read-only user.')
95 with conn.cursor() as cur:
96 cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
97 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
99 postgis_version = conn.postgis_version_tuple()
100 if postgis_version[0] >= 3:
101 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis_raster')
105 _require_version('PostGIS',
106 conn.postgis_version_tuple(),
107 POSTGIS_REQUIRED_VERSION)
110 def import_osm_data(osm_files: Union[Path, Sequence[Path]],
111 options: MutableMapping[str, Any],
112 drop: bool = False, ignore_errors: bool = False) -> None:
113 """ Import the given OSM files. 'options' contains the list of
114 default settings for osm2pgsql.
116 options['import_file'] = osm_files
117 options['append'] = False
118 options['threads'] = 1
120 if not options['flatnode_file'] and options['osm2pgsql_cache'] == 0:
121 # Make some educated guesses about cache size based on the size
122 # of the import file and the available memory.
123 mem = psutil.virtual_memory()
125 if isinstance(osm_files, list):
126 for fname in osm_files:
127 fsize += os.stat(str(fname)).st_size
129 fsize = os.stat(str(osm_files)).st_size
130 options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
131 fsize * 2) / 1024 / 1024) + 1
133 run_osm2pgsql(options)
135 with connect(options['dsn']) as conn:
136 if not ignore_errors:
137 with conn.cursor() as cur:
138 cur.execute('SELECT * FROM place LIMIT 1')
139 if cur.rowcount == 0:
140 raise UsageError('No data imported by osm2pgsql.')
143 conn.drop_table('planet_osm_nodes')
145 if drop and options['flatnode_file']:
146 Path(options['flatnode_file']).unlink()
149 def create_tables(conn: Connection, config: Configuration, reverse_only: bool = False) -> None:
150 """ Create the set of basic tables.
151 When `reverse_only` is True, then the main table for searching will
152 be skipped and only reverse search is possible.
154 sql = SQLPreprocessor(conn, config)
155 sql.env.globals['db']['reverse_only'] = reverse_only
157 sql.run_sql_file(conn, 'tables.sql')
160 def create_table_triggers(conn: Connection, config: Configuration) -> None:
161 """ Create the triggers for the tables. The trigger functions must already
162 have been imported with refresh.create_functions().
164 sql = SQLPreprocessor(conn, config)
165 sql.run_sql_file(conn, 'table-triggers.sql')
168 def create_partition_tables(conn: Connection, config: Configuration) -> None:
169 """ Create tables that have explicit partitioning.
171 sql = SQLPreprocessor(conn, config)
172 sql.run_sql_file(conn, 'partition-tables.src.sql')
175 def truncate_data_tables(conn: Connection) -> None:
176 """ Truncate all data tables to prepare for a fresh load.
178 with conn.cursor() as cur:
179 cur.execute('TRUNCATE placex')
180 cur.execute('TRUNCATE place_addressline')
181 cur.execute('TRUNCATE location_area')
182 cur.execute('TRUNCATE location_area_country')
183 cur.execute('TRUNCATE location_property_tiger')
184 cur.execute('TRUNCATE location_property_osmline')
185 cur.execute('TRUNCATE location_postcode')
186 if conn.table_exists('search_name'):
187 cur.execute('TRUNCATE search_name')
188 cur.execute('DROP SEQUENCE IF EXISTS seq_place')
189 cur.execute('CREATE SEQUENCE seq_place start 100000')
191 cur.execute("""SELECT tablename FROM pg_tables
192 WHERE tablename LIKE 'location_road_%'""")
194 for table in [r[0] for r in list(cur)]:
195 cur.execute('TRUNCATE ' + table)
200 _COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
201 ('osm_type', 'osm_id', 'class', 'type',
202 'name', 'admin_level', 'address',
203 'extratags', 'geometry')))
206 def load_data(dsn: str, threads: int) -> None:
207 """ Copy data into the word and placex table.
209 sel = selectors.DefaultSelector()
210 # Then copy data from place to placex in <threads - 1> chunks.
211 place_threads = max(1, threads - 1)
212 for imod in range(place_threads):
213 conn = DBConnection(dsn)
216 pysql.SQL("""INSERT INTO placex ({columns})
217 SELECT {columns} FROM place
218 WHERE osm_id % {total} = {mod}
219 AND NOT (class='place' and (type='houses' or type='postcode'))
220 AND ST_IsValid(geometry)
221 """).format(columns=_COPY_COLUMNS,
222 total=pysql.Literal(place_threads),
223 mod=pysql.Literal(imod)))
224 sel.register(conn, selectors.EVENT_READ, conn)
226 # Address interpolations go into another table.
227 conn = DBConnection(dsn)
229 conn.perform("""INSERT INTO location_property_osmline (osm_id, address, linegeo)
230 SELECT osm_id, address, geometry FROM place
231 WHERE class='place' and type='houses' and osm_type='W'
232 and ST_GeometryType(geometry) = 'ST_LineString'
234 sel.register(conn, selectors.EVENT_READ, conn)
236 # Now wait for all of them to finish.
237 todo = place_threads + 1
239 for key, _ in sel.select(1):
245 print('.', end='', flush=True)
248 with connect(dsn) as syn_conn:
249 with syn_conn.cursor() as cur:
250 cur.execute('ANALYSE')
253 def create_search_indices(conn: Connection, config: Configuration,
254 drop: bool = False, threads: int = 1) -> None:
255 """ Create tables that have explicit partitioning.
258 # If index creation failed and left an index invalid, they need to be
259 # cleaned out first, so that the script recreates them.
260 with conn.cursor() as cur:
261 cur.execute("""SELECT relname FROM pg_class, pg_index
262 WHERE pg_index.indisvalid = false
263 AND pg_index.indexrelid = pg_class.oid""")
264 bad_indices = [row[0] for row in list(cur)]
265 for idx in bad_indices:
266 LOG.info("Drop invalid index %s.", idx)
267 cur.execute(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
270 sql = SQLPreprocessor(conn, config)
272 sql.run_parallel_sql_file(config.get_libpq_dsn(),
273 'indices.sql', min(8, threads), drop=drop)