1 # SPDX-License-Identifier: GPL-3.0-or-later
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2024 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 ..errors import UsageError
21 from ..config import Configuration
22 from ..db.connection import connect, get_pg_env, Connection
23 from ..db.async_connection import DBConnection
24 from ..db.sql_preprocessor import SQLPreprocessor
25 from .exec_utils import run_osm2pgsql
26 from ..version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
28 LOG = logging.getLogger()
30 def _require_version(module: str, actual: Tuple[int, int], expected: Tuple[int, int]) -> None:
31 """ Compares the version for the given module and raises an exception
32 if the actual version is too old.
35 LOG.fatal('Minimum supported version of %s is %d.%d. '
36 'Found version %d.%d.',
37 module, expected[0], expected[1], actual[0], actual[1])
38 raise UsageError(f'{module} is too old.')
41 def _require_loaded(extension_name: str, conn: Connection) -> None:
42 """ Check that the given extension is loaded. """
43 with conn.cursor() as cur:
44 cur.execute('SELECT * FROM pg_extension WHERE extname = %s', (extension_name, ))
46 LOG.fatal('Required module %s is not loaded.', extension_name)
47 raise UsageError(f'{extension_name} is not loaded.')
50 def check_existing_database_plugins(dsn: str) -> None:
51 """ Check that the database has the required plugins installed."""
52 with connect(dsn) as conn:
53 _require_version('PostgreSQL server',
54 conn.server_version_tuple(),
55 POSTGRESQL_REQUIRED_VERSION)
56 _require_version('PostGIS',
57 conn.postgis_version_tuple(),
58 POSTGIS_REQUIRED_VERSION)
59 _require_loaded('hstore', conn)
62 def setup_database_skeleton(dsn: str, rouser: Optional[str] = None) -> None:
63 """ Create a new database for Nominatim and populate it with the
66 The function fails when the database already exists or Postgresql or
67 PostGIS versions are too old.
69 Uses `createdb` to create the database.
71 If 'rouser' is given, then the function also checks that the user
72 with that given name exists.
74 Requires superuser rights by the caller.
76 proc = subprocess.run(['createdb'], env=get_pg_env(dsn), check=False)
78 if proc.returncode != 0:
79 raise UsageError('Creating new database failed.')
81 with connect(dsn) as conn:
82 _require_version('PostgreSQL server',
83 conn.server_version_tuple(),
84 POSTGRESQL_REQUIRED_VERSION)
86 if rouser is not None:
87 with conn.cursor() as cur:
88 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
91 LOG.fatal("Web user '%s' does not exist. Create it with:\n"
92 "\n createuser %s", rouser, rouser)
93 raise UsageError('Missing read-only user.')
96 with conn.cursor() as cur:
97 cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
98 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
100 postgis_version = conn.postgis_version_tuple()
101 if postgis_version[0] >= 3:
102 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis_raster')
106 _require_version('PostGIS',
107 conn.postgis_version_tuple(),
108 POSTGIS_REQUIRED_VERSION)
111 def import_osm_data(osm_files: Union[Path, Sequence[Path]],
112 options: MutableMapping[str, Any],
113 drop: bool = False, ignore_errors: bool = False) -> None:
114 """ Import the given OSM files. 'options' contains the list of
115 default settings for osm2pgsql.
117 options['import_file'] = osm_files
118 options['append'] = False
119 options['threads'] = 1
121 if not options['flatnode_file'] and options['osm2pgsql_cache'] == 0:
122 # Make some educated guesses about cache size based on the size
123 # of the import file and the available memory.
124 mem = psutil.virtual_memory()
126 if isinstance(osm_files, list):
127 for fname in osm_files:
128 fsize += os.stat(str(fname)).st_size
130 fsize = os.stat(str(osm_files)).st_size
131 options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
132 fsize * 2) / 1024 / 1024) + 1
134 run_osm2pgsql(options)
136 with connect(options['dsn']) as conn:
137 if not ignore_errors:
138 with conn.cursor() as cur:
139 cur.execute('SELECT * FROM place LIMIT 1')
140 if cur.rowcount == 0:
141 raise UsageError('No data imported by osm2pgsql.')
144 conn.drop_table('planet_osm_nodes')
146 if drop and options['flatnode_file']:
147 Path(options['flatnode_file']).unlink()
150 def create_tables(conn: Connection, config: Configuration, reverse_only: bool = False) -> None:
151 """ Create the set of basic tables.
152 When `reverse_only` is True, then the main table for searching will
153 be skipped and only reverse search is possible.
155 sql = SQLPreprocessor(conn, config)
156 sql.env.globals['db']['reverse_only'] = reverse_only
158 sql.run_sql_file(conn, 'tables.sql')
161 def create_table_triggers(conn: Connection, config: Configuration) -> None:
162 """ Create the triggers for the tables. The trigger functions must already
163 have been imported with refresh.create_functions().
165 sql = SQLPreprocessor(conn, config)
166 sql.run_sql_file(conn, 'table-triggers.sql')
169 def create_partition_tables(conn: Connection, config: Configuration) -> None:
170 """ Create tables that have explicit partitioning.
172 sql = SQLPreprocessor(conn, config)
173 sql.run_sql_file(conn, 'partition-tables.src.sql')
176 def truncate_data_tables(conn: Connection) -> None:
177 """ Truncate all data tables to prepare for a fresh load.
179 with conn.cursor() as cur:
180 cur.execute('TRUNCATE placex')
181 cur.execute('TRUNCATE place_addressline')
182 cur.execute('TRUNCATE location_area')
183 cur.execute('TRUNCATE location_area_country')
184 cur.execute('TRUNCATE location_property_tiger')
185 cur.execute('TRUNCATE location_property_osmline')
186 cur.execute('TRUNCATE location_postcode')
187 if conn.table_exists('search_name'):
188 cur.execute('TRUNCATE search_name')
189 cur.execute('DROP SEQUENCE IF EXISTS seq_place')
190 cur.execute('CREATE SEQUENCE seq_place start 100000')
192 cur.execute("""SELECT tablename FROM pg_tables
193 WHERE tablename LIKE 'location_road_%'""")
195 for table in [r[0] for r in list(cur)]:
196 cur.execute('TRUNCATE ' + table)
201 _COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
202 ('osm_type', 'osm_id', 'class', 'type',
203 'name', 'admin_level', 'address',
204 'extratags', 'geometry')))
207 def load_data(dsn: str, threads: int) -> None:
208 """ Copy data into the word and placex table.
210 sel = selectors.DefaultSelector()
211 # Then copy data from place to placex in <threads - 1> chunks.
212 place_threads = max(1, threads - 1)
213 for imod in range(place_threads):
214 conn = DBConnection(dsn)
217 pysql.SQL("""INSERT INTO placex ({columns})
218 SELECT {columns} FROM place
219 WHERE osm_id % {total} = {mod}
220 AND NOT (class='place' and (type='houses' or type='postcode'))
221 AND ST_IsValid(geometry)
222 """).format(columns=_COPY_COLUMNS,
223 total=pysql.Literal(place_threads),
224 mod=pysql.Literal(imod)))
225 sel.register(conn, selectors.EVENT_READ, conn)
227 # Address interpolations go into another table.
228 conn = DBConnection(dsn)
230 conn.perform("""INSERT INTO location_property_osmline (osm_id, address, linegeo)
231 SELECT osm_id, address, geometry FROM place
232 WHERE class='place' and type='houses' and osm_type='W'
233 and ST_GeometryType(geometry) = 'ST_LineString'
235 sel.register(conn, selectors.EVENT_READ, conn)
237 # Now wait for all of them to finish.
238 todo = place_threads + 1
240 for key, _ in sel.select(1):
246 print('.', end='', flush=True)
249 with connect(dsn) as syn_conn:
250 with syn_conn.cursor() as cur:
251 cur.execute('ANALYSE')
254 def create_search_indices(conn: Connection, config: Configuration,
255 drop: bool = False, threads: int = 1) -> None:
256 """ Create tables that have explicit partitioning.
259 # If index creation failed and left an index invalid, they need to be
260 # cleaned out first, so that the script recreates them.
261 with conn.cursor() as cur:
262 cur.execute("""SELECT relname FROM pg_class, pg_index
263 WHERE pg_index.indisvalid = false
264 AND pg_index.indexrelid = pg_class.oid""")
265 bad_indices = [row[0] for row in list(cur)]
266 for idx in bad_indices:
267 LOG.info("Drop invalid index %s.", idx)
268 cur.execute(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
271 sql = SQLPreprocessor(conn, config)
273 sql.run_parallel_sql_file(config.get_libpq_dsn(),
274 'indices.sql', min(8, threads), drop=drop)