1 # SPDX-License-Identifier: GPL-3.0-or-later
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2026 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
19 from psycopg import sql as pysql
21 from ..errors import UsageError
22 from ..config import Configuration
23 from ..db.connection import connect, get_pg_env, Connection, server_version_tuple, \
24 postgis_version_tuple, drop_tables, table_exists, execute_scalar
25 from ..db.sql_preprocessor import SQLPreprocessor
26 from ..db.query_pool import QueryPool
27 from .exec_utils import run_osm2pgsql
28 from ..version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
30 LOG = logging.getLogger()
33 def _require_version(module: str, actual: Tuple[int, int], expected: Tuple[int, int]) -> None:
34 """ Compares the version for the given module and raises an exception
35 if the actual version is too old.
38 LOG.fatal('Minimum supported version of %s is %d.%d. '
39 'Found version %d.%d.',
40 module, expected[0], expected[1], actual[0], actual[1])
41 raise UsageError(f'{module} is too old.')
44 def _require_loaded(extension_name: str, conn: Connection) -> None:
45 """ Check that the given extension is loaded. """
46 with conn.cursor() as cur:
47 cur.execute('SELECT * FROM pg_extension WHERE extname = %s', (extension_name, ))
49 LOG.fatal('Required module %s is not loaded.', extension_name)
50 raise UsageError(f'{extension_name} is not loaded.')
53 def check_existing_database_plugins(dsn: str) -> None:
54 """ Check that the database has the required plugins installed."""
55 with connect(dsn) as conn:
56 _require_version('PostgreSQL server',
57 server_version_tuple(conn),
58 POSTGRESQL_REQUIRED_VERSION)
59 _require_version('PostGIS',
60 postgis_version_tuple(conn),
61 POSTGIS_REQUIRED_VERSION)
62 _require_loaded('hstore', conn)
65 def setup_database_skeleton(dsn: str, rouser: Optional[str] = None) -> None:
66 """ Create a new database for Nominatim and populate it with the
69 The function fails when the database already exists or Postgresql or
70 PostGIS versions are too old.
72 Uses `createdb` to create the database.
74 If 'rouser' is given, then the function also checks that the user
75 with that given name exists.
77 Requires superuser rights by the caller.
79 proc = subprocess.run(['createdb'], env=get_pg_env(dsn), check=False)
81 if proc.returncode != 0:
82 raise UsageError('Creating new database failed.')
84 with connect(dsn) as conn:
85 _require_version('PostgreSQL server',
86 server_version_tuple(conn),
87 POSTGRESQL_REQUIRED_VERSION)
89 if rouser is not None:
90 cnt = execute_scalar(conn, 'SELECT count(*) FROM pg_user where usename = %s',
93 LOG.fatal("Web user '%s' does not exist. Create it with:\n"
94 "\n createuser %s", rouser, rouser)
95 raise UsageError('Missing read-only user.')
98 with conn.cursor() as cur:
99 cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
100 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
101 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis_raster')
102 cur.execute('CREATE EXTENSION IF NOT EXISTS ltree')
106 _require_version('PostGIS',
107 postgis_version_tuple(conn),
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 + getattr(mem, 'cached', 0)) * 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 true FROM place LIMIT 1')
140 if cur.rowcount == 0:
141 raise UsageError('No data imported by osm2pgsql.')
144 drop_tables(conn, 'planet_osm_nodes')
147 if drop and options['flatnode_file']:
148 Path(options['flatnode_file']).unlink()
151 def create_tables(conn: Connection, config: Configuration, reverse_only: bool = False) -> None:
152 """ Create the set of basic tables.
153 When `reverse_only` is True, then the main table for searching will
154 be skipped and only reverse search is possible.
156 SQLPreprocessor(conn, config).run_sql_file(conn, 'tables.sql',
157 create_reverse_only=reverse_only)
159 # reinitiate the preprocessor to get all the newly created tables
160 SQLPreprocessor(conn, config).run_sql_file(conn, 'grants.sql')
163 def create_table_triggers(conn: Connection, config: Configuration) -> None:
164 """ Create the triggers for the tables. The trigger functions must already
165 have been imported with refresh.create_functions().
167 sql = SQLPreprocessor(conn, config)
168 sql.run_sql_file(conn, 'table-triggers.sql')
171 def create_partition_tables(conn: Connection, config: Configuration) -> None:
172 """ Create tables that have explicit partitioning.
174 sql = SQLPreprocessor(conn, config)
175 sql.run_sql_file(conn, 'partition-tables.src.sql')
178 def truncate_data_tables(conn: Connection) -> None:
179 """ Truncate all data tables to prepare for a fresh load.
181 with conn.cursor() as cur:
182 cur.execute('TRUNCATE placex')
183 cur.execute('TRUNCATE place_addressline')
184 cur.execute('TRUNCATE location_area')
185 cur.execute('TRUNCATE location_area_country')
186 cur.execute('TRUNCATE location_property_tiger')
187 cur.execute('TRUNCATE location_property_osmline')
188 cur.execute('TRUNCATE location_postcodes')
189 if table_exists(conn, 'search_name'):
190 cur.execute('TRUNCATE search_name')
191 cur.execute('DROP SEQUENCE IF EXISTS seq_place')
192 cur.execute('CREATE SEQUENCE seq_place start 100000')
194 cur.execute("""SELECT tablename FROM pg_tables
195 WHERE tablename LIKE 'location_road_%'""")
197 for table in [r[0] for r in list(cur)]:
198 cur.execute(pysql.SQL('TRUNCATE {}').format(pysql.Identifier(table)))
203 _COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
204 ('osm_type', 'osm_id', 'class', 'type',
205 'name', 'admin_level', 'address',
206 'extratags', 'geometry', 'categories')))
209 async def load_data(dsn: str, threads: int) -> None:
210 """ Copy data into the word and placex table.
212 placex_threads = max(1, threads - 1)
214 progress = asyncio.create_task(_progress_print())
216 async with QueryPool(dsn, placex_threads + 1) as pool:
217 # Copy data from place to placex in <threads - 1> chunks.
218 for imod in range(placex_threads):
219 await pool.put_query(
220 pysql.SQL("""INSERT INTO placex ({columns})
221 SELECT {columns} FROM place
222 WHERE osm_id % {total} = {mod}
223 """).format(columns=_COPY_COLUMNS,
224 total=pysql.Literal(placex_threads),
225 mod=pysql.Literal(imod)), None)
227 # Interpolations need to be copied separately
228 await pool.put_query("""
229 INSERT INTO location_property_osmline (osm_id, type, address, linegeo)
230 SELECT osm_id, type, address, geometry
231 FROM place_interpolation
236 async with await psycopg.AsyncConnection.connect(dsn) as aconn:
237 await aconn.execute('ANALYSE')
240 async def _progress_print() -> None:
243 await asyncio.sleep(1)
244 except asyncio.CancelledError:
245 print('', flush=True)
247 print('.', end='', flush=True)
250 async def create_search_indices(conn: Connection, config: Configuration,
251 drop: bool = False, threads: int = 1) -> None:
252 """ Create tables that have explicit partitioning.
255 # If index creation failed and left an index invalid, they need to be
256 # cleaned out first, so that the script recreates them.
257 with conn.cursor() as cur:
258 cur.execute("""SELECT relname FROM pg_class, pg_index
259 WHERE pg_index.indisvalid = false
260 AND pg_index.indexrelid = pg_class.oid""")
261 bad_indices = [row[0] for row in list(cur)]
262 for idx in bad_indices:
263 LOG.info("Drop invalid index %s.", idx)
264 cur.execute(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
267 sql = SQLPreprocessor(conn, config)
269 await sql.run_parallel_sql_file(config.get_libpq_dsn(),
270 'indices.sql', min(8, threads), drop=drop)