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
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')
105 _require_version('PostGIS',
106 postgis_version_tuple(conn),
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 true FROM place LIMIT 1')
139 if cur.rowcount == 0:
140 raise UsageError('No data imported by osm2pgsql.')
143 drop_tables(conn, '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 table_exists(conn, '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 async def load_data(dsn: str, threads: int) -> None:
208 """ Copy data into the word and placex table.
210 placex_threads = max(1, threads - 1)
212 progress = asyncio.create_task(_progress_print())
214 async with QueryPool(dsn, placex_threads + 1) as pool:
215 # Copy data from place to placex in <threads - 1> chunks.
216 for imod in range(placex_threads):
217 await pool.put_query(
218 pysql.SQL("""INSERT INTO placex ({columns})
219 SELECT {columns} FROM place
220 WHERE osm_id % {total} = {mod}
221 AND NOT (class='place'
222 and (type='houses' or type='postcode'))
223 AND ST_IsValid(geometry)
224 """).format(columns=_COPY_COLUMNS,
225 total=pysql.Literal(placex_threads),
226 mod=pysql.Literal(imod)), None)
228 # Interpolations need to be copied seperately
229 await pool.put_query("""
230 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' """, None)
237 async with await psycopg.AsyncConnection.connect(dsn) as aconn:
238 await aconn.execute('ANALYSE')
241 async def _progress_print() -> None:
244 await asyncio.sleep(1)
245 except asyncio.CancelledError:
246 print('', flush=True)
248 print('.', end='', flush=True)
251 async def create_search_indices(conn: Connection, config: Configuration,
252 drop: bool = False, threads: int = 1) -> None:
253 """ Create tables that have explicit partitioning.
256 # If index creation failed and left an index invalid, they need to be
257 # cleaned out first, so that the script recreates them.
258 with conn.cursor() as cur:
259 cur.execute("""SELECT relname FROM pg_class, pg_index
260 WHERE pg_index.indisvalid = false
261 AND pg_index.indexrelid = pg_class.oid""")
262 bad_indices = [row[0] for row in list(cur)]
263 for idx in bad_indices:
264 LOG.info("Drop invalid index %s.", idx)
265 cur.execute(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
268 sql = SQLPreprocessor(conn, config)
270 await sql.run_parallel_sql_file(config.get_libpq_dsn(),
271 'indices.sql', min(8, threads), drop=drop)